Datasets you can actually have today
Environmental datasets for an investigation, each one downloaded and counted. Bringing your own is better than taking one from here: check anything you find against the four requirements.
81 datasets across 13 scales46 ready for a question on the three lines14 portalsall checked between 10 and 28 September 2026, and re-checked every year
A dataset is one file of numbers that you download and open; a portal is a website you search to find datasets.
A research question here has three lines: what people do, what happens to nature, and somebody's plan.
What a dataset has to be
- 1It has to be numbers you can change, not a picture of numbers.
- 2You have to be able to say what one row is, in a sentence.
- 3It has to cover your years and your places. Both, not one.
- 4You have to be able to have it on screen in ten minutes. Today, not in two weeks.
The catalogue
Checked portalNot one dataset but a place to search for one. Free, working when checked, and useful for the issues on this page. Whatever you find there, check it against the four requirements yourself.
Switzerland2
opendata.swiss, the federal open data portal
Everything the Confederation, the cantons and the communes publish, in one searchable place, including all the Geneva water datasets above. Search by keyword and filter the format to CSV, or you will spend the morning opening map layers.
A catalogue to search; most entries offer CSV
River discharge and lake levels
Discharge, water level and water temperature at gauging stations across Switzerland, including the Rhône and the lake itself. Excellent data, but with an access problem: anything older than the recent period has to be requested, which is free but not immediate, so check before you build a question on it.
⚠️ Recent data self-service; long historical series must be ordered on a form
The Alps1
EnviDat, the Swiss environmental research data portal
Snow, forest, avalanche and landscape datasets published by WSL and SLF researchers. Research-grade rather than official statistics, so read the documentation first, and the reward is data nobody else in your class will have.
A catalogue of research datasets; formats vary
France3
Hub'Eau, France's water data as web addresses
Thirteen APIs for the whole of France: river chemistry (289 million analyses at 25,613 stations), river biology, flow, water temperature, fish, and pesticide sales. The Haute-Savoie cards show how to build the address; change the département code to study your own region.
Free web addresses that return CSV directly, no account; each address gives at most 20,000 rows
data.gouv.fr, the French open data portal
The French equivalent of opendata.swiss, and the place to find départemental and communal data for the areas across the border. Useful when a Geneva question needs its French half.
A catalogue to search; most entries offer CSV
Air quality across the border, in Haute-Savoie and Ain
The French half of a Geneva air question: particulates, nitrogen dioxide and ozone for the départements the canton shares a border with. Filter by format first, because a fair share of the catalogue is modelled annual map layers with no table behind them, and what you want is the measurement series.
A catalogue of 186 datasets; the useful ones carry CSV and GeoJSON
Europe2
European air quality, station by station
Measured concentrations from monitoring stations across Europe, with the historical archive alongside recent years. Comparing cities means thinking hard about what else differs between them, which is exactly the controlling work sprint 4 asks for.
A download service; choose country, pollutant and years, get CSV
The EEA data hub
Bathing water quality, land cover, emissions, waste and protected areas, compiled to a common standard across member states. It gives you the same definitions in every country, which most global datasets cannot promise.
A catalogue of European environmental datasets
United Kingdom1
Water quality sampling across England
Laboratory results from samples taken at sampling points across England since 2000, tens of millions of them, which makes this the nearest thing here to the counted Geneva river files. Two catches: older guides link to environment.data.gov.uk/water-quality/view/..., which now returns a 404, and the area has to be narrowed before the file is small enough for a spreadsheet.
⚠️ Choose an area, a type and a year; the whole-of-England file is too big to open
United States1
US commercial fish and shellfish landings, by state and species
Pounds and dollars landed each year, for every coastal state and hundreds of species, from Maine lobster to Alaska pollock. The Gulf shrimp are counted on their own card; for anything else, check the species name and the Confidentiality column before you rely on it.
Query form with CSV download, no account; 1950 to 2024
Global4
Our World in Data
Energy, emissions, food, land use and biodiversity, compiled from original sources into long comparable series that you cite alongside the original. Add .csv to any chart's URL and the whole series downloads, so your extraction is one line long. But check the entity column before you average anything: World, Africa and High-income countries sit in it beside the countries.
Direct CSV for any chart, no account
World Bank Open Data
Environmental and socio-economic indicators for every country; GDP per person, protected land, rainfall and farming's share of water are counted on their own cards. One indicator is one file, countries down and years across: reshape it to one row per country-year, and take out World, the EU and the income groups.
Direct CSV or Excel per indicator, all countries and years
FAOSTAT
Food production, livestock, fertiliser and pesticide use, and agricultural emissions, country by country and year by year. Fertiliser application per hectare against water quality is a classic ESS pairing that this file makes possible.
Bulk downloads and a query builder, both giving CSV
GBIF, the Global Biodiversity Information Facility
Occurrence records with coordinates and dates for almost anything alive, pooled from museums, surveys and citizen science: about 418,000 for the European beaver, 167,000 of them in Switzerland alone, and 27,000 for the sea otter. Records show where people looked rather than where a species is, so abundance needs care, and every download carries its own DOI to cite.
Filtered downloads with a citable DOI for every one; a free GBIF account is needed to download
What the badges mean
Opened and countedEvery dataset here. Downloaded and taken apart. The card says what one row is, how many rows are really in the file, and what went wrong when somebody opened it.
Water pollution12
Nutrients and major ions in Geneva's rivers, 1969 onwards
Nutrient enrichment of the watercourses draining into Lake Geneva, across the whole period in which phosphate left detergents, treatment plants were built and farming practice changed.
- One row:
- One monitoring station in one year: the 90th percentile of that year's samples for about twenty parameters, plus the number of samples behind them.
- Covers:
- 198 stations, 1969 to 2025. The longest series in this family by twenty-six years.
Full record7 traps
- Usable rows
- 950
- Coverage
- 198 stations, 1969 to 2025. The longest series in this family by twenty-six years.
- How it arrives
- CSV inside a zip, semicolon-separated
- Language
- The page is in English; the file itself is in French
- Natural· Diatom index (DI-CH)· by station code and year
- Natural· Invertebrate index (IBGN)· by station code and year
- Natural· Dissolved oxygenagainstHuman· Phosphate
- Natural· Dissolved oxygenagainstHuman· Total phosphorus
- Natural· Dissolved oxygenagainstHuman· Nitrate
- Natural· Dissolved oxygenagainstHuman· Ammonium
- Natural· Dissolved oxygenagainstHuman· Biochemical oxygen demand
- Natural· Dissolved oxygenagainstHuman· Chloride
- 1Missing values are recorded as −99, not as blanks, and nothing in the file says so. Twenty parameters carry them. Average silica without removing them and you get a large negative concentration, which is impossible, and a spreadsheet will calculate it without any warning.
- 2Silica is −99 in 856 of 950 rows, so that column is 90% missing rather than 90% present. Check how much of a column is real before you build a question on it.
- 31973, 1989 and 1990 are absent from the series entirely.
- 4The network grew from 6 stations in 1969 to about 30 now, so early and late years describe very different sets of places. Any before-and-after comparison needs restricting to the stations present in both.
- 5Station names are not unique. Twenty-three different stations are called "Embouchure", because every river has a mouth. Group by CODEMESURE, never by name, or you will silently merge twenty rivers into one.
- 6There are no unit columns here, unlike the physico-chemistry file. The units are in the documentation, and you have to go and read it.
- 7Every value is the year's 90th percentile, not its mean: nine samples in ten were at or below it, so it reports a station's bad days. The documentation in the zip says so for every parameter. Describe your variable that way, and don't average it with a file that gives means.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has phosphate (each year's 90th percentile) at Geneva's river monitoring stations changed between 1969 and 2025?
- A comparison of nitrate between stations on predominantly agricultural catchments and predominantly urban ones.
- Which parameter most often downgrades a station's grade, and has that changed over fifty years?
- Phosphate and total phosphorus, nitrate, nitrite, ammonium, total nitrogen
- Dissolved oxygen and saturation, biochemical oxygen demand, dissolved organic carbon
- Conductivity, pH, alkalinity
- Chloride, sulphate, sodium, potassium, calcium, magnesium, silica
- DECLASSANT: which parameter dragged that station-year's grade down
Do Geneva's river stations with more phosphate have worse diatom index scores?
Watch outThe diatom index runs backwards: 1.4 is the cleanest water in the file and 8.0 the worst, so a positive correlation with phosphate means the diatoms get worse as phosphate rises. Say so in your write-up or your reader will take it the wrong way round. 420 station-years appear in both files; match on CODEMESURE and year, and remove the −99s.
Do Geneva's river stations with more phosphate have lower invertebrate index scores?
Watch outHere a higher index is better water, the opposite of the diatom index. The invertebrate series stops in 2022, so the nutrient file's last three years have nothing to match; 408 station-years appear in both. Remove the −99s from the nutrient file first.
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileThe contract names assainissement, and nutrients are what wastewater treatment removes. Its years fall well inside this series, which is less common than it sounds.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileThe most direct link of any file in this family. Sewer separation and treatment capacity are what nutrient concentrations respond to, and this series is long enough to see the response happen.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileRenaturation changes how a channel holds and processes nutrients rather than how much arrives. The connection is real but indirect, so you have to argue it, not just state it.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceNitrate carried down the Mississippi to the Gulf of Mexico, every year from 1979 to 2025
Fertiliser and manure from farms across the Mississippi basin wash into the rivers as nitrate. The Mississippi and the Atchafalaya carry it to the Gulf, where it feeds the algae whose decay takes the oxygen from the seabed each summer.
- One row:
- One site, one substance, one water year and one kind of estimate: the tons the river carried. For this card, the row where SITE_ABB is GULF, CONSTIT is NO3_NO2 and MODTYPE is WRTDS_K: the nitrate that the Mississippi and the Atchafalaya together carried to the Gulf from October to September.
- Covers:
- 47 water years, 1979 to 2025, for the Mississippi near St. Francisville, the Atchafalaya at Melville, and the two added together
Full record12 traps, 2 on the three lines
- Usable rows
- 47
- Coverage
- 47 water years, 1979 to 2025, for the Mississippi near St. Francisville, the Atchafalaya at Melville, and the two added together
- How it arrives
- One CSV of about 290 KB, WRTDS_loads.csv, one of six files on the USGS data release page. 14 lines of notes come before the headings
- Language
- Page and file in English
- Natural· Dead zone area· by year: the USGS water year (October to September) against the July survey of the same year
- Natural· Brown shrimp landed in Louisiana· by year: the calendar year of the landings against the USGS water year (October to September) of the same number
- 1Many rows for every year. The file has 4,041 rows: three sites, ten substances and three kinds of estimate for every water year. Filter to SITE_ABB GULF, CONSTIT NO3_NO2 and MODTYPE WRTDS_K, and you have 47 rows, one per year. If you add up rows without filtering, you count each year many times.
- 2Fourteen lines of notes first. The file starts with 14 lines beginning with #, which explain every column. The headings are on line 15. A spreadsheet reads the notes as data, so delete them, or tell the import to start at line 15. Excel also drops the zeros at the start of the site numbers (07373420 becomes 7373420).
- 3A water year is not a calendar year. WY 2019 runs from 1 October 2018 to 30 September 2019. So the load you set against the July 2019 dead zone includes October to December 2018, before the spring, and August and September 2019, after the survey. From 1979 to 2025, about 17% of a water year's river flow came in October to December, and about 10% in August and September (from mflow.csv). Say in your method that the two years are not the same length of time.
- 4The forecasters use May. Each June, NOAA forecasts the dead zone from the May nitrate load, the river's load in the weeks just before the summer. USGS publishes the May figure only one year at a time, as a preliminary spreadsheet on its Gulf page, and it is not in this release. A yearly total is what this file can give you. Say so in your evaluation, and say why May might matter more.
- 5Wet years carry more nitrate. More rain washes more nitrate off the fields and down the rivers, so the load goes up and down with the flow. 2019 had both the highest mean flow in the file (1,380,000 cubic feet per second, in aflow.csv) and the largest load (1,650,000 tons). USGS names the fresh water and the nutrients the rivers bring as "two primary controls" on the size of the dead zone. So a link with the dead zone may come from the river's water as well as from the farms' nitrate. Look at the flow too, and say what you think.
- 6Actual or flow-normalised. WRTDS_K is what actually reached the Gulf. The flow-normalised rows (WRTDS_FNG and WRTDS_FNS) take out the wet and dry years. That is right for a trend and for the Task Force's target, but it is not what the algae received. Choose one before you look, and say why.
- 7US tons of nitrogen. TONS is in US tons (907 kg), not metric tonnes, and it is the weight of the nitrogen in the nitrate, not of the nitrate itself. NOAA's announcements give the May load in metric tons, and 1 metric ton is 1.1 US tons. Convert once, and say which you used.
- 8Use the GULF row. Loads are rounded to three figures (984,000; 1,650,000), so STFR plus MELV can differ from GULF by up to half a percent. Take GULF as it is, and do not add the two yourself.
- 9The whole series is estimated again every year. Each May, USGS publishes a new version, with a new DOI and a new download link, and runs the model again. Earlier years change: between version 2 (February 2025) and this version 3.0 (May 2026), 37 of the 45 years they share changed, and 1989 fell by 15%. Name the version and the date you downloaded it, and keep your copy.
- 10Search for the Gulf of America. The release's title says Gulf of America, the name US federal agencies have used since 2025. Its first version, from 2022, was called Gulf of Mexico. Say in your method that the two names mean the same sea.
- 11The download page checks your browser. sciencebase.gov shows a "Just a moment" page for a few seconds before the file downloads. A normal browser passes it on its own, but a script, or a strict school network, may be stopped.
- 12The dead zone stops in 2021. Its file, on the EPA card, runs from 1985 to 2021, with no survey in 1989 or 2016, so 35 years have both. Later summers are in NOAA's yearly announcements, each a separate source to cite. WY 2026 will come in next May's version.
- Across the summers from 1985 to 2021, is the area of the Gulf dead zone related to the nitrate load that the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
- Was the Gulf dead zone larger in the years when the rivers carried more nitrate than the median than in the years when they carried less, 1985 to 2021?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the flow-normalised nitrate load to the Gulf changed from 1979 to 2025?
- What share of the nitrate reaching the Gulf comes down the Atchafalaya, and has that share changed since 1979?
- Did the flow-normalised total nitrogen and total phosphorus loads to the Gulf reach the Hypoxia Task Force's interim target of 20% below their 1980 to 1996 average by 2025?
- Is the nitrate load related to the rivers' mean flow in the same water year, 1979 to 2025?
- WY: the water year, 1 October to 30 September, named after the year it ends in. WY 2019 is October 2018 to September 2019
- SITE_ABB: STFR (Mississippi River near St. Francisville), MELV (Atchafalaya River at Melville), and GULF, the two added together, which is what reaches the Gulf
- CONSTIT: what the river carried. NO3_NO2 is nitrate plus nitrite, weighed as nitrogen. TN is total nitrogen and TP total phosphorus, among ten substances
- MODTYPE: WRTDS_K is the load that actually arrived that year. WRTDS_FNG and WRTDS_FNS are flow-normalised: the load as if every year had the same river flow, which is how the Hypoxia Task Force follows change
- TONS: the load, in US tons of nitrogen. TONS_L90 and TONS_U90 are the 90% confidence limits, and read NA on every WRTDS_K row
- In the same release, aflow.csv and mflow.csv give the rivers' mean flow for each water year and each month, in cubic feet per second, and WRTDS_trends.csv gives the Task Force's percentage changes in total nitrogen and phosphorus since 1980 to 1996
Across the summers from 1985 to 2021, is the area of the Gulf dead zone related to the nitrate load that the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
Watch outMatch the USGS WY to the EPA Year: 35 years have both, 1985 to 2021. A water year runs from October to September, so each load includes three months of the autumn before and two months after the July survey. The scientists who forecast the dead zone use the May load, which is not in the USGS file: say so in your evaluation. Wet years carry more nitrate and more fresh water, and USGS names both as controls on the dead zone, so look at the flow in aflow.csv as well. A storm can mix oxygen down just before the survey, as Hurricane Barry did in 2019: keep those years, and say what they do. Both series may also drift over the years, so correlate the change from one year to the next as well. The missing summers leave 32 changes.
Across the years 1979 to 2024, is the weight of brown shrimp landed in Louisiana related to the nitrate the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
Watch outMatch the landings Year to the USGS WY: 46 years have both, 1979 to 2024. A landing year runs from January to December and a water year from October to September, so they share nine months. Nitrate does not act on shrimp directly: any link runs through the dead zone, and the same river water also feeds the plankton and the seabed life that shrimp eat, so more nitrate could mean more food as well as less oxygen. Landings also follow the boats, fuel and the price of imported shrimp, the storms of 2005, 2008, 2020 and 2021 and the oil spill closures of 2010, and neither file can separate them. Scientists who studied this found the effect hard to see in landings. Much of Louisiana's brown shrimp is caught in the bays in May and June. Both series may drift over the decades, so correlate the change from one year to the next as well: that gives 45 changes.
The Mississippi River/Gulf of America Watershed Nutrient Task Force: federal agencies led by the EPA, one agency from each of 12 states (ten along the Mississippi and two along the Ohio), and the tribes, through the National Tribal Water Council. In 2026 the EPA and Iowa's Secretary of Agriculture are its co-chairs.
Link to this fileDirect. The interim target is measured on loads from this release: the Task Force compares USGS's flow-normalised total nitrogen and total phosphorus loads with their 1980 to 1996 average. Nitrate is most of the nitrogen (between 50% and 74% of total nitrogen at the Gulf in this file), but the target is set on total nitrogen, so say which you use and why. The plan acts on farms and towns upstream, while the load in any one year also rises and falls with the rain, which no plan controls.
The tensionVoluntary action or legal limits. In 2008 thirteen groups petitioned the EPA to set legal limits for nitrogen and phosphorus in the basin's waters. The groups, the Mississippi River Collaborative, said the petition "chronicled EPA’s failed attempts to address the problem through workgroups, collaboration, guidance to states, and 'action plans' that never resulted in any action" (MRC). The EPA said no on 29 July 2011: "We do not believe that the comprehensive use of federal rulemaking authority is the most effective or practical means of addressing these concerns at this time" (EPA letter). A federal court upheld the EPA's refusal in 2016.
SourceEach Task Force state wrote its own: Arkansas, Illinois, Indiana, Iowa, Kentucky, Louisiana, Minnesota, Mississippi, Missouri, Ohio, Tennessee and Wisconsin. Iowa's was written by the Iowa Department of Agriculture and Land Stewardship, the Iowa Department of Natural Resources and Iowa State University.
Link to this fileUpstream, and only all together. The strategies act on farms and sewage works in each state, but this file measures what reaches the river mouth, with every state's nitrate added together. It can show how the total changed, not which state's strategy changed it.
The tensionVoluntary or required, for farms. Iowa's strategy says "Iowa’s many successes can be duplicated using the tools known to work, such as targeted, voluntary conservation measures" (2025 update). The Iowa Environmental Council replied in 2022: "The NRS includes required actions for some sources of nutrient pollution, but only voluntary measures for agriculture. This voluntary policy has not resulted in progress" (report). Des Moines Water Works, which supplies about 500,000 people, sued 13 drainage districts upstream over nitrate in its river water. A federal court dismissed the case on 17 March 2017, holding that "drainage districts lack the ability to redress that injury" (Penn State).
SourceNutrients and weedkiller in Haute-Savoie's rivers, from Hub'Eau
Nutrients from sewage and farms, and weedkiller from farms, gardens and roads, in the rivers on the French side of the Geneva region: the Arve, the rivers that run into Lake Geneva, and the rivers round Lake Annecy.
- One row:
- One measurement of one substance in one water sample, taken at one station on one date. A station usually has about four samples a year.
- Covers:
- Haute-Savoie. Orthophosphate at 598 stations and nitrate at 612, 1971 to 2026, but never more than 44 stations a year before 2002. Glyphosate and AMPA at 109 stations, 2001 to 2026
Full record9 traps
- Usable rows
- 10,171
- Coverage
- Haute-Savoie. Orthophosphate at 598 stations and nitrate at 612, 1971 to 2026, but never more than 44 stations a year before 2002. Glyphosate and AMPA at 109 stations, 2001 to 2026
- How it arrives
- CSV, semicolon-separated. Open the web address and the file downloads, about 1 MB, always called analyse-pc.csv. One address for each substance
- Language
- Page and file in French
- Natural· Invertebrate index (I2M2), Haute-Savoie· by station code and year
- 1A value below the limit is written as the limit. When code_remarque is 10, resultat holds the limit of quantification (0.02, say), not a measured value. For AMPA that is 1,286 of 2,211 rows, and for glyphosate 1,833 of 2,208. Average resultat as it downloads and every "less than 0.02" counts as 0.02. Choose before you look: the share of samples above the limit, the mean with each value below the limit set to half the limit, or the highest value of the year. Then say which, and why.
- 2The limit changed. For AMPA it was 0.1 or 0.05 µg/l up to 2013 and 0.02 from 2014. A lower limit finds more, so years with different limits cannot be compared as they are. Start AMPA in 2014, or count only values above 0.05 in every year.
- 3AMPA does not only come from glyphosate. It also forms when phosphonates break down, and a review of AMPA in natural waters found that in towns it comes "from phosphonates and glyphosate in wastewater" (Grandcoin, Piel and Baurès, Water Research, 2017). So AMPA below a treatment works is not a measure of weedkiller alone. Glyphosate is, and it is in the same file.
- 4Keep size=20000 in the address. Without it you get the first 1,000 rows and no warning. With it, one address gives at most 20,000 rows, and a file cut at that limit looks exactly like a complete one. Each address on this card is well under it. If you change an address and get exactly 20,000 rows, ask for fewer years.
- 5One row per sample, about four samples a year. Where a station also has an invertebrate score, the middle number of samples in a year is 4, and some years have 36. Make one value per station and year before you match the biology, which has one score a year. First remove rows that are exact copies of another row: there are 197 in the orthophosphate file.
- 6Look at the highest values before anything else. The two highest orthophosphate values, 26.0 and 25.0 mg/l on the Arve on 4 December 2012, are marked Incorrecte: remove them. 43 nitrate values are above 100 mg/l, all at small streams round Annecy and Cluses, and the highest is 1,290. Those are not marked. Decide what to do with them before you look at any pattern, and say so.
- 7The units are the whole ion. Geneva's files give nitrate as mg N/l and phosphate as mg P/l. To compare with them, multiply nitrate by 0.226 and orthophosphate by 0.326.
- 8Open it through the import command. The file is separated by semicolons and has no marker to say it is UTF-8, so a double-click can put every column into one and turn é into é. In Excel use Data, then From Text/CSV, and choose UTF-8 and semicolon. Every download is called analyse-pc.csv: rename each one as it arrives.
- 9The file grows. Hub'Eau copies new results from Naïades all the time, so the same address can give more rows next month. Write down the date and the number of rows you got. On 27 September 2026 the orthophosphate file had 10,171.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has orthophosphate at the Arve's monitoring stations changed between 2002 and 2025?
- In what share of samples was AMPA found above the limit at Haute-Savoie's river stations each year, 2014 to 2025?
- Is nitrate higher at the Arve's stations below Cluses than at Chamonix?
- code_station: the station's national (SANDRE) code, such as 06063900 for the Arve at Arthaz-Pont-Notre-Dame. libelle_station is its name
- date_prelevement: the day the sample was taken, written 2023-07-18
- libelle_parametre: what was measured: Orthophosphates (PO4), Nitrates, Ammonium, Glyphosate or AMPA. AMPA is a substance glyphosate breaks down into
- resultat and symbole_unite: the value and its unit. Orthophosphate is in mg(PO4)/L and nitrate in mg(NO3)/L, milligrams of the whole ion in a litre. Glyphosate and AMPA are in µg/L, micrograms in a litre
- code_remarque: 1 means the value was measured. 10 means it was below the limit of quantification, the smallest amount the laboratory can measure, and then resultat is that limit, not a measurement
- limite_quantification: that limit, when the laboratory gave it
- libelle_qualification: the water agency's check on the value: Correcte, Incertaine (uncertain) or Incorrecte (wrong)
One address for each substance, Haute-Savoie (département 74) (5)
Do Haute-Savoie's river stations with more orthophosphate have lower invertebrate index (I2M2) scores, 2008 to 2024?
Watch outMatch code_station in the chemistry to code_station_hydrobio in the biology: they are the same national codes. Take the year from the first four characters of each date, because the biology writes 2025-06-25T00:00:00Z. The chemistry has about four samples a year at a station and the biology one score, so make one chemistry value per station and year first, and when code_remarque is 10 remember that resultat is the limit, not a measurement. 913 station-years have both an I2M2 score and an orthophosphate value, from 191 stations, and 25 of those stations appear in ten years or more: average each station's years into one value as well, or the stations surveyed most often count most. With the diatom index it is 1,059 station-years from 2005, and with AMPA 341. For both indices a higher score is better water.
The préfet coordonnateur du bassin Rhône-Méditerranée, for the French State, under the EU Nitrates Directive (91/676/EEC, 12 December 1991). In a vulnerable zone, farmers must follow rules on how they use nitrogen fertiliser and how they manage their land.
Link to this fileDirect for nitrate, and it has a place and a date. The file has a station on each of the three rivers, each with nitrate in 13 to 20 years since 2002: the Aire at Saint-Julien-en-Genevois 2 (06999125), the Vion at Excenevex 1 (06830246) and the Hermance at Chens-sur-Léman (06830258). Compare them with stations outside the zones, or compare years before and after 2021. Only a few years have passed since 2021, so you can describe where nitrate stood, not show that the zones worked. The same report judged that the Aire's nitrate comes mainly from sewage, and says a new treatment works at Neydens should be fully working by 2027.
The tensionFarms or sewage. In the 2026 review, requests to take the Vion and the Hermance off the list argued that their pollution may come from sewage as well as from farms. The State kept both. For the Vion its report answers that "la contribution agricole ne peut être ignorée" (the part from farming cannot be ignored), because there are large cereal fields right beside the monitoring point (DREAL Auvergne-Rhône-Alpes, July 2026).
SourceThe comité de bassin Rhône-Méditerranée, which adopted it on 18 March 2022, and the préfet coordonnateur de bassin, who approved it on 21 March 2022. It is how France applies the Water Framework Directive (2000/60/EC, 23 October 2000) to the Rhône basin, which includes every river in Haute-Savoie.
Link to this fileDirect in aim. Its orientation 5B, "Lutter contre l'eutrophisation des milieux aquatiques", asks for less phosphorus to reach rivers and lakes, and less nitrogen, mainly for lagoons. Orientation 5D is about pesticides. The plan covers the whole basin, so connect it to your stations through what it asks of your rivers, not through a basin total.
The tensionThe plan names something it cannot control. On pesticides it says it has no hold over "la compétitivité du prix des produits agricoles", the price competition that leads some farmers to use inputs such as pesticides "pour améliorer les rendements et sécuriser leurs revenus" (to raise their yields and protect their income).
SourceThe Commission locale de l'eau (CLE) of the Arve, with the SM3A, the Syndicat mixte d'aménagement de l'Arve et de ses affluents, as its technical body. The préfet of Haute-Savoie approved it.
Link to this fileDirect for orthophosphate and ammonium on the Arve and its tributaries, because sewage carries both. The plan records that 86% of the basin's people are on mains sewers, with 39 treatment works in 2012. Of the homes with their own systems that were inspected by 2012, 40% did not comply and another 26% did not comply with a proven risk. It calls pesticide discharges in the basin "globalement limités" (limited overall), found mostly on the crop and market-garden land of the lower valley.
The tensionTourists against the river. The plan says treatment works must keep up with fast population growth and treat tourists' wastewater "en période de pointe hivernale, période de forte sensibilité des milieux récepteurs": most visitors come in winter, when the rivers that receive the treated water are at their lowest.
SourceThe French Parliament: loi n° 2014-110 of 6 February 2014, with its dates brought forward by the energy transition law (loi n° 2015-992 of 17 August 2015, article 68).
Link to this fileDirect for glyphosate in towns, and not for farms, which it does not cover. Glyphosate and AMPA run from 2001 in this file, so there are years before and after 2017 and 2019. But the limit of quantification changed in 2014, and AMPA has other sources (see the traps), so a change below a town is not only a change in weedkiller.
SourceThe French government, led by the ministry of agriculture.
Link to this fileIndirect. It is about how much pesticide French farms use, not about any river, and it sets no target for Haute-Savoie or for glyphosate. Any link to your stations has to go through the farmland upstream of them, and you would have to show it.
The tensionWhether it worked, and how to measure it. In 2019 the Cour des comptes, France's public audit court, found that use measured by the NODU indicator had risen by 12% between 2009 and 2016, and wrote that "les effets des plans Écophyto demeurent très en deçà des objectifs fixés" (the effects remain far below the targets). In early 2024, after protests by farmers, the government paused the plan, then replaced the NODU with the EU's HRI1 indicator. Members of the plan's own scientific committee wrote that the new indicator's weightings "sont arbitraires et ne sont étayées par aucun résultat scientifique" (are arbitrary and backed by no scientific result) (The Conversation, 21 February 2024).
SourceForever chemicals (PFAS) in Haute-Savoie's rivers and the Rhône at Lyon, from Hub'Eau
PFAS, the forever chemicals that do not break down, in the rivers on the French side of the Geneva region, including the rivers at Rumilly, where a Tefal factory stands, and in the Rhône below the two chemical plants at Pierre-Bénite, just south of Lyon.
- One row:
- One measurement of one PFAS in one water sample, taken at one station on one date. Most stations are sampled 4 or 12 times a year.
- Covers:
- Haute-Savoie. PFOS at 80 stations and PFOA and PFHxA at 34, 2014 to 2026, with 10 to 41 stations a year. The same address for the Rhône département covers Lyon and the Rhône below it, 2011 to 2026
Full record10 traps, 2 on the three lines
- Usable rows
- 6,423
- Coverage
- Haute-Savoie. PFOS at 80 stations and PFOA and PFHxA at 34, 2014 to 2026, with 10 to 41 stations a year. The same address for the Rhône département covers Lyon and the Rhône below it, 2011 to 2026
- How it arrives
- CSV, semicolon-separated. Open the web address and the file downloads, about 1 MB, always called analyse-pc.csv. One address gives all three PFAS for one département
- Language
- Page and file in French
- Natural· Invertebrate index (I2M2), Haute-Savoie· by station code and year
- 1A value below the limit is written as the limit. When code_remarque is 10, resultat holds the limit of quantification, not a measured value. That is 2,133 of the 2,502 PFOS rows, 1,531 of the 2,163 PFOA rows and 1,853 of the 2,209 PFHxA rows. Choose before you look: the share of samples where the PFAS was found, or the mean with each value below the limit set to half the limit. Then say which, and why.
- 2The limit fell, at different times for each PFAS. PFOS was mostly 0.02 µg/l up to 2021, with some samples at 0.001 from 2018, then 0.002 or 0.001 from 2022. PFOA was 0.02 up to 2016, 0.01 in 2017, 0.002 from 2018 to 2024 and 0.001 from 2025. PFHxA was 0.01 up to 2017 and 0.002 from 2018. A lower limit finds more, so keep to years with one limit, or count as found only values above the highest limit in your years.
- 3The river standard is below every limit in the file. The EU standard for PFOS is an average of 0.00065 µg/l over a year, and the lowest limit of quantification here is 0.001. So a result below the limit cannot show that a river meets the standard. Say so in your evaluation.
- 4Remove exact copies first. 451 rows are exact copies of another row, most of them from 2014 to 2016: 6,423 rows remain. One PFHxA sample at the Fier at Motz on 1 September 2021 appears twice with the same number, once measured and once below the limit. Decide which to keep.
- 5PFOS has two codes. This address asks for 6561, the one in use. An older code, 6560, marked gelé (frozen), holds 289 more PFOS results at 7 stations from 2017 to 2022, and 216 of its station-dates are samples already in 6561. Do not add them, or you count samples twice.
- 6The stations change from year to year. PFOS was measured at 10 to 17 stations a year up to 2017, 28 in 2018 and 41 in 2023 and 2024. A yearly average across all stations compares different places each year: keep to the stations measured in every year you use.
- 7One row per sample, 4 or 12 samples a year. Of the 298 station-years in the file, 147 have 12 sampling dates and 112 have 4. Make one value per station and year before you match the biology, which has one score a year.
- 8Keep size=20000 in the address. Without it you get the first 1,000 rows and no warning. Asking for more PFAS in one address can pass the cap: the 20 PFAS of the drinking-water limit, one by one, are 31,648 rows for Haute-Savoie, and a file cut at 20,000 rows looks exactly like a complete one.
- 9Open it through the import command, and match on the code. The file is separated by semicolons and has no marker to say it is UTF-8, so in Excel use Data, then From Text/CSV, and choose UTF-8 and semicolon. 17 of the 80 station names end with a space, so match on code_station, never the name. Both addresses download as analyse-pc.csv: rename each one as it arrives.
- 10The Rhône file has traps of its own, and both files grow. In the Rhône file (département 69), 162 rows are exact copies, and 47 times one station, one date and one PFAS have two different results, 41 of them at two stations whose codes do not start with 06 (Eaux Bleues and Anse). 2026 is not complete in either file. Write down the date and the number of rows you got: on 28 September 2026 the Haute-Savoie file had 6,874.
- Do Haute-Savoie's river stations where PFOS was found have lower invertebrate index (I2M2) scores than stations where it was not, 2022 to 2024?
- Are diatom index (IBD) scores lower at Haute-Savoie's river stations where PFOA was found above the limit of quantification, 2018 to 2024?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has PFHxA in the Rhône at Chasse-sur-Rhône, below the Pierre-Bénite plants, changed between 2018 and 2025?
- Is PFOA found more often in the Chéran at Rumilly than in the Arve at Ayse, 2018 to 2024?
- In what share of samples was PFOS found at Haute-Savoie's river stations each year, 2022 to 2025?
- code_station: the station's national (SANDRE) code, such as 06071000 for the Chéran at Rumilly. libelle_station is its name
- date_prelevement: the day the sample was taken, written 2023-07-18
- libelle_parametre: which PFAS, by its French chemical name. Acide sulfonique de perfluorooctane is PFOS (code 6561), Acide perfluoro-octanoïque is PFOA (5347) and Acide perfluoro-n-hexanoïque is PFHxA (5978)
- resultat and symbole_unite: the value and its unit, always µg/L, micrograms in a litre. 0.001 µg/l is 1 nanogram (ng) in a litre
- code_remarque: 1 means the PFAS was measured. 10 means it was below the limit of quantification, the smallest amount the laboratory can measure, and then resultat is that limit, not a measurement
- limite_quantification: that limit. It fell from 0.02 or 0.01 µg/l to 0.002 or 0.001 over these years
- libelle_qualification: the water agency's check on the value. Every row in this file is Correcte
One address for PFOS, PFOA and PFHxA together (4)
Do Haute-Savoie's river stations where PFOS was found have lower invertebrate index (I2M2) scores than stations where it was not, 2022 to 2024?
Watch outMatch code_station in the PFAS file to code_station_hydrobio in the biology, and take the year from the first four characters of each date. 211 station-years have a PFOS value and an I2M2 score, from 58 stations; with the diatom index it is 211 from 62 stations, and PFOA or PFHxA meet the I2M2 in 159. The limit of quantification fell at different times for each PFAS (for PFOS mostly 0.02 µg/l before 2022, then 0.002 or 0.001; for PFOA 0.002 from 2018 to 2024), so keep to years with one limit: for PFOS that leaves 85 station-years from 48 stations, 2022 to 2024, and for PFOA 98 with the I2M2, 2018 to 2024. Most results are below the limit, so compare two groups, found and not found, rather than drawing a line through the values, and count as found only results at or above the highest limit in your years (0.002 for PFOS in 2022 to 2024). Average each station's years into one value first. Many surveys were in February (41 of the 86 in the PFOS years): compare the same season, or say why you did not. Then read what level harms river life: the EU dossier behind the PFOS standard set 0.23 µg/l, from a test on midge larvae, insects of the kind the I2M2 counts. Count how many of your values reach it. PFAS reach rivers with other pollution from towns and factories, so a difference between the groups need not be caused by PFAS. For both indices a higher score is better water.
The European Parliament and the Council of the EU. Directive 2013/39/EU of 12 August 2013 made PFOS a priority substance under the Water Framework Directive, with a standard every river has to meet. Directive (EU) 2026/805 of 30 March 2026 puts PFOS into one standard for 25 PFAS added together. France checks its rivers for PFOS under a monitoring order of 26 April 2022.
Link to this fileDirect: this is the limit the regulator sets for a river. For PFOS it is an average of 0.00065 µg/l over a year in rivers and lakes, and 36 µg/l at any one time. The new standard is 0.0044 µg/l for the 25 PFAS added together, each weighted by how harmful it is compared with PFOA: PFOS counts twice, PFHxA a hundredth. This file has 3 of the 25. Both standards are set mainly to protect people who eat fish, and no limit of quantification in this file is as low as 0.00065 (see the traps).
The tensionWater or the economy. The European Commission says the same revision "allows more flexible application of the principle of non-deterioration of water status to facilitate important economic activities" (European Commission, 11 May 2026).
SourceThe French Parliament: loi n° 2025-188, proposed by the deputy Nicolas Thierry and the Écologiste group. The ministries for the environment and for health apply it, and the money from its charge goes to the water agencies.
Link to this fileDirect for what factories release into rivers, through the charge and the path to zero, and indirect for the product bans, which act on what people buy. Its measures start in 2025 and 2026, at the end of this file, so you can describe the years before it, not show whether it worked.
The tensionJobs and imports, or health. The first text also banned PFAS in cookware, and the Assemblée nationale took that out in April 2024. In the Senate's committee, the rapporteur Bernard Pillefer feared that a national ban would end "par des délocalisations de productions ou par un accroissement des importations" (in production moving abroad or in more imports). Jacques Fernique reported the message that many scientists repeated in the hearings: "Il faut fermer le robinet" (we have to turn off the tap) (Senate report, 22 May 2024).
SourceThe préfète of Haute-Savoie, with the DREAL (the regional environment office, which inspects factories), the regional health agency (ARS) and the community of communes Rumilly Terre de Savoie, which runs the drinking water.
Link to this fileDirect in place, and it has dates. The file has the Chéran at Rumilly 1 from 2016, the Néphaz and the Dadon at Rumilly from 2021, and the Fier at Motz from 2014. But most of the measures are about groundwater, drinking water and air, not the rivers. The DREAL names five possible sources, not one, and says Tefal stopped using PFOA in 2012. So a change in these rivers would not show that any one measure worked.
SourceThe préfet of the Auvergne-Rhône-Alpes region, with the DREAL, which inspects the two plants on the Rhône just south of Lyon. Both make fluorinated plastics (fluoropolymers) and handle PFAS, and Arkema's site is classed Seveso high threshold.
Link to this fileDirect for PFHxA in the Rhône, in the second address (département 69). PFHxA is the PFAS Daikin used. The Rhône at Chasse-sur-Rhône, below the plants, has PFHxA in every year from 2011 to 2026, the Rhône at Vernaison in 2014, 2015, 2018, 2021 and 2024, and the Rhône at Lyon, above them, only in 2018, 2021 and 2024. Arkema's PFAS, 6:2 FTS, is not in this file. Daikin stopped at the end of 2025, so 2026, which is not complete, is the only year after.
The tensionWhether the State did enough. In the Assemblée nationale's committee, Anne Stambach-Terrenoir said that "Le plan PFAS du Gouvernement n’impose aucune contrainte aux industriels" (the Government's PFAS plan puts no constraint on industry). Jean-Luc Fugit said that Arkema had cut its discharges by 80%, and called it "un premier pas intéressant, mais il faudra atteindre dès que possible les 100 %" (an interesting first step, but it must reach 100% as soon as possible) (Assemblée nationale report, 27 March 2024).
SourceThe European Commission, under the EU's chemicals law (REACH): Commission Regulation (EU) 2024/2462 of 19 September 2024.
Link to this fileDirect for one column, because PFHxA is in this file. But the first date is in 2026, so the file shows the years before the restriction, not after it. It acts on products, so any link to a river runs through what those products release.
SourceDenmark, Germany, the Netherlands, Norway and Sweden proposed it under REACH. The European Chemicals Agency (ECHA) and its two scientific committees assess it. Then the European Commission drafts a law and the member states decide.
Link to this fileIndirect. It is not law yet, and it would act on products and uses across Europe, not on any river in this file. Use it to discuss what could change after the years in this file.
The tensionIndustry against campaigners. Cefic, the European chemical industry council, said the proposal "will eliminate key applications in batteries, semi-conductors, electric vehicles and renewable energy production" (Cefic, 9 February 2023). ChemSec, a campaign group, called anything less than a ban on all uses "a concession to industry pressure" (ChemSec, 10 July 2025).
SourceE. coli in Geneva's rivers and at the lake shore
Faecal contamination of the watercourses draining into Lake Geneva, and whether thirty years of investment in wastewater infrastructure shows up in the water.
- One row:
- One monitoring station in one year: the mean E. coli count in CFU/ml, and the number of sampling campaigns that mean was calculated from.
- Covers:
- 184 stations across 78 watercourses plus one lake bathing site, 1995 to 2026
Full record8 traps
- Usable rows
- 757
- Coverage
- 184 stations across 78 watercourses plus one lake bathing site, 1995 to 2026
- How it arrives
- CSV inside a zip, semicolon-separated, with a documentation folder alongside it
- Language
- Page and file in French
- Natural· Invertebrate index (IBGN)Natural· Number of invertebrate taxa· by station code and year
- 1The file has 37,009 rows and 757 observations. Every station-year is repeated about fifty times, identical apart from the object id. Run a test on the file as downloaded and your sample size is wrong by a factor of fifty, so everything comes out significant.
- 22007 is missing from the dataset entirely.
- 3The number of campaigns behind each annual mean ranges from 1 to 12. A mean from a single visit is not comparable with a mean from twelve, so filter before you compare.
- 4The current year is present but incomplete, with a single campaign behind it.
- 5The station roster grows over time, from 19 stations in 1995 to around 30 now. Comparing early years with late years compares different sets of places unless you restrict to the stations present in both.
- 6One value sits at exactly 1000, which looks like the highest value the lab reports (a ceiling), not a real measurement.
- 7Zeros and values of 0.8 recur, which suggests a detection limit rather than genuinely no bacteria.
- 8The only lake bathing station, Pâquis, reads zero in 24 of its 28 years. The lake itself is clean. The pollution is in the rivers flowing into it (the tributaries), so a study of the lake shore has nothing to analyse.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has mean annual E. coli at Geneva's river monitoring stations changed between 1995 and 2024?
- A comparison of E. coli between stations on predominantly agricultural catchments and predominantly urban ones.
- Do stations downstream of a wastewater treatment plant differ from stations upstream of one?
- E. coli, CFU/ml (annual mean)
- Sanitary class, five ordered categories from Très bon to Mauvais
- Number of sampling campaigns behind that year's mean
- Station name and code, watercourse, year
- Swiss coordinates, so it maps
Do Geneva's river stations with more E. coli have lower invertebrate scores, or fewer kinds of invertebrates, 1995 to 2022?
Watch out387 station-years appear in both files, but they come from only 142 stations: most stations were surveyed in two or three of those years. Average each station's years into one value, or you count the same station several times over. Remove the repeated rows from the E. coli file first (each station-year appears about fifty times) and match on CODEMESURE and year, never the station name. A higher invertebrate score is cleaner water, so more E. coli with fewer invertebrates is a negative correlation. Choose the score (IBGN_MOYENNE) or the number of taxa (TAXON_SOMME) before you look. 39 station-years have E. coli of 0, probably too few to detect rather than none, and one reads 1,000, far above the rest: decide what to do with both and say so. The invertebrates stop in 2022.
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileNames improvement of water quality through assainissement among its stated objectives, and covers 160 km² and nine watercourses. Three of them, the Aire, the Drize and the Laire, carry monitoring stations in this dataset. Of the three, it fits a question about bacteria most closely, and you can compare the years before and after it.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileThe most direct of the three. Separating foul water from surface water is the mechanism by which sewage stops reaching a river when it rains, which is exactly what an E. coli series is measuring.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileIt aims at the shape of the river channel (its morphology) and how it floods, not at bacteria, so you have to argue the link to an E. coli question, not assume it. Sprint 3 allows exactly this: the strategy need not be what you measured, as long as you state the connection and it is believable. Doing this well is worth more than picking the obvious one.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourcePhosphate in the River Wye and the Lugg, sampled since 2000
Phosphate from chicken manure and sewage feeding algal blooms in the Wye, a protected river, and its tributary the Lugg.
- One row:
- One measurement of one substance in one water sample, with the sampling point, date and time, the result and its unit.
- Covers:
- Monthly spot samples from 2000 to 2026 at dozens of points on the English Wye and its tributaries. Orthophosphate on the Lugg at Mordiford alone is 311 samples
Full record7 traps
- Usable rows
- 311
- Coverage
- Monthly spot samples from 2000 to 2026 at dozens of points on the English Wye and its tributaries. Orthophosphate on the Lugg at Mordiford alone is 311 samples
- How it arrives
- One CSV per year: on the Downloads page choose Local Authority, then Herefordshire, County of, then the year. Each file holds every measurement in the county that year
- Language
- Page and file in English
- Natural· Salmon rod catch on the Wye· by year, for the Wye
- 1Results below the detection limit. Many results read <0.02 or <0.004: too little to measure, not zero. At Whitney Toll Bridge, where the Wye enters England, 39% of results are like this, and the limit itself changes over the years. Decide how to treat them before you look at the pattern, and say what you did.
- 2Gaps. Whitney Toll Bridge has no samples from 2011 to 2018, and Ross drops to four a year from 2015 to 2018. Count samples per year before you compare years.
- 3There are two River Wyes. Searching for Wye also finds the Derbyshire Wye at Buxton and Rowsley. The Herefordshire points have codes starting MD-500.
- 4One year per file, and everything in it. Each download is every measurement of every substance at every point in Herefordshire for one year. Filter to your sampling points and to one determinand before anything else, then stack the years. For a long series, that is up to 26 files: plan it.
- 5Monthly spot samples miss the blooms. Algal blooms happen in hot, low-flow weeks, and a sample once a month can miss them. Yearly averages look flat for that reason. Summer against winter, or low flow against high flow, shows more.
- 6England only. The upper Wye is in Wales and is monitored by Natural Resources Wales, published separately. Volunteers in the Wye Alliance have taken over 50,000 samples since 2021, but their test kits are less precise: a good comparison, not a replacement.
- 7Orthophosphate is not total phosphorus. It is the dissolved form algae can use straight away; total phosphorus includes what is bound to soil. Say which you used, and use one throughout.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is the orthophosphate concentration in the River Lugg at Mordiford higher than in the River Wye at Wilton Bridge, 2021 to 2025?
- Is orthophosphate in the Wye higher below Hereford's sewage works (Carrots Pool) than above them (Victoria Bridge)?
- How does orthophosphate in the Wye at Ross-on-Wye vary between summer and winter?
- samplingPoint.notation and samplingPoint.prefLabel: the point's code and name, such as MD-50050, R LUGG AT MORDIFORD BRIDGE
- phenomenonTime: when the sample was taken
- determinand.prefLabel: what was measured. Orthophosphate, reactive as P is code 0180
- result and unit: the value, in milligrams per litre for phosphate. Some results start with <
- samplingPurpose: why the sample was taken; most are statutory monitoring
Were salmon rod catches on the Wye lower in years when summer orthophosphate in the Wye was higher, 2008 to 2025?
Watch outEighteen years is a small sample, and both series change over the period, so the raw values can correlate whether or not one affects the other. Correlate the change from one year to the next instead. Salmon spend one to three years at sea, where most of them die, so the river is one influence among several. You have to average the phosphate samples into one value per summer yourself.
Herefordshire Council, on the advice of Natural England.
Link to this fileDirect. The restriction exists because the Lugg fails its phosphate target, which is what this data measures. Compare the Lugg with the main Wye, before and after the wetlands, or above and below them.
The tension199 planning applications for 2,217 homes were held up. Housebuilders and people needing homes on one side, and a protected river on the other.
SourceDefra, with a river champion and a taskforce for the catchment.
Link to this fileDirect in aim: it targets phosphate from poultry manure, which is what these samples measure. Only a few years of data exist since it began, so you can describe where phosphate stood when it started rather than whether it worked.
The tensionThe Soil Association said it was likely to shift the problem elsewhere, since the manure and the number of birds stay the same.
SourceSewage spills from every storm overflow in England, 2020 to 2025
When the sewers are overwhelmed, storm overflows release rainwater and sewage into rivers and the sea. The water companies in England now report how often and for how long each overflow spilled.
- One row:
- One storm overflow in one year: where it is, which water body and river or sea it runs into, how many hours it spilled, how many spills that makes by the 12-24 hour count, and for how much of the year its monitor was working.
- Covers:
- Every storm overflow of the ten water and sewerage companies in England: 12,788 listed in 2020 and 14,303 in 2025, one file a year. A separate table gives national totals from 2016
Full record13 traps, 2 on the three lines
- Usable rows
- 80,041
- Coverage
- Every storm overflow of the ten water and sewerage companies in England: 12,788 listed in 2020 and 14,303 in 2025, one file a year. A separate table gives national totals from 2016
- How it arrives
- One zip for each year. From 2021 it holds one Excel workbook with a sheet for each of the ten companies (2025 adds a sheet with all of them together), a summary workbook and a guide as a PDF. 2020 is ten separate workbooks, one for each company
- Language
- Page and file in English
- Natural· River invertebrate score (WHPT ASPT), England· by water body code and year
- 12020 is different. It is ten separate workbooks, one for each company, with no grid reference, no water body and no receiving water. Only from 2021 can you tell where an overflow runs into. From 2021 the companies also had to list every overflow, with a monitor or not, so 2021 lists 14,470 against 12,788 in 2020.
- 2Missing is not zero. A blank, N/A or #N/A spill count means there is no reliable data, not that nothing spilled. In 2021, 2,077 of the 14,470 overflows have no number; in 2025, 64 of 14,303. Leave them out and say how many. Never count them as 0.
- 3A monitor that was off saw nothing. Spills while the monitor was not working are not counted. In 2021, 2,705 overflows had a monitor working for less than 90% of the year; in 2025, 1,039. Choose a rule, for example only overflows at 90% or more, and use it for every year.
- 4Is 90% written 90 or 0.9? From 2024 the monitor column is a percentage. Before that, some companies wrote a fraction: six of the ten in 2020, and still Southern Water and Welsh Water in 2023. Excel can show 0.95 as 95%, so look at the value in the cell, not how it is displayed, before you compare it with 90.
- 5The hours change format. Up to 2023 they are decimal hours, such as 12.5. From 2024 they are written hh:mm:ss, and in 2025 all of Anglian Water's are text, such as 139:02:00. Some Severn Trent and Thames Water cells are stored as a time of day. Turn every value into a number of hours before you add them up.
- 6A spill count is not how often the overflow opened. Under the 12-24 method, everything in the first 12 hours is one spill, and each 24 hours after that is one more. A spill that runs for three days without stopping counts as 4, and several short spills in one morning count as 1. The hours are measured before this counting, so the two columns tell different stories.
- 7The columns move. In 2024 Unique ID became column A, so every other column moved one to the right, and the water body code changed from Cycle 2 (column I) to Cycle 3 (column J). Match columns by their names, never by their letters. Of 12,909 overflows with a code in both 2023 and 2024, 97 changed code.
- 8Not every water body code is a code. In 2025, 668 overflows have no proper code (GB and twelve digits). Most are South West Water's (527), which writes numbers such as 73, and many of those run into an estuary or the sea. River water bodies are the codes that start GB1.
- 9Follow an overflow by its Unique ID. The permit number is not enough: in 2025, 862 rows share a permit number and activity reference with another row. Unique ID starts in 2023, changed format in 2024 (AnW0001 became AWS00009), and the 2024 file gives the old code in its last column.
- 10Emergency overflows are mostly not here. They open when a pump or the power fails, not when it rains. Southern Water's 2020 file includes 495 of them, so Southern lists 1,477 overflows in 2020 and 978 in 2021.
- 11A few values are impossible. In 2020, five United Utilities overflows count 371 to 378 spills, more than a whole year of spilling can make. Six Northumbrian cells in 2020 hold two numbers, such as 84/74. Find them and say what you did.
- 12No year column, and no volume. The year is only in the sheet name or the title row, so add a year column before you stack files. The file says how often and for how long, not how much sewage: a short spill from a large overflow can carry more than a long one from a small overflow.
- 13Spills follow the rain. A wet year raises most overflows at once. Compare overflows within one year, or bring in rainfall, before you say that something changed.
- Do England's river water bodies where storm overflows spilled for more hours in 2025 have lower invertebrate scores (WHPT ASPT) in the Environment Agency's spring samples?
- Across the river water bodies where Yorkshire Water's storm overflows discharge, is the number of spills in 2024 related to the invertebrate score (WHPT ASPT) of the same water body that year?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of spills per storm overflow in England change from 2021 to 2025, counting only overflows whose monitor worked for at least 90% of each year?
- What share of each company's storm overflows spilled more than 20 times in 2025, the level at which Ofwat says a company should investigate?
- Do overflows named in the Bathing Water(s) column spill less often than other overflows of the same company, 2021 to 2025?
- Do storm tanks at sewage works spill for more hours a year than overflows on the sewer network?
- Counted spills using 12-24h count method: how many spills in the year, counted in blocks of 12 and 24 hours (column Q from 2022)
- Total Duration: how many hours the overflow spilled in the year, measured before the spills were counted (column P from 2022)
- EDM Operation - % of reporting period EDM operational: for how much of the year the monitor was working (column T from 2022)
- WFD Waterbody ID and Receiving Water: the water body code and the river, estuary or sea the overflow runs into (from 2021)
- Water Company Name, Site Name, EA Permit Reference and Outlet Discharge NGR (the grid reference, from 2021)
- Bathing Water(s) and Shellfish Water(s): filled in only when the overflow must be monitored for a bathing water or a shellfish water
- Storm Discharge Asset Type: on the sewer network, at a pumping station, or at a sewage works
- Unique ID: one code for each overflow, from 2023. The code changed in 2024, and the 2024 file also gives the old one
One zip for each year, from the Environment Agency (7)
Each link downloads that year's zip. Open the workbook called all water and sewerage companies (2021 to 2025), or the ten company workbooks (2020). Add a year column before you stack years, and match columns by name, not by letter.
- 2025Zip, 6.0 MB
- 2024Zip, 3.3 MB
- 2023Zip, 2.9 MB
- 2022Zip, 2.7 MB
- 2021Zip, 2.5 MB
- 2020Zip, 1.5 MB
- National totals, 2016 to 2025Long-term trends zip
Do England's river water bodies where storm overflows spilled for more hours have lower invertebrate scores (WHPT ASPT), 2021 to 2025?
Watch outAdd up the hours of every overflow in a water body first: one water body can have dozens. Match WFD Waterbody ID in the overflow file to WFD_WATERBODY_ID in the invertebrate Sites table, reach the Metrics table through SITE_ID, and take the year from SAMPLE_DATE. Keep river codes only (GB1 and eleven digits). 870 river water bodies have both in 2025, and 3,658 water body-years from 2021 to 2025. In 693 of the 2025 ones, every overflow's monitor worked for at least 90% of the year: keep to those, or say why not. Average one season's samples (spring or autumn) in each water body, because the score changes with the season. A higher score is better water. The match only sees the water body the overflow runs into, but sewage also flows on downstream into the next one. A water body with samples and no overflow listed is not always a zero: some overflows have no proper code.
Defra (the Department for Environment, Food and Rural Affairs), because the Environment Act 2021 requires a plan. The water companies carry it out, and the Environment Agency and Ofwat hold them to it.
Link to this fileDirect. Its targets are counted in spills a year, which is the spill count in this file. But the 2050 target is an average over many years, so one year of this file cannot show that an overflow meets it: the Long-term average spill count column (from 2022) comes closer. Most of the work is planned for after 2025, so this file shows where each overflow started, not whether the plan worked.
The tensionSpeed against bills. In the government's response to its consultation, many who disagreed with the targets "were of the view that there should be greater ambition in setting the timescales and strengthening the targets. However, a significant number of respondents indicated that they would not be willing to pay more in their monthly water bill to tackle storm sewage discharges". Most responses came through campaigns run by 38 Degrees and Surfers Against Sewage.
SourceParliament. The Act adds new duties to the Water Industry Act 1991. The water companies must carry them out, and the Secretary of State and Ofwat can enforce them.
Link to this fileDirect, and it explains the file. From 2021 the companies had to list every overflow, with a monitor or not, which is why 2021 lists 14,470 overflows and 2020 lists 12,788. Section 82 will measure oxygen, temperature, pH, cloudiness (turbidity) and ammonia above and below overflows, but those readings are not in this file.
SourceParliament, on a bill from Defra. The Environment Agency and Ofwat use the new powers.
Link to this fileIndirect. It changes the penalties, not the overflows. Emergency overflows open when a pump or the power fails, not when it rains, and after 2020 few are listed on their own in this file. Use it to explain why spills are now watched so closely, not as something this file can test.
The tensionSurfers Against Sewage called the bill "welcome but flawed": "We can't allow legislation to pass that will mean the public foot the bill for failing water companies and we can't stand by as the industry regulator prioritises profit over environmental protections" (Dani Jordan, Big Issue, 30 December 2024).
SourceOfwat, which regulates the water companies of England and Wales, alongside a separate Environment Agency investigation.
Link to this fileDirect. Ofwat's line is a spill count, which is a column in this file: in its final decisions a company should investigate "when an asset has spilled more than 20 times a year" (Defra's progress report, December 2025). You can count how many of a company's overflows cross that line each year. The file cannot say why an overflow spilled, which is what Ofwat investigated.
The tensionWho pays. Ofwat says the Yorkshire package "will not be passed on to Yorkshire Water's customers in their bills". Surfers Against Sewage said it could not "stand by as the industry regulator prioritises profit over environmental protections" (Big Issue, 30 December 2024).
SourceDefra designates bathing waters in England, when local groups or councils apply, under the Bathing Water Regulations 2013. The Environment Agency then tests the water each bathing season.
Link to this fileDirect for the overflows in the Bathing Water(s) column: they must be monitored because they discharge near a designated bathing water. A new designation can add overflows to that column in later years, so compare the same overflows from year to year.
The tensionFarmers against swimmers. The National Farmers Union "did not support any of the designations and said they would need a longer amount of time to consult their local members", while Surfers Against Sewage "supported the designation of all 27 sites" (Defra's summary of responses, 13 May 2024).
SourceThe Gulf of Mexico dead zone, every summer from 1985 to 2021
Fertiliser washed off farms across the Mississippi basin feeds algae in the Gulf, and their decay strips the oxygen from the seabed every summer: the dead zone.
- One row:
- One summer: the area of seabed water with less than 2 mg/L of oxygen, mapped on a survey cruise of 5 to 7 days in late July. Exhibit 2's unit is one station on one cruise.
- Covers:
- Every summer from 1985 to 2021 except 1989 and 2016, which were only half mapped: 35 years. Exhibit 2 adds 94 stations from the 2021 cruise
Full record8 traps, 1 on the three lines
- Usable rows
- 35
- Coverage
- Every summer from 1985 to 2021 except 1989 and 2016, which were only half mapped: 35 years. Exhibit 2 adds 94 stations from the 2021 cruise
- How it arrives
- CSV files on one page, a download link under each exhibit. Exhibit 1 is the dead zone's area every year; Exhibit 2 is the oxygen at every station on the July 2021 survey
- Language
- Page and file in English
- Human· Nitrate load to the Gulf· by year: the USGS water year (October to September) against the July survey of the same year
- 1Search for the Gulf of America. The EPA renamed the Gulf in 2025, so the page, the file headings and the citation all say Gulf of America. Searching the EPA site for "Gulf of Mexico" misses it. Say in your method that the two names mean the same sea.
- 2Six lines of notes first. Each CSV opens with a title, the data source, a note and the units, then a blank line, then the headings. A spreadsheet reads them as data: delete them, or tell the import to start at row 7.
- 3Null is not zero. 1989 and 2016 read null: only half the area was mapped. Leave them out. 1988 is real: 15 sq mi, a drought year.
- 4One cruise a year. The area is what one survey in late July found. It says nothing about how long the dead zone lasted or how big it was in June or August, and the page says so in its Limitations tab.
- 5Weather breaks the pattern. 2019 was forecast to be one of the largest on record, but Hurricane Barry mixed oxygen down days before the survey. 2017 is the record. Look at these years rather than deleting them.
- 6Square miles against square kilometres. The file is in square miles, the Task Force goal is 5,000 km² (about 1,930 sq mi). Convert once and state it.
- 7Exhibit 2 is one week in 2021. Its 94 stations are enough for a question about where the oxygen is lowest, not about change over time. You work out each station's distance from the river mouth yourself, from its latitude and longitude, and that calculation is part of your method.
- 8It stops in 2021. Later summers are only in NOAA's yearly announcements, each a source in its own right: 2022 (3,275 sq mi), 2023 (3,058), 2024 (6,705) and 2025 (4,402). Add them only if you cite each one, and note that the EPA's 2017 figure (8,772) and NOAA's (8,776) differ slightly.
- Was the Gulf dead zone larger in the summers after the Hypoxia Task Force's 2001 Action Plan (2002 to 2021) than before it (1985 to 2001)?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- On the July 2021 survey, is the oxygen in the bottom water related to a station's distance from the mouth of the Mississippi?
- Year and Square miles in Exhibit 1: the area of bottom water below 2 mg/L of oxygen that July
- Latitude, Longitude and Dissolved oxygen concentration (mg/L) in Exhibit 2: one row per station, measured near the seabed, 25 to 31 July 2021
- The page's own text gives the thresholds that matter for marine life: bottom fish start to leave below about 3 mg/L, crustacean larvae die below 2.5, and fish avoid or die below 1
Across the summers from 1985 to 2021, is the area of the Gulf dead zone related to the nitrate load that the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
Watch outMatch the USGS WY to the EPA Year: 35 years have both, 1985 to 2021. A water year runs from October to September, so each load includes three months of the autumn before and two months after the July survey. The scientists who forecast the dead zone use the May load, which is not in the USGS file: say so in your evaluation. Wet years carry more nitrate and more fresh water, and USGS names both as controls on the dead zone, so look at the flow in aflow.csv as well. A storm can mix oxygen down just before the survey, as Hurricane Barry did in 2019: keep those years, and say what they do. Both series may also drift over the years, so correlate the change from one year to the next as well. The missing summers leave 32 changes.
The Mississippi River/Gulf of America Watershed Nutrient Task Force: federal agencies led by the EPA, and twelve states along the river, each with its own nutrient reduction strategy.
Link to this fileDirect: the dead zone's area is how the plan measures success, and it is the number in Exhibit 1. The 2001 plan falls in the middle of the file, so there is a before and an after. But the plan acts on farms upstream through voluntary state strategies, and the dead zone also moves with rain, river flow and storms, so a before-and-after cannot pin a change on the plan alone: say so.
The tensionThe deadline has already been moved once, by twenty years, and the states' strategies rely on farmers volunteering. See the Iowa lawsuit for what happened when a city tried to make them compulsory.
SourceIowa's agriculture and natural resources departments and Iowa State University wrote the strategy. Des Moines Water Works, which supplies drinking water to about 500,000 people, sued the drainage districts of three upstream counties.
Link to this fileUpstream and indirect. Iowa is one of the biggest nitrate sources in the basin, but the dead zone is about 1,500 km downstream and fed by a dozen states, so this data cannot separate Iowa from the rest. Use it for the argument about who should act, not for a before-and-after.
The tensionA city paying to remove farm nitrate from its drinking water, against drainage districts and farmers who argued that field drains are not a regulated source of pollution. The court left it to the state, and the strategy stayed voluntary.
SourceFull physico-chemistry of Geneva's rivers, 1962 to 2017
The chemistry of the watercourses draining into Lake Geneva across the whole period of post-war industrialisation, sewage treatment and agricultural change.
- One row:
- One monitoring station in one year: annual values for about fifty parameters, most with their own unit column beside them.
- Covers:
- 165 stations, 1962 to 2017
Full record6 traps, 1 on the three lines
- Usable rows
- 836
- Coverage
- 165 stations, 1962 to 2017
- How it arrives
- CSV inside a zip, semicolon-separated, about 120 columns
- Language
- The page is in English; the file itself is in French
- 1It stops in 2017. It is the biggest and most detailed file in the family, so it looks like the main one, but it is the one that was discontinued. The major-elements file carries most of the same parameters through to 2025. Work out which one your question needs before you commit a day to this.
- 2Missing values are marked differently here from the rest of the family: small negative numbers such as −0.005 or −0.05, rather than −99. Same publisher, same subject, two conventions, and no warning in either file.
- 3Discharge is present in only 502 of 836 rows, so a question that normalises by flow loses 40% of the data before it starts.
- 4About 120 columns, half of them unit columns. So many columns is a real hazard: it makes it very easy to test pair after pair until one is significant.
- 5Station names are not unique. Group by CODEMESURE.
- 6The overlap with the major-elements file is large but not exact. If you use both, make the parameter names match and check the units, because only one of the two carries unit columns.
- A comparison of dissolved oxygen before and after a named wastewater treatment plant opened.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did nitrate at Geneva's river monitoring stations change between 1962 and 2017?
- Which parameters improved over the period, and which did not improve at all?
- Discharge, temperature, conductivity, pH, oxygen and saturation
- Nitrogen and phosphorus in several forms, BOD and dissolved organic carbon
- About thirty metals
- A unit column beside almost every parameter, and quality classes for the regulated ones
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileThe contract ran 2003 to 2010, comfortably inside this series, so there is data either side of it. That is the one thing this file does better than the file that replaced it.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileFifty-five years of chemistry against a programme of network investment. The longest view of the mechanism available anywhere in this family.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileThe Aire renaturation began in the 2000s and this series covers its early phases but ends before the later ones. Check your dates against the works before building on it.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceForever chemicals (PFAS) in tap water round Lyon and in Haute-Savoie, from Hub'Eau
The sum of 20 PFAS in the tap water of each commune, against the legal limit of 0.1 µg/l, in the Rhône département, where wells below the Pierre-Bénite plants supply several towns, and in Haute-Savoie, including Rumilly.
- One row:
- One tap-water sample from one supply network in one commune on one date, with the sum of 20 PFAS in µg/l.
- Covers:
- Rhône département: 499 samples from 119 communes, July 2022 to July 2026. Haute-Savoie: 635 samples from 201 communes, August 2022 to June 2026, most of them from 2025
Full record9 traps
- Usable rows
- 498
- Coverage
- Rhône département: 499 samples from 119 communes, July 2022 to July 2026. Haute-Savoie: 635 samples from 201 communes, August 2022 to June 2026, most of them from 2025
- How it arrives
- CSV, semicolon-separated, 31 columns. Open the web address and the file downloads, always called resultats_dis.csv. One address for each département
- Language
- Page and file in French
- 1Below the limit is written 0.0. In resultat_numerique, every result below the limit of quantification is 0.0: 29 of the 499 in the Rhône file and 364 of the 635 in Haute-Savoie. Read resultat_alphanumerique, where it is written <0,029, and decide what to do with those before you average anything.
- 2Some limits of quantification are too high to judge the water. 3 results in the Rhône file and 66 in Haute-Savoie are below a limit of 0.1 or more (<0,100 or <0,175). They cannot tell you whether the water met the 0.1 limit. The Commission's guidelines ask laboratories for 0.03 or lower. Two Rumilly samples of 22 August 2022 read <0,175: the DREAL gives that day's results for single PFAS on its Rumilly page.
- 3The C and N flags judge the whole sample, not the PFAS. Some samples above 0.1 µg/l are flagged C, all of them in 2022, and in Haute-Savoie three samples with almost no PFAS are flagged N because something else failed. Compare resultat_numerique with 0.1 yourself.
- 4More samples each year, in more places. The Rhône file has 44 samples in 2022 and 148 in 2025; Haute-Savoie has 11 in 2022 and 367 in 2025, from 3 communes to 168. The first campaigns looked where a problem was suspected, below Pierre-Bénite and at Rumilly. A change in the share above the limit can come from where the samples were taken, not from the water.
- 5Some communes have many samples, most have one. Up to 32 in one commune in the Rhône file and 54 (Rumilly) in Haute-Savoie, while 60 of 119 and 74 of 201 communes have one. Make one value per commune and year, and say how many samples it rests on.
- 6A commune is not one water supply. The Rhône file has 128 networks for 119 communes and Haute-Savoie 362 for 201. Two samples from one commune on one day can be from different networks.
- 7One result is not a number. One Rhône sample reads N.M. (not measured) and has no resultat_numerique: take it out. That leaves 498.
- 8Look at the highest values first. The DREAL calls one result, at Grigny on 1 December 2022, surprising ("Ce résultat est surprenant"): the other samples in that area that month did not agree with it, and the extra checks that followed gave ordinary results. It suggests a laboratory error as one explanation. Decide what to do with it before you look at any pattern, and say so.
- 9Open it through the import command. Semicolons, UTF-8 with no marker, and a decimal comma in resultat_alphanumerique: in Excel use Data, then From Text/CSV. Both addresses download as resultats_dis.csv: rename each one. The file grows every month: on 28 September 2026 the Rhône file had 499 rows.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the share of tap-water samples above the 0.1 µg/l limit for 20 PFAS changed in the Rhône département between 2023 and 2025?
- Is the sum of 20 PFAS in tap water higher in Ternay, Grigny and Givors, south of the Pierre-Bénite plants, than in the rest of the Rhône département, 2023 to 2025?
- How did the sum of 20 PFAS in Rumilly's tap water compare before and after two wells were cut off from the supply on 14 November 2022?
- nom_commune and code_commune: the commune (town or village) and its INSEE code. nom_reseau and code_reseau: the supply network the water came from. A commune can have several
- date_prelevement: when the sample was taken, written 2024-05-14T09:30:00Z
- resultat_alphanumerique: the result as the laboratory wrote it, with a decimal comma, such as 0,078, or <0,029 when it was below the limit of quantification
- resultat_numerique: the same result as a number with a decimal point. A result below the limit of quantification is written 0.0 here
- limite_qualite_parametre: the legal limit, <=0,1 µg/L in every row
- conformite_limites_pc_prelevement: C (conforms) or N (does not) for the whole sample, all its chemical limits together
- libelle_parametre: always Somme de 20 substances perfluoroalkylées (PFAS), SANDRE code 8847
The European Parliament and the Council: Directive (EU) 2020/2184 of 16 December 2020 on drinking water. France wrote the limit into its own rules (the arrêté of 11 January 2007) by an arrêté of 30 December 2022. The regional health agencies (ARS) take the samples.
Link to this fileDirect: it is the limit in the file's own column, limite_qualite_parametre. The limit is on the sum of 20 PFAS. The directive's second limit, 0.5 µg/l for all PFAS together (PFAS Total), is not in this file. The years before 2026 show where tap water stood before the deadline, not whether the directive worked.
SourceThe préfet of the Auvergne-Rhône-Alpes region, with the DREAL, which inspects the two plants on the Rhône just south of Lyon. Both make fluorinated plastics (fluoropolymers) and handle PFAS, and Arkema's site is classed Seveso high threshold.
Link to this fileDirect for the communes south of the plants, whose tap water comes from wells in the Rhône's gravels (its alluvial aquifer), so the link runs through the groundwater. The DREAL says one well field, at Ternay, did not conform and that a treatment plant is being built there. The file has 31 samples from Ternay, 25 from Grigny and 10 from Givors, from 2022.
The tensionWhether the State did enough. In the Assemblée nationale's committee, Anne Stambach-Terrenoir said that "Le plan PFAS du Gouvernement n’impose aucune contrainte aux industriels" (the Government's PFAS plan puts no constraint on industry). Jean-Luc Fugit said that Arkema had cut its discharges by 80%, and called it "un premier pas intéressant, mais il faudra atteindre dès que possible les 100 %" (an interesting first step, but it must reach 100% as soon as possible) (Assemblée nationale report, 27 March 2024).
SourceThe préfète of Haute-Savoie, with the DREAL (the regional environment office, which inspects factories), the regional health agency (ARS) and the community of communes Rumilly Terre de Savoie, which runs the drinking water.
Link to this fileDirect, and it has a date. Rumilly has 54 samples in the Haute-Savoie file, from 22 August 2022, and two contaminated wells were cut off on 14 November 2022. But only six of those samples are from before that day, so the time before the change is very short. The DREAL publishes the ARS's own 2022 results on the same page: use them as a check.
SourceThe French Parliament: loi n° 2025-188, proposed by the deputy Nicolas Thierry and the Écologiste group. The ministries for the environment and for health apply it, and the money from its charge goes to the water agencies.
Link to this fileDirect for the checks: the law makes PFAS part of the health checks on tap water, the checks this file comes from, and asks each ARS to publish its results every year. The product bans and the charge on factories act further up, on what reaches rivers and groundwater.
The tensionJobs and imports, or health. The first text also banned PFAS in cookware, and the Assemblée nationale took that out in April 2024. In the Senate's committee, the rapporteur Bernard Pillefer feared that a national ban would end "par des délocalisations de productions ou par un accroissement des importations" (in production moving abroad or in more imports). Jacques Fernique reported the message that many scientists repeated in the hearings: "Il faut fermer le robinet" (we have to turn off the tap) (Senate report, 22 May 2024).
SourceMetals in Geneva's rivers
Contamination of watercourses by metals from road runoff, industry and the urban surface, and whether the places that exceed the standards are the places you would expect.
- One row:
- One monitoring station in one year: annual values for about thirty metals, plus the number of samples, usually twelve.
- Covers:
- 180 stations, 1995 to 2025
Full record6 traps
- Usable rows
- 683
- Coverage
- 180 stations, 1995 to 2025
- How it arrives
- CSV inside a zip, semicolon-separated
- Language
- The page is in English; the file itself is in French
- 1The same −99 convention as the major-elements file, and worse. 21% of mercury values are −99. The mean of the column as it downloads is −20.555; with those values removed it is 0.403. A fiftyfold error, in the wrong direction, producing a negative concentration that cannot exist.
- 2Gadolinium and vanadium are −99 in about two thirds of rows, so those columns are mostly absent rather than mostly present.
- 3Copper is the downgrading parameter in 255 of 683 station-years, which makes it the obvious variable and also the one everybody else in your class will pick.
- 4With thirty metal columns, it is tempting to test pair after pair until one comes out significant. Decide which metal your question is about before you open the file, and say in your method that you did.
- 5Station names are not unique. Group by CODEMESURE.
- 6These are small numbers with real detection limits, so a value of exactly zero is likelier to mean "below the limit" than "none present".
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- A comparison of copper concentrations between stations on urban catchments and rural ones.
- How has zinc at Geneva's river monitoring stations changed between 1995 and 2025?
- Which metal most often determines a station's grade, and does that differ between river types?
- Copper, zinc, lead, cadmium, chromium, nickel, mercury
- About twenty more, from aluminium and iron to platinum and uranium
- DECLASSANT: which metal downgraded that station-year, most often copper
- Number of samples behind each annual value
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileMetals cross the border in the same water as everything else, and the contract covers the whole basin rather than the cantonal half of it.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileA weaker link than for nutrients, and worth saying so. Separating foul water from surface water moves road runoff around rather than treating it, so the mechanism is different and you have to make the argument clearly.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileRenaturation changes how a channel traps sediment, and metals travel attached to sediment. The connection is real, but you would have to explain it, not just state it.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceThe Asse, nutrients from the Jura foothills to the lake
What a small river picks up between its source and Lake Geneva, and whether a treatment works on its upper reach is still the reason it fails on nitrite.
- One row:
- One measurement: one parameter in one water sample, taken on one date at one station.
- Covers:
- 193 samples at 5 stations on the Asse, 2003 to 2025, in seven campaign years; only Calèves and Nyon Lac run across the whole period
Full record9 traps
- Usable rows
- 1,700
- Coverage
- 193 samples at 5 stations on the Asse, 2003 to 2025, in seven campaign years; only Calèves and Nyon Lac run across the whole period
- How it arrives
- One Excel .xls per station, from the Export button on each station page. Five stations, five files
- Language
- Page and file in French
- 1There is no bulk download. Each station has its own Export button, so the river is five exports. The page warns that more than a year or two at once may fail; twenty years at a time worked when this was checked.
- 2The export needs the page, not the link. The button submits a form, so a copied link or a script gets a page asking for your screen size instead of the file. Click Export on the station page itself.
- 3Campaigns, not years. The secondary network is sampled every four or five years: 2003 (Calèves), 2006 (Nyon Lac), 2009, 2013, 2016, 2020 and 2025. Chéserex and Moulin Velliet were sampled once, in 2009, so they can show where along the river, never when.
- 42009 at Nyon Lac is three samples. February, May and July. The same year's figures elsewhere are monthly, so a 2009 average for Nyon Lac is not the same kind of number.
- 5The sampling method changed. In 2003 (Calèves), 2006 (Nyon Lac) and 2009 (Embouchure) the samples were 24-hour composites, `Durée [j]` = 1. Every later sample is a single grab, `Durée [j]` = 0. A composite smooths out peaks a grab can catch, so part of any before-and-after is the method.
- 6The criterion moves. The ammonium threshold is 0.2 or 0.4 depending on water temperature, and the nitrite threshold is 0.02, 0.05 or 0.1 depending on chloride. Judge failures with the `QR` column (value divided by criterion; above 1 fails), not against one fixed number.
- 7Text in the number column. Twelve values read `<LOQ` (below the limit of quantification) and one reads `nd`, which turns the column into text in a spreadsheet. Decide what to substitute and say so in your method; half the limit is the usual choice, and the limit itself is only recorded from 2016.
- 8The station summary table uses a different statistic. It shows the 90th percentile of a year, except where fewer than twelve samples were taken, when it shows the maximum instead and adds a `w`. That is why 2020 at Nyon Lac looks about three times worse than its neighbours on the summary page.
- 9It is an old-format .xls with French headings, and dates are written 04.12.2025, day first.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How do nitrite and ammonium change between Chéserex and Moulin Velliet, and what does that suggest about a point source between them?
- Has the share of samples at Calèves that fail the nitrite criterion changed between 2003 and 2025?
- Is nitrate at Calèves higher in autumn than in spring, and has that difference changed since 2003?
- Nitrate, nitrite, ammonium, total nitrogen, total phosphorus and dissolved reactive phosphorus, all in mg per litre as N or P
- Dissolved organic carbon, chloride, pH and water temperature
- The Swiss quality criterion for each measurement, and QR: the value divided by that criterion, so anything above 1 fails
- The limit of quantification (from 2016 only), station coordinates in the Swiss LV95 grid, and the sampling network (secondary or passive)
The communes of Gingins and Chéserex, whose treatment works discharges into the Asse, and the City of Nyon, whose plant is to take their wastewater, under the canton's plan for micropollutants (DGE).
Link to this fileThe strongest pairing in the file. In 2009, the one year every station was sampled, the median nitrite was 0.003 mg/L at Chéserex and 0.103 at Moulin Velliet. That is more than thirty times higher, about two and a half kilometres downstream, and ammonium failed in 13 of 15 samples there. Calèves, the station still sampled, failed on nitrite in 7 of 12 samples in 2003 and again in 2025. Nyon's own 2022 report says the nitrite is "certainly linked" to a treatment works upstream. The 2025 campaign is the before. The after is the first campaign once the connection is made.
The tensionWho treats whose water. A single regional plant for 30 communes at Gland was stopped in November 2020 by a conditional vote of Nyon's council. The Gland association went ahead without Nyon, and Nyon is now enlarging its own plant to take Gingins, Chéserex and Prangins. That is the background to a 1973 works still discharging into the Asse in 2022.
SourceThe City of Nyon, with the canton paying 60% and the Confederation 35% of the CHF 1.93 million estimated for studies and works.
Link to this fileThe whole course of the river within Nyon, about 4.5 km, for flood capacity, fish migration and habitat. It changes the shape of the channel, not what flows into it, so do not expect it in the nutrient columns. Its first measure is a new reach-by-reach water quality survey, which suggests that five stations are too few to find where the problems start.
The tensionThe sewer is in the river. The council's own report notes a wastewater main running along the bank or in the bed through town, which is part of why that reach is so heavily engineered, and a renaturation has to work around it.
SourceThe Swiss Confederation, through the Federal Office for Agriculture, with target agreements negotiated with the farming sector.
Link to this fileThe upper Asse runs through fields, and it shows: nitrate at Calèves has a median of 4.18 mg N/L in October against 1.77 in April, the autumn peak typical of a farmed catchment. The 2016 campaign falls inside the law's reference period and 2025 is the latest, so the comparison the law asks for can be made here, with one catchment and two campaigns. State that limit clearly.
The tensionThe law was Parliament's alternative to the Drinking Water Initiative, which would have cut direct payments to farms using pesticides and which voters rejected on 13 June 2021. The reduction is now negotiated with farmers rather than imposed on them.
SourceRivers and lakes3
Plankton in Lake Geneva, every month from 1974 to 2010
What happened to the lake's algae and the water fleas that graze them as the phosphorus that had choked Lake Geneva in the 1970s was brought down, and as the water warmed.
- One row:
- One month at the deepest point in the middle of the lake: the average of the one or two samples taken that month.
- Covers:
- Every month from January 1974 to December 2010: 444 months, with phytoplankton missing in 9 and Daphnia in 31
Full record8 traps, 1 on the three lines
- Usable rows
- 435
- Coverage
- Every month from January 1974 to December 2010: 444 months, with phytoplankton missing in 9 and Daphnia in 31
- How it arrives
- One CSV, semicolon-separated, in the Latin-1 encoding
- Language
- Page and file in English
- Natural· Phytoplankton biomass, Lake GenevaagainstHuman· Dissolved phosphate, Lake Geneva
- Natural· Daphnia (water fleas), Lake GenevaagainstHuman· Dissolved phosphate, Lake Geneva
- 1Check the phosphate units. The PO4 heading says µgP/L, but the values are in mg/L: 0.034 is 34 µg per litre. Multiply by 1,000 and say so in your method.
- 2Semicolons and Latin-1. Open it as UTF-8 and "µg/L" in every heading turns into garbage characters. Tell your spreadsheet the file is semicolon-separated and Latin-1 (or Western European).
- 3NaN means missing. Phytoplankton is missing in 9 months and Daphnia in 31. Leave those months out, never read NaN as 0, and count how many months each year really has before you average it.
- 4The sampling depth changed. Phytoplankton was sampled in the top 10 m until 2001 and the top 18 m after. A step in biomass around 2001 could be the method, not the lake. The thermometer was replaced by a probe in 1998.
- 5Months are not independent. Every year rises and falls with the seasons, so a correlation across all 444 months is mostly summer against winter. Compare the same month across years, or annual means, and say which.
- 6Different amounts of water. Phytoplankton is per litre of the upper layer; Daphnia is per square metre of lake surface, down to 50 m. They are not the same kind of number, so compare how they change, not their sizes.
- 7A1 to A5 are only names here. The file does not say which algae are in each group. That is in Anneville et al. (2018), Hydrobiologia 824: 121-141. Read it before you build a question on one group.
- 8It stops in 2010. Later years exist but need an account and a data request from the observatory, which will not arrive in a lesson.
- Is annual mean phytoplankton biomass related to annual mean dissolved phosphate in Lake Geneva, 1974 to 2010?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is the number of Daphnia related to phytoplankton biomass in the same month in Lake Geneva, 1974 to 2010?
- Is the biomass of the summer phytoplankton group related to summer water temperature across the years 1974 to 2010?
- Year, Month
- PhytoTot: total phytoplankton biomass, µg per litre, in the top 10 m of water (top 18 m after 2001)
- A1 to A5: the biomass of five groups of phytoplankton that grow at different times of year, defined in Anneville et al. (2018)
- Daphnia: water fleas per square metre of lake surface, counted in a net pulled up from 50 m
- PO4: dissolved phosphate. The heading says µgP/L; the values are mg/L
- WaterTemp in °C and SolarRad in joules per cm²
The Swiss Confederation, for all laundry detergent sold in Switzerland.
Link to this fileDirect on the PO4 column and indirect on the plankton, which respond to phosphate through a food web. 1986 falls in the middle of this file, so there is a before and an after. But the ban came with sewage-works improvements and with France's own measures on the other shore, so a before-and-after cannot pin the change on the ban alone: say so.
SourceThe International Commission for the Protection of the Waters of Lake Geneva (CIPEL), for Switzerland and France together.
Link to this fileDirect in aim: the target is written for phosphorus because phosphorus drives the algae in this file. The file ends in 2010, so it describes the lake the 2011 target was set for, not whether the target was met. CIPEL's yearly reports carry the later years as tables in PDFs.
The tensionFish against clear water. CIPEL's own scientific report of 2024 found that managing phosphorus divides the lake's stakeholders, because low phosphorus does not generally keep fish populations strong. It proposed 15 µg per litre as a compromise between fish and the lake's other uses, and left CIPEL to reconsider its 10 to 15 µg target in 2025.
SourceZooplankton and water quality in Lake Geneva, 1959 to 2018
The water fleas and copepods at the base of the lake's food web, which fish feed on, through six decades in which the phosphorus in Lake Geneva rose and was brought back down.
- One row:
- One kind of zooplankton at one stage of its life on one sampling date at SHL2, the deepest point of the lake: its density in the top 50 m, in individuals per litre. The chemistry sheet is one measurement on one date.
- Covers:
- Lake Geneva: zooplankton on 973 dates from 1959 to 2018, and total phosphorus, water temperature, transparency and chlorophyll from 1974 to 2015. Lakes Annecy and Bourget are in the same file
Full record9 traps, 1 on the three lines
- Usable rows
- 32,032
- Coverage
- Lake Geneva: zooplankton on 973 dates from 1959 to 2018, and total phosphorus, water temperature, transparency and chlorophyll from 1974 to 2015. Lakes Annecy and Bourget are in the same file
- How it arrives
- One Excel workbook of 7.5 MB with five sheets. On the page choose Original File Format: the tab-delimited version holds only the first sheet
- Language
- Page and file in English
- Natural· Whitefish caught· by year
- Natural· Daphnia (water fleas)againstHuman· Total phosphorus
- 1Choose Original File Format. The page's default download is a tab-delimited file holding only the first sheet, a list of authors (3.8 KB). The Excel original has all five sheets.
- 2Three lakes in one file. Filter waterbody_name to Geneva first: Annecy and Bourget are mixed in.
- 3The headings are on row 2, under a title row, and the first column is a row number with no heading.
- 4One row per stage, under several names. Daphnia arrive as adult females, egg-carrying females and juveniles, named DAPHNIA SP until 1990 and HYALINA, LONGISPINA and GALEATA from 1984, as the counters began naming species. Add every row that starts DAPHNIA on a date to get all the water fleas.
- 5No row means none counted. The file has no zeros: a date with no Daphnia row had none recorded. Put the zeros in before you average, or those dates vanish and the mean goes up.
- 6Two nets. Some dates used a 212 µm net and others a 64 µm one, which also catches smaller animals. No date has Daphnia from both, but the mix changes over the years: keep zoop_mesh_um and say how you handled it.
- 7Uneven sampling. 2001 has a single sampling date and 1981 has 37. Average within each year before comparing years, and say how many dates each year had.
- 8Chemistry is its own sheet, one measurement per row. Total phosphorus (TP) is sampled at the surface about 18 times a year; temperature comes twice per date, surface and deep, so choose one.
- 9Not the same unit as the monthly plankton card. Daphnia here are per litre of the top 50 m; the 1974 to 2010 file counts them per square metre of lake surface.
- Is the annual mean density of Daphnia related to annual mean total phosphorus in Lake Geneva, 1974 to 2015?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is the density of Daphnia in spring related to spring surface water temperature in Lake Geneva, 1974 to 2015?
- Zooplankton_Abundance sheet: waterbody_name, year_yyyy, month_mm, day_of_month_dd, taxa_name_orig_upper (the animal and its stage, in French) and zoop_value, individuals per litre
- zoop_mesh_um: the mesh of the net, 212 or 64 µm
- Water_Chemistry sheet: W_parameter (TP is total phosphorus in µg per litre; also temperature, secchi and chlorophyll_a), W_value, and W_zone (surface, or hypolimnion for the deep water)
Across the years 1974 to 2015, does the professional whitefish catch in Lake Geneva's Swiss waters go with the lake's annual mean total phosphorus?
Watch outAverage the phosphorus to one value a year first (about 18 surface samples each), and 42 years meet the catch. Use the professional catch, which runs the whole period: the total adds recreational fishing from 1994 and jumps there. The catch is Swiss waters only. Both series move steadily over these decades, so correlate the change from one year to the next as well.
The International Commission for the Protection of the Waters of Lake Geneva (CIPEL), for Switzerland and France together.
Link to this fileDirect in aim: the target is set for phosphorus, which this file measures, and the zooplankton sit between the algae the target controls and the fish the tension is about. The file ends in 2015 for phosphorus and 2018 for zooplankton, so it shows the lake the targets were set for, not whether the latest one is met.
The tensionFish against clear water. CIPEL's own scientific report of 2024 found that managing phosphorus divides the lake's stakeholders, because low phosphorus does not generally keep fish populations strong. It proposed 15 µg per litre as a compromise between fish and the lake's other uses, and left CIPEL to reconsider its 10 to 15 µg target in 2025.
SourceThe Swiss Confederation, for all laundry detergent sold in Switzerland.
Link to this fileDirect on the phosphorus and indirect on the zooplankton, which respond through the algae they eat. 1986 falls in the middle of the phosphorus series, but the ban came with sewage-works improvements and France's own measures, so a before-and-after cannot pin the change on the ban alone.
SourceAQUASTAT, how much water each country takes and what it takes it for
Water scarcity: how much freshwater a country withdraws for farming, industry and towns, set against how much it actually has.
- One row:
- One variable for one country in one year. The file holds 199 variables in the same three columns, so every question starts by filtering to the variable it needs.
- Covers:
- 182 countries and 19 regional aggregates, 1962 to 2022, for total water withdrawal; 936,332 rows across all 199 variables
Full record8 traps
- Usable rows
- 6,305
- Coverage
- 182 countries and 19 regional aggregates, 1962 to 2022, for total water withdrawal; 936,332 rows across all 199 variables
- How it arrives
- One CSV of 88 MB, the whole database in long form, encoded Latin-1 rather than UTF-8
- Language
- Page and file in English
- 1It is not UTF-8. The file is Latin-1, so a spreadsheet that assumes UTF-8 turns Côte d'Ivoire and Türkiye into nonsense or refuses to open it. Choose the encoding on import. Everything else in the file is plain ASCII, which is why the fault is easy to miss.
- 2Most of the numbers were never measured. Of 6,305 country values for total water withdrawal, 499 are marked A, official. The rest are estimated, imputed or borrowed. India has 48 years and not one of them is official.
- 3Imputed means drawn as a straight line. Between two real values the gaps are filled with a straight line, and after the last one the value is copied forward: India reads exactly 761 every year from 2010 to 2022. About 23% of the withdrawal values equal the year before. A flat line here usually means nobody measured, not that nothing changed.
- 4Nineteen of the areas are not countries. World, Europe, Sub-Saharan Africa, Least Developed Countries and fifteen more sit in the same column, so a sum or an average across areas counts places twice. Filter on the numeric code, not the name: Central African Republic and South Africa look like regions to a text search.
- 5Percentages above 100 are real. Water stress is withdrawal as a share of renewable water, and 17 countries were over 100 in 2022, Kuwait at 3,850. Countries that pump fossil groundwater or desalinate seawater take more than rain replaces. One Kuwait makes an average across countries meaningless, so use the median.
- 6Coverage grows over time. Total withdrawal has values for 2 countries in 1965, 54 in 1980 and 182 from 2015, so a total or an average drawn across the whole period is mostly measuring how many countries joined.
- 7The unit sits inside the variable name, `[10^9 m3/year]`, which is cubic kilometres. The header also repeats itself: two columns called aquastatElement and two called timePointYears. Both year columns are identical; delete one.
- 8The last year is 2022. Before you use a country's most recent years, check that they are not an earlier value copied forward.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How far did Australia's agricultural water withdrawal fall during the Millennium Drought, and how quickly did it recover?
- Which countries withdraw more freshwater than they renew each year, and what do they have in common?
- Did Spain's total water withdrawal fall after 2015, using only the values marked official?
- Total water withdrawal, in 10^9 m³ a year (cubic kilometres): India 761, Spain 29.0, Australia 16.7 in 2022
- The same split three ways: agricultural, industrial and municipal withdrawal, each also per person
- SDG 6.4.2 water stress: withdrawal as a percentage of renewable freshwater, and each sector's share of it
- Total renewable water resources, per country and per person
- A status letter on every value: A official, E estimated, I imputed, X taken from another international organisation
Every UN member state, reporting through FAO, which is the custodian of both indicators and publishes them in this file.
Link to this fileThe closest pairing on the page: indicator 6.4.2 is a column in this file, so the strategy and the measurement are literally the same number. Read the status letters before you rely on that. Not one water stress value for a country is marked official; all 6,248 are estimated or imputed. The companion indicator, 6.4.1, is measured in US dollars of output per cubic metre, so it rises when an economy grows even if not a litre of water is saved.
The tensionIrrigation. Farming is the largest withdrawal almost everywhere, and the same agenda asks countries to end hunger (SDG 2), which in dry countries has usually meant irrigating more land rather than less.
SourceThe Murray-Darling Basin Authority under the Commonwealth Water Act 2007, with the four basin states and the Australian Capital Territory.
Link to this fileAustralia's figures are among the few in the file marked official year after year. They show agricultural withdrawal halving from 15.0 cubic kilometres in 2001 to 7.3 in 2009, then back to 13.0 by 2013. That swing is the Millennium Drought, and it is nearly three times the 2.75 km³ the plan set out to recover. The plan covers one basin and this is the whole country, so the honest finding is how small the policy looks next to the weather.
The tensionWater for rivers against water for farms, bought from irrigators. In January 2019 a South Australian Royal Commission found that the Authority had set the recovery figure unlawfully and had ignored its own scientists' climate advice.
SourceEvery EU member state, through river basin management plans renewed in six-year cycles, with the European Commission checking them.
Link to this fileIndirect, and say so. The directive is aimed at the condition of each river, lake and aquifer, not at a national total. Spain is still the case to look at: withdrawal falls from 36.6 km³ in 2012 to 29.0 in 2020, and the values from 2015 onwards are marked official. Stop at 2020, because 2021 and 2022 are the 2020 figure copied forward.
The tensionWho pays. The Commission took Germany to the Court of Justice, arguing that hydropower, navigation and flood protection were water services that had to pay their costs. The Court dismissed the case on 11 September 2014 (C-525/12), which left each country to decide what counts.
SourceBiodiversity27
Kelp forest surveys off California, inside and outside marine reserves, every year since 1999
Kelp forests and what grazes them: sea urchins, the predators that eat them, and whether marine reserves, where fishing is banned, change the balance.
- One row:
- One transect on one day: the number of one species counted on a strip of reef 30 m long and 2 m wide (60 m²), at one site, in one depth zone. The site table adds each site's position and whether it is inside a marine protected area.
- Covers:
- Kelp forest sites from Oregon to southern California, 1999 to 2024. The northern Channel Islands alone have 80 site codes and 5,253 transects
Full record8 traps, 2 on the three lines
- Usable rows
- 5,253
- Coverage
- Kelp forest sites from Oregon to southern California, 1999 to 2024. The northern Channel Islands alone have 80 site codes and 5,253 transects
- How it arrives
- A data package of eight files. The swath counts are a 33 MB CSV and the site table a 1 MB one, both in Windows-1252 rather than UTF-8
- Language
- Page and file in English
- Natural· Purple urchin densityagainstHuman· Protection from fishing
- Natural· Red urchin densityagainstHuman· Protection from fishing
- Natural· Giant kelp densityagainstHuman· Protection from fishing
- 1A species that was not seen has no row. 599 of the 5,253 transects in the northern Channel Islands have no purple urchin row at all. List every transect surveyed, from all of its rows, and give it 0 before you average, or every mean comes out too high.
- 2Group by site_status, never by MPA_Name. An outside site carries the name of the reserve it is paired with: SMI_CUYLER is listed under Harris Point SMR, but it is a reference site.
- 3Two halves of one place can sit either side of a boundary. SCI_SCORPION_W is inside the Scorpion reserve and SCI_SCORPION_E is outside it. Treat every site code as its own site, as PISCO does.
- 4A conservation area is not a reserve. SMCA sites (Painted Cave, west Anacapa) allow some fishing, so they fit neither side of an inside-or-outside comparison.
- 5The site table has many rows per site, one for each year and method, and some are blank. Filter method to SBTL_SWATH_PISCO and drop the rows with a blank site_status before you look anything up. Five island sites (SCI_VALLEY_CEN, _E and _W, SCI_YELLOWBANKS_CEN and _E) have no status at all.
- 62013 is thin. Only 19 island sites were surveyed that year, against 34 to 69 in every other year from 2008 to 2024.
- 7Giant kelp comes in several rows per transect, one for each number of stipes (the size column), so add them up. Urchins are counted only over 2.5 cm across, and kelp only over 1 m tall.
- 8The package is updated. This card counted version 1.11 (January 2026). A later version can add years and move every count.
- In the northern Channel Islands, did purple urchin density differ between kelp forest sites inside and outside no-take reserves before (2010 to 2012) and after (2014 to 2024) sea star wasting disease removed the sunflower star?
- Are red urchins, which are fished, denser at kelp forest sites inside the Channel Islands reserves than at sites outside them?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Across the northern Channel Islands' kelp forest sites, is purple urchin density related to the density of giant kelp?
- site: the site code, such as SCI_GULL_ISLE_E, which is also the key to the site table
- year, month, day, zone and transect: together, one transect on one day
- classcode and count: the species and how many were seen. STRPURAD is purple urchin, MESFRAAD red urchin, MACPYRAD giant kelp and PYCHEL sunflower star; the taxon table in the same package lists the rest
- In the site table, site_status (mpa or reference) and mpa_type (SMR, a state marine reserve where all fishing is banned, or SMCA, which allows some)
The California Fish and Game Commission, for state waters, and NOAA's Channel Islands National Marine Sanctuary, for federal waters, around Channel Islands National Park.
Link to this fileDirect on protection: the site table records which sites are inside the reserves. Indirect on anything alive, which responds through the food web: the reserves protect the fished predators of sea urchins, California sheephead and spiny lobster, not the urchins themselves.
The tensionFishers against the closures. When the reserves opened in April 2003, fishing was banned over about 175 square miles around the islands, and commercial and recreational fishers fought the restrictions (NPR, April 2003).
SourceRiver invertebrates and diatoms in Haute-Savoie, from Hub'Eau
Whether the insect larvae, worms and algae living on the river bed, on the French side of the Geneva region, are in the state the water law asks for, and where they are not.
- One row:
- One index score from one survey at one station on one date. Most stations have one survey in a year, and many are not surveyed every year.
- Covers:
- Haute-Savoie. The invertebrate index I2M2 at 199 stations, 2008 to 2025, and the diatom index IBD at 251 stations, 2005 to 2025
Full record9 traps, 2 on the three lines
- Usable rows
- 987
- Coverage
- Haute-Savoie. The invertebrate index I2M2 at 199 stations, 2008 to 2025, and the diatom index IBD at 251 stations, 2005 to 2025
- How it arrives
- CSV, semicolon-separated. Open the web address and the file downloads, always called indices.csv. One address for each index
- Language
- Page and file in French
- Human· Orthophosphate in Haute-Savoie's rivers· by station code and year
- Human· PFOS in Haute-Savoie's rivers· by station code and year
- 1Here a higher score is better, for both indices. Geneva's diatom index (DI-CH) runs the other way. Never put the French IBD and the Swiss DI-CH in one column.
- 2I2M2 starts in 2008. Before that, surveys here were scored with the IBGN (code 1000, 1992 to 2006), a different formula with scores out of 20. Never join the two into one series.
- 3Almost a third of the surveys were in February or March. 306 of the 987 I2M2 surveys were, and most of the rest in June, July or August. Insects hatch and leave the river at different times of year, so a February survey and a July survey can find different animals. Compare surveys from the same season, or say why you did not.
- 42025 is not complete in the file. The last survey in either file is 25 June 2025, and 2025 has 30 I2M2 scores against 84 in 2024. Stop at 2024, or say what you did with 2025.
- 5Keep size=10000 in the address. Without it you get the first 1,000 rows and no warning. The biology refuses more than 10,000 rows in one address.
- 6Some station-years have two scores, and some are not checked. 14 station-years have two I2M2 surveys on different dates, and 33 scores are non qualifié. Decide what to do with both before you match.
- 7River names are not unique. Five different rivers in the file are called Le Foron, and three are called La Morge. Group by code_station_hydrobio, never by the name of the station or the river.
- 8Take the year from the first four characters. The date is written 2025-06-25T00:00:00Z, which a spreadsheet may not read as a date, and the chemistry writes 2025-06-25. Make a year column in both files before you match.
- 9The index is calculated, not counted. I2M2 combines five measures of the invertebrates found, and IBD scores the diatom species by how much pollution each one tolerates. Read what each formula gives most weight to before you base a conclusion on it.
- Do Haute-Savoie's river stations with more AMPA, a breakdown product of glyphosate, have lower invertebrate index (I2M2) scores, 2014 to 2024?
- Do the Arve's stations with more nitrate have lower diatom index (IBD) scores, from Chamonix to the Swiss border, 2008 to 2024?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the invertebrate index (I2M2) changed at the Arve's four long-running stations, from Les Houches to Arthaz, between 2008 and 2024?
- Do the invertebrate index and the diatom index agree about which stations are in the worst state?
- code_station_hydrobio: the station's national (SANDRE) code, the same code as code_station in the chemistry files. libelle_station_hydrobio is its name and libelle_cours_eau its river
- date_prelevement: the day of the survey, written 2025-06-25T00:00:00Z
- resultat_indice: the score. I2M2, the Indice Invertébrés Multimétrique, runs from 0 to 1 (0 to 0.918 here). IBD, the Indice Biologique Diatomées, is a score out of 20 (8 to 20 here). For both, a higher score means better water
- code_indice and libelle_indice: which index the row is. 7613 is I2M2 and 5856 is IBD
- libelle_qualification: correcte (checked by the agency) or non qualifié (not checked)
- latitude and longitude: where the station is
Do Haute-Savoie's river stations with more orthophosphate have lower invertebrate index (I2M2) scores, 2008 to 2024?
Watch outMatch code_station in the chemistry to code_station_hydrobio in the biology: they are the same national codes. Take the year from the first four characters of each date, because the biology writes 2025-06-25T00:00:00Z. The chemistry has about four samples a year at a station and the biology one score, so make one chemistry value per station and year first, and when code_remarque is 10 remember that resultat is the limit, not a measurement. 913 station-years have both an I2M2 score and an orthophosphate value, from 191 stations, and 25 of those stations appear in ten years or more: average each station's years into one value as well, or the stations surveyed most often count most. With the diatom index it is 1,059 station-years from 2005, and with AMPA 341. For both indices a higher score is better water.
Do Haute-Savoie's river stations where PFOS was found have lower invertebrate index (I2M2) scores than stations where it was not, 2022 to 2024?
Watch outMatch code_station in the PFAS file to code_station_hydrobio in the biology, and take the year from the first four characters of each date. 211 station-years have a PFOS value and an I2M2 score, from 58 stations; with the diatom index it is 211 from 62 stations, and PFOA or PFHxA meet the I2M2 in 159. The limit of quantification fell at different times for each PFAS (for PFOS mostly 0.02 µg/l before 2022, then 0.002 or 0.001; for PFOA 0.002 from 2018 to 2024), so keep to years with one limit: for PFOS that leaves 85 station-years from 48 stations, 2022 to 2024, and for PFOA 98 with the I2M2, 2018 to 2024. Most results are below the limit, so compare two groups, found and not found, rather than drawing a line through the values, and count as found only results at or above the highest limit in your years (0.002 for PFOS in 2022 to 2024). Average each station's years into one value first. Many surveys were in February (41 of the 86 in the PFOS years): compare the same season, or say why you did not. Then read what level harms river life: the EU dossier behind the PFOS standard set 0.23 µg/l, from a test on midge larvae, insects of the kind the I2M2 counts. Count how many of your values reach it. PFAS reach rivers with other pollution from towns and factories, so a difference between the groups need not be caused by PFAS. For both indices a higher score is better water.
The comité de bassin Rhône-Méditerranée, which adopted it on 18 March 2022, and the préfet coordonnateur de bassin, who approved it on 21 March 2022. It is how France applies the Water Framework Directive (2000/60/EC, 23 October 2000) to the Rhône basin, which includes every river in Haute-Savoie.
Link to this fileThe most direct link here. The plan aims for good ecological status in 67.2% of the basin's rivers by 2027, and ecological status is judged partly on these two indices: an arrêté of 27 July 2018 made the I2M2 the invertebrate index for it, and the arrêté of 25 January 2010 names the IBD for diatoms. So these scores are part of how the plan's success is measured.
The tensionThe plan names something it cannot control. On pesticides it says it has no hold over "la compétitivité du prix des produits agricoles", the price competition that leads some farmers to use inputs such as pesticides "pour améliorer les rendements et sécuriser leurs revenus" (to raise their yields and protect their income).
SourceThe Commission locale de l'eau (CLE) of the Arve, with the SM3A, the Syndicat mixte d'aménagement de l'Arve et de ses affluents, as its technical body. The préfet of Haute-Savoie approved it.
Link to this fileDirect for the Arve and its tributaries, and indirect for the rest of the file. Four Arve stations (Les Houches, Magland, Ayse and Arthaz) have I2M2 scores in 17 or 18 of the years 2008 to 2025, nine or ten of them before 2018. Its measures on sewage are the ones most likely to reach the animals and algae on the river bed, but you would have to argue that link.
The tensionTourists against the river. The plan says treatment works must keep up with fast population growth and treat tourists' wastewater "en période de pointe hivernale, période de forte sensibilité des milieux récepteurs": most visitors come in winter, when the rivers that receive the treated water are at their lowest.
SourceThe préfet coordonnateur du bassin Rhône-Méditerranée, for the French State, under the EU Nitrates Directive (91/676/EEC, 12 December 1991). In a vulnerable zone, farmers must follow rules on how they use nitrogen fertiliser and how they manage their land.
Link to this fileIndirect. The zones act on nitrate from farms, not on the river bed. But the SDAGE's annex gives these three rivers a less strict objective for 2027 and names the invertebrates and the diatoms among the reasons, with nutrients as the pressure on the Aire and pesticides on the Vion and the Hermance. The file has I2M2 scores for the Aire at Saint-Julien-en-Genevois 2 in 17 years from 2008, the Vion at Excenevex 1 in 11 years from 2015, and the Hermance at Chens-sur-Léman in 6 years from 2019.
The tensionFarms or sewage. In the 2026 review, requests to take the Vion and the Hermance off the list argued that their pollution may come from sewage as well as from farms. The State kept both. For the Vion its report answers that "la contribution agricole ne peut être ignorée" (the part from farming cannot be ignored), because there are large cereal fields right beside the monitoring point (DREAL Auvergne-Rhône-Alpes, July 2026).
SourceBenthic invertebrates in Geneva's rivers
Whether the insect larvae, worms and crustaceans living on the river bed recovered as the water got cleaner, and where they did not.
- One row:
- One monitoring station in one year: a mean biological index score, the number of taxa found, and how many samples it rests on.
- Covers:
- 168 stations, 1995 to 2022
Full record5 traps, 1 on the three lines
- Usable rows
- 554
- Coverage
- 168 stations, 1995 to 2022
- How it arrives
- CSV inside a zip, semicolon-separated
- Language
- Page and file in French
- Human· Phosphate· by station code and year
- Human· E. coli· by station code and year
- 1The cleanest file in the family: 554 rows and 554 observations, no duplication at all. That is itself the lesson: two other files in this family are duplicated fiftyfold, and nothing on the outside tells them apart. Every file has to be checked on its own.
- 2The series stops in 2022. It sits alongside files running to 2025, so a comparison across the family silently ends three years early unless you notice.
- 3Scores rest on between 1 and 4 samples, and 46 of them rest on a single sample. Filter, or say why you did not.
- 4The index has a ceiling of 20, so improvement at an already-good station is compressed. A station moving from 17 to 18 is not comparable with one moving from 4 to 5.
- 5Station names are not unique. Group by CODEMESURE.
- A comparison of invertebrate index scores between renatured reaches and channelised ones.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the biological index at Geneva's river monitoring stations changed between 1995 and 2022?
- Does the number of taxa tell the same story as the index score?
- The mean biological index, running 2.0 to 18.25
- The number of taxa recorded
- A five-level quality class from Mauvaise to Très bonne
- The number of samples behind the mean, between 1 and 4
Do Geneva's river stations with more phosphate have lower invertebrate index scores?
Watch outHere a higher index is better water, the opposite of the diatom index. The invertebrate series stops in 2022, so the nutrient file's last three years have nothing to match; 408 station-years appear in both. Remove the −99s from the nutrient file first.
Do Geneva's river stations with more E. coli have lower invertebrate scores, or fewer kinds of invertebrates, 1995 to 2022?
Watch out387 station-years appear in both files, but they come from only 142 stations: most stations were surveyed in two or three of those years. Average each station's years into one value, or you count the same station several times over. Remove the repeated rows from the E. coli file first (each station-year appears about fifty times) and match on CODEMESURE and year, never the station name. A higher invertebrate score is cleaner water, so more E. coli with fewer invertebrates is a negative correlation. Choose the score (IBGN_MOYENNE) or the number of taxa (TAXON_SOMME) before you look. 39 station-years have E. coli of 0, probably too few to detect rather than none, and one reads 1,000, far above the rest: decide what to do with both and say so. The invertebrates stop in 2022.
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileThe contract's objectives are about water quality, and an invertebrate community reflects water quality over months, not just on the morning a sample is taken. It gives a slower answer to the same question.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileSewer separation reduces the pollution shocks that kill sensitive invertebrates, so a recovery in the community is one of the things the programme is for.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileThe best fit of any file in this family. Renaturation rebuilds the physical bed these animals live in, and the Aire is the most documented example of it in the canton.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceUK pollinators counted on flowers, in farmed and semi-natural squares, 2017 to 2022
Pollinating insects and farming. Volunteers watch one patch of flowers for ten minutes and count every insect that lands on it, in 1 km squares chosen at random from mainly farmed and mainly semi-natural land across the UK.
- One row:
- One ten-minute count: the insects that landed on one 50 cm by 50 cm patch of one kind of flower, in ten groups, with the square's land cover, the habitat, the flower and the weather.
- Covers:
- 92 randomly chosen 1 km squares in England, Scotland, Wales and Northern Ireland, 2017 to 2022
Full record8 traps, 2 on the three lines
- Usable rows
- 2,502
- Coverage
- 92 randomly chosen 1 km squares in England, Scotland, Wales and Northern Ireland, 2017 to 2022
- How it arrives
- A zip with two CSVs, the 1 km square counts and the counts by the public, and a Word document explaining every column
- Language
- Page and file in English
- Natural· Insects landing on flowersagainstHuman· Farmed or semi-natural land
- Natural· Bumblebees landing on flowersagainstHuman· Farmed or semi-natural land
- 1The flower decides a lot. Different flowers draw different insects: a count on hogweed is not a count on heather. 817 of the 2,502 counts are on a flower outside the scheme's list of 14. Choose one flower, or a few, before you look, and say which.
- 2Land cover goes with country. Scotland's 24 squares are mostly semi-natural (18) and England's 37 are mostly farmland (31). A difference between the two land covers may partly be a difference between Scotland and England: check it inside one country too, or say that you could not.
- 3The square is the unit, not the count. Counts in the same square are not independent of each other. Average each square's counts first, which gives you up to 92 values, not 2,502.
- 4land_cover and habitat_type are different things. land_cover is the whole square, taken from the Land Cover Map 2007 when the square was chosen. habitat_type is the spot where the patch was. 967 of the 1,457 counts in farmed squares were made in a semi-natural spot, such as a verge or a hedge. Only land_cover is the random design.
- 5The season and the weather change the count. Counts are only made when it is warm enough, but the month, the sunshine and the wind still vary. Compare squares over the same months, and look at the weather columns before you trust a difference.
- 62017 was a pilot, and 2020 was cut short. The pilot used slightly different categories, translated since, and covid cut 2020 to 212 counts in the squares. Start at 2018, and say what you did with 2020.
- 7Zeros are real. 215 counts saw no insects at all. Keep them: taking them out makes every average too high.
- 8The public file is bigger, but people chose where to count. It has 9,633 counts, 5,071 of them in gardens, and no land_cover column. A few characters, such as curly quotes, may come out garbled when you open it.
- Is the number of insects landing on flowers in a ten-minute count different in 1 km squares that are mainly farmland and squares that are mainly semi-natural, across the UK Pollinator Monitoring Scheme's squares, 2018 to 2022?
- Is the number of bumblebees landing on thistles in a ten-minute count different in mainly farmed and mainly semi-natural 1 km squares across the UK, 2018 to 2022?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Which insect groups make up most of the visitors to hogweed, and which to bramble?
- Do counts made in full sunshine record more insects than counts made under cloud?
- How does the number of insects in a ten-minute count change from May to September?
- land_cover: the main land cover of the whole 1 km square, agricultural or semi-natural (in the 1 km square file only)
- habitat_type: the habitat right around the patch: agricultural, garden, semi-natural or urban
- target_flower_corrected: the flower watched, as checked by PoMS staff from the count's photo
- bumblebees, honeybees, solitary_bees, hoverflies and six more groups: insects counted in the ten minutes
- all_insects_total: the ten groups added up
- cloud_cover, sunshine, wind_speed: the weather during the count
- date, year, country, x1km_square
The UK Government, through Defra, for England. Scotland, Wales and Northern Ireland have pollinator plans of their own.
Link to this fileDirect. It aims to give pollinators more food and nesting places, on farmland above all, and it committed the government to long-term pollinator monitoring; this scheme exists to provide evidence to the UK's four pollinator strategies. What the file cannot tell you is which squares had Countryside Stewardship options, so it can compare farmed and semi-natural land, not show whether the strategy worked.
The tensionPesticides. The strategy accepts that some pesticides, including neonicotinoids, "have the potential to damage pollinators", but relies on regulation and on pesticides being used "responsibly and sustainably". Parliament's Environmental Audit Committee, reporting on the draft in July 2014, said the precautionary principle should be "the overriding issue when it comes to permitting the use of neonicotinoids".
SourceDiatom algae in Geneva's rivers
Whether the microscopic algae living on the river bed report the same water quality the chemistry does, and what it means when they disagree.
- One row:
- One monitoring station in one year: a diatom index score, its quality class, and the number of sampling campaigns behind it.
- Covers:
- 180 stations, 1998 to 2025, with no missing years
Full record5 traps, 1 on the three lines
- Usable rows
- 503
- Coverage
- 180 stations, 1998 to 2025, with no missing years
- How it arrives
- CSV inside a zip, semicolon-separated
- Language
- Page and file in French
- Human· Phosphate· by station code and year
- 1The file publishes 24,759 rows and holds 503 observations. Every station-year is repeated, usually 50 times but sometimes 28, 32, 33 or 34, identical apart from an internal id. The duplication is not even consistent, so checking one group and assuming the rest match will still leave you wrong.
- 2Most scores rest on one or two campaigns, not twelve. That is far less underlying sampling than the chemistry files, so a single year's score is a noisier number than it looks.
- 3The index is calculated, not measured: a published formula turns a whole community into one value. Read what the formula gives most weight to before you base a conclusion on it.
- 4Because the scores come already sorted into five bands, it is tempting to analyse the bands. They are ordinal, so a mean of them means nothing. Use the numeric index and keep the classes for description.
- 5Station names are not unique. Group by CODEMESURE.
- A comparison of the diatom index between stations upstream and downstream of urban areas.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Do the diatom index and the chemistry agree about which stations are the worst?
- How has the diatom index at Geneva's river monitoring stations changed since 1998?
- DI-CH, the Swiss diatom index, running about 1.4 to 8.0
- A five-level quality class from Mauvaise to Très bonne
- Number of campaigns behind the score, usually one or two
- Station, watercourse, year and coordinates
Do Geneva's river stations with more phosphate have worse diatom index scores?
Watch outThe diatom index runs backwards: 1.4 is the cleanest water in the file and 8.0 the worst, so a positive correlation with phosphate means the diatoms get worse as phosphate rises. Say so in your write-up or your reader will take it the wrong way round. 420 station-years appear in both files; match on CODEMESURE and year, and remove the −99s.
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileA biological answer to the question the contract set itself. If assainissement worked, the community living on the river bed should say so as well as the chemistry does.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileDiatoms respond to nutrient enrichment, which is what treatment removes, so the link is a direct biological one.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileOne of the two best fits in this family. Renaturation rebuilds the physical habitat a bed community lives in, which is closer to what it acts on than a chemical concentration is.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceWolves in France, every piece of evidence since 2013
A protected top predator recolonising France from the Alps, and reaching farming country it had been wiped out of.
- One row:
- One piece of evidence: a sighting, droppings, tracks, a howl, a carcass. Each has a date, a department, a commune and an expert verdict on whether it was a wolf.
- Covers:
- 67,212 records, 1 January 2013 to August 2026, from 90 departments. 46,953 are confirmed as wolf, in 2,662 communes
Full record7 traps
- Usable rows
- 46,908
- Coverage
- 67,212 records, 1 January 2013 to August 2026, from 90 departments. 46,953 are confirmed as wolf, in 2,662 communes
- How it arrives
- One CSV of about 14 MB, returned by a map service query rather than a download button
- Language
- Page and file in French
- Human· Sheep on farms· by department and year
- 1Records measure searching as well as wolves. Confirmed records fell from 6,376 in 2024 to 4,408 in 2025, while the OFB's population estimate rose from 1,013 to 1,082. A count of evidence is not a count of animals, and the OFB's own estimate comes from genetics, not from this file.
- 2Keep only Retenu. Three records in ten are not a confirmed wolf: 7,206 rejected, 9,854 impossible to tell, 3,199 still being analysed. Every year has some still being analysed, so counts can change after you download.
- 3The location is blurred on purpose. Every point sits at the centre of its commune, and nothing appears until a month after it is found. Count by commune or department, never by distance.
- 4The evidence types are single letters, and the file does not say what they mean. From the genetics and the OFB's own list: V is a sighting (half the file), F droppings, T tracks, H a howl, U urine, P hair, S blood. C looks like the carcass of wild prey and D a dead wolf, whose yearly count tracks the culling. E is unexplained. Check against the OFB before you rely on any of them.
- 545 confirmed records are duplicates: the same reference code, date and commune under two ids. Remove them by the r column, which is how 46,953 becomes 46,908.
- 62026 is a part year, running to late August at the time of checking.
- 7Livestock attacks are not in this file. Those are a separate state system; this one is evidence of the wolf itself. The attacks are published by canton and year, 2010 to 2022: they have their own card. Joining the two needs INSEE's 2021 commune list, since this file names the commune but not its canton.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the number of French communes with confirmed wolf evidence changed between 2013 and 2024?
- When did confirmed wolf evidence become regular in the Ain, Doubs and Jura departments, and how fast did it grow after that?
- Is evidence reported by members of the public confirmed as wolf less often than evidence reported by OFB staff?
- date_indice: when the evidence was found
- d and c: department and commune. The map point is the centre of the commune, never the find itself
- f: the verdict. Retenu (confirmed wolf), Non retenu (not a wolf), Invérifiable (cannot tell), En cours d'expertise (still being analysed)
- ti: the type of evidence, as a single letter
- o: who reported it: OFB staff, national parks, hunters' federations, the forestry office, or a member of the public
- g_sp: the species a genetic test found, for evidence that was tested
Across French departments, is the number of confirmed wolf records related to the number of sheep kept on farms, 2013 to 2025?
Watch outStrip the leading zero from Agreste's three-character codes (004 becomes 04) and turn its year columns into one column first. 516 department-years appear in both, from 83 departments. The 2025 sheep figures are provisional, and 25 matched rows are in departments whose figure is quality 3. Sheep are counted at the farm in December, not on the summer pastures where wolves meet them, and a department is a coarse unit: say so.
The French government, through the Ministry of Ecological Transition and the Ministry of Agriculture, coordinated by the prefect of Auvergne-Rhône-Alpes.
Link to this fileDirect. The plan pays for herd protection and sets culling rules by where the wolf is established, and this file is part of the monitoring that decides where that is. Dead wolves also appear in it, so the culling shows up in the records.
The tensionFarmers' organisations argue the ceiling is too low to protect flocks; conservation groups argue that shooting a fifth of a recovering population each year is too high. Published after a public consultation, it was widely reported as satisfying neither.
SourceThe Council of the EU and the European Parliament, amending the Habitats Directive, after the Bern Convention's Standing Committee made the same change.
Link to this fileIndirect and recent. It gives member states more room to manage wolf numbers, but France's own rules decide what happens, and there is little data after it yet.
The tensionThe Commission cited growing populations and livestock losses. More than 200 environmental organisations wrote to EU environment ministers in December 2025 urging them not to use the lower status.
SourceUK butterflies, species by species, every year since 1976
Insect decline in farmed countryside. Trained volunteers have walked the same routes every week of the summer since 1976, counting every butterfly.
- One row:
- One species in one year in one country: its collated index of abundance, and how many sites it came from.
- Covers:
- 59 species for the UK, 1976 to 2023, plus England, Wales, Scotland and Northern Ireland separately
Full record7 traps
- Usable rows
- 2,649
- Coverage
- 59 species for the UK, 1976 to 2023, plus England, Wales, Scotland and Northern Ireland separately
- How it arrives
- A zip with one CSV of about 8,400 rows, and a Word document explaining every column
- Language
- Page and file in English
- Human· Dwellings added in England· by year
- 1It is not the Big Butterfly Count. That is the survey everyone has heard of, but it publishes only yearly headlines, such as 10.3 butterflies per 15-minute count in 2025. It is a good way into the issue. This file is the one you analyse.
- 2The index is not a count. It is the logarithm (base 10) of a relative index, scaled so that each species' average over the whole series is 2. An index of 2.3 means about twice the average year, not 2.3 butterflies. To compare species, compare their change, never their index values.
- 3Five countries in one file. Every species appears for the UK and again for England, Wales, Scotland and Northern Ireland. Filter to one COUNTRY first, or the UK is counted alongside its own parts.
- 4The number of sites grew enormously. The Peacock was indexed from 37 sites in 1976 and 2,859 in 2023. Early years rest on far fewer places, so say how many sites stand behind your first years.
- 5Species start in different years. 36 of the 59 start in 1976; others start as late as 2003 and 2009. Compare species over years they all share.
- 6Habitat specialist and wider countryside are JNCC's groups, not yours. Take the lists from JNCC's UK Biodiversity Indicator C6 (26 specialists, 24 wider countryside species) and cite it. Sorting species yourself would decide your result.
- 7Two versions. This file ends in 2023. The UKBMS official statistics spreadsheet covers 2025 but gives one trend per species, and its significance stars sit inside the numbers (-48***), which turns the cell into text.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Have UK habitat specialist butterflies changed in abundance between 1976 and 2023 more than wider countryside species?
- Is the change in abundance since 1976 different in England and in Scotland, for the species recorded in both?
- How has the abundance of one farmland butterfly, such as the Small Tortoiseshell, changed between 1990 and 2023?
- COMMON_NAME and SPECIES: the butterfly, in English and Latin
- YEAR
- COLLATED_INDEX: how abundant the species was that year, on a log scale where the species' own average is 2
- N_SITES: how many sites the index was calculated from
- YEAR_RANK: the year's rank for that species, 1 being its best year
- COUNTRY: UK, England, Wales, Scotland or Northern Ireland
Is the change in a butterfly species' England index from one year to the next related to the number of dwellings added in England that year, 1976 to 2023?
Watch outUse the England rows of the butterfly file and the All dwellings column on the sheet LT_104 of Live Table 104. Work out dwellings added as this year's count minus last year's, and look before you trust it: 1971 and 1981 have an extra census row (1 April) that splits a year in two, so drop those rows, and in 1991 the count moved from 31 December to 31 March, so the 1990 to 1991 step covers three months, not twelve. The move also changes which months a year means: up to 1990 a step runs January to December and includes that summer's butterfly counts, but from 1992 it runs April to March and ends before they begin. Check your result on 1992 to 2023 alone, where every step is the same twelve months: 32 changes. The steps for 1980 and 1987 are very different from the homes MHCLG counts as completed those years: decide what to do with them before you look. Stock figures are rounded to thousands before 2002, each census's correction is spread over the ten years before it, and 2022 and 2023 are provisional. Both series change steadily, so use the change against the change, not the level against the level. And a dwelling is not a hectare: homes built on land that was already built on take no new habitat, so say what you are assuming.
The UK Government, through Defra, for England only.
Link to this fileDirect. The target is measured by an index of 1,195 species that includes 55 butterflies, from monitoring schemes like this one. What you can test is the direction butterflies were heading when the target was set; the 2030 result is not in the data yet.
The tensionThe Office for Environmental Protection, the government's own watchdog, has reported that the government is off track to meet it. Environmental groups, through Wildlife and Countryside Link, argue that the target itself is too weak.
SourceDefra and the Rural Payments Agency, paying farmers for actions such as flower-rich margins and hedgerow management.
Link to this fileIndirect. The actions it pays for are meant to help insects on farmland, and a farmland butterfly index exists. But this file cannot tell you which sites were in the scheme, so it can show how farmland butterflies are doing, not whether the scheme is the reason.
The tensionThe National Farmers' Union's president called the closure "another shattering blow to English farms delivered, yet again, with no warning", and the NFU threatened legal action. The government said the money had all been committed.
SourceSalmon caught on every river in England and Wales, 2008 to 2025
Atlantic salmon disappearing from British rivers, including the Wye, where chicken manure and algal blooms have become a national story.
- One row:
- One river in one year: the number of salmon anglers declared they caught, including fish they released.
- Covers:
- 84 rivers, 2008 to 2025, in Table 34. Other tables give the latest season by month, weight, sea age and fishing effort
Full record7 traps
- Usable rows
- 1,512
- Coverage
- 84 rivers, 2008 to 2025, in Table 34. Other tables give the latest season by month, weight, sea age and fishing effort
- How it arrives
- One Excel workbook of 40 tables. Google Sheets opens it directly
- Language
- Page and file in English
- Human· Orthophosphate in the Wye· by year, for the Wye
- 1A catch is not a population. It depends on how many people fished, the rules, and the weather: in a dry summer rivers run low and salmon are hard to catch. Say that your data measures catch, and that the population is estimated separately in the annual salmon stock assessment.
- 2The Wye has not fallen most. Its average catch for 2021 to 2025 is 23% of its 2008 to 2012 average; for Wales as a whole it is 16%, and for the Usk 8%. If your question assumes the Wye stands out, this table may disagree. That is a finding, not a failure.
- 3Salmon are falling everywhere. Survival at sea has dropped across the North Atlantic, so a decline on one river does not show a cause on that river. Compare with other rivers before you blame anything local, including chicken farms.
- 4Effort is only for the latest season. Table 37 gives days fished for 2025 alone. To compare catch per day across years, take Table 37 from earlier editions of the same publication.
- 5Released fish are included. The figures count every salmon caught, kept or not, and on the Wye every salmon must now be released. A fish caught twice may be counted twice.
- 6The Wye is listed under Wales, although much of it runs through England. Net catches (the Wye lave nets) are in a separate table, Table 32.
- 7Do not let one year carry the trend. 2025's 60 fish is the lowest in the series. Use averages over several years, as in the comparison above, rather than first year against last.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the declared salmon rod catch on the River Wye changed between 2008 and 2025, compared with the total for Wales?
- Has the salmon rod catch fallen more on the Wye than on other Welsh rivers, comparing 2021 to 2025 with 2008 to 2012?
- Is the share of multi-sea-winter salmon in the 2025 rod catch different on the Wye from other large rivers?
- River, with the Environment Agency area or Wales it is listed under
- One column per year, 2008 to 2025: salmon caught by rod, released fish included
- Days fished and catch per licence day (Table 37): how hard anglers fished, for the latest season only
- Grilse and multi-sea-winter salmon (Table 38): how many returned after one winter at sea and how many after two or more
Were salmon rod catches on the Wye lower in years when summer orthophosphate in the Wye was higher, 2008 to 2025?
Watch outEighteen years is a small sample, and both series change over the period, so the raw values can correlate whether or not one affects the other. Correlate the change from one year to the next instead. Salmon spend one to three years at sea, where most of them die, so the river is one influence among several. You have to average the phosphate samples into one value per summer yourself.
Defra, announced by the farming minister, with a river champion and a taskforce for the catchment.
Link to this fileIndirect. The plan targets phosphate from poultry manure, which feeds the algal blooms blamed for the river's decline. Salmon catches cannot show whether it works, and they fall for other reasons too. Comparing the Wye with other rivers is the only way this file can separate a Wye problem from a national one.
The tensionThe Soil Association said the plan was likely to shift the problem elsewhere rather than reduce it, since the manure and the number of birds stay the same. Campaigners and residents have taken legal action against the poultry firms and the water company.
SourceNatural Resources Wales, with the Environment Agency applying the same rules on the English Wye.
Link to this fileDirect, and a trap. It is a response to the decline this table shows, and it changes what the table measures: every fish in the recent Wye figures was released.
SourceSignal crayfish in England's river invertebrate samples, 1965 to 2026
An invasive crayfish from North America, brought to England in the 1970s to be farmed for food. It escaped into rivers, and it carries crayfish plague, which kills the native white-clawed crayfish.
- One row:
- One sample: three minutes of kick sampling and a one-minute hand search at one river site on one day. The Taxa table has one row for each kind of animal found in it. The Metrics table has one row for the sample itself, whatever was found.
- Covers:
- 280,833 samples at 30,715 sites across England, 1965 to 2026. Signal crayfish is in 3,047 of them, at 1,056 sites, from 1985. The Loddon catchment alone is 1,564 samples
Full record12 traps
- Usable rows
- 255,699
- Coverage
- 280,833 samples at 30,715 sites across England, 1965 to 2026. Signal crayfish is in 3,047 of them, at 1,056 sites, from 1985. The Loddon catchment alone is 1,564 samples
- How it arrives
- A zip of three CSVs for the catchment you choose: Sites, Metrics (one row for every sample) and Taxa (one row for every kind of animal found). The whole of England is a 62 MB zip that no spreadsheet can open
- Language
- Page and file in English
- Human· Hours of storm overflow spills· by water body code and year
- 1Never type in the Taxon Name box. In the explorer it filters every table, not only Taxa. For the Loddon catchment, the Metrics table fell from 1,564 samples to the 103 with signal crayfish, so every sample without one disappeared. Download the catchment with no filter, then filter TAXON_NAME yourself in the spreadsheet.
- 2A crayfish that was not found has no row. The Taxa table lists only what was found. The Metrics table has one row for every sample, so it is your list of samples. Mark each sample 1 or 0 by looking up its ANALYSIS_ID among the crayfish rows, then work out the share. Do not match on site and date: a few sites have two samples on one day.
- 3Before 2014, many crayfish were named only as a family. For years most samples were identified only to family, so a crayfish could be written Astacidae, which covers both species. 2,142 samples have an Astacidae row, 1,990 of them before 2006, and 215 have both an Astacidae row and a signal crayfish row. Names to species were required everywhere only from 2014: the Invertebrate timeline in the documentation zip lists the changes. Start in 2014, or count Astacidae too and say why.
- 4Each crayfish has more than one name. 561 samples, all in the Midlands from 1985 to 1999, name the native crayfish only as Austropotamobius, and 8 name signal crayfish only as Pacifastacus. Search for the first word of the name, not the full species name, and say what you counted.
- 5The number of samples changed a lot. There were 16,505 samples in 1991, 12,665 in 2000 (a national survey), 1,378 in 2020 (no spring sampling, because of COVID-19) and 3,986 in 2025. Always divide by the samples taken. Never compare the number of crayfish rows between years.
- 6A kick sample is not a crayfish survey. Three minutes of netting catches what is on the riverbed. Crayfish hide under stones, in burrows and in deep water, and researchers who used these samples say they detect crayfish "at moderate densities" (Mathers and others, 2020). Not found does not mean not there. Crayfish are also easier to catch in some months than others, so compare spring with spring and autumn with autumn.
- 7Abundance is what the net caught. TOTAL_ABUNDANCE is the number found in the sample, counted when there are fewer than about 10 and estimated above that. Before 2002, some numbers were made from bands: 3 stands for A (1 to 9) and 33 for B (10 to 99). Use found or not found, not how many.
- 8All of England is too big for a spreadsheet. The bulk zip is 62 MB, but its Taxa file has 7.6 million rows (805 MB), which Excel cannot open. It holds no species names, only a code (TAXON_LIST_ITEM_KEY) to look up in a fourth file. Work one catchment at a time in the explorer. A drawn area can be up to 3,000,000 hectares.
- 9The API leaves some out. The Ecology and Fish Data API lists signal crayfish in 2,997 samples, and the bulk file has 3,047. The 50 missing are the ones with no abundance recorded. It also answers in JSON, not as a table.
- 10Two kinds of catchment. The explorer's list is the Water Framework Directive operational catchments, drawn on the map. The CATCHMENT column in the Sites table is an older name for each site, so a Loddon download also holds sites marked BLACKWATER (THAMES) and WEY (THAMES). Say which you used.
- 11The explorer opens on 2020. Change Year From before you press Load data, or you only get the last few years. The data are refreshed every month, so record the date you downloaded.
- 12The water body code can be an old one. The documentation warns that WFD_WATERBODY_ID "may also relate to the WFD cycle 1 set of waterbodies or cycle 2 waterbody IDs". Other files, such as the storm overflow returns from 2024, use Cycle 3 codes. Most codes are the same in every cycle, but a few will not match: count how many sites you lose.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In what share of the Environment Agency's river invertebrate samples in the Loddon catchment was signal crayfish found, in each year from 2014 to 2025?
- Is signal crayfish found in a larger share of river samples in the Loddon catchment than in the Wensum catchment, 2014 to 2025?
- In the Wensum catchment in Norfolk, in which years and at which sites were white-clawed crayfish and signal crayfish found in river samples, 1989 to 2025?
- TAXON_NAME, in the Taxa table: the animal found. Signal crayfish is Pacifastacus leniusculus and white-clawed crayfish is Austropotamobius pallipes
- ANALYSIS_ID: the number that links each Taxa row to its sample in the Metrics table
- SAMPLE_DATE and SAMPLE_DATE_YEAR, in the Metrics table: when the sample was taken
- SITE_ID, WATER_BODY (the river's name) and NGR_10_FIG (the grid reference). There is no site name
- CATCHMENT and REPORTING_AREA, in the Sites table: where the site is
- TOTAL_ABUNDANCE and ABUNDANCE_CATEGORY: how many were found in the sample, and the band that number falls in (A is 1 to 9, B is 10 to 99)
- SAMPLE_METHOD_DESCRIPTION and ANALYSIS_METHOD_DESCRIPTION: how the sample was taken and how it was counted
- WHPT_ASPT, in the Metrics table: the sample's invertebrate score, the average score of the kinds of animal found. Animals that need clean water score high, so a higher score means cleaner water
- WFD_WATERBODY_ID, in the Sites table: the code of the Water Framework Directive water body the site is in
Do England's river water bodies where storm overflows spilled for more hours have lower invertebrate scores (WHPT ASPT), 2021 to 2025?
Watch outAdd up the hours of every overflow in a water body first: one water body can have dozens. Match WFD Waterbody ID in the overflow file to WFD_WATERBODY_ID in the invertebrate Sites table, reach the Metrics table through SITE_ID, and take the year from SAMPLE_DATE. Keep river codes only (GB1 and eleven digits). 870 river water bodies have both in 2025, and 3,658 water body-years from 2021 to 2025. In 693 of the 2025 ones, every overflow's monitor worked for at least 90% of the year: keep to those, or say why not. Average one season's samples (spring or autumn) in each water body, because the score changes with the season. A higher score is better water. The match only sees the water body the overflow runs into, but sewage also flows on downstream into the next one. A water body with samples and no overflow listed is not always a zero: some overflows have no proper code.
The UK Government (Defra) and the Welsh Government, for England and Wales. In England, Natural England grants the licences the Order allows, and the Environment Agency gives permission to trap crayfish.
Link to this fileIndirect. It controls what people do with live crayfish, which is how the species reaches new rivers, but it does not remove crayfish that are already there. It began in December 2019, and 2020 has few samples, because spring sampling stopped for COVID-19. The file can show where signal crayfish were found before and after it began. It cannot show that the Order caused a change.
The tensionEating them. Celebrity chefs and conservation charities have promoted trapping signal crayfish to eat, and licensed commercial trapping is still allowed. Researchers at UCL and King's College London found that trapping does not control them, because most are too small to trap. They warned that "trapping to eat" "inadvertently incentivises members of the public to spread the species to new habitats" (British Ecological Society, 13 October 2020).
SourceDefra, the Scottish Government and the Welsh Government, co-ordinated by the GB Non-native Species Secretariat.
Link to this fileIndirect. It covers every invasive species in Great Britain and sets no target for signal crayfish, which it names as one whose impact is "well known". Its target is about new species becoming established, not about one that is already here. It says prevention matters most "in aquatic environments where control and eradication are technically challenging", which is the case for crayfish in rivers.
SourceThe Ministry of Agriculture, Fisheries and Food, for England, and the Welsh Office, under the Import of Live Fish (England and Wales) Act 1980.
Link to this fileDirect in aim, but hard to test with this file. Ministers made the Order because these crayfish "might compete with, displace, prey on or harm the habitat of" native fish and shellfish, and the areas limited where signal crayfish could be kept. But the areas are postcode districts, and this file gives grid references and catchments, not postcodes. Matching samples to zones is work you would have to do yourself and explain.
SourceThe Environment Agency, with Yorkshire Water, Yorkshire Wildlife Trust and the University of Leeds.
Link to this fileDirect for the native crayfish, but outside this file. The arks are a pond and a reservoir, and this file holds river samples. What it can show is where white-clawed crayfish were still being found in rivers, and where signal crayfish had been found, which is the problem the arks respond to.
SourceBrown and white shrimp landed in Louisiana and Texas, year by year since 1962
The Gulf of Mexico shrimp fishery and the dead zone. Brown shrimp live on the seabed of the shelf where the oxygen runs out each summer, and the boats that catch them land their shrimp in Louisiana and Texas.
- One row:
- One kind of shrimp landed in one state in one calendar year: its live weight in pounds, and what the boats were paid for it at the dock, in dollars of that year.
- Covers:
- Brown and white shrimp landed in Louisiana and in Texas, 57 years each: 1962 to 1971 and 1978 to 2024. The same query gives Mississippi, Alabama and Florida's Gulf coast
Full record12 traps, 1 on the three lines
- Usable rows
- 228
- Coverage
- Brown and white shrimp landed in Louisiana and in Texas, 57 years each: 1962 to 1971 and 1978 to 2024. The same query gives Mississippi, Alabama and Florida's Gulf coast
- How it arrives
- A query form with a CSV download, no account. Choose Commercial, every year, your states and your species, press RUN REPORT, then Actions, Download, CSV. The file arrives as FOSS_landings.csv, one row for each species, state and year
- Language
- Page and file in English
- Human· Nitrate load to the Gulf· by year: the calendar year of the landings against the USGS water year (October to September) of the same number
- 1A landing is not a catch, and a year is not a season. FOSS counts shrimp in the state where they crossed the dock, not where they were caught, so a Texas boat landing in Louisiana counts for Louisiana. Shrimp from the bays and from the open sea are added together, and FOSS gives only whole calendar years: no months, no gear, no fishing area.
- 2No species in 1972 to 1977, and none before 1962. In those years FOSS gives the Gulf states only SHRIMPS, PENAEOID (1972 to 1977) and SHRIMP, MARINE (1950 to 1961): every kind of shrimp added together. Brown shrimp starts in 1962 and has a six-year gap. Leave the gap, say so, and never share out a total yourself.
- 3Search for the right name. Type shrimp in Search Species and press Search. Brown shrimp is Shrimp, Brown, but white shrimp is Shrimp, Northern White: there is no Shrimp, White. A name ending in ** is a group, shrimp that were not named to species. Many papers call brown shrimp Farfantepenaeus aztecus, and FOSS calls it Penaeus aztecus. They are the same animal.
- 4The first line is empty, and the numbers have commas. FOSS_landings.csv starts with a line of commas, and the headings are on line 2. Pounds are written with commas between the thousands (8,033,120). If your computer uses a comma for decimals, as Swiss and French settings do, import these columns as text and take the commas out, or the numbers come out wrong.
- 5Live weight, not tails. FOSS gives pounds of live weight, with the heads on. The Gulf Council counts shrimp in pounds of tails, with the heads off, so the figures in its documents are lighter. Never mix the two, and say which you used.
- 6Pounds are the shrimp; dollars are the market. Dollars are what the boats were paid, in money of that year. They move with inflation, with the size of the shrimp and with the price of imported farmed shrimp. For nature, use pounds. The Report Format TOTALS BY YEAR/SPECIES DEFLATED VALUE gives dollars in 2024 money, but FOSS does not say which price index it uses, and that format drops the State column and adds all the states you chose together. Run it for one state at a time.
- 7Fewer boats, not only fewer shrimp. Landings also follow how hard people fished, and FOSS has no column for that. Federal Gulf shrimp permits are capped by a moratorium, extended in 2016 for ten more years. Fuel is expensive, and imported farmed shrimp keep prices low: in 2025 Texas's agriculture commissioner said that 90% of the shrimp sold in the US was farm raised abroad (TPR). Say this in your evaluation.
- 8Storms and the oil spill. Hurricanes Katrina and Rita (2005) wrecked boats, docks and processing plants, and Gustav and Ike (2008), Laura (2020) and Ida (2021) hit the same coast. In 2010 large parts of the Gulf, and of Louisiana's waters, were closed to fishing after the Deepwater Horizon oil spill. Look at these years rather than deleting them, and say what you did with them.
- 9Scientists have found the dead zone hard to see in landings. Brown shrimp move away from low oxygen and gather at its edges (Craig, 2012), and the boats follow them (Purcell and colleagues, 2017). Smith and colleagues (2017) wrote that "quantifying effects of hypoxia on fisheries using quantity data has been inconclusive". Read them before you choose your state and your years, and use them in your evaluation.
- 10Each year is one point, and landings drift. Over six decades the fleet, the prices and the rules changed as well as the shrimp. Correlate the change from one year to the next as well as the values.
- 11Florida has two coasts. FOSS's Metadata and Caveats page says Florida comes as one state, but the form now lists FLORIDA-EAST (the Atlantic) and FLORIDA-WEST (the Gulf). For the Gulf, use FLORIDA-WEST only.
- 12If the page is blocked or busy. In an ordinary browser, www.fisheries.noaa.gov/foss opens the Landings form. If it says Too Many Requests, the server is busy: wait a minute and try again. The same form is also at apps-st.fisheries.noaa.gov. FOSS updates its data every week, so record the date you ran the query and keep your file.
- Across the years 1979 to 2024, is the weight of brown shrimp landed in Louisiana related to the nitrate the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Were fewer brown shrimp landed in Louisiana and Texas in the summers when the Gulf dead zone was larger, 1985 to 2021?
- Has the weight of brown shrimp landed in Louisiana changed since 1978 in the same way as the weight of white shrimp?
- Which Gulf state lands the most brown shrimp, and has that changed since 1962?
- Year, State and NMFS Name: the calendar year, the state where the shrimp were landed, and the kind of shrimp, SHRIMP, BROWN or SHRIMP, NORTHERN WHITE
- Pounds: the live weight landed, with the heads on. Metric Tons is the same weight in tonnes, rounded to whole tonnes
- Dollars: what the boats were paid at the dock (the ex-vessel value), in dollars of that year, not adjusted for inflation
- Confidentiality: Public on all 228 rows. Confidential marks a value that is hidden or added into a group, which happens to some other shrimp names
- Scientific Name and Tsn: Penaeus aztecus (brown, 95605) and Penaeus setiferus (white, 95610). Source is the database the row came from: GULFFINHIST up to 2003, GULFFIN from 2004
Across the years 1979 to 2024, is the weight of brown shrimp landed in Louisiana related to the nitrate the Mississippi and the Atchafalaya carried to the Gulf in the same water year?
Watch outMatch the landings Year to the USGS WY: 46 years have both, 1979 to 2024. A landing year runs from January to December and a water year from October to September, so they share nine months. Nitrate does not act on shrimp directly: any link runs through the dead zone, and the same river water also feeds the plankton and the seabed life that shrimp eat, so more nitrate could mean more food as well as less oxygen. Landings also follow the boats, fuel and the price of imported shrimp, the storms of 2005, 2008, 2020 and 2021 and the oil spill closures of 2010, and neither file can separate them. Scientists who studied this found the effect hard to see in landings. Much of Louisiana's brown shrimp is caught in the bays in May and June. Both series may drift over the decades, so correlate the change from one year to the next as well: that gives 45 changes.
The Gulf of Mexico Fishery Management Council (the Gulf Council) wrote the plan. NOAA Fisheries closes the federal waters off Texas, from 9 to 200 nautical miles out, and the Texas Parks and Wildlife Department closes Texas's own waters. The department's sampling of young brown shrimp decides the dates.
Link to this fileDirect: it is aimed at the animal on this card. NOAA closes the waters "to allow brown shrimp to reach a larger and more valuable size prior to harvest, and to prevent waste of brown shrimp that might otherwise be discarded due to their small size". But it changes when Texas's brown shrimp are caught more than how many there are, and this file only gives whole years. Louisiana's waters do not close, so Louisiana can be your comparison. Before 1981 the file has only 13 years of brown shrimp (1962 to 1971 and 1978 to 1980), and much else changed after 1981, so say what else could explain a difference.
SourceThe Mississippi River/Gulf of America Watershed Nutrient Task Force: federal agencies led by the EPA, one agency from each of 12 states (ten along the Mississippi and two along the Ohio), and the tribes, through the National Tribal Water Council. In 2026 the EPA and Iowa's Secretary of Agriculture are its co-chairs.
Link to this fileIndirect: two steps away from the shrimp. The plan acts on farms and towns upstream, on the nitrate in the USGS card's file, to shrink the dead zone. It sets no target for shrimp, although its Quality of Life goal names "in particular the agriculture, fisheries and recreation sectors" among those it hopes to help. So it belongs on line 3 because it changes line 1, the nitrate. Landings could only show it through the nitrate and then the dead zone, and they also follow boats, prices and storms: say so.
The tensionVoluntary action or legal limits. In 2008 thirteen groups petitioned the EPA to set legal limits for nitrogen and phosphorus in the basin's waters. The groups, the Mississippi River Collaborative, said the petition "chronicled EPA’s failed attempts to address the problem through workgroups, collaboration, guidance to states, and 'action plans' that never resulted in any action" (MRC). The EPA said no on 29 July 2011: "We do not believe that the comprehensive use of federal rulemaking authority is the most effective or practical means of addressing these concerns at this time" (EPA letter). A federal court upheld the EPA's refusal in 2016.
SourceThe Louisiana Wildlife and Fisheries Commission sets the seasons for the state's waters, on the advice of the biologists of the Louisiana Department of Wildlife and Fisheries (LDWF). It also hears the Louisiana Shrimp Task Force and other stakeholders.
Link to this fileDirect on when Louisiana's brown shrimp are landed, not on how many there are. An early or a late opening moves the catch within the year, and this file only gives whole years. The season is in the bays and marshes, not on the shelf where the dead zone forms: keep that in mind when you explain what you find.
The tensionBiologists against shrimpers, and shrimpers against each other. On 5 May 2016 the Commission chose one opening date, 23 May, for the whole coast. SeafoodSource reported that this "opposes Louisiana Department of Wildlife and Fisheries data", which showed shrimp in the Barataria, Timbalier and Terrebonne basins reaching harvestable size by 16 May. Dean Blanchard, who owns a dock on Grand Isle, said "The brown-shrimp season should never be opened until the first two crops leave Grand Isle". David Chauvin, in Dulac, wanted to open on 9 May: "We would like to start earlier" (SeafoodSource, 9 May 2016).
SourcePossums, rats and forest birds before and after a 1080 poison drop
Introduced possums and rats eating the eggs, chicks and food of New Zealand's forest birds, and the aerial poison used to kill them.
- One row:
- Birds: one species in one five-minute count at one station. Mammals: one species on one line of ten chew cards on one visit.
- Covers:
- Four South Island forests, two poisoned with 1080 in 2012 and two left alone, counted in winter 2012 and the three summers after: 2,006 bird counts and 240 chew-card lines
Full record8 traps, 3 on the three lines
- Usable rows
- 2,006
- Coverage
- Four South Island forests, two poisoned with 1080 in 2012 and two left alone, counted in winter 2012 and the three summers after: 2,006 bird counts and 240 chew-card lines
- How it arrives
- Two .txt files that are really CSV: the link is the bird counts; the mammal chew cards (pestrelabund.txt) are on the same page
- Language
- Page and file in English
- Natural· Birds counted, one speciesagainstHuman· 1080 poison drop (poisoned or not)
- 1Most rows are zero. Every count lists all 38 species, so 67,148 of 76,228 rows read 0. That is real data (no birds of that species seen), not a gap. Average across counts, don't drop the zeros.
- 2Count the counts, not the rows. One five-minute count is one Site, Transect, Station, Date, Time and Observer together. The number of counts varies from 40 to 179 per forest per season, so compare birds per count, never totals.
- 3Before is winter, after is summer. Birds are easier to hear in summer, so every forest goes up after winter 2012, poisoned or not. Compare the treated forest with its untreated partner, not with its own winter.
- 41080 kills rats and possums. Rats are on the chew cards too, and on the West Coast they were back on 48% of cards by 2014/15. You can't say the birds responded to possums alone. And watch the mice: on the East Coast treated site they rose to 82% of cards.
- 5Two names, one bird. dunnock and hedge_sparrow are the same species, entered under both names. Add them together. finch and unknown are birds nobody could identify: leave them out and say so.
- 6Which birds are native? Silvereye is the most counted bird in the file (3,308) and arrived from Australia by itself in the 1850s, so it is usually treated as native. On the West Coast most of the treated forest's late gain is silvereye. Decide before you count, and check your answer both ways.
- 7Two forests per side. Hundreds of counts come from just two poisoned and two untreated forests, and most stations were counted twice on the same day. The counts are not independent forests. Say so in your evaluation.
- 8Dates are day/month/two-digit year (7/12/12 is 7 December 2012), and the files end .txt but are comma-separated.
- Is the number of native birds per five-minute count higher in the 1080-treated forests than in the untreated forests, in the third summer after the drop?
- How did the share of chew cards bitten by possums change in treated and untreated forests between winter 2012 and summer 2014/15?
- Did bellbirds increase more in the treated forests than in the untreated forests after the 1080 drop?
- species, count: the bird and how many were seen or heard in the five minutes. Every count lists all 38 species, so most rows are 0
- Area, Site, treatment: East or West Coast, the forest, and whether it was poisoned (treatment) or not (control)
- Season: Winter 2012 (before), then Summer 2012/13, 2013/14 and 2014/15
- Transect, Station, Date, Time, Observer: which count it was; together they identify one five-minute count
- Sun, Wind, Rain, Temp, Noise: conditions during the count, which change how many birds you hear
- Chew cards: species (possum, rat, mouse), total_cards and cards_marked, so the share bitten is marked divided by total
The Department of Conservation and OSPRI (for bovine TB), with regional councils. Cereal baits containing sodium fluoroacetate are spread from helicopters over forest.
Link to this fileDirect: this dataset is one of those drops, with forests left untreated for comparison. In the poisoned forests possum chew cards fell to 0-2% while the untreated forests rose to 68-96%. It kills rats too, so anything the birds do after a drop is a response to both.
The tensionIt is one of New Zealand's most argued-over environmental policies. The Commissioner backed more of it; opponents object to poison spread from the air over land, water and hunting areas, and about 40% of New Zealanders have been reported as opposed to using poison on invasive animals.
SourceThe New Zealand government, through the Department of Conservation and the company Predator Free 2050 Ltd, with community trapping groups and iwi.
Link to this fileAimed at exactly the animals on these chew cards. This dataset shows the problem it faces: one drop, and rats on the West Coast site were back on 48% of cards within three summers. A one-off operation suppresses; it does not eradicate.
The tensionWhether it can be done. The Prime Minister said in 2016 that it would need a scientific breakthrough, and some ecologists have called it a flawed policy that draws money from better-evidenced conservation.
SourceOSPRI, funded by farmers and the government under the national bovine tuberculosis pest management plan. Possums are the main wild carrier of TB to cattle and deer.
Link to this fileIndirect. It is the same possums and often the same 1080, but its goal is farm disease, not birds. Use it to show that possum control is paid for by farming as well as conservation, which is part of why it happens at this scale.
SourceLionfish spreading across the Mediterranean, every published record since 1991
A venomous Red Sea predator that came through the Suez Canal and has spread west and north across the Mediterranean since 2012, eating small native fish, into seas that are warming.
- One row:
- One published record of a lionfish at one place and date, with the paper it came from.
- Covers:
- 2,019 lionfish records in 15 countries, 1991 to 2025, all with coordinates
Full record6 traps
- Usable rows
- 2,019
- Coverage
- 2,019 lionfish records in 15 countries, 1991 to 2025, all with coordinates
- How it arrives
- A zip: the records are occurrence.txt, tab-separated. Filter scientificName to Pterois miles
- Language
- Page and file in English
- Human· Vessels arriving in main ports· by country, and the year of the first lionfish record
- 1Records are not population. There are 605 records for 2020 and 9 for 2023, because of when scientists published, not how many lionfish there were. Use where and when lionfish were first recorded, never the number of records in a year.
- 2No record is not proof of absence. Presence-only data can show that lionfish reached a place, not that they are missing from one. For your "not recorded" places, choose coasts with plenty of divers and scientists, and say that is what you did.
- 3Memories, not sightings. Records from Spain (2002), Turkey (2005) and northern Greece (2008) come from studies that asked dive centres and locals when they first saw lionfish. Check associatedReferences, and decide whether to keep them before you start.
- 4Twenty years of nothing. The first record is Israel, 1991, then nothing solid until Lebanon in 2012. A lag like this is common in invasions, so don't draw a straight line from 1991.
- 5Year-only dates. 1,178 of the 2,019 records give only a year. That is fine for first records by country, not for anything by month.
- 6Use this copy. The same database on SEANOE is the older 2022 version: 4,015 records up to 2020, with commas as decimal points. The GBIF archive is updated to January 2026.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is the February sea surface temperature higher at places where lionfish have been recorded than at places in the Mediterranean where they have not?
- Did each country's first lionfish record come when its February sea surface temperature had passed about 15 °C?
- How far north has the northernmost lionfish record in the Mediterranean moved each year since 2012?
- scientificName: the species. The file holds 234 exotic fish; keep Pterois miles
- eventDate: when it was seen. Year only in 1,178 rows, year and month or a full date in the rest
- country, locality, decimalLatitude, decimalLongitude: where
- individualCount: how many were seen, blank in 375 rows
- associatedReferences: the paper each record comes from. Read it for any record that looks wrong
Is the year lionfish were first recorded in a Mediterranean country related to the number of vessels arriving at its main ports?
Watch outOnly 8 of the 15 lionfish countries have a traffic series: Israel, Lebanon, Egypt, Tunisia, Libya, Albania and Syria are not in Eurostat. Match by hand, because Eurostat writes Greece as EL and Turkey as Türkiye. The first records of Greece (1995) and Turkey (2005) fall before their traffic series begin, so only 6 first-record years have a traffic figure. That is a small sample: describe it rather than lean on a test, and say that first-record years rest on single early records. Lionfish swam in through the Suez Canal; shipping is not how they arrived.
An EU LIFE project led from Cyprus with the University of Plymouth, the Marine and Environmental Research Lab and Enalia Physis, working with Cyprus's Department of Fisheries and Marine Research.
Link to this fileLocal, and only indirectly about temperature: removal holds numbers down at a few sites but cannot stop a spread that ORMEF shows reaching Croatia by 2021. Use it to argue what control can and can't do against an invader the warming sea keeps favouring.
The tensionSpearfishing with scuba is forbidden by law across the EU, to protect native fish, so the removal teams needed a special supervised permit. Removing an invader meant relaxing a rule written to protect the species it threatens.
SourceThe European Commission and member states. A species on the Union list must be prevented, detected early and managed in every member state.
Link to this fileDirect in aim, but slow. The linked 2021 paper argues that the law, built for species that arrive by trade, struggles with a fish that swims in through a canal. Check the current Union list before you write about it.
The tensionListing requires a detailed risk assessment and a check that the costs are not disproportionate, which takes years; the fish spread from Lebanon to Italy in four. Kleitou and colleagues (2021) call this a limitation of EU law for marine invasions.
SourceSharks caught off Queensland's beaches, month by month since 1996
Sharks moving along the coast with the seasons, following the water temperature they prefer, and what a warming ocean could do to when and where they turn up.
- One row:
- One kind of animal caught on one kind of gear at one beach in one month, with how many were caught.
- Covers:
- Every month from January 1996 to May 2026, at 181 beaches in 12 areas from Cairns to the Gold Coast: 23,118 rows, 14,776 of them sharks up to 2023
Full record6 traps, 1 on the three lines
- Usable rows
- 14,776
- Coverage
- Every month from January 1996 to May 2026, at 181 beaches in 12 areas from Cairns to the Gold Coast: 23,118 rows, 14,776 of them sharks up to 2023
- How it arrives
- One Excel file, one sheet, headings in the first row
- Language
- Page and file in English
- Natural· White sharks among the sharks caughtagainstHuman· Gear: net or drumline
- 1The gear changed in 2024. All sharks caught ran at about 500-850 a year until 2022, then 948, 1,497 and 3,430 in 2023-25, as new locations, daily servicing and new drumlines came in. That is effort, not more sharks. Stop long-term trends at 2023, or compare months and species within the same years.
- 2Catch is not the number of sharks. It depends on how much gear is in the water, where and how often it is checked. Say what your variable is: sharks caught.
- 3Add up NumberCaught, don't count rows. 4,003 rows record more than one animal.
- 4White sharks are rare. 214 in 30 years, almost all south of Bundaberg. Pool the years for monthly patterns rather than testing year by year.
- 5Names with a star are a group, not a species. HAMMERHEAD SHARK * means a hammerhead nobody identified to species; there are also 81 UNKNOWN SHARK. Decide whether to include them.
- 6The 2026 rows stop in May. A part year: leave it out of any yearly total.
- Do nets and drumlines differ in the share of their shark catch that is white sharks, in the four areas that set both gears, 1996 to 2023?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Do white sharks and bull sharks peak in different months on the southern beaches, and does that match the water temperature each prefers?
- Are bull sharks caught in a higher share of months in the southern areas in the warmest years than in the coolest, 1996 to 2023?
- Year, Month: when
- Area, BeachName: where, in 12 areas from Cairns (about 17°S) to the Gold Coast (28°S)
- Gear: Net or Drum (a baited hook on a buoy)
- SpeciesGroup: SHARK, TURTLE, MAMMAL or OTHER. CommonName and ScientificName: the species
- NumberCaught: how many. Alive/Deceased and Fate: whether the animal was released, died or was killed
The Queensland Government, through the Department of Primary Industries: nets and drumlines off about 85 beaches, serviced by contractors.
Link to this fileDirect: it produces this dataset, and which gear to set, nets or drumlines, is its own decision. It aims at bather safety, not at sharks' movements, so it bears on a temperature question only as the way the data is collected: when the gear changes, the catch changes, whatever the sharks do.
The tensionSwimmers' safety and a beach economy against wildlife: the file records 1,456 turtles and 590 marine mammals. The government's own KPMG review advised moving away from nets and drumlines; the 2025 plan expanded them, and scientists said culling does not make beaches safer.
SourceHumane Society International with the Environmental Defenders Office brought the case; the Administrative Appeals Tribunal decided it; Queensland now runs catch alert drumlines that send a satellite alert so a team can release the shark.
Link to this fileIndirect for temperature, but a trap for your data: more sharks are now released alive and the new gear may catch more, which Queensland itself says needs analysis. It is a reason to stop your trend at 2023.
The tensionThe tribunal found the evidence "overwhelming" that killing sharks does not reduce the risk of bites, against a state government that fought the ruling in court.
SourceSharks caught in the Atlantic by every fishing nation since 1950, and the ban on landing makos
Sharks caught as bycatch by longline fleets fishing for swordfish and tuna, which countries' fleets catch them, and whether banning the landing of an overfished shark reduces how many die.
- One row:
- The tonnes of one species that one fleet caught with one kind of gear in one area in one year, either landed or thrown back dead.
- Covers:
- Every year from 1950 to 2024, 80 flags and 99 kinds of shark: 21,516 shark rows, 15,556 of them with a catch above zero
Full record8 traps, 3 on the three lines
- Usable rows
- 15,556
- Coverage
- Every year from 1950 to 2024, 80 flags and 99 kinds of shark: 21,516 shark rows, 15,556 of them with a catch above zero
- How it arrives
- A zip holding one Excel workbook. The data is on the sheet called Data, headings in the first row; the other sheets are notes and code lists
- Language
- Page and file in English
- Human· GDP per person (PPP)· by country and year
- 1Don't read the Pivot sheet. It is a ready-made summary that opens filtered to one small fleet (UK-Turks and Caicos). Work from the Data sheet.
- 2Choose the stock. Shortfin mako has three stocks in the file (ATN, ATS, MED) and the 2022 ban applies to ATN, the North Atlantic, only. Filter on Stock before you add anything up.
- 3Add the dead discards to the landings. When keeping a species is banned, a fleet's catch can move from L to DD rather than stop: Spain's North Atlantic mako is 870 t landed in 2020 and 585 t discarded dead in 2021. Reading L alone mistakes a change of label for a change in fishing. Sharks released alive are not recorded here.
- 4More countries reported over time. Two flags reported North Atlantic mako in 1975 and 19 by 2020, and in the 2000s over 90,000 t of sharks were recorded only as a group (SHX, SKH). Some of the rise before 2000 is better reporting, not more fishing: compare fleets that reported throughout, or start later.
- 5Tonnes caught is not the number of sharks alive. It depends on how much fishing was done: most of these makos are caught on longlines set for swordfish (Spain's biggest mako fleet is coded EU.ESP-ES-SWO). A country with a big longline fleet catches more, whatever its wealth.
- 65,960 rows are zero. They are reports of no catch, not missing data. Keep them for counting who reported; leave them out of averages.
- 7Estimated rows. CatchSource marks 1,113 shark rows as Estimated: ICCAT's scientists filled a gap, often by carrying the last reported value forward. Decide whether to keep them and say so.
- 82024 is the latest year and may still be revised. Download the current version and cite its date.
- Across the countries whose fleets catch North Atlantic shortfin mako, is the tonnage caught related to the country's GDP per person, 1990 to 2021?
- After the ban on keeping North Atlantic shortfin mako in 2022, did the total dead catch (landed plus dead discards) fall below ICCAT's 250 t limit, and what share of it was thrown back dead?
- Did porbeagle landings fall more in the North-east Atlantic, after the EU's zero catch limit in 2010, than in the North-west Atlantic over the same years?
- Species, ScieName, SpeciesGrp: what was caught. Sharks are the groups 4-Sharks (major) and 5-Sharks (other). SMA is shortfin mako, BSH blue shark, POR porbeagle
- Stock: which population. ATN is the North Atlantic, ATS the South Atlantic, MED the Mediterranean (porbeagle is split four ways)
- YearC: the year
- FlagName, FleetCode: which country's boats, in words rather than a standard code. EU members appear separately, e.g. EU-España
- GearGrp: LL is longline, the gear that catches most of these sharks
- CatchTypeCode: L for landed, DD for dead discards (thrown back dead)
- Qty_t: the catch in tonnes
Across the countries whose fleets catch North Atlantic shortfin mako, is the tonnage caught related to the country's GDP per person, 1990 to 2021?
Watch outICCAT names each fleet's flag in words (EU-España, Maroc, UK-Bermuda) and the World Bank uses three-letter codes, so write the code beside each flag yourself. Three flags have no GDP to match: Chinese Taipei (not in the World Bank's list), Venezuela (no value in this series) and St Pierre et Miquelon. That leaves 413 country-years with a mako catch above zero and a GDP value, from 30 countries, but only 4 of them in 1990 and 11 to 22 a year from 2006, so pool the years or compare within one. Add dead discards to landings, and stop at 2021 or say how you handled the 2022 ban. GDP per person does not catch sharks: any link runs through the size and reach of a country's fleet, and a big fleet from a middle-income country can catch more than a small one from a rich country. Say what you think the link runs through.
ICCAT, whose member governments set the rules for tuna, swordfish and shark fishing across the Atlantic and Mediterranean.
Link to this fileDirect: it controls the species and stock in this file. But it bans keeping the shark, not catching it, so its effect shows in how the catch is recorded as well as in how much there is. Makos released alive are not in the file at all.
The tensionThe makos are caught on longlines set for swordfish, a valuable fishery nobody proposed closing. A ban on landing can turn a fish that was sold into one that dies and is thrown away, which is why the rule also set a limit on total deaths.
SourceThe parties to CITES, the treaty on international trade in endangered species, whose secretariat is in Geneva.
Link to this fileIndirect: CITES controls international trade, including sharks landed from the high seas, and requires a finding that the trade will not harm the species. It sets no catch limit, so compare it with the ICCAT ban, which does act on the catch.
The tensionFishing nations argued that fisheries bodies such as ICCAT, not a trade treaty, should manage sharks; conservationists argued those bodies had failed to. The mako listing was carried by a vote, not by agreement.
SourceQuagga mussels spreading through Swiss lakes, every record since 2016
An invasive mussel, native to the rivers around the Black Sea, that has colonised Lake Geneva and is spreading from lake to lake, carried on boats.
- One row:
- One record of one mollusc species on one date at one point, placed on a 5 km grid.
- Covers:
- 344,823 records of all molluscs, 1840 to 2025; 373 quagga mussel records in Switzerland, 2016 to 2025, in 18 cantons
Full record6 traps
- Usable rows
- 373
- Coverage
- 344,823 records of all molluscs, 1840 to 2025; 373 quagga mussel records in Switzerland, 2016 to 2025, in 18 cantons
- How it arrives
- A zip (25 MB) holding a tab-separated file of every mollusc record: filter it before you open it
- Language
- Page and file in English
- Human· Registered boats· by canton
- 1You download every Swiss snail and mussel. 344,823 rows and 72 columns: filter to the quagga before opening it in a spreadsheet.
- 2Two mussels with similar names. 2,888 rows are zebra mussel; the quagga is Dreissena rostriformis bugensis.
- 3Not only Switzerland. 212 of the 585 quagga rows are in Germany, Austria or France, and stateProvince mixes canton codes with names like Baden-Württemberg.
- 4No lake column. The lake is not named in any Swiss row, and every point is a 5 km square. A canton is not a lake.
- 5A record is not an arrival. The earliest Swiss row is December 2016, while RTS reports Lake Geneva colonised since 2015.
- 6Rows pile up. 277 of the 373 Swiss rows share a date and grid point with another. Do not read the number of rows as how many mussels there are.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In which order were quagga mussels first recorded in each Swiss canton, 2016 to 2025?
- How many Swiss cantons had a quagga mussel record in each year from 2016 to 2025?
- scientificName: the species. Quagga mussel is Dreissena rostriformis bugensis
- eventDate and year: when it was recorded
- countryCode and stateProvince: the country, and the canton as a two-letter code (VD, GE)
- decimalLatitude, decimalLongitude: where, on a 5 km grid
Do Swiss cantons where quagga mussels have been recorded have more registered boats than cantons where they have not?
Watch outThe mussel file writes cantons as codes (VD) and the boat file as names (Vaud): add one to the other. All 18 cantons with a record match a boat row; 8 cantons have none (AI, AR, GL, GR, JU, NW, TI, UR). Choose one year's boat sheet, take out the regional subtotals, and say which year. Boats are counted where they are registered and mussels on 5 km squares with no lake named, and the Rhine cantons' records may come from the river rather than a lake.
Each canton sets its own rules.
Link to this fileDirect: moving boats between lakes is how the mussel spreads, and the rules target exactly that. The records cannot show whether cleaning works yet, because most of the rules are newer than the spread.
The tensionCleaning costs boat owners 300 to 1,000 francs, according to RTS, and a Biel harbour guard doubts that ducks and swans, which can also carry larvae, can be cleaned at all.
SourceAsian hornets spreading across Switzerland, every record since 2017
An invasive hornet from Asia that hunts honey bees and other insects. It was first found in Switzerland in 2017, and the cantons now search for its nests and destroy them.
- One row:
- One report of the Asian hornet on one date, placed on a 5 km square. It may be one hornet or a nest: the file does not say which.
- Covers:
- 103,997 records of wasps and ants, 1855 to 2026; 19,950 Asian hornet records in Switzerland, 2017 to 2026, in 22 cantons
Full record10 traps, 1 on the three lines
- Usable rows
- 19,942
- Coverage
- 103,997 records of wasps and ants, 1855 to 2026; 19,950 Asian hornet records in Switzerland, 2017 to 2026, in 22 cantons
- How it arrives
- A zip (8 MB) holding a tab-separated file of every wasp and ant record, 102 MB once unzipped: filter it before you open it
- Language
- Page and file in English
- Human· Settlement area· by canton
- 1You download every Swiss wasp and ant. 103,997 rows and 72 columns, 102 MB once unzipped, which may be too big to import into Google Sheets. Open it in Excel or LibreOffice, filter to the Asian hornet (21,373 rows), and save those rows as a file of their own.
- 2Two hornets. 2,763 rows are the native European hornet, Vespa crabro. The Asian hornet is written Vespa velutina Lepeletier, 1836, with its author and year, so filter on velutina.
- 3Not only Switzerland, and FR means two things. 1,423 of the Asian hornet rows are in France (1,179) or Germany (244), where stateProvince is a region such as Rhône-Alpes or Baden-Württemberg. Filter countryCode to CH first. Then take care: in countryCode, FR is France; in stateProvince, FR is the canton of Fribourg (499 rows).
- 4A row is a report, not a hornet or a nest. individualCount is empty in every row, and no column says whether someone saw one hornet or found a nest. The file has 14,990 Swiss rows for 2025, while the Cercle Exotique counts about 3,220 nests reported that year. Never call your variable nests.
- 5Reporting changed as well as the hornet. Swiss rows went from 25 in 2022 to 14,990 in 2025, and how reports arrived changed in the same years. In 2023, 984 Swiss rows came in to info fauna as anonymous records and 208 through the reporting platform; from 2024 almost every row is the platform's (recordedBy reads PSAFA: import frelon asiatique). Part of any rise is more people knowing how to report.
- 6Every point is a 5 km square. The coordinates are the centre of a 5 km square, and coordinateUncertaintyInMeters is 3535 in every row. You cannot map a nest, and a square cannot tell a town from a forest. No square carries two cantons, so a report near a border may be counted in the canton of its square.
- 7Many rows repeat. 5,790 Swiss rows share a date and a square with another row. 1,173 numbers in fieldNumber appear on two to four rows each, almost always in the same square and a few days apart, and the file does not say why. Count 5 km squares as well as rows.
- 8One record is not a colony. Ticino's only row is a single report in October 2020, and Uri's only row is at 2,084 m. minimumElevationInMeters is the real height, not the square's: 15 rows are above 1,200 m, one at 2,648 m, while the Cercle Exotique counts land below 1,000 m as where the hornet can settle. Check rows like these before you use them.
- 9Stop at 2025. The data were exported on 8 January 2026. The 8 Swiss rows dated 2026 are an unfinished year, and three of them (4 May, 28 August and 9 October 2026) are dated after the export.
- 10Most reports come in late summer and autumn. 12,583 of the 14,990 Swiss rows for 2025 are from August to November, when the colonies are biggest. Compare the same months across years.
- Are there more Asian hornet records per km² in the Swiss cantons where settlements cover a larger share of the land, 2025?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In which order was the Asian hornet first recorded in each Swiss canton, 2017 to 2025?
- How many Swiss cantons had an Asian hornet record in each year from 2017 to 2025?
- How did the number of 5 km squares with an Asian hornet record change from 2023 to 2025?
- scientificName: the species. The Asian hornet is written Vespa velutina Lepeletier, 1836
- eventDate, year and month: when it was reported
- countryCode and stateProvince: the country, and the canton as a two-letter code (VD, GE)
- decimalLatitude, decimalLongitude: the centre of a 5 km square
- minimumElevationInMeters: the height where it was reported, not rounded to the square
- higherGeography: the biogeographical region, such as Bassin lémanique or Jura und Randen
- recordedBy: where the record came from. Most rows read PSAFA: import frelon asiatique, the Swiss reporting platform for the hornet
Are there more Asian hornet records per km² in the Swiss cantons where settlements cover a larger share of the land, 2025?
Watch outBoth files write the canton as the same two-letter code, and Fribourg is FR in both, so all 26 cantons match: 22 have a hornet record, 20 of them in 2025, and 4 have none (AI, AR, GL, GR). Filter the hornet file to countryCode CH first. Divide by each canton's area (NOAS 1 to 4 added up), and use the 2013/18 survey, the last one every canton has. More settlement also means more people to see a hornet and report it, and the file does not say how many people looked: count the 5 km squares with a record as well as the rows, and say that neither is a count of nests. A small canton has few squares (Basel-Stadt has one), so its count of squares soon reaches the top. The hornet did not reach every canton at the same time, so check your result on the 15 cantons with a record by 2023 as well.
The Federal Council, through the Federal Office for the Environment (FOEN). Under the release ordinance, fighting an invasive species is the cantons' job; the Confederation watches the spread and can coordinate.
Link to this fileIndirect, but it pays for this file. One of its aims is to watch how invasive species spread, and the FOEN has paid for info fauna's hornet map since the first find in 2017 and for the reporting platform since 2025. It sets no measures for the hornet itself, and the hornet still has no priority level under it.
The tensionWho leads, and who pays. The strategy says the fight against invasive species should be coordinated across Switzerland. In December 2024 National Councillor Maya Graf told the Federal Council that, from the cantons' point of view, coordination was not at all good, and asked for a national plan. In February 2025 the Federal Council replied that fighting the hornet is in principle the cantons' task, and that its plan to save money from 2027 would remove the legal basis for its training grants (100,000 francs a year each to two bodies, 2023 to 2025).
SourceThe Cercle Exotique, the invasive species group of the conference of the cantons' environment offices (KVU), with the FOEN. Its hornet working group brings together federal and cantonal officials, beekeeping associations and researchers. The cantons are not bound by it.
Link to this fileDirect. It asks the cantons to find and destroy nests, above all the small nests of spring, and it counts about 220 nests reported in 2023, 780 in 2024 and 3,220 in 2025. It also shapes this file: once the hornet spreads fast, it asks people to report nests rather than thousands of single hornets, so what gets reported changes as the plan is followed.
The tensionTrapping queens in spring. Many beekeepers do it, but the recommendations advise against it outside supervised trials, because the traps catch many other insects: in Swiss trials, only 0.2% of the insects caught were Asian hornets. Vaud and Jura have banned it outside such trials. The document itself says that opinions on trapping differ widely, inside the working group and outside it.
SourceThe Canton of Geneva's Office for Agriculture and Nature (OCAN), with the Pôle invertébrés du bassin genevois (PIBG), which since 2023 has handled reports, organised the search for nests and trained volunteers for the canton. Dangerous nests in public places are destroyed by the fire and rescue service, and others by the association Pollinea-action.
Link to this fileDirect: it searches for nests and destroys them. It also feeds this file, because the canton asks beekeepers to report nests and attacked hives on the national platform. A canton that searches hard finds more, so Geneva's rows measure the search as well as the hornet.
The tensionWho pays. The Cercle Exotique estimates that about 300,000 francs went into fighting the hornet in Geneva in 2025: about 130,000 of it from beekeepers' unpaid work and associations, 80,000 from the canton, 60,000 from the fire service, 20,000 from owners and communes and 10,000 from the Confederation. The same recommendations warn against leaving most of the cost to beekeepers.
SourceA cantonal steering group set up in 2024: the Office de l'environnement, the Fondation Rurale Interjurassienne, the Jura beekeepers' federation, the research centre CABI and the association Dare-Dard, which now has a larger part in organising and destroying nests.
Link to this fileDirect in aim, but it starts after this file ends. The canton no longer tries to destroy every nest, and puts its effort into the small spring nests near houses. The file stops in January 2026, so it shows Jura before the change, not what the change did.
The tensionOwners pay. Jura asks the owner of a nest near a house to hire a specialist, and says insurance usually covers the cost. In March 2026 National Councillor Yvan Pahud told the Federal Council that the cost, usually 300 to 400 francs a nest, stops some owners from reporting nests or having them destroyed. In May 2026 the Federal Council replied that destroying nests is the cantons' task. Fewer reports would change this file as well.
SourceWolves in the canton of Schwyz, every record since 2023
Wolves returning to central Switzerland. In 2025 the canton of Schwyz confirmed its first pack, in the Wägital, where farms keep sheep.
- One row:
- One record of a wolf: a camera-trap photo, a sighting, droppings, tracks or a kill. Each has a date, a commune, a map point and how sure the canton is that it was a wolf.
- Covers:
- 409 records, 14 January 2023 to 23 September 2026, from 21 of the canton's 30 communes, and 8 from outside the canton
Full record10 traps
- Usable rows
- 400
- Coverage
- 409 records, 14 January 2023 to 23 September 2026, from 21 of the canton's 30 communes, and 8 from outside the canton
- How it arrives
- One Excel file with one sheet, called Wolf. The canton rebuilds it every night
- Language
- Page and file in German
- Human· Sheep on farms, by commune· by commune and year
- 1One canton, and only wolves. The file is called Hinweise Grossraubtiere (large carnivore records), but every row is a wolf and every commune is in the canton of Schwyz. It cannot compare cantons.
- 2Leave out ausserkantonal. 8 rows are marked ausserkantonal (outside the canton), and 5 of them have map points that cannot be right, with a digit wrong. Take them out before you count.
- 3Everything is text. The dates (2026-09-23) and the map points are stored as text, not as numbers. Convert them before you sort, count by month or draw a map.
- 4It changes almost every day. The canton rebuilds the file every night: it had 393 rows on 15 September 2026 and 409 on 27 September. 2026 is a part year. Keep your copy, and give the date you downloaded it.
- 5Records measure searching as well as wolves. 229 rows are camera-trap photos (Fotofalle): 4 in 2023, none in 2024 and 111 in 2025, when the canton set up more than 30 cameras to follow its first pack. More records can mean more cameras, not more wolves.
- 6Sure or probable. 305 rows are sicher (certain) and 104 wahrscheinlich (probable). The file does not define them. The canton's wolf plan of 2018 counts a dead wolf, a photo or a genetic test as a sure record, and a kill or tracks checked by a trained person as a probable one. Decide which you keep before you look, and say why.
- 7One event can be several rows. In 26 cases, two or more rows share a date, a commune and a kind of record (55 rows in all): one wolf passing three cameras in a night can be three rows. Rows 371 and 374 are the same record twice, so remove one. Numbers with a letter (173a, 173b) are separate records, not copies. No column says how many wolves were seen.
- 8It starts in 2023. A wolf called M107 has lived in the canton since 2020, and others passed through from 2009, but none of that is in this file. The pack's range also crosses into the canton of Glarus, and this file is Schwyz's.
- 9Swiss map points. E_Koordinate and N_Koordinate are metres on the Swiss grid (LV95), about 2,700,000 east and 1,220,000 north. A mapping tool has to be told that, or every point lands in the wrong place.
- 10Kills are records too. Riss: Nutztier (10 rows) is a farm animal killed and Riss: Wildtier (18 rows) a wild one. Each is evidence of a wolf, like a photo. They are not a list of every attack, and they say nothing about what was paid.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In which months of the year was a wolf recorded most often in the canton of Schwyz, 2023 to 2025?
- Are camera-trap records rated sicher (certain) more often than sightings reported by people, 2023 to 2026?
- Datum Sichtung: the date of the record, written 2026-09-23. Sichtung means sighting, but every kind of record has this date
- Ereignis: what was found. Fotofalle (a camera-trap photo), Sichtung (seen by a person), Kotfund (droppings), Spuren (tracks), Riss: Wildtier (a wild animal killed), Riss: Nutztier (a farm animal killed), anderes (other)
- Einschätzung: how sure the canton is. sicher (certain) or wahrscheinlich (probable)
- Gemeinde: the commune, as a name with no number, or ausserkantonal (outside the canton)
- E_Koordinate and N_Koordinate: where, in metres on the Swiss map grid, not latitude and longitude
- Kennung and Tierart: the record's number, and the species, which is always Wolf
Across the 30 communes of the canton of Schwyz, is the number of wolf records related to the number of sheep kept on farms, 2023 to 2025?
Watch outTake the sheep from the commune rows, which start with six dots and a number (......1348 Vorderthal): strip those off, write the wolf file's Küssnacht as Küssnacht (SZ), and every commune matches. Leave out the 8 ausserkantonal rows and the one record that appears twice. A commune-year with no wolf record has no row, so give it 0: that makes 90 commune-years, and 36 of them have at least one record, from 20 communes. The 2026 records have no sheep count to meet yet. Most records are from 2025, when the canton set up more than 30 cameras to follow its first pack, so records show where people looked as well as where wolves were. Sheep are counted in January in the farm's home commune, not on the summer pastures where they meet wolves. Communes differ greatly in size, so think about sheep per hectare of farmland (the same table has it). With many zeros and only 30 communes, choose your test and your years before you look, and say why.
The canton's government (Regierungsrat) orders it, and the Amt für Wald und Natur carries it out with the canton's wardens and trained hunters. The Federal Office for the Environment (FOEN) has to agree first.
Link to this fileDirect, and in the same valley as most of the records. In 2025 the canton set up more than 30 camera traps to follow the pack, and says the data it collected were the basis of its request to the Confederation. So the records and the cull are not independent: the same cameras served both. The file has no column for wolves removed.
SourceParliament changed the law and the Federal Council wrote the ordinance. A canton asks FOEN to agree to a cull, and the canton carries it out.
Link to this fileDirect: it is the law under which Schwyz culled the Chöpfenberg pack, and the reason its May 2025 order had to stop. It acts on wolves, not on this file: the file records evidence of wolves, and says nothing about culls.
The tensionPro Natura, WWF and BirdLife appealed against culls in December 2023, supported by Gruppe Wolf Schweiz. They wrote: "Damit missachten Bund und Kantone jede Verhältnismässigkeit und ignorieren die wichtige Rolle des Wolfs im Lebensraum Wald" (the Confederation and the cantons ignore any sense of proportion, and the wolf's important role in the forest). The farmers' union (Schweizer Bauernverband) answered the next day: "Die Umweltverbände gefährden die Alpwirtschaft" (the environmental groups put alpine farming in danger). The Federal Administrative Court paused the disputed culls on 5 January 2024, then ruled on 28 June 2024 that the groups could not appeal against FOEN's agreement.
SourceFOEN and the cantons. The money pays for electric fences, night pens, guard dogs and similar measures, on farms and on summer pastures.
Link to this fileIndirect for this file. It protects flocks, not wolves, and the file does not say whether a farm animal that was killed had been protected. Where the protection was not good enough, a kill neither counts towards a cull nor is paid for by the Confederation.
The tensionThe farmers' union (3 September 2024): "Der SBV unterstützt die Herdenschutzmassnahmen. Deren Wirkung hängt mit der wirksamen Begrenzung der Wolfspopulation zusammen" (protection only works if the number of wolves is limited). Four conservation groups (23 September 2026): "Denn Abschüsse in unauffälligen Rudeln können diese auseinanderreissen und dadurch zu mehr Problemen mit Nutztieren führen statt weniger" (culling packs that cause no trouble can break them up, and lead to more problems with farm animals, not fewer).
SourceThe Standing Committee of the Bern Convention, a Council of Europe treaty that Switzerland has signed, on a proposal from the EU.
Link to this fileIndirect. It changes the international rule that Swiss law has to respect, but the Schwyz culls were made under Swiss law. The pack formed in the same year as the change, so a before and after in this file cannot separate the two.
The tensionPro Natura (3 December 2024): "Die Herabstufung ist zudem kein Patentrezept für weniger Nutztierrisse" (the downgrade is no simple answer to farm animals being killed).
SourceProtected land in every country, as a share of its land, every year since 2013
Protected areas: how much of each country's land is protected or conserved, the measure the world's 30 per cent target is counted in.
- One row:
- One country in one year: land in protected areas and other conserved areas, as a percentage of its total land area.
- Covers:
- 213 countries with values, 2013 to 2025, among 265 rows that include 47 aggregates such as World and Low income. 209 countries have a value for 2015
Full record5 traps, 1 on the three lines
- Usable rows
- 2,746
- Coverage
- 213 countries with values, 2013 to 2025, among 265 rows that include 47 aggregates such as World and Low income. 209 countries have a value for 2015
- How it arrives
- A zip of three CSVs: the data is the one whose name begins API_, and the two Metadata files say which rows are countries
- Language
- Page and file in English
- Natural· Tree cover loss· by country, by name
- 1Four lines of notes sit above the header. The header is the fifth line and begins Country Name, and there is an empty column after the last year.
- 2Aggregates sit among the countries: 47 rows such as World, Euro area and Low income. The zip's Metadata_Country file leaves Region blank for every one of them, so use it to take them out.
- 3Years run across. One column per year: turn them into one column before you match the file to anything with a year column.
- 4Values jump when the database changes. Switzerland reads 24.8% in 2013 and 9.7% in 2015: a change in what was counted, not land losing its protection. Check any country that jumps before you rely on its year.
- 5Protected is not managed. The figure counts land designated as protected or conserved, whether or not anyone enforces it.
- Across the countries with at least 1 million hectares of tree cover, does protecting more land in 2015 go with losing less tree cover from 2016 to 2025?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Across countries, is the share of land that is protected related to GDP per person?
- Country Name, and Country Code, the three-letter ISO code
- 1960 to 2025: one column per year, empty before 2013
Across the countries with at least 1 million hectares of tree cover, does protecting more land in 2015 go with losing less of that tree cover from 2016 to 2025?
Watch outGFW names countries and never codes them, so match by hand: 13 of the 109 countries with 1 million hectares of tree cover are spelled differently in the World Bank file (México is Mexico, Russia is Russian Federation, Vietnam is Viet Nam, both Congos and both Koreas are written Congo, Dem. Rep. and so on), and French Guiana and Taiwan are not in it at all. That leaves 107. Filter GFW to one canopy threshold first, and add together the extra rows China, India and Pakistan carry for disputed areas.
The 196 parties to the Convention on Biological Diversity.
Link to this fileDirect: the target is counted in protected areas and other effective area-based conservation measures, which is what this file measures. It says nothing about whether the land is "effectively conserved and managed", which is the other half of the target.
The tensionSurvival International has campaigned against it as a land grab, warning that new protected areas will fall on Indigenous peoples' land in the global South. The final text added recognition of Indigenous and traditional territories.
SourceThe parties to the Convention on Biological Diversity.
Link to this fileThe target in force in 2015, so a country's share that year is partly a record of how far it had got towards 17 per cent. Useful when a question starts from 2015.
SourceFish caught in Lake Geneva's Swiss waters, every year since 1904
Fishing on Lake Geneva: the perch and whitefish landed each year, and the fish side of the argument about how much phosphorus the lake should hold.
- One row:
- One year: the catch of one species in Lake Geneva's Swiss waters by one kind of fishing, as tonnes or as a number of fish.
- Covers:
- 1904 to 2024 for professional fishing, with 1911 to 1920 recorded as 0; recreational fishing from 1994
Full record6 traps, 2 on the three lines
- Usable rows
- 111
- Coverage
- 1904 to 2024 for professional fishing, with 1911 to 1920 recorded as 0; recreational fishing from 1994
- How it arrives
- One small CSV per table, from the CSV button under it. Use a browser: the page is built there
- Language
- Page and file in French
- Human· Total phosphorus· by year
- 1Swiss waters only. The cantons report their own fishers; the French catch on the south shore is not in it. Say that your variable is the Swiss catch.
- 21911 to 1920 read 0. Those years are missing, not empty nets. Leave them out.
- 3Use the professional series for anything long. Recreational catches start in 1994, so the total jumps there when they are added in.
- 4The file does not say what it is. It is named after the species alone (Corégones, 1904-2024.csv), with no lake, fishing or unit in it. Rename it at once.
- 5A catch is not a population. It depends on how many fishers there are, how many nets they set and the rules they fish under. The official sheet records the nets, but this table does not.
- 6Use a browser. The table is built in the page, so it cannot be fetched by other programs.
- Is the professional whitefish catch in Lake Geneva's Swiss waters related to the lake's total phosphorus, 1974 to 2015?
- Is the professional perch catch related to total phosphorus in Lake Geneva, 1974 to 2015?
- Année and one column named after the species: Corégones (whitefish) or Perche (perch)
- In the page, before you download: the fishing (Rendement de la pêche professionnelle, de loisir or totale), Poids (tonnes) or Nombre (fish), and the species
Lake Geneva, professional fishing, 1904 to 2024 (2)
Each opens as a table with a CSV button under it. The file is named only after the species, so rename it at once with the fishery and the unit.
- Whitefish (Corégones)TonnesNumber of fish
- Perch (Perche)TonnesNumber of fish
Across the years 1974 to 2015, does the professional whitefish catch in Lake Geneva's Swiss waters go with the lake's annual mean total phosphorus?
Watch outAverage the phosphorus to one value a year first (about 18 surface samples each), and 42 years meet the catch. Use the professional catch, which runs the whole period: the total adds recreational fishing from 1994 and jumps there. The catch is Swiss waters only. Both series move steadily over these decades, so correlate the change from one year to the next as well.
The Swiss Federal Council and the French government, applied by the cantons of Vaud, Valais and Geneva and the French state through a joint advisory commission.
Link to this fileDirect: it sets who may fish, where, when and with what, and its rules make every professional fisher record each day's catch on an official sheet (article 52), which is where these numbers come from. A change in the rules can move the catch on its own.
SourceThe International Commission for the Protection of the Waters of Lake Geneva (CIPEL), for Switzerland and France together.
Link to this fileIndirect: the plans set phosphorus, not catches, but CIPEL's own review ties the two together through the food fish depend on. This file is the fish side of that argument; the phosphorus is in French3lakes.
The tensionFish against clear water. CIPEL's own scientific report of 2024 found that managing phosphorus divides the lake's stakeholders, because low phosphorus does not generally keep fish populations strong. It proposed 15 µg per litre as a compromise between fish and the lake's other uses, and left CIPEL to reconsider its 10 to 15 µg target in 2025.
SourceFish in Geneva's rivers
Whether the fish communities of the watercourses draining into Lake Geneva recovered as the water got cleaner and the channels were rebuilt, and which reaches never did.
- One row:
- One station on one survey date: four component scores, a final score and a quality class for the fish community found there.
- Covers:
- 55 stations on 26 watercourses, 2009 to 2025, and only 15 years carry any surveys at all
Full record6 traps, 1 on the three lines
- Usable rows
- 92
- Coverage
- 55 stations on 26 watercourses, 2009 to 2025, and only 15 years carry any surveys at all
- How it arrives
- CSV inside a zip, semicolon-separated, with a documentation folder beside it
- Language
- Page and file in French
- 192 rows is the whole file. 55 stations, 26 watercourses, and 28 of those stations were surveyed exactly once. This is too small for a t-test on its own. Use it alongside one of the chemistry files, not as the basis of an investigation.
- 2The file is not UTF-8. It is Latin-1, and opening it as UTF-8 turns "Très bonne" into "Très bonne" in every row of the class column. The first header also carries a byte-order mark, so the first column arrives named `OBJECTID` unless your reader strips it.
- 3Semicolon-separated, like the rest of this family.
- 4CLASSE is ordinal: five ordered bands, so a mean of them means nothing. Use SCORE_FINAL for arithmetic and keep the classes for description. Seven rows have no class at all.
- 5The four component columns are called PARAMETRE_1 to PARAMETRE_4 and nothing in the file says what they measure. The documentation folder inside the zip does, and it is the difference between a variable and a column heading.
- 6Fifteen distinct years span 2009 to 2025, so the gaps are real. A station's two visits may be a decade apart, which is a long time in a river and a short time in a fish population.
- Do renatured reaches carry higher fish index scores than channelised ones?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Which watercourses have never scored above Moyenne, and what do they have in common?
- Has the fish index at the stations surveyed more than once changed between visits?
- SCORE_FINAL, the fish index, running 0 to 12
- CLASSE, five ordered bands from Médiocre to Très bonne
- PARAMETRE_1 to PARAMETRE_4, the components behind the score, named only in the documentation
- Station code and name, watercourse, survey date, Swiss coordinates
The French State, the Rhône-Alpes region, the Canton of Geneva, the Haute-Savoie department, the Agence de l'eau Rhône-Méditerranée-Corse and basin users, with operational lead shared between the Communauté de communes du Genevois and the canton.
Link to this fileFish move, and the contract covers the whole basin rather than the cantonal half of it. A barrier or a polluted stretch on the French side shows up in what a Swiss station catches. That is why a cross-border plan is the right scale for this indicator, though not for every one.
The tensionThe contract exists because two sides of a border had to agree on goals. Geneva makes the laws for its own network and pays for it, but part of the water arriving comes from France, where Geneva cannot make laws. Several of the stations that got worse between the two periods are cross-border ones.
SourceThe canton sets the PGEE, which decides sector by sector whether the system is combined or separate. The secondary network is owned by the communes: more than 1,300 km of foul and surface-water sewers and 28 pumping stations. Since 1 January 2015 the FIA shares the cost across all of them.
Link to this fileSewer separation removes the pollution shocks that kill fish outright. But a fish community rebuilds over years, not seasons, so this file cannot date the effect of this strategy precisely. Say that, rather than claiming a year.
The tensionWho pays, and on what basis. The FIA is funded by a one-off connection charge plus two annual ones: the canton and the communes pay on the impermeable public road surface connected to the network, and property owners pay on their drinking-water consumption. Because every commune shares the cost, the communes that pay are not always the ones that benefit. I could not find any documented opposition to it, so if you choose this one you have to find sourced positions, not just claim that people disagree.
SourceThe Grand Conseil, which in April 1997 added seven articles on renaturation to the cantonal water law of 5 July 1961 and created the fund. It is fed mainly by the hydraulic royalties paid by SIG and the Société des Forces Motrices de Chancy-Pougny, by pumping taxes, by federal subsidies and by donations, at around 11.8 million francs a year with a floor of 6 million.
Link to this fileThe best fit in the family. Renaturation rebuilds the physical channel and removes the barriers that stop fish moving, which is exactly what a fish index responds to, and the Aire is the most documented example of it in the canton.
The tensionThe best-documented of the three. Local farmers said the project took too much land (one description called it science fiction). They argued that it ignored the realities of farming, and that farmland is a farmer's main working tool and inheritance. Part of the land at stake sits in the surfaces d'assolement, the protected cantonal quota of the best arable land. This sets an economic and a cultural claim against an environmental one: a tension along three of the five lines the criterion names.
SourceSea surface temperature and coral bleaching alerts, reef by reef
Heat stress on coral reefs: how often the water goes above what the corals at a given reef are used to, and for how long.
- One row:
- One reef region on one day: daily minimum, maximum and representative sea surface temperature, the anomaly, the HotSpot, accumulated heat stress and an alert level.
- Covers:
- 214 named reef regions worldwide, daily from 1 January 1985. The Aden file counted here holds 15,231 days.
Full record6 traps
- Usable rows
- 15,231
- Coverage
- 214 named reef regions worldwide, daily from 1 January 1985. The Aden file counted here holds 15,231 days.
- How it arrives
- One plain text file per reef, space-separated, with a 22-line header block
- Language
- Page and file in English
- 1Twenty-two lines of header before the column names, carrying the reef's coordinates and its own maximum monthly mean. Those are not decoration: the threshold is what everything else is measured against.
- 2The header dates and the table dates are in different orders. The header says `First Valid BAA Date: 1985 31 03`, which can only be 31 March, while the table below is `YYYY MM DD`. Two date formats in one file, twelve lines apart.
- 3The alert level is derived, not measured. BAA comes from DHW, DHW comes from HotSpot, and HotSpot is the SST minus this reef's own maximum monthly mean. The observation is the temperature; everything else is arithmetic on it, and a conclusion built on the alert level is a conclusion about a formula.
- 4The threshold is local, so raw temperature does not compare across reefs. Aden's maximum monthly mean is 30.48 °C. A reef at 26 °C may be in severe stress and Aden at 29 °C may be perfectly comfortable. Compare HotSpot or DHW, never SST.
- 5One reef file is 15,231 daily rows and there are 214 of them, so a question about several reefs is three million rows. Pick your reefs before you start downloading.
- 6The series runs to yesterday, so the current year is always partial and any annual figure for it is a part-year, which matters most in the months when bleaching actually happens.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How many days a year has one reef spent above its bleaching threshold, and has that changed since 1985?
- Do two reefs at similar latitudes accumulate heat stress at the same rate?
- In the worst year on record for one reef, how long did the heat stress last?
- SST_MIN and SST_MAX, the day's range in °C
- SSTA, how far the day sits from normal for that reef
- HotSpot, degrees above the reef's own maximum monthly mean
- DHW, degree heating weeks: heat stress accumulated over the previous twelve weeks
- BAA, the bleaching alert level from 0 to 4, which is derived from the two above
The parties to the Convention on Biological Diversity, committing to protect 30% of coastal and marine areas by 2030, with coral reefs among the ecosystems named.
Link to this fileProtection is the strategy most often proposed for reefs and it does not act on heat, which is what this file measures. That mismatch is the investigation: a marine park stops fishing and anchoring and cannot stop a marine heatwave, so a reef inside one and a reef outside it should show the same DHW. Testing that is a real finding either way.
The tensionProtected on paper against protected in practice, and whose fishing grounds are inside the line. Coastal communities carry the cost of a designation that cannot protect the reef from the thing currently killing it.
SourceAn intergovernmental partnership of countries and organisations, which coordinates reef monitoring worldwide and publishes the regular global status reports that governments quote in their arguments.
Link to this fileIt acts on knowledge rather than on water, which makes it the strategy this dataset is part of rather than one it can test. Worth choosing when your question is about how bleaching came to be measured the same way everywhere, which is itself a decision somebody made.
The tensionA voluntary partnership with no power over the emissions that heat the water. It publishes more and more serious reports to governments that paid for the monitoring but not for the solution.
SourceThe United States agency that funds reef management, restoration and monitoring in US waters and territories, and which runs the satellite product this file comes from.
Link to this fileThe producer of your data is also a party with a programme, which is worth noticing rather than avoiding. Its restoration work is concentrated in Florida and the Caribbean, so those reefs appear in this file alongside reefs nobody is restoring, and you can compare them.
The tensionRestoration against mitigation: growing and replanting coral is visible, fundable and local, and it does not change the temperature that killed the last colony.
SourceWhat UK boats land, species by species, and totals since 1938
Overfishing and changing seas around the UK: which fish are still landed, which are disappearing from the catch, and which are arriving as the water warms.
- One row:
- One species in one year: the weight (thousand tonnes) and value (£ million) that UK vessels landed. Other tables split this by port, sea area, gear and month.
- Covers:
- 36 species and their group totals, 2020 to 2024 (Table 2.2); demersal, pelagic and shellfish totals from 1938 to 2024 (Table 2.7)
Full record7 traps
- Usable rows
- 180
- Coverage
- 36 species and their group totals, 2020 to 2024 (Table 2.2); demersal, pelagic and shellfish totals from 1938 to 2024 (Table 2.7)
- How it arrives
- One spreadsheet (.ods) of about 30 tables. Google Sheets opens it directly
- Language
- Page and file in English
- 1Pollack is not pollock. Pollock (Alaska pollock) is the certified-sustainable fish in fish fingers, caught in the North Pacific and imported. Pollack, listed here as Pollack (Lythe), is the UK's own inshore fish, and scientists advised zero catch for it in 2025. Coley (saithe) is a third species. Check which one a source means before you use it.
- 2Only five years per species. Table 2.2 covers 2020 to 2024. For a longer species series you have to copy the same table from earlier years' reports and stack them, checking that the species names and units have not changed.
- 3Octopus has no row. The 2025 octopus bloom off Devon and Cornwall, the largest of four in 125 years, is real, but octopus sits inside Other Shellfish here. The MMO publishes octopus figures only as monthly summaries on its octopus bloom page. Use the bloom as a way into the issue, not as your dataset.
- 4Thousands of tonnes, to one decimal place. Pollack falls from 1.4 to 0.8. The table cannot see a change smaller than 100 tonnes, which is a large share of a small fishery.
- 5Landings are not catches, and not the stock. Fish thrown back, fish caught by anglers and fish landed abroad are not in the landings into UK ports. A fall in landings may mean fewer fish, a lower limit or fewer boats.
- 6Value and quantity move differently. Value is in cash of each year and is not adjusted for inflation. Compare weights, or adjust the values and say how.
- 7Do not download the underlying data file. The species-by-port-by-month version on the same page is 108 MB, and over 3 GB once opened. The summary tables in this spreadsheet are the usable form.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How have UK landings of demersal, pelagic and shellfish species changed between 1938 and 2024?
- How have UK landings of pollack changed between 2020 and 2024, compared with the catch limit set for it?
- Which ports landed the most shellfish in 2024, and how much of it was crab and lobster?
- Species: 36 named species, grouped as demersal (bottom-living fish), pelagic (open-water fish) and shellfish
- Quantity: live weight landed, in thousand tonnes to one decimal place
- Value: first-sale value, in £ million
- Year: five years side by side, as columns
The UK and the EU, agreed each December in their annual fisheries consultations under the Trade and Cooperation Agreement.
Link to this fileDirect. The limit sets how much pollack UK boats may land, and Pollack (Lythe) has its own row in the landings tables. Recreational catches, which are not in these tables, are thought to be large.
The tensionThe scientists advised zero catch: ICES found no level of fishing that would rebuild the stock above its safe limit by 2026. The UK and EU set 766 tonnes anyway, to avoid a choke that would close other fisheries where pollack is caught by accident. Conservation against keeping mixed fisheries open, in the agreement's own words.
SourceThe UK Government and the devolved governments, through fisheries management plans.
Link to this fileIndirect. It is the framework every catch decision sits inside, but it sets no number you can test against this file. Use it to explain why limits exist, and the pollack limit to test one.
SourceWild boar crop damage in every Swiss canton since 1992
Wild boar numbers rising across Switzerland, and the crop damage they cause: an animal that lives in forest and feeds in fields.
- One row:
- The wild boar damage paid out in one canton in one hunting year, in Swiss francs.
- Covers:
- 25 cantons with boar damage at some point, hunting years 1992 to 2025: 854 canton-years. The same site gives boar shot per canton since 1933 and found dead since 1992
Full record10 traps
- Usable rows
- 854
- Coverage
- 25 cantons with boar damage at some point, hunting years 1992 to 2025: 854 canton-years. The same site gives boar shot per canton since 1933 and found dead since 1992
- How it arrives
- One table per year, every canton in it, exported with the CSV button under the table. The link opens 2024; move the year slider for the others. There is no single file of every canton and every year: you build that table from the yearly exports, and say how
- Language
- Page and file in French
- 1One year per file, and no year in it. Each export is one hunting year for every canton, and the CSV has no year column. Add one as you stack the files, or 34 files become one undated pile.
- 2A missing canton is not a zero. Cantons with no report are left out of that year's table, not listed as 0. Between 2004 and 2025, 30 canton-years are missing, almost all from cantons with few or no boar (Nidwalden, Obwalden, Glarus, Uri, Schwyz). Treat them as no data and say so.
- 3The names won't join. This file says Argovie, Genève, Berne and Soleure; the land-use file says Aargau, Geneva, Bern and Solothurn. Add a two-letter canton code to both before you match them.
- 4Divide by area. Vaud pays more than Basel-Landschaft because it is six times bigger. Compare damage per km² of canton, or per hectare of farmland, never raw francs.
- 5Francs paid are not damage done. The figure is compensation, and each canton sets its own rules: small claims are not paid, and a farmer without an electric fence may get nothing. Say that your variable is compensation.
- 6Hunting years, not calendar years. In most cantons the hunting year runs from 1 April to 31 March, so "2018" ends in March 2019. The land-use survey uses the calendar year the photographs were taken.
- 7The range is still spreading. Bern and Fribourg have plenty of arable land but little damage (CHF 12 and 56 per km² in 2014-18), because boar spread in from the north-west and are only now reaching the central plain. Part of your pattern is where the boar have got to, not only what the land is like.
- 8Big swings between years. Damage was CHF 4.28 million in 2024 and 2.74 million in 2023. Good years for beech and oak nuts, and snowy winters, move boar numbers. Average several years before comparing cantons.
- 9Geneva shoots no boar. In the Chasse table Geneva is always 0, because hunting is banned there. Its culls are under Tir spécial. Add Chasse and Tir spécial together for every canton, or Geneva looks empty.
- 10Sum the columns. The shot and found-dead tables split into males, females, young and undetermined. Before 1992 everything is in undetermined, so a total needs all four columns.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is wild boar damage per km² related to the share of arable land across the 26 Swiss cantons, 2014 to 2018?
- Is wild boar damage per km² more closely related to forest cover or to arable land across Swiss cantons?
- Has wild boar crop damage in Switzerland risen between 1992 and 2025, and do the years with high damage match the years with many boar shot?
- Area: the canton, in French (Argovie, Genève, Vaud)
- Dégats en francs: damage caused by wild boar in that hunting year, in Swiss francs
- On the Faune sauvage tab, for Sanglier: Chasse (shot by hunters), Tir spécial (special culls, including Geneva's wardens) and Gibier péri (found dead), each split into males, females and young from 1992
- Pair it with the land-use card above for each canton's arable land, farmland and forest
Crop damage: one table per year, every canton in it (34)
Export each year with the CSV button and add a year column as you stack them: the file itself does not say which year it is.
- 1992Damage in francs
- 1993Damage in francs
- 1994Damage in francs
- 1995Damage in francs
- 1996Damage in francs
- 1997Damage in francs
- 1998Damage in francs
- 1999Damage in francs
- 2000Damage in francs
- 2001Damage in francs
- 2002Damage in francs
- 2003Damage in francs
- 2004Damage in francs
- 2005Damage in francs
- 2006Damage in francs
- 2007Damage in francs
- 2008Damage in francs
- 2009Damage in francs
- 2010Damage in francs
- 2011Damage in francs
- 2012Damage in francs
- 2013Damage in francs
- 2014Damage in francs
- 2015Damage in francs
- 2016Damage in francs
- 2017Damage in francs
- 2018Damage in francs
- 2019Damage in francs
- 2020Damage in francs
- 2021Damage in francs
- 2022Damage in francs
- 2023Damage in francs
- 2024Damage in francs
- 2025Damage in francs
Boar removed and found dead: one table per canton, 1992 to 2025 (26)
Add shot and special culls together for boar removed (Geneva has no hunting, so all of its boar are special culls). Obwalden has 28 years, not 34, and Nidwalden's found-dead table 32.
- Aargau (AG)ShotSpecial cullsFound dead
- Appenzell Ausserrhoden (AR)ShotSpecial cullsFound dead
- Appenzell Innerrhoden (AI)ShotSpecial cullsFound dead
- Basel-Landschaft (BL)ShotSpecial cullsFound dead
- Basel-Stadt (BS)ShotSpecial cullsFound dead
- Bern (BE)ShotSpecial cullsFound dead
- Fribourg (FR)ShotSpecial cullsFound dead
- Geneva (GE)ShotSpecial cullsFound dead
- Glarus (GL)ShotSpecial cullsFound dead
- Graubünden (GR)ShotSpecial cullsFound dead
- Jura (JU)ShotSpecial cullsFound dead
- Lucerne (LU)ShotSpecial cullsFound dead
- Neuchâtel (NE)ShotSpecial cullsFound dead
- Nidwalden (NW)ShotSpecial cullsFound dead
- Obwalden (OW)ShotSpecial cullsFound dead
- Schaffhausen (SH)ShotSpecial cullsFound dead
- Schwyz (SZ)ShotSpecial cullsFound dead
- Solothurn (SO)ShotSpecial cullsFound dead
- St. Gallen (SG)ShotSpecial cullsFound dead
- Ticino (TI)ShotSpecial cullsFound dead
- Thurgau (TG)ShotSpecial cullsFound dead
- Uri (UR)ShotSpecial cullsFound dead
- Vaud (VD)ShotSpecial cullsFound dead
- Valais (VS)ShotSpecial cullsFound dead
- Zug (ZG)ShotSpecial cullsFound dead
- Zurich (ZH)ShotSpecial cullsFound dead
The Canton of Geneva. Its environment wardens (gardes de l'environnement), under the cantonal nature office, carry out all the culling that hunters do elsewhere.
Link to this fileDirect: it decides who removes boar and how many. Geneva's culls are in Tir spécial, not Chasse: 40 in 1990, an average of 189 a year in 2014-18, 419 in 2025. Its damage in 2014-18 was about CHF 100 per km² against 161 in Vaud next door. Two cantons are an example, not a test, so use Geneva to explain your pattern, not to prove it.
The tensionGeneva's hunters' association argues that crop damage is rising and that taxpayers pay for a state hunt. Animal protection groups defend the ban. The hunters are a party to the argument, so cite them as a position, not as a fact.
SourceThe cantons, under the Federal Act on Hunting (JSG) of 20 June 1986, Article 13: damage by game to crops, forest and livestock is to be compensated appropriately.
Link to this fileDirect, and a limit on the data: the francs in this file are compensation paid, so the rules decide the numbers. Small claims are not paid, and a farmer who did not take reasonable precautions such as an electric fence may get nothing. A canton with strict rules can look as if it has fewer boar.
The tensionWho pays for damage from a wild animal: the public, the hunters, or the farmer who has to fence? Cantons answer differently, which is part of why their figures differ.
SourceThe Federal Food Safety and Veterinary Office (FSVO) with the cantons, who test boar found dead or shot sick.
Link to this fileIndirect: it does not aim at crop damage. It is why found-dead boar are now looked for and tested, which may itself raise the Gibier péri counts. Use it to explain a change in the found-dead series, not the damage.
SourceEuropean green crabs in Washington State's traps, every WDFW trap from 2018 to 2025
An invasive crab from Europe that eats clams, oysters and young Dungeness crabs and destroys eelgrass beds. It is spreading along Washington's coast and into the Salish Sea, and the state sets traps to find it and remove it.
- One row:
- One trap set by WDFW: where and when it went in and came out, the kind of trap, why it was set, and how many green crabs were in it, male and female. A trap with no crabs has a row with 0.
- Covers:
- 36,048 traps at about 310 sites in 12 management areas on Washington's outer coast and in the Salish Sea, set from July 2018 to October 2025
Full record11 traps
- Usable rows
- 35,764
- Coverage
- 36,048 traps at about 310 sites in 12 management areas on Washington's outer coast and in the Salish Sea, set from July 2018 to October 2025
- How it arrives
- One CSV (7.4 MB) with headings in the first row, one row per trap. The map position is in metres (Web Mercator), not latitude and longitude
- Language
- Page and file in English
- 1The traps follow the crabs. Assessment traps, the largest group (22,229 rows), are set where green crabs have already been found, and WDFW says crews move traps to new places when catches fall. So more traps in an area can be a result of more crabs, not a cause of fewer. Say which way you think the link runs, and why.
- 2WDFW's own traps only. In 2023 this file holds 11,846 green crabs. WDFW reports 361,418 caught in Washington that year, most of them by tribes, shellfish growers and other partners whose traps are not here. The file shows where WDFW trapped and what its traps caught, not how many crabs were removed.
- 3Count trap-days, not traps. A trap stayed in the water for one to four nights, and a trap left longer has more time to catch crabs. Take the hours from Set Date and Time to Date and time of Trap being removed., divide by 24 and round, with 1 as the smallest. These trap-days match WDFW's own count of trap checks in 97 of every 100 rows. WDFW reports crabs per 100 trap checks; crabs per 100 trap-days is the nearest you can get from this file.
- 4Keep to one kind of trap. WDFW's plan says Fukui traps miss the smallest crabs and minnow traps keep out the larger ones. Shrimp traps went from 5 in 2020 to more than one trap in five from 2023. A change in the mix of traps changes the catch. Every area-year with 50 or more traps from 2020 to 2025 has at least 20 Fukui traps: use those, and say so.
- 5Zeros are real. 28,322 of the 36,048 traps caught no green crabs. WDFW's plan says "traps that do not contain green crab are just as valuable as traps that do." Keep them, or every catch rate comes out too high.
- 6Tidy the names before you group. Management Area is written 15 ways for 12 areas: Eastern Strait & Admiralty Inlet and Eastern Strait and Admiralty Inlet, North Puget Sound with a space after it, and North central Puget Sound. Effort Type has Assessment with a space after it 193 times. Site Name has 360 spellings of about 310 sites.
- 7Some traps appear two or three times. 272 rows repeat another row in every column except the map position and the row's ID, and none of them caught a crab. 175 of them are one batch at the site South (North Puget Sound) on 22 August 2022, where 89 traps were entered three times. Take the copies out and say how many.
- 8Twelve rows have impossible times. Two traps were taken out before they were set (one of them in 2013), one in the same minute, one has no removal time, and eight stayed in for between 7 and 367 days. Leave all 12 out of any catch per trap-day.
- 9Two places hold almost every crab. Willapa Bay and Grays Harbor, on the outer coast, hold 49,979 of the 52,680 crabs in the file. South Puget Sound and South Central Puget Sound have 1,540 traps between them and no crabs. Use a rank correlation (Spearman), and check whether your answer holds without the two coastal areas.
- 10Few crabs before 2020, and no 2026 yet. The file has 2 green crabs in 2018, from 3 areas, and 42 in 2019, from 7; Willapa Bay and Grays Harbor start in 2020. The last trap was set on 20 October 2025, because the public file only shows rows that WDFW has checked.
- 11The hub's page for this dataset will not open. On the European Green Crab Hub, the dataset's page says "Can't access this dataset?" The download link on this card works with no account, and the ArcGIS page describes the file.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In which months do WDFW's Fukui traps in Willapa Bay catch the most green crabs per 100 trap-days?
- Do Fukui, minnow and shrimp traps set at the same site on the same days catch different numbers of green crabs per trap-day?
- What share of the green crabs in WDFW's traps were female, on the outer coast and in the Salish Sea?
- How did the number of traps WDFW set in each management area change after the emergency proclamation of January 2022?
- Management Area and Site Name: where. 12 management areas, written in 15 ways, and about 310 sites
- Set Date and Time and Date and time of Trap being removed. (the full stop is part of the heading): when the trap went in and came out. The hours between them give the trap-days
- Trap Type: Minnow, Fukui, four kinds of shrimp trap, and a few Other and Pitfall-Large
- Effort Type: why the trap was set. Early Detection is where no green crabs have been found in the last five years, Assessment is where they have, Monitoring (from 2024) is a fixed routine, and Control is trapping to reduce them
- Total Male, Total Female and Total Catch: the green crabs in the trap. 0 is a real count
- Total Gravid: females carrying eggs, blank before 2023
- Trap Number, Version Used, Acquisition Date, x and y: the trap's label, the form it was entered on, when it was entered, and its position in metres
Governor Jay Inslee. It orders the Washington Department of Fish and Wildlife to act, and asks the state's other land and water agencies to make green crab a high priority on the shores they own.
Link to this fileDirect for what people do: it is the order WDFW has trapped under since 2022, so it is the plan behind the number of traps. It cannot be judged from this file alone. This file holds 11,846 of the 361,418 green crabs WDFW reports caught in Washington in 2023; tribes, shellfish growers and other partners caught most of the rest, and their traps are not here.
The tensionThe state's own goals do not agree. The proclamation orders measures "to effect the eradication of or to prevent the permanent establishment and expansion of European green crab". WDFW's management plan of 2024 says "Eradication is not a recommended goal for green crab" unless a population can be cut off from nearby waters, because larvae drift in from other populations.
SourceThe Washington State Legislature, at WDFW's request. The budget tells WDFW to "implement eradication and control measures on European green crabs through coordination and grants with partner organizations".
Link to this fileDirect for what people do, with a limit. The money pays for traps, but most of it is passed to partners whose traps are not in this file. So more money need not mean more rows here.
The tensionCost, and where the traps go. In its request of 14 December 2021 WDFW said the $2.3 million the Legislature gave in 2021 "is not sufficient to successfully control these populations", and that it could not keep its trappers at the Lummi Sea Pond "without reducing effort elsewhere, which would significantly enhance EGC invasion risks statewide".
SourceWDFW, written with more than 40 tribes, agencies, shellfish growers and other partners, because the Governor and the Legislature asked for a long-term plan.
Link to this fileDirect. It decides what kind of trapping happens where: early detection where no crabs have been found, control where they are established. That shapes where WDFW's traps go, and so which crabs they meet.
The tensionWhere the effort goes. WDFW's report for July 2025 to June 2026 says that "management capacity and funding remain insufficient to address every need simultaneously", and that "increasing trapping intensity in known populations can reduce capacity for broader surveillance and vice versa."
SourceThe Lummi Indian Business Council, the Lummi Nation's governing body. It set up a Lummi task force to lead the Nation's response.
Link to this fileIndirect. The Lummi Nation traps on its own tidelands, and its catch is not in this file: no site in it is the Sea Pond. Use it to explain why the state declared an emergency, and that tribes run their own programmes alongside WDFW's.
The tensionTreaty rights, and who pays. Chairman William Jones Jr. said: "The European green crab is a serious threat to our treaty fishing rights" and "We need a huge increase in effort and resources to properly address the green crab invasion, and it's going to take a coordinated response from tribal, state and federal partners."
SourceEvery recorded shark bite in Australia since 1791
Shark bites on people in Australian waters, whether they follow the warmth of the sea, and the nets, drumlines and deterrents used to prevent them, which also kill sharks, turtles and dolphins.
- One row:
- One shark bite on a person, or on their board or boat, with where and when it happened, the shark thought responsible and what the person was doing.
- Covers:
- 1,304 incidents with a year and month, 1791 to June 2026, in every state and the Northern Territory; 874 unprovoked
Full record8 traps
- Usable rows
- 1,304
- Coverage
- 1,304 incidents with a year and month, 1791 to June 2026, in every state and the Northern Territory; 874 unprovoked
- How it arrives
- One Excel file, one sheet, headings in the first row
- Language
- Page and file in English
- 1Warm months are also busy months. More people swim and surf in summer, and this file has no count of people in the water. A link between bites and temperature may be a link between bites and beach use: say so, and think about how you could separate them (the same months in warmer and cooler years, for instance).
- 2The nets are seasonal. In New South Wales they are in the water from September to April, the warmer half of the year.
- 3A month with no bite has no row. Count bites per state and month yourself, and put in the zeros, or you will only ever compare months that had a bite. Most months have none: 528 months off NSW from 1982 to 2025 and 211 unprovoked bites.
- 4Decide on provoked bites before you look. 423 incidents were provoked (spearfishing, handling a caught shark, feeding). Their timing follows the people, not the sharks, so most questions use unprovoked bites only.
- 5Tidy the text first. State has one "Qld" beside 378 "QLD", and Site.category and Victim.injury have stray capitals and a trailing space. Make them consistent or your counts split in two.
- 6The shark is often a judgement. Most species were identified from the bite marks and where it happened, not seen, and 66 incidents name no species or "unknown". Wobbegong bites (216) are small and nearly half provoked: decide whether they belong in your question.
- 7The file's own water temperature is nearly empty. Water.temperature.°C is filled in for only 93 incidents, so use NOAA's sea temperature instead and say why.
- 8The record grew with the coast's population. Recorded bites rise decade by decade as more people use the sea and records improve. Compare the same months across years, or the share of bites in each month, rather than raw totals across decades. 2026 stops in June.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Are unprovoked shark bites in New South Wales more frequent in the calendar months when the sea off Sydney is warmer, 1982 to 2025?
- Is the sea off each state warmer, on average, in the months when white sharks are blamed for unprovoked bites than in the months when bull and tiger sharks are?
- Were there more unprovoked bites off New South Wales in warmer summers (December to February) than in cooler ones, 1982 to 2025?
- Incident.year, Incident.month: when
- State, Location, Latitude, Longitude: where
- Provoked/unprovoked: whether the person touched, fed, speared or caught the shark first
- Shark.common.name, Shark.identification.method: which shark, and how anyone knows
- Victim.activity, Victim.injury: swimming, boarding, diving and so on; injured, uninjured or fatal
- Site.category: coastal, estuary/harbour, river, island open ocean or ocean/pelagic
The NSW Department of Primary Industries, with Surf Life Saving NSW flying the drones.
Link to this fileDirect for bites, but it matters for temperature too: the nets are in during the warm months, so any comparison of warm and cool months in NSW is also a comparison of months with and without nets.
The tensionNets catch turtles, dolphins, rays and harmless sharks as well as the sharks they are meant for, and much of what they catch dies. The planned trial of removing them was stopped by one fatal bite, which shows how a single event can outweigh years of catch records in the argument.
SourceThe Western Australian Government. The drum lines were assessed by the state's Environmental Protection Authority; the rebate is run through SharkSmart WA.
Link to this fileDirect for bites. One policy kills sharks near busy beaches; the other protects the person wherever they swim, surf or dive. Neither changes the sea, so in a temperature question each is something to account for, not the thing you explain.
The tensionThe 2014 trial killed 68 sharks and not one was a white shark, the species this file blames for 22 of Western Australia's 25 fatal bites since 1990. The EPA cited a high degree of uncertainty about the effect on the south-western white shark population.
SourceWolf attacks on livestock in France, canton by canton, 2010 to 2022
How often wolves kill or injure farm animals, and where, as recorded by the compensation claims the State accepted: the other side of the argument about wolves returning to the Alps and the Jura.
- One row:
- One canton (or one piece of a canton) in one year: the number of compensated attack reports, and the animals killed or injured, by species.
- Covers:
- 2010 to 2022, 408 cantons in 58 departments: 34,995 compensated reports, rising from 1,079 in 2010 to 3,984 in 2022, and 109,673 sheep killed or injured
Full record8 traps
- Usable rows
- 1,687
- Coverage
- 2010 to 2022, 408 cantons in 58 departments: 34,995 compensated reports, rising from 1,079 in 2010 to 3,984 in 2022, and 109,673 sheep killed or injured
- How it arrives
- One ODS spreadsheet: the data are on the sheet Feuille1, the notes on Feuille2
- Language
- Page and file in French
- 1Blank cells mean "same as above". It is a pivot-table export: region, department, arrondissement, canton and id appear only on the first row of each group. Fill them down before you sort or filter.
- 2The last row is a grand total. Total Résultat (34,995 reports) sits at the bottom. Delete it, or every sum doubles.
- 3Three rows have no place. Donnée géographique non renseignée covers 23 reports with no canton.
- 4One canton can take several rows in one year. Fifty cantons are cut by an arrondissement boundary. Add the rows up by department, canton and year: 1,862 rows become 1,687 canton-years.
- 5A missing canton-year means no paid claim, not missing data. Only canton-years with at least one compensated report are listed, so fill the gaps with zero when you join.
- 6It counts paid claims, not attacks. Only cases where compensation was decided and the wolf was not ruled out are listed, and the compensation rules changed in 2019.
- 7Today's cantons for every year. The boundaries are those drawn in 2015, even for 2010 to 2014, and ten codes are towns rather than real cantons (0599 is the whole of Gap).
- 8It stops in 2022. The note promises 2023 after consolidation; in September 2026 later years exist only as PDFs.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Across French cantons, is the number of compensated wolf attacks on livestock related to the number of confirmed wolf records, 2013 to 2022?
- How did the number of sheep killed or injured for each compensated report change between 2010 and 2022?
- INSEE_DEP and INSEE_CAN: department and canton number. Together they make the INSEE canton code: 04 + 12 = 0412
- Annee: the year
- Compter - numero_constat: the number of compensated attack reports
- Somme - retenues_ovin: sheep killed or injured. Beside it, cattle, goats, dogs, horses and other animals
- INSEE_REG, INSEE_ARR, ID: region, arrondissement and the map id of the canton piece
The ecology and agriculture ministries, run by the coordinating prefect, the prefect of the Auvergne-Rhône-Alpes region.
Link to this fileDirect on the argument, indirect on this file: shooting is meant to protect flocks, but neither this file nor the wolf evidence records culling, and the shooting totals are published as PDFs only.
The tensionOn 19 October 2023 the national nature protection council (CNPN) gave the draft plan an unfavourable opinion, unanimously, objecting to making lethal shooting easier, including during the breeding season with no scaring first.
SourceThe State, through the environment ministry, paying through the ASP (the State payments agency).
Link to this fileIt defines this file: only claims that were paid appear. Since 2019 an unprotected sheep or goat flock in the most exposed zone is not compensated from its third attack in 12 months, so a rule change can move the count without any change in the wolves.
The tensionThe plan itself says the payments are EU state aid, so they must be neither so high that they distort competition nor so low that they harm farmers.
SourceFarmers apply through their department's DDT(M); shepherds are funded by the agriculture ministry and the EU.
Link to this fileDirect in aim: protection is meant to cut attacks. But neither file says which flocks were protected, so you cannot separate protection from attacks, only say that both rose.
The tensionThe plan says the budget will not cope if the growth continues, and farmers keep asking for shepherd costs to be paid in advance, which EU farm-policy rules do not allow.
SourceSwiss breeding birds by group, migrants included, every year since 1990
Whether Switzerland's breeding birds are declining, and which kinds: migrants that spend the winter far away, farmland birds, birds of the Alps. The Swiss Ornithological Institute combines the trends of up to 180 species into one index for each group.
- One row:
- One group of bird species in one year: its index, where 1990 is 100, with its lower and upper 95% limits and the number of species in the group that year.
- Covers:
- 15 groups of Swiss breeding birds, 1990 to 2025
Full record8 traps
- Usable rows
- 540
- Coverage
- 15 groups of Swiss breeding birds, 1990 to 2025
- How it arrives
- One Excel file: the index for 15 groups, a sheet listing which species are in each group, and a sheet giving where each group comes from
- Language
- Page and file in English
- 1An index, not a count. 100 is the group's level in 1990, not 100 birds. An index of 120 means the group is about a fifth above its 1990 level. Compare groups by how much they changed, never by numbers of birds.
- 2The whole series is recalculated every year. Each new file re-estimates every year back to 1990. Compared with last year's file, 302 of 525 group-years moved by more than half a point, and some by up to 10. Record the file name and keep your copy.
- 3The last year is the least sure. In the 2024 file, 2024 held 51 long-distance migrants; this file has 57 for the same year. Look at the 95% limits, which are wider at both ends of the series than in the middle, before you lean on the latest year.
- 4The groups overlap. One species can be in several groups: the Barn Swallow is a long-distance migrant and a farmland characteristic species. 20 of the 57 long-distance migrants are farmland target or characteristic species. Two groups that share species are not independent, so do not treat them as separate evidence.
- 5The species in a group change. Long-distance migrants number 53 in 1990 and 57 later. Check the Species-Composition sheet and say which species your group holds.
- 6Choose your group before you look. There are 15 indices. Testing one after another until one looks interesting decides your result for you.
- 7Migrants are counted only while they breed here. A long-distance migrant spends most of its year on its route and in its winter quarters, so a change in its index can come from anywhere along the way.
- 8Groups, not species. The file gives one index per group. It does not give each species' own trend.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has the index of long-distance migrants changed differently from the index of short-distance migrants between 1990 and 2025?
- Have the farmland species named in the environmental objectives for agriculture changed differently from Swiss breeding birds as a whole since 2008, when the objectives were published?
- Have the species expected to gain from climate change and those expected to lose from it moved apart since 1990?
- e: the group's name in English (d, f and i give it in German, French and Italian)
- year
- mean: the group's index, 100 in 1990
- lower95 and upper95: the range the index probably lies in
- Nspecies: how many species the group held that year
- The Species-Composition sheet: every species, with a 1 under each group it belongs to
The Federal Office for the Environment (FOEN) and the Federal Office for Agriculture (FOAG). Farmers meet them mainly through direct payments, including payments for biodiversity areas on their land.
Link to this fileDirect for three of the groups. The objectives name target and characteristic species for farmland, and the file's three EOA groups are exactly those species. For the migrant groups it is partial: some migrants are farmland species (the Barn Swallow is both), but most of a migrant's year is spent outside Switzerland.
The tensionWhen the 2016 report came out, Pro Natura criticised "the constant glossing over of the sometimes serious environmental damage done by the farming sector" (Pro Natura, 9 December 2016, translated from German).
SourceThe Federal Council, with the Federal Office for the Environment and the cantons.
Link to this fileIndirect for the migrants: it can only act on Switzerland, where they breed, and not on their routes or where they spend the winter. Closer for the groups that live here all year. Say which of your groups it can reach.
The tensionOn 22 September 2024 about 63% of voters rejected the biodiversity initiative, which asked for stronger protection. Farmers and business groups called it too extreme and said the current law was enough. The farmer and SVP member of parliament Katja Riem said: "A balance must be struck between protection and benefit" (swissinfo, 22 September 2024).
SourceAir quality3
Fine particles (PM2.5) day by day at 19 stations across Nouvelle-Aquitaine, including the Gironde fires of July 2026
Wildfire smoke travelling: how much fine dust is in the air at stations near to and far from a large fire, on the days of the fire and on ordinary days. The Gironde fires of late July 2026 are inside the file.
- One row:
- One monitoring station on one day: the day's mean PM2.5 in µg/m³, with a flag saying whether the value was validated.
- Covers:
- 19 stations in the 12 departments of Nouvelle-Aquitaine, 18 of them with values, for the 12 months up to the day you download (28 September 2025 to 27 September 2026 when counted). Three are in Gironde: Talence, Bassens and Bordeaux - Gautier
Full record8 traps
- Usable rows
- 6,363
- Coverage
- 19 stations in the 12 departments of Nouvelle-Aquitaine, 18 of them with values, for the 12 months up to the day you download (28 September 2025 to 27 September 2026 when counted). Three are in Gironde: Talence, Bassens and Bordeaux - Gautier
- How it arrives
- One CSV, separated by semicolons, about 1.6 MB, from the csv button after choosing PM2,5 and Moyenne journalière. Also as xlsx or json
- Language
- Page and file in French
- 1Choose the smoke days from a source, not from the numbers. The Copernicus Atmosphere Monitoring Service (CAMS) reported fire emissions in France on 24 July 2026 higher than ever before at that time of year, smoke from the Gironde fires over Bordeaux on Sunday 26 July, and firefighters calling the fires stabilised on Tuesday 28 July (CAMS news, 30 July 2026). Fix your smoke days from a report like this before you look. Days chosen because their values are high will always look high.
- 2The form reaches back only 12 months. The date boxes stop a year before today, so the July 2026 fires leave the form in late July 2027. The csv button is a plain web address with the dates in it (date_deb and date): changing them brings back earlier years, from 1 January 2017 at Talence and Bassens. Say in your method that you did this.
- 3The day runs on UTC, not French time. Each daily value is the mean from midnight to midnight UTC, which is 2 am to 2 am in a French summer. The portal's description says local time, but the daily values match the UTC day. The hourly file is in French time, so daily means you build from it will not match this file.
- 4Missing days are NULL or blank. In the 12 months, 144 values are NULL and 428 are blank, all with statut_valid False. One station, La Rochelle - Saint-François-La Pallice, is blank on every day. Keep only rows where statut_valid is True.
- 5Gautier is beside a road. Bordeaux - Gautier is a traffic station; Talence and Bassens measure the general air of the city. Compare stations of the same kind, or say why you mixed them. Gautier also has only 277 validated days in the 12 months.
- 6PM2.5 does not only come from smoke. Traffic, wood stoves in winter and dust raise it too. Compare smoke days with other days of the same season, not with the whole year.
- 7Older years have holes. If you go back further: Talence has no validated PM2.5 from 24 February 2022 to the end of that year, so the Gironde fires of summer 2022 are missing there. Bassens has three days of -1 marked True (20 and 21 August and 1 October 2022): a concentration cannot be below zero, so take them out. Bordeaux - Gautier starts on 6 July 2020.
- 8French in the file. Particules fines PM2,5 is PM2.5, written with a decimal comma; journaliere means daily; Fond means the general air of an area. The values are whole numbers.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much higher was the daily PM2.5 at Talence, Bassens and Bordeaux - Gautier on the smoke days of 24 to 28 July 2026 than on the other days of July and August 2026?
- Did the stations farther from the Gironde fires have a smaller rise in PM2.5 from 24 to 28 July 2026 than the stations near them?
- On how many days in the 12 months was each station's daily PM2.5 above 25 µg/m³, the EU's daily limit from 2030, and how many of those days fell during the fires?
- nom_dept, nom_com, nom_station and code_station: the department, the town, the station and its code, such as FR31002 for Talence
- valeur: the day's mean PM2.5, in whole µg/m³ (unite reads ug.m-3)
- date_debut: the day, written 2026-07-26 00:00:00+00:00
- statut_valid: True if the value was validated, False if not
- typologie and influence: whether the station is urban (Urbaine), on the edge of a town (Périurbaine) or rural, and whether it measures the general air (Fond), traffic (Trafic) or industry (Industrielle)
- x_reglementaire and y_reglementaire: where the station is, in metres on the French Lambert 93 grid, so you can work out its distance from a fire
The European Parliament and the Council of the EU. France has to meet it, and the European Commission checks.
Link to this fileDirect: the limit is a daily mean of PM2.5 at a station, which is exactly one row of this file. But the directive also lets a country leave out days when it can show the pollution came from natural sources, and it names wild-land fires among them. So a smoke day may not count against France at all. Decide whether your question is about the air people breathe or about the law's count.
SourceThe French Parliament. The law asks the interior, forest, environment and planning ministries to write a national strategy against fires in forests and other vegetation. It also extends the legal duty to clear undergrowth (the obligations légales de débroussaillement) to parts of France that did not have it before.
Link to this fileIndirect. It aims to prevent fires, which is where the smoke comes from, but this file measures the air, not the fires. One year of air cannot judge a law about fires; use it to explain why the smoke matters.
The tensionNot a clash between groups, but a gap between the law and what followed. In April 2025 three members of the National Assembly who reviewed the law regretted that the national strategy it asked for within one year had still not been adopted, so it was not yet working.
SourceOzone-depleting substances, country by country, since 1986
Whether the Montreal Protocol actually worked, and what happened to the chemicals brought in to replace the ones it banned.
- One row:
- One country in one year: consumption of each chemical family in ODP tonnes, meaning tonnes weighted by ozone-depleting potential rather than tonnes of gas.
- Covers:
- 177 entities, 169 of them countries, 1986 to 2022
Full record9 traps
- Usable rows
- 5,380
- Coverage
- 177 entities, 169 of them countries, 1986 to 2022
- How it arrives
- One CSV, one column per chemical family, from the chart URL with .csv on the end
- Language
- Page and file in English
- 1Not one EU member state is in this file as a country. They report through a single "European Union (28)" row, because the Union ratified the treaty in its own right. Germany, France, Italy, Spain and Poland are simply absent, and nothing says so: you search for Germany, find nothing, and think you made a mistake. No other problem in this file limits your questions more: it cuts the developed countries you can compare from fifty down to eighteen.
- 2Removing the aggregates takes two rules, not one. Only the European Union row has a blank code. The other seven, the six continents and World, carry codes beginning OWID_, so the obvious "delete the rows with no code" leaves seven aggregates sitting among your countries and your country count reads 176 instead of 169. Unlike the CO2 file, no real country here carries an OWID_ code, so the two-part rule is safe.
- 3Consumption can be negative, and 211 country rows are. UNEP defines it as production plus imports minus exports, so a country running down a stockpile posts a negative figure: Argentina is −528 ODP tonnes of carbon tetrachloride in 1990. It is a real accounting entry rather than an error, and either way it will distort a mean.
- 4The zeros are filled in, not measured. Our World in Data's companion series on these same 5,642 rows is labelled "zero-filled", and 297 country rows read zero for every one of the six chemicals. Zero here mixes "none consumed" with "nothing reported", and the difference is the whole story of compliance.
- 5The last year is a cliff, not a collapse. 2022 carries 14 countries. 2021 carries 168. Plot a global total to 2022 and it shows a huge fall that is really 92% of the reporting countries missing.
- 6Whether a party is an Article 5 party is not in the file, and it is the one column a question about the treaty's fairness needs. You have to bring it: the Handbook for the Montreal Protocol defines the two groups, and the Ozone Secretariat's list of parties says which group each country is in. Check it country by country: seven former Soviet states, including Azerbaijan and Kazakhstan, are on the developed side.
- 7ODP tonnes are weighted by how much damage a substance does, not by its mass. A tonne of halon counts for many times a tonne of HCFC, so these numbers cannot be added to anything measured in real tonnes, and part of any trend you find comes from the weighting.
- 8Eight aggregates share the entity column with the 169 countries: the continents, the EU-28 and World.
- 9The download has one column per chemical family and no total. If your question is about ozone-depleting substances as a whole, you are making that total yourself, and the sibling chart that already has it is a different file.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In 1996, the deadline for developed parties, how much of its own peak was each party still using?
- In one country, did HCFC consumption rise as CFC consumption fell, and by how much of the gap?
- Compare the phase-out paths of two countries at different income levels against the dates the treaty set for each.
- Chlorofluorocarbons (CFCs), the original problem
- Hydrochlorofluorocarbons (HCFCs), the substitute that became the next problem
- Halons, carbon tetrachloride, methyl chloroform, methyl bromide
- Entity, code and year
Every country in the world, which is why people give it as the example of an environmental treaty that worked. Each party reports its consumption annually under Article 7, and those reports are this dataset.
Link to this fileAs direct as it gets: the treaty controls the substances, and the numbers in this file are the parties' own returns on those substances. The warning here is the opposite of the usual one. The link is so close that the interesting question is not whether consumption fell, but which countries cut first and what it cost them.
The tensionDeveloping countries argued that they were being asked to give up technologies the rich countries had already used to grow. To get every country to sign, the treaty gave them extra time under Article 5 and a fund to pay for compliance.
SourceIndustrialised parties pay in; developing (Article 5) parties draw on it. It exists to pay the agreed extra costs of compliance, so that a country is not asked to choose between the treaty and its own industry. It has approved over 10,000 projects worth about $4.3 billion across 144 developing countries.
Link to this fileIt does not act on the chemicals. It acts on who can afford to stop using them. So it affects how a country phased them out, not whether it did. That makes it the right strategy for a question comparing two countries, and the wrong one for a question about a global total.
The tensionWho pays, and how much is enough. The fund is replenished by negotiation every three years, so the level of support is argued over on a cycle, and the argument is precisely the economic-versus-environmental one Criterion B is asking about.
SourceThe parties to the Montreal Protocol, in decision XIX/6 and then at Kigali.
Link to this fileThis is the story the file tells best, and you can see it in a single country's rows: CFCs fall away and HCFCs climb to replace them, then HCFCs turn down in their turn. China's HCFC consumption runs 621 ODP tonnes in 1989, 5,165 in 2000 and 19,935 in 2010 while its CFCs collapse from 39,124 to 969 over the same decade.
The tensionA solution that created the next problem, twice. HCFCs damaged the ozone layer less than CFCs and still damaged it; HFCs do not damage it at all and are powerful greenhouse gases. Kigali is an environmental goal in conflict with a different environmental goal, and behind both is the demand for cooling in the hottest and poorest places.
SourceUK air quality, hour by hour at every monitoring site
Roadside air pollution in a city that has been charging drivers to clean it up, and whether the air at the kerb changed when the charges did.
- One row:
- One hour at one monitoring site, with a measured value, a status letter and a unit for each pollutant the site runs.
- Covers:
- Hundreds of sites. One site-year is 8,760 rows, one per hour; the London Marylebone Road file counted here holds 8,353 usable nitrogen dioxide values for 2023.
Full record10 traps
- Usable rows
- 8,353
- Coverage
- Hundreds of sites. One site-year is 8,760 rows, one per hour; the London Marylebone Road file counted here holds 8,353 usable nitrogen dioxide values for 2023.
- How it arrives
- One CSV per site per year, about 6 MB, 134 columns
- Language
- Page and file in English
- 1Four lines of preamble, and the site's name sits on its own row. Point a spreadsheet at the file and your header is "Data supplied by UK-AIR on 29/6/2024". The real header is row five.
- 2134 columns for 44 pollutants, because every pollutant is followed by its own status column and its own unit column. Select a pollutant by its position and you may pick up a status column instead.
- 3Column names carry HTML. Particulates arrive as `PM<sub>10</sub> particulate matter`, which no lookup you type by hand will ever match.
- 4Missing hours are blank, and there are hundreds. 120 hours of nitrogen dioxide are missing in 2015, 474 in 2019 and 430 in 2024. An annual mean is a mean of the hours the instrument was working, and the number of those changes year to year.
- 52024 has 8,784 rows rather than 8,760. It is a leap year, and anything that assumes a fixed row count breaks on it quietly.
- 6The status letter matters: R is ratified, P and P* are provisional. Recent years arrive provisional and the numbers can still change.
- 7There is unusually heavy confounding here, and ignoring it is the main way to lose marks. The zone began in April 2019, the pandemic emptied the road in 2020, the zone expanded twice afterwards, and a congestion charge has been operating on the same street since 2003. A before-and-after across 2020 compares two very different situations.
- 8For wildfire smoke, use Manchester Piccadilly (MAN3), not Marylebone Road. The Saddleworth Moor fires began on 24 June 2018, and Graham and others (2020, Environmental Research Communications) show their smoke over Manchester on 27 to 30 June and use this station. The file is MAN3_2018.csv, in the same place as the Marylebone Road files.
- 9No PM2.5 at Manchester Piccadilly from 24 June to 12 July 2016, and 381 of the 456 hours in 2019. Leave 2016 out, and say how you treated 2019.
- 10The PM2.5 unit changes name in 2020. At Manchester Piccadilly the unit column reads ugm-3 (TEOM FDMS) up to 2019 and ugm-3 (Ref.eq) from 2020. Keep a comparison on one side of the change, or say why you crossed it. The number of columns changes too (26 up to 2018, 29 in 2019, 23 from 2020), so find the PM2.5 column by its name in each year.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Did nitrogen dioxide at Marylebone Road fall faster in the five years after April 2019 than in the five years before it?
- How does the weekday and weekend difference in nitrogen dioxide at one site compare before and after the zone began?
- Which pollutant at one site fell most between 2015 and 2024, and which barely moved?
- How did daily PM2.5 at Manchester Piccadilly from 24 June to 12 July 2018, while the Saddleworth Moor fires burned, compare with the same weeks of 2014, 2015 and 2017?
- Nitrogen dioxide, nitric oxide and NOx, µg/m³, hourly
- PM10 and PM2.5, ozone, carbon monoxide, sulphur dioxide
- At this site, about thirty hydrocarbons and six black-carbon channels as well
- A status column and a unit column after every single pollutant
The Mayor of London and Transport for London. Drivers of vehicles that miss the emission standards pay £12.50 a day to drive inside the zone: Euro 4 for petrol cars, Euro 6 for diesel, Euro 3 for motorcycles.
Link to this fileA strategy and a dataset rarely fit this closely. Marylebone Road sits inside the original central zone, so the April 2019 boundary falls in the middle of a continuous hourly series that starts long before it and runs long after. The annual mean nitrogen dioxide at that site reads 84.7 µg/m³ in 2018 and 62.7 in 2019.
The tensionWho pays to clean the air. The charge falls on people who cannot afford to replace a vehicle, which is why a £110 million scrappage scheme was attached to the last expansion. Five authorities, the London boroughs of Bexley, Bromley, Harrow and Hillingdon with Surrey County Council, took the expansion to judicial review and lost in July 2023, having spent £730,941 between them.
SourceTransport for London, operating a Greater London zone aimed at lorries, buses, coaches and larger vans rather than at cars.
Link to this fileHeavy diesel vehicles produce much of the nitrogen oxide in a street canyon (a street lined with tall buildings), so this acts on the biggest single source at a kerbside site. It also overlaps the zone above, and that is the problem: with two schemes on the same street, neither one can take all the credit for a change.
The tensionThe cost lands on hauliers and small operators who cannot pass it on, and on the trades whose vans are their livelihood, against the health of people who live on the road.
SourceTransport for London, charging vehicles to enter central London on weekdays. Its purpose is traffic, not emissions, and it covers almost exactly the area of the original ULEZ.
Link to this fileHere for the opposite reason to the other two: this is the confounder. An older instrument with a different aim has been acting on the same streets for the whole of your series, so any change you credit to the 2019 boundary has to survive the question of what else was already changing traffic on that road.
SourceClimate and weather13
Geneva's official chestnut tree: the first leaf of spring since 1818
Spring arriving earlier in a warming city. The date one horse chestnut on the Treille opens its first leaf has been written down every year since 1818.
- One row:
- One spring: the date the sautier saw the first leaf open on the official tree, with the tree's number (1 to 4) and the sautier's name.
- Covers:
- 209 springs, 1818 to 2026, one date each, observed on four trees in turn
Full record8 traps, 1 on the three lines
- Usable rows
- 209
- Coverage
- 209 springs, 1818 to 2026, one date each, observed on four trees in turn
- How it arrives
- A table on a web page, not a file. It is folded away at first: open it, select the whole table, copy it and paste it into Google Sheets
- Language
- Page and file in French
- Natural· Change in when the horse chestnut comes into leaf· by nothing to match: each site gives one number, how far its leafing moved
- 12003 is labelled 2002. The table has two rows for 2002: 7 February 2002, and 29 December 2002, which the sautier counted as the first leaf of spring 2003. There is no row for 2003. Copy it naively and you lose a year and put the earliest date on record in the wrong one. On a day-of-year scale, 29 December before the spring is day −2.
- 2Autumn leaves are not spring. In 2006 and 2011 the old tree also opened a leaf in October or November, under stress. Those dates sit in the notes column and are not the spring date.
- 3The dates are French text. A spreadsheet reads 16 mars as words, not a date. Split it into day and month, then turn it into a day of the year with =DATE(year, month, day) - DATE(year, 1, 1) + 1. That day of the year is the quantity you analyse, and making it is your processing.
- 4The tree number and the sautier are merged cells. Each appears once at the top of its block and pastes as blanks below. Fill them down before you group by tree.
- 5Four trees, not one. They were replaced in 1905, 1928 and 2016, and each successor was chosen to leaf close to the last, but not identically. The third tree was early (median about day 68) and the fourth has leafed later, which can look like the trend reversing. Name the change of tree as a confounding factor, or compare within one tree.
- 6One tree in a city centre. It stands beside the warm stone of the Tour Baudet, and the Grand Conseil's own booklet puts part of the trend down to the city growing hotter, not only the climate. What you find is true of this tree, not of Geneva's trees.
- 7Footnote numbers come with the copy. Five date cells carry a reference, which pastes as a number in square brackets straight after the date. Delete them.
- 8With 209 years, any trend is significant. Year against day of year gives Spearman's rs of about −0.78. Report how strong the change is, in days per century or between periods, not only that p is small.
- Between 1955 to 1984 and 1986 to 2015, did the official tree come into leaf earlier by more than the horse chestnuts at MeteoSwiss's countryside stations? (MeteoSwiss phenology, one file per station.)
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the date of the first leaf on Geneva's official chestnut tree changed between 1818 and 2026?
- How much earlier was the first leaf in the last forty springs on tree 3 (1976 to 2015) than in the first forty on tree 1 (1818 to 1857)?
- Date: the day and month in French, such as 16 mars. Text, not a date a spreadsheet can read
- Année: the year
- N°: which of the four official trees was observed. Tree 1 until 1905, tree 2 from 1906 to 1928, tree 3 from 1929 to 2015, tree 4 since 2016
- Sautier: who made the observation. Fourteen people over two centuries
- Remarque: notes, including the weather and the years a leaf also appeared in autumn
Between 1955 to 1984 and 1986 to 2015, did Geneva's official chestnut, in the city centre, come into leaf earlier by more than the horse chestnuts at the countryside stations?
Watch outThe official tree records the first bud and the stations record half the leaves unfolded, about six weeks later, so compare how far each moved and never the days. Stop at 2015: the tree was replaced in 2016. Keep the stations below 600 m and out of cities, and it is one tree against fourteen places.
The Canton of Geneva's Conseil d'État, with funding for the communes and for private owners, who hold two thirds of the urban area.
Link to this fileIndirect. The strategy exists to cool the city, and part of this tree's earlier spring is urban heat, not only the climate. But one tree cannot show a canopy effect, and the plan's target lies decades beyond the data, so it frames the issue rather than being measurable here. The credit passed unanimously: find a documented disagreement before you write about tension.
SourceThe Canton of Geneva's Conseil d'État.
Link to this fileIndirect. It addresses the warming that is one cause of earlier leafing, and its heat-island measures address the other. Neither can be seen in one tree's dates, so state the connection and do not claim the data measures the plan.
SourceWhen Switzerland's trees come into leaf: MeteoSwiss's phenology network
Spring arriving earlier as the climate warms. Observers at 175 stations have written down when 26 plants leaf, flower and change colour, some since 1951.
- One row:
- One station in one year: the day each stage was seen there, written as a number like 19540330, one column per plant and stage.
- Covers:
- 175 stations, 1951 to 2026; horse chestnut leaf unfolding at 149 of them
Full record7 traps, 1 on the three lines
- Usable rows
- 759
- Coverage
- 175 stations, 1951 to 2026; horse chestnut leaf unfolding at 149 of them
- How it arrives
- One CSV per station, semicolon-separated, in the Latin-1 encoding, plus a station list and a list of what each column means
- Language
- Page and file in English
- Human· City centre or countryside· by nothing to match: each site gives one number, how far its leafing moved
- 1Not the same stage as Geneva's official tree. The stations record the day half the leaves have unfolded; the official tree's register records the first bud. In any year they are about six weeks apart, so never compare the days themselves: compare how far each has moved.
- 2One file per station. There is no single table. Download each station's file and keep its year and its horse chestnut column.
- 3The days are numbers, not dates. 19540330 is 30 March 1954. Turn it into a day of the year with =DATE(LEFT(B2,4), MID(B2,5,2), RIGHT(B2,2)) - DATE(LEFT(B2,4), 1, 1) + 1.
- 4Blank means not recorded, never zero. Stations skip years and change observers. Count the springs you have in each period, and leave out a station with fewer than 20 in either.
- 5Height sets the season. A station at 1,000 m comes into leaf long after one at 400 m. Keep to one band of height, such as below 600 m, so you compare like with like.
- 6Some stations swing a long way. Between 1955 to 1984 and 1986 to 2015, Murg moved 13 days later and Wädenswil 13 days earlier, far more than their neighbours. A new observer or a new tree can do that: say so rather than quietly dropping them.
- 7Three low stations are in cities. Zürich / Fluntern, Zürich / Albisgüetli and Biel. Leave them out of a countryside group, or make them a group of their own.
- Between 1955 to 1984 and 1986 to 2015, did Geneva's official chestnut, in the city centre, come into leaf earlier by more than the horse chestnuts at the countryside stations below 600 m?
- maesh13d: horse chestnut, half its leaves unfolded, written as YYYYMMDD (19540330 is 30 March 1954)
- maesh09d: horse chestnut budbreak, the stage Geneva's official tree records. No station has a value in it
- reference_timestamp: the year, written 01.01.1954 00:00
- station_height_masl, station_name and station_canton, in the station list: how high the station is, and where
- Every other column is another plant or stage, named in the parameter list
The 14 countryside stations (14)
Below 600 m, with at least 20 springs of horse chestnut leaf unfolding in both 1955 to 1984 and 1986 to 2015, and not in a city.
- Cartigny (GE, 400 m)CSV
- Versoix (GE, 440 m)CSV
- Nyon / Changins (VD, 435 m)CSV
- Möhlin (AG, 305 m)CSV
- Liestal (BL, 350 m)CSV
- Aurigeno / Ronchini (TI, 350 m)CSV
- Sargans II (SG, 480 m)CSV
- Wädenswil (ZH, 480 m)CSV
- Murg (SG, 500 m)CSV
- Sarnen (OW, 500 m)CSV
- Rafz (ZH, 515 m)CSV
- Moutier (BE, 530 m)CSV
- Merishausen (SH, 540 m)CSV
- Seon (AG, 550 m)CSV
Between 1955 to 1984 and 1986 to 2015, did Geneva's official chestnut, in the city centre, come into leaf earlier by more than the horse chestnuts at the countryside stations?
Watch outThe official tree records the first bud and the stations record half the leaves unfolded, about six weeks later, so compare how far each moved and never the days. Stop at 2015: the tree was replaced in 2016. Keep the stations below 600 m and out of cities, and it is one tree against fourteen places.
The Swiss Federal Council, with the Federal Office for the Environment (FOEN) coordinating the federal offices.
Link to this fileIndirect. It plans for the consequences of a warming climate, of which earlier springs are one; it does not act on when trees come into leaf. State the connection and do not claim the data measures the strategy.
SourceJapan's cherry blossom dates at every weather station since 1953
Spring coming earlier across Japan, and faster in some places than others. The same cherry is watched at weather stations from Hokkaido to Okinawa.
- One row:
- One station in one year: the month and day its standard cherry tree first flowered (開花, kaika) or reached full bloom (満開, mankai).
- Covers:
- 102 stations have appeared since 1953; 58 were still observed in 2026, and 45 have every year from 1953 to 2026
Full record7 traps, 1 on the three lines
- Usable rows
- 3,330
- Coverage
- 102 stations have appeared since 1953; 58 were still observed in 2026, and 45 have every year from 1953 to 2026
- How it arrives
- Eight web pages of text, one per decade, not a file. Copy each page's block into Google Sheets and use Data › Split text to columns
- Language
- Page and file in Japanese
- 1Not every station watches the same tree. The main tree is Somei-yoshino, but 10 of the stations still observed, five in Hokkaido and five in Okinawa and the Amami islands, watch another species, named in the 代替種目 column. Compare like with like: drop them, or name the difference.
- 2Stations have stopped. 102 stations have appeared since 1953, but only 58 were observed in 2026: the rest stopped between 1991 and 2013, most of them between 2005 and 2009. Use the 45 with every year, or your early and late groups will be different places.
- 3It arrives as text. Paste a page into Google Sheets and each station is one line of text in one cell. Data › Split text to columns, with a space as the separator, gives you the columns. The dates then have to be turned into a day of the year: =DATE(year, month, day) - DATE(year, 1, 1) + 1.
- 4Eight pages, eight years' headings. Each page covers a decade, from 1953 to 1960 up to 2021 to 2026, with its own columns. Stack them carefully, and check that the station order has not changed between pages.
- 5First flowering and full bloom are separate pages. sakura003 pages are first flowering and sakura004 pages are full bloom. Pick one and say which.
- 6The 平年値 column is an average, not a year. It sits at the end of every row. Delete it before you calculate.
- 7Station names are kanji. You need latitude and whether a station is in a city for most questions, and neither is on the page. JMA's station list gives both; record where you got them.
- Is the change in first-flowering date since the 1950s larger at stations in large cities than at stations in small towns?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How many days earlier did cherry trees first flower in 2017 to 2026 than in 1953 to 1962, across the 45 stations with every year?
- What is the relationship between a station's latitude and how much earlier its cherry trees now flower?
- 地点名: the station name, in Japanese. Kyoto is 京都 and Tokyo is 東京. Keep a list of the ones you use with their English names
- Each year: two numbers, the month (月) then the day (日). - means no observation that year
- 平年値: the normal, JMA's average for 1991 to 2020. Not a year: leave it out
- 代替種目: a substitute species, filled in only where the station watches a different cherry
A committee of national ministries, set up in 2002, with local governments, businesses and residents expected to act on it.
Link to this fileMore direct than most. The framework is about cities warming faster than the country around them, and this network has large cities and small towns side by side, so comparing how much earlier each now flowers is a question the policy is about. It cannot show whether the policy has worked.
SourceThe Government of Japan, through the Cabinet and the Ministry of the Environment.
Link to this fileIndirect. The targets address the warming that moves flowering earlier, but no flowering date can show whether a target is working. State the connection; do not claim the data measures the plan.
The tensionMore than 3,000 public comments were received and more than 80% asked for higher targets; environmental groups and Climate Action Tracker called the 2035 goal far too weak. The government kept it, and InfluenceMap reports that policy follows the Japan Business Federation (Keidanren), whose position was the least ambitious of the industry groups.
SourceAverage rainfall in every country, in millimetres a year
How much rain each country receives in a normal year: the supply that every country's water use is set against.
- One row:
- One country: its long-term average rainfall in millimetres a year, copied into every year's column.
- Covers:
- 182 countries, from Egypt's 18 mm to Colombia's 3,240 mm, in the columns for 1961 to 2022
Full record6 traps, 1 on the three lines
- Usable rows
- 182
- Coverage
- 182 countries, from Egypt's 18 mm to Colombia's 3,240 mm, in the columns for 1961 to 2022
- How it arrives
- A zip of three CSVs: the data is the one whose name begins API_, and the two Metadata files say which rows are countries
- Language
- Page and file in English
- 1A long-term average, not the year's rain. The same number is copied into every year for all but nine countries. It describes a country's climate: compare countries with it, never one country's years.
- 2One number for a whole country. Brazil's covers the Amazon and the dry north-east, and Australia's covers its tropical north and its desert.
- 3Four lines of notes sit above the header. The header is the fifth line and begins Country Name, and there is an empty column after the last year.
- 4Aggregates sit among the countries: 47 rows such as World, Euro area and Low income. The zip's Metadata_Country file leaves Region blank for every one of them, so use it to take them out.
- 5Years run across. One column per year: turn them into one column before you match the file to anything with a year column.
- 6It stops in 2022, and 2022 has two countries fewer than 2021.
- Is the relationship between protected land and tree cover loss the same in dry countries as in wet ones? Rainfall is the weather, so it is neither of the three lines: split the countries at a rainfall you fix before you look, then test each group (the starter investigation).
- Country Name, and Country Code, the three-letter ISO code
- 1961 to 2022: the same average in each year's column
Its 197 parties, 196 countries and the European Union. 169 of the countries say they are affected by desertification, land degradation or drought.
Link to this fileIndirect: the convention's drylands are defined by how much rain falls against how much evaporates, and this file is the rain half of that comparison. It cannot say where land is degrading.
SourceEvery UN member state, reporting through FAO, which is the custodian of the target's indicators and the source of the World Bank's water figures.
Link to this fileIndirect: no strategy changes the rain. It is the supply side of indicator 6.4.2, which divides what a country withdraws by what its rain and rivers renew, so a dry country reaches high water stress on far less use.
The tensionIrrigation. Farming is the largest withdrawal almost everywhere, and the same agenda asks countries to end hunger (SDG 2), which in dry countries has usually meant irrigating more land rather than less.
SourceGISTEMP, global surface temperature anomalies since 1880
How much warmer the surface of the planet is than it used to be, which is the number every climate target is written in terms of.
- One row:
- One year: twelve monthly anomalies in °C against the 1951 to 1980 average, plus annual and seasonal summaries of the same year.
- Covers:
- 147 years, 1880 to 2026, for the globe; the same page carries hemispheres and zones
Full record6 traps
- Usable rows
- 147
- Coverage
- 147 years, 1880 to 2026, for the globe; the same page carries hemispheres and zones
- How it arrives
- One CSV of about 13 KB, with a title line above the header row
- Language
- Page and file in English
- 1The header is on the second row. The first line is a title, `Land-Ocean: Global Means`, so a reader that assumes row one is the header treats the column names as data.
- 2Missing values are three asterisks. Nine cells read `***`, all of them the current year's months and the two overlapping-year columns for 1880. Any tool that converts the columns to numbers will either fail or quietly leave them as text.
- 3Monthly and summary columns sit in the same row. J-D, D-N and the four seasons are averages of the months beside them, so if you reshape the file into one row per month without removing them, you count every year twice.
- 4J-D and D-N are not the same year. One is January to December, the other December to November, and they differ by a tenth of a degree in places. Choose one and say which.
- 5An anomaly is a difference from the 1951 to 1980 average, not a temperature. Every number in this file is relative to a baseline, and that baseline is not the one the 1.5 °C target uses.
- 6The last row is a part-year. 2026 has values to August and asterisks after, so an annual figure for the current year does not exist and a chart drawn to it ends on a month rather than a year.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How does the rate of warming since 1980 compare with the rate before it?
- Which season has warmed fastest, and does that hold across the whole record?
- How many of the ten warmest years in this file are in the last fifteen years of it?
- Twelve monthly anomalies, in degrees Celsius against the 1951 to 1980 mean
- J-D and D-N, two different ways of averaging a year
- DJF, MAM, JJA and SON, the four seasons
- Nothing else. It is a small file that says one thing very well
Industrialised countries, taking binding emission targets while developing countries took none: the first climate treaty with numbers in it.
Link to this fileWith a temperature series, the honest question is not whether a treaty worked. Temperature is what all the treaties are about, and this file is the record you can compare each of them with. Kyoto is the useful first one because it has a clear start and end, and the series shows what happened either side without claiming why.
The tensionThe split between Annex I countries and everybody else. The United States never ratified it and Canada withdrew in 2011, which is a documented disagreement about who owes what, with dates.
SourceThe IPCC, asked by governments at Paris to assess what 1.5 °C of warming would mean and what reaching it would take. The number is now the headline of climate politics.
Link to this fileThis dataset measures that number directly, which makes it the rare case where the strategy and the measurement use the same units. Be careful about the baseline: the limit is measured against 1850 to 1900 and this file is anomalies against 1951 to 1980, so the two do not share the same zero. Mixing them up is the commonest error a student can make here.
The tensionA scientific assessment that became a political target. What counts as exceeding 1.5 °C, a single year or a twenty-year average, decides whether it has already happened.
SourceNearly every country, agreeing to hold warming well below 2 °C and to pursue efforts towards 1.5 °C, with each setting its own contribution.
Link to this fileUse it for the gap rather than the effect: what this series says about where the number stands against the goal is answerable, and whether the agreement moved it is not.
The tensionA collective target with no enforcement, so the argument is about ambition rather than compliance, and every country's answer depends on what it thinks the others will do.
SourceMean sea level, tide gauge by tide gauge
How far the sea has risen against the land at places people actually live, measured by instruments in harbours rather than from orbit.
- One row:
- One station in one month: mean sea level in millimetres above that station's own Revised Local Reference datum, which sits about 7,000 mm below the local mean.
- Covers:
- 1,620 stations worldwide. Station 1 is Brest and it carries 2,622 monthly values from 1807 to 2025, which is the count behind this card.
Full record6 traps
- Usable rows
- 2,310
- Coverage
- 1,620 stations worldwide. Station 1 is Brest and it carries 2,622 monthly values from 1807 to 2025, which is the count behind this card.
- How it arrives
- One plain text file per station, semicolon-separated, no header row at all
- Language
- Page and file in English
- 1Missing months are −99999. Brest has 312 of them in 2,622 rows, 12% of the record, and there is no header to warn you. Averaging the column as it downloads produces a number thousands of millimetres wrong, in the wrong direction, and it looks like a plausible sea level.
- 2There is no header row. Four columns separated by semicolons, and the only way to know what they are is the documentation page. A reader who guesses will guess wrong about the fourth.
- 3The date is a decimal year. 1807.0417 is mid-January 1807. Nothing will parse that as a date for you, and rounding it to 1807 puts twelve months on top of each other.
- 4The datum is local, and this is the trap that matters. Values are millimetres above a Revised Local Reference set roughly 7,000 mm below that station's own mean sea level. Brest reading 6,900 and another station reading 6,900 are not the same height above anything. Only the change within a single station means something; the difference between two stations means nothing at all.
- 5A tide gauge measures the sea against the land, and the land moves. Scandinavian stations show sea level falling because the land is still rising after the ice sheets left. That is not an error in the data, it is what the instrument measures, and any conclusion about the ocean has to say so.
- 6348 of Brest's values carry an attention flag in the fourth column. They are not wrong, they are marked, and saying whether you kept them is part of the method.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much has mean sea level risen at one named station, and has the rate changed over its record?
- Do two stations on the same coast agree, and if not, what could make them differ?
- How does the trend at a station in a subsiding city compare with one on stable ground?
- Mean sea level for the month, in millimetres
- The date, as a decimal year: 1807.0417 is mid-January 1807
- A missing-data indicator, and a flag marking values the compilers want you to look at
- The station list adds latitude, longitude, name and country for all 1,620
The Netherlands, through a Delta Commissioner with a statutory budget. It sets flood protection standards for every stretch of defence in the country and funds the work to meet them, on a rolling programme rather than a project.
Link to this fileThe clearest pairing on this card, because the Dutch gauges are in this database and the programme is explicitly designed against the number they produce. Its planning scenarios are stated in centimetres of sea level rise, so a student can hold a station's measured trend against the range the country is budgeting for.
The tensionMoney and time against a number nobody can predict exactly. Protecting to a higher scenario costs billions now for a rise that may arrive later or sooner, and the same land is wanted for housing and farming in the meantime.
SourceThe Environment Agency, managing tidal flood risk for London: the Thames Barrier, the walls behind it, and a decision about when to replace them.
Link to this fileA single-city plan against a single gauge, and Sheerness is station 3 in this database. The plan's whole design is adaptive: it names the sea level triggers at which each next step is taken, so the measured series is the thing that decides when money is spent.
The tensionOne of the most valuable floodplains in the world sits behind a barrier that was built for a 1980s estimate. Deciding when to commit to a new one is a choice between spending early and being caught late, argued in public.
SourceThe Italian state, through a consortium, building mobile gates across the three inlets of the Venice lagoon and raising them when a high tide is forecast.
Link to this fileThe case where the strategy and the gauge argue with each other. Venice's relative sea level is rising partly because the city is sinking, which a tide gauge cannot separate from the sea rising, and the barrier is designed against the combined number.
The tensionDecades late, billions over budget, and interrupted by a corruption prosecution. Each closure also cuts the lagoon off from the sea, which is how the lagoon flushes, so protecting the city has a cost paid by the water it sits in.
SourceSea surface temperature over the Great Barrier Reef, every year since 1900
How much warmer the sea over the Great Barrier Reef has become: the heat behind the mass coral bleaching that has struck the reef again and again since 1998.
- One row:
- One year over the Great Barrier Reef: how far its mean sea surface temperature sat above or below the 1961 to 1990 average, across seven grid points in and beside the Marine Park.
- Covers:
- 126 complete years, 1900 to 2025, none missing. The same page gives seven other Australian ocean regions in the same format
Full record8 traps
- Usable rows
- 126
- Coverage
- 126 complete years, 1900 to 2025, none missing. The same page gives seven other Australian ocean regions in the same format
- How it arrives
- A plain text file, two columns separated by spaces, no headings. The page's Download > Data link gives the same series
- Language
- Page and file in English
- 1The file is anomalies, not temperatures. Each value is how far that year sat above or below the 1961 to 1990 average. For the mean annual temperature in °C, add the average printed beside the Download link on the page: 25.8 °C. For a trend or a correlation you do not need to: adding the same number to every year moves the line without changing its slope.
- 2Two averages, and they do not quite agree. Tick 91-20 on the page and every anomaly drops by exactly 0.25 °C, but the printed average rises from 25.8 to 26.0 °C, only 0.2. Both averages are rounded to one decimal place, so the two routes to °C give years 0.05 °C apart. Choose one, say which, and never mix them.
- 3No headings, and the first column is two dates run together. `199301199312` is January 1993 to December 1993, so the year is the first four characters: in Excel, =LEFT(A1,4). The columns are separated by spaces: use Data › Text to Columns, Delimited, Space, and tick "Treat consecutive delimiters as one".
- 4Seven points, 220 km apart. The series averages the seven 2° grid points of NOAA's ERSST v5 that lie in or near the Marine Park. It describes the reef region, not any one reef, and NOAA says ERSST suits long-term, basin-wide studies best. For a single reef, use Coral Reef Watch.
- 5Ships, buoys and floats, and fewer of them early on. ERSST is built from measurements at sea and filled in statistically where there are none, so the early 1900s rest on far fewer observations than recent decades.
- 6The numbers can change. The Bureau recalculates the series whenever NOAA updates ERSST: in March 2020 every value from 2008 onwards changed when missing buoy reports were restored. Record the date you downloaded and keep your copy.
- 7Each year is one point, and it trends. Almost anything that has changed steadily since 1993 will correlate with it. Correlate the change from one year to the next as well as the values.
- 8Old links to the graph are dead. Links of the form bom.gov.au/climate/change/#tabs=Tracker now open a general climate change page without the graph. Use the page this card links to.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has the sea over the Great Barrier Reef warmed faster or slower than the globe between 1993 and 2023?
- Has the Great Barrier Reef warmed as fast as the Coral Sea as a whole between 1993 and 2023?
- Has the tropical Great Barrier Reef warmed faster or slower than the temperate Tasman Sea between 1993 and 2023?
- First column: the twelve months the value covers, start and end run together, so `199301199312` is January to December 1993
- Second column: the sea surface temperature anomaly in °C against 1961 to 1990, to two decimal places
- On the page, beside Download: the 1961 to 1990 average, 25.8 °C, which turns an anomaly into a temperature
- Change Region on the page for the Coral Sea, the Tasman Sea and five other Australian regions
The Australian and Queensland governments jointly, as the overarching plan for protecting the reef as a World Heritage property.
Link to this fileThe plan says that "climate change is the single, biggest threat to the Reef", and then acts mostly on the pressures Australia controls: run-off from farms, crown-of-thorns starfish, fishing and coastal development. None of those changes the temperature of the sea, so do not look for the plan in this line. What this file measures is the pressure the plan is trying to make the reef resilient against.
The tensionIn 2021 UNESCO's World Heritage Centre and IUCN recommended listing the reef as World Heritage in Danger. The Australian Government lobbied against it and the Committee did not list it, and the question has come back every year since. In July 2026 the Committee again declined, while warning that the reef's "capacity to tolerate and recover from such events is increasingly compromised" after the mass bleaching of 2024 and 2025.
SourceThe Australian Parliament, with the Climate Change Authority advising the government on targets and on progress towards them.
Link to this fileIndirect, and say so. The water over the reef warms with the whole world's emissions, of which Australia's are a small share, so the Act cannot show in this file. Use it to explain what Australia is doing about the cause, and a global series such as GISTEMP to set the reef's warming beside the world's.
The tensionAustralia is one of the world's largest exporters of coal and gas, and in September 2025 the same government gave final approval to extend Woodside's North West Shelf gas project to 2070. Cutting emissions at home while selling fossil fuels abroad is the argument this strategy cannot avoid.
SourceFunded by the partnership between the Australian Government's Reef Trust and the Great Barrier Reef Foundation, and run by AIMS, CSIRO, Southern Cross University and other universities.
Link to this fileThe only strategy here that aims at the temperature itself: spraying fine sea salt into low clouds to brighten them and shade the reef during a marine heatwave. It is still an experiment, over areas far smaller than one of this file's 220 km grid squares, so it cannot show in this line, and explaining why is part of evaluating it.
The tensionFriends of the Earth Australia called schemes like it "a convenient smokescreen for the fossil fuel industry" in 2021, and scientists questioned field trials going ahead before much research had been published. The Great Barrier Reef Foundation's own explainer says emissions cuts are essential and that the technology does not replace them.
SourceEurope's land temperature, month by month since 1850
How fast Europe is warming compared with the planet as a whole, which decides how much heat a European strategy has to plan for.
- One row:
- One month over the land area of Europe: how far its average temperature was above or below that month's 1901 to 2000 average, in °C.
- Covers:
- 2,120 monthly values, January 1850 to August 2026. The same tool gives the globe, each hemisphere, the Arctic and the other continents in the same format
Full record7 traps
- Usable rows
- 2,120
- Coverage
- 2,120 monthly values, January 1850 to August 2026. The same tool gives the globe, each hemisphere, the Arctic and the other continents in the same format
- How it arrives
- One CSV of about 26 KB, with three comment lines above the header
- Language
- Page and file in English
- 1Three comment lines before the header. Title, units and base period come first, each starting with #, so a spreadsheet that assumes row one is the header reads the title as a column name.
- 2The date is one number. 185001 is January 1850, and plotted as a number the series jumps 89 units between every December and January. Split it into year and month (divide by 100) before you draw anything.
- 3The warmest month is a February. February 2024 (+3.95) tops the file, and the top eight are all winter months. An anomaly measures how far a month was from its own normal, and European winters swing further (standard deviation 1.31 against 0.75 for summers). "Hottest" and "most unusual" are different questions; say which you are asking.
- 4Monthly values jump around; yearly ones do not. Since 1990 Europe's monthly values vary about three times as much as the global series. Take annual or seasonal means before you fit a trend, or your trend will just show the weather.
- 5Three baselines, three zeros. This file uses 1901 to 2000. NASA's GISTEMP uses 1951 to 1980, and the Paris limit uses 1850 to 1900. On this file Europe's 1850 to 1900 average is −0.34, so warming since pre-industrial is the anomaly plus 0.34.
- 6The end year is part of the web address. The download URL ends 1850-2026. Ask for a year that has not started and you get an HTML error page saved with a .csv name, not a shorter file. In 2027, edit the year in the URL.
- 72026 has eight months, January to August. An annual mean for 2026 is a mean of eight months, and it will not be the same number once winter is in.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much faster has Europe's land warmed since 1975 than the globe as a whole?
- Have European summers or winters warmed faster since 1975, and which season varies more from year to year?
- How many of Europe's ten warmest complete years came after 2010?
- The month, written as one six-digit number: 185001 is January 1850
- The departure from the 1901 to 2000 average for that calendar month, in °C: from −4.60 to +3.95
- Nothing else, which is what makes it quick to pair with the global series from the same page
The European Commission, with member states writing their own national adaptation plans. The 2021 European Climate Law makes adaptation a legal duty.
Link to this fileAdaptation responds to warming and does not reduce it, so do not look for the strategy in the line. What this file can do is size the problem: Europe's land has warmed at 0.41 °C a decade since 1975, twice the global rate in NOAA's own global series (0.20). Whether the strategy's pace matches that is a question you can argue from the data.
The tensionAdaptation that makes things worse. The European Court of Auditors' 2024 review (special report 15/2024) found EU-funded measures that were maladaptive: irrigating more land instead of switching crops, and snow cannons, which about 70% of Austria's ski slopes are now equipped with.
SourceThe French health ministry, with Santé publique France running the heat-health warning system and Météo-France issuing the alerts.
Link to this filePair these carefully, because that care is the point. A heatwave is days in one place, and this file is a month across a continent: August 2003 reads +2.00, only the 65th warmest month in the record. The series cannot show the event the plan was built for. It can show that European summers are warming, 0.48 °C a decade since 1975, which is the background every heatwave now arrives on.
SourceNearly every country, agreeing to hold global warming well below 2 °C and to pursue 1.5 °C, each setting its own contribution.
Link to this fileThe limit is a global average measured against 1850 to 1900. On this file Europe's land was 2.6 °C above its own 1850 to 1900 average in 2024. The global series from the same tool was 1.4 °C above. A continent can pass any regional threshold long before the world passes the global one, and that gap is a finding in itself.
SourceMonthly temperature at Geneva since 1753 and Sion since 1864
How the climate has changed at two Swiss places, month by month: Geneva, beside the lake, and Sion, in the Rhône valley in the Alps. This is the heat that Switzerland's plans for adapting to climate change prepare for.
- One row:
- One station in one month: the mean air temperature 2 m above the ground, in °C, corrected by MeteoSwiss to today's site and instruments, with 29 more columns about the same month.
- Covers:
- Two stations: Geneva, January 1753 to August 2026, and Sion, January 1864 to August 2026
Full record10 traps
- Usable rows
- 5,236
- Coverage
- Two stations: Geneva, January 1753 to August 2026, and Sion, January 1864 to August 2026
- How it arrives
- One CSV per station, one row per month, separated by semicolons: ogd-nbcn_gve_m.csv for Geneva (about 280 KB) and ogd-nbcn_sio_m.csv for Sion (about 190 KB). The same files exist for all 29 of MeteoSwiss's climate stations, and its open data portal has every other station
- Language
- Page and file in English
- 1Semicolons, not commas. Excel in English puts each whole row into column A. Use Data › From Text/CSV and choose Semicolon as the delimiter. In Google Sheets use File › Import and choose a custom separator of ;.
- 2The month is text, and Excel cannot hold a date before 1900. reference_timestamp reads 01.02.1753 00:00. Take the year with =VALUE(MID(B2,7,4)) and the month with =VALUE(MID(B2,4,2)), and do not let the spreadsheet turn it into a date.
- 32026 stops at August. On 27 September 2026 the last row was August 2026. February to March and June to August 2026 are complete, but the average of all the 2026 rows is not a year. For whole years, stop at 2025, or use the yearly file, which stops there.
- 4Every column is corrected (homogenised). MeteoSwiss adjusts old values to today's site and instruments. Its uncorrected monthly means for the same stations are a different download (ogd-smn, column tre200m0). They differ from these by up to 0.8 °C in a month at Geneva and 1.0 °C at Sion. Never join the two into one series.
- 5Both thermometers moved. Geneva's stood at the Observatoire in the city centre until 1962, then moved 5.5 km to the airport at Cointrin. Sion's stood in the Capuchin monastery in the town from 1873 to 1977, and moved 2.5 km to the airfield in 1978. The corrections allow for this, but they are estimates. Name the moves as a limitation.
- 6The airport is not the city. Cointrin stands in open fields on the edge of Geneva, so it does not measure the extra heat of the city centre.
- 7Geneva before 1768 is not all Geneva. MeteoSwiss's series fills long gaps in the early years with measurements from Neuchâtel. Researchers who checked the old records again in 2022 left out everything before 1768, because it was not good enough. Start at 1768 or later, or say why you did not.
- 8Use the monthly file, not the daily one. MeteoSwiss corrects the months, and it works out the daily values from them with more uncertainty. Geneva's daily values only start in December 1864.
- 9Each year is one point, and it trends. Almost anything that has changed steadily since 1864 will correlate with temperature. Correlate the change from one year to the next as well as the values.
- 10The numbers can change, and old links are dead. The series is matched to today's station, so MeteoSwiss changes past values when a station changes again. Write down the date you downloaded and keep your copy. Older guides link to homog_mo_GVE.txt on MeteoSwiss's website, which now gives an error: use the links on this card.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has Geneva's February to March mean temperature changed more than its June to August mean temperature, between 1864 to 1893 and 1996 to 2025?
- Has the June to August mean temperature changed more at Sion, in the Alps, than at Geneva, between 1864 to 1893 and 1996 to 2025?
- Since 1864, has the mean of Geneva's daily minimum temperatures changed more than the mean of its daily maximum temperatures?
- ths200m0: the month's mean air temperature 2 m above the ground, homogenised, in °C to one decimal place. The column to use
- reference_timestamp: the month, written 01.02.1753 00:00 for February 1753. The value is for the whole month
- station_abbr: GVE for Geneva, SIO for Sion
- th9120mv: how far the month was above or below its 1991 to 2020 average, in °C
- ths2dymn and ths2dymx: the month's mean daily minimum and mean daily maximum. Geneva has them from 1864, Sion from 1959 and 1958
- rhs150m0: the month's precipitation in mm, at both stations from 1864
- Every other column (sunshine, frost days, wind and more) is named in the parameter list. Most start in 1959 or later. A blank means it was not measured then, never zero
The two station files (2)
What the columns mean, and the other stations (2)
The other 27 climate stations have the same files: change the three letters in both places in the link (bas for Basel, sma for Zürich). The station list gives every code. It is saved in the Windows-1252 encoding, so a name such as Genève can show a strange character if it is opened as UTF-8.
- What each column meansParameters
- Every station, with its height and coordinatesStations
The Swiss Federal Council, with the Federal Office for the Environment (FOEN) leading the committee of federal offices that carries it out. Since 2025 the Climate and Innovation Act also requires the Confederation and the cantons to take adaptation measures.
Link to this fileAdaptation responds to warming and does not reduce it, so do not look for the strategy in the line. It is built around heat in towns, summer drought and natural hazards. What these files can do is measure the change it plans for, in spring and in summer, at two real places.
SourceThe Swiss Confederation. The Act binds Switzerland to net zero greenhouse gas emissions by 2050. Parliament wrote it as a counter-proposal to the Glacier Initiative.
Link to this fileIndirect, and say so. The temperature at Geneva and Sion follows the air over the whole region, which Swiss emissions hardly change. Nothing in these files can show whether the Act works. Use it to explain what Switzerland is doing about the cause.
The tensionCost against climate, and it is on the public record. In 2021 the CO2 Act's critics, swissinfo reported, "viewed the additional taxes as too expensive and unfair as part of the population has no choice but to use a car". In 2023 the Swiss People's Party called the new law an "electricity guzzler". The two places in this file voted differently: in 2021, 61.4% of voters in the canton of Geneva said yes, and 39.1% in Valais.
SourceThe Canton of Geneva's Conseil d'État.
Link to this fileIndirect, and only for the Geneva file. Its targets are about emissions, which cannot show in one station's temperature. Its measure 4.5 is against urban heat islands, but the station stands in open fields at the airport, so it does not record the city's extra heat. State the connection and do not claim the data measures the plan.
The tensionThe canton's own advisers doubt it. The Conseil pour le climat, which the canton set up in October 2021, reported in January 2024 that the plan's targets for 2030 and 2050, and its targets for adapting to climate change, were "inatteignables en l'état": unattainable as things stand.
SourceKyoto's cherry blossom, peak bloom since the year 812
Spring coming earlier in Japan. The day Kyoto's cherry trees reach full bloom has been recorded, on and off, for twelve centuries.
- One row:
- One year: the day of the year on which Kyoto's cherry trees reached full bloom.
- Covers:
- 838 years with a date, from 812 to 2026. Every year since 1953 is present
Full record7 traps
- Usable rows
- 838
- Coverage
- 838 years with a date, from 812 to 2026. Every year since 1953 is present
- How it arrives
- One CSV, about 30 KB, from the chart's download button
- Language
- Page and file in English
- 1Most early years are missing. Only about one year in four has a date before 1100, and one in two before 1400. After 1400 the record is nearly complete. A trend drawn across all twelve centuries rests mostly on the later ones: choose a period where the gaps are small, and say how many years it has.
- 2The early dates were not measured. They come from court diaries, chronicles and records of blossom-viewing parties, read by historians. The modern dates are observations. The way the date was recorded changes along the series, which is a limitation to name.
- 3Thirty-year average is not your data. It is a smoothed column Our World in Data calculated. Analyse the yearly column and do your own averaging.
- 4One city, and a warm one. Kyoto has grown since the Middle Ages, and Japan's heat island policy records six large cities warming 2 to 3 °C in the twentieth century, against 0.6 °C for the world. Part of the recent trend is the city, not only the climate.
- 5This is the wild mountain cherry, Prunus jamasakura. Japan's weather agency watches a different cherry, Somei-yoshino, so its Kyoto dates will not match these.
- 6Leap years shift the day of the year by one after 28 February. Across centuries that is noise, but do not read a one-day change as real.
- 7The record is 2023, at day 84 (25 March), the earliest in over 1,200 years. With this many years, any trend will be significant: report how many days earlier, not only the p-value.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much earlier is Kyoto's peak cherry blossom in 1990 to 2026 than in 1850 to 1889?
- Did the day of peak cherry blossom in Kyoto change between 1400 and 1850, before industrial warming?
- Has peak bloom in Kyoto become more or less variable from year to year since 1950 than it was between 1850 and 1949?
- Year
- Day of the year with peak cherry blossom: already a day of the year, so 95 is 5 April in most years
- Thirty-year average: Our World in Data's own moving average. Not data: do not analyse it as if it were
The Government of Japan, through the Cabinet and the Ministry of the Environment.
Link to this fileIndirect. The targets address the warming that moves flowering earlier, but no flowering date can show whether a target is working. State the connection; do not claim the data measures the plan.
The tensionMore than 3,000 public comments were received and more than 80% asked for higher targets; environmental groups and Climate Action Tracker called the 2035 goal far too weak. The government kept it, and InfluenceMap reports that policy follows the Japan Business Federation (Keidanren), whose position was the least ambitious of the industry groups.
SourceA committee of national ministries, set up in 2002, with local governments, businesses and residents expected to act on it.
Link to this fileIndirect for this card. It addresses the city part of the warming, and Kyoto is a large city, but one city's flowering date cannot separate urban heat from climate. The JMA network can: it has small towns as well as cities.
SourceUK mean temperature, every month and year since 1884
How much the UK has warmed, in which seasons and in which of its four nations, which is the heat that the UK's adaptation plans have to prepare for.
- One row:
- One year for the whole of the UK: the mean air temperature of each month, each season and the year, averaged over a 1 km grid built from the weather stations.
- Covers:
- 142 complete years, 1884 to 2025, plus 2026 so far; the same for England, Scotland, Wales and Northern Ireland
Full record8 traps
- Usable rows
- 142
- Coverage
- 142 complete years, 1884 to 2025, plus 2026 so far; the same for England, Scotland, Wales and Northern Ireland
- How it arrives
- A plain text file, columns separated by spaces, with five lines of notes above the headings. England, Scotland, Wales and Northern_Ireland are the same file with the name changed in the link
- Language
- Page and file in English
- 1Five lines of notes come first. The headings are on line 6. Delete the notes, or start the import at line 6.
- 2It is not a CSV. The columns are separated by runs of spaces. In Excel use Data › Text to Columns, Delimited, Space, and tick "Treat consecutive delimiters as one". In Google Sheets, paste it into column A and use =SPLIT(A1, " ") beside it.
- 3The last row is this year, and it is not finished. 2026 stops in August and has no annual value. Stop at 2025.
- 4Winter starts in the year before. win is December of the previous year with January and February, so the 1884 winter is --- (missing) and 2025's winter includes December 2024. Say which winter you mean.
- 5Use the ann column, not your own average. The months are rounded to one decimal place and ann is not, so averaging the twelve months gives a slightly different number.
- 6It is not Central England Temperature. This series covers the whole UK, Scottish hills included, so it runs about 1 °C cooler than CET in every year. Never mix the two in one series; if you compare them, compare their change.
- 7The numbers can change. The Met Office regenerates HadUK-Grid every year, adding newly digitised records and improving the method, and the latest months are provisional until then. Record the "Last updated" date and keep your downloaded copy.
- 8Each year is one point, and it trends. Almost anything that has changed steadily since 1976 will correlate with temperature. Correlate the change from one year to the next as well as the values.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has the UK's winter mean temperature risen more than its summer mean temperature between 1976 and 2022?
- Has Scotland's annual mean temperature risen as much as England's between 1976 and 2022?
- Is the UK's annual mean temperature related to a butterfly species' UK index across the years 1976 to 2022?
- year
- jan to dec: each month's mean temperature in °C, to one decimal place
- win, spr, sum, aut: winter (December to February), spring, summer and autumn, to two decimal places
- ann: the year's mean temperature in °C
The UK Government, through Defra, for England and the UK-wide matters it controls. Scotland, Wales and Northern Ireland write their own adaptation plans.
Link to this fileAdaptation responds to warming and does not reduce it, so do not look for the programme in the line. What this file can do is size the problem it plans for: how much warmer UK years, summers and winters have become, and whether one nation or season has changed more than another.
The tensionThe Climate Change Committee, the government's own statutory adviser, reported on 30 April 2025 that "NAP3 has been ineffective in driving a shift towards adaptation delivery" and that it did "not find a single outcome with evidence of 'good' delivery on adaptation".
SourceThe UK Parliament, with the Climate Change Committee advising on five-yearly carbon budgets and reporting on progress.
Link to this fileIndirect, and say so. The UK's temperature is set by the world's emissions, of which the UK's are a small share, so the Act cannot show in this file. Use it to explain what the UK is doing about the cause, and a global series such as GISTEMP to set the UK's warming beside the world's.
The tensionIn October 2025 the Conservative leader, Kemi Badenoch, pledged to repeal the Act, ending a consensus the main parties had held since 2008. Former Prime Minister Theresa May, whose government set the net zero target, called it a retrograde step.
SourceCentral England Temperature, every season and year since 1659
How England's climate has changed, in what the Met Office calls the longest temperature record in the world. It is the change that England's adaptation plan prepares for, and an English temperature to check a wildlife series against.
- One row:
- One year in central England: the mean air temperature of each season and of the whole year, from a few weather stations between Lancashire, London and Bristol.
- Covers:
- 367 complete years, 1659 to 2025, plus 2026 so far. Daily values from 1 January 1772
Full record10 traps
- Usable rows
- 367
- Coverage
- 367 complete years, 1659 to 2025, plus 2026 so far. Daily values from 1 January 1772
- How it arrives
- A plain text file, columns separated by spaces, with nine lines of notes above the headings. The monthly file and the daily file are on the same page, in the same format
- Language
- Page and file in English
- 1Sum means summer. Nothing is added up. The headings Win, Spr, Sum and Aut are winter, spring, summer and autumn. The file name says "totals", but every value is a mean temperature in °C.
- 2Nine lines of notes come first, and it is not a CSV. In the seasonal file the headings are on line 10; in the monthly file, on line 5. The columns are separated by runs of spaces, and every row starts with two spaces. In Excel use Data › Text to Columns, Delimited, Space, and tick "Treat consecutive delimiters as one"; if the years land in column B, delete the empty column A. In Google Sheets, paste it into column A and use =SPLIT(A1, " ") beside it.
- 3-99.90 means missing, and the last row is not finished. It appears in three cells: the 1659 winter, and the 2026 autumn and year. 2026 has a winter, spring and summer, but they are provisional until the Met Office's next yearly version, usually in February or March. For whole years, stop at 2025.
- 4Winter starts in the year before. The file says "Winter: year refers to January", so Win for 2026 is December 2025 with January and February 2026. That is why the 1659 winter is missing. Say which winter you mean.
- 5Use the Annual column, not your own average. Annual is January to December, worked out from more exact values than the file shows. In 1947 the twelve months average 9.58 °C, but Annual is 9.65. The four seasons are worse: Win holds the December before, so their average can be 0.6 °C away from Annual.
- 6The early years are rounded. The first years come from Gordon Manley's work on old records. Every month from 1659 to 1670 is a whole degree, and from 1671 to 1698 a whole or half degree. The two decimals in the seasonal file do not make those years more exact. The daily file starts in 1772.
- 7It is central England, not England or the UK. It is the average of a few stations in lowland England: since 2006, Rothamsted, Pershore College and Stonyhurst. In every year from 1884 to 2025 it is 0.8 to 1.4 °C warmer than the Met Office's UK series, and 0.2 to 0.6 °C warmer than its England series. Never join two of them into one series; if you compare them, compare their change.
- 8The warmth of towns has been taken out. Since 1974 the Met Office lowers the mean temperatures to allow for urban warming, by 0.2 °C at present. So this file cannot show what towns and building do to temperature.
- 9The numbers have been revised. Version 2 (May 2022) changed the stations used for 1878 to 1930 and for 2004 to 2005. Version 2.1 (July 2024) corrected an error that made the series slightly too cool: about one monthly value in eight changed, by up to 0.2 °C, most of them upwards and after 1973. The files carry no date or version, and the page's own text can be out of date. Write down the version on the download page (v2.1.1.0 when checked) and the date you downloaded, and keep your copy.
- 10Each year is one point, and it trends. Almost anything that has changed steadily over the years will correlate with temperature. Correlate the change from one year to the next as well as the values.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has central England's winter mean temperature changed more than its summer mean temperature between 1901 to 1930 and 1996 to 2025?
- Has the number of days a year with a mean temperature above 20 °C in central England changed between 1772 to 1801 and 1996 to 2025?
- Is a butterfly species' England index related to the mean temperature of the summer before, across the years 1976 to 2023?
- Year
- Win, Spr, Sum, Aut: the mean temperature of winter (December to February), spring (March to May), summer (June to August) and autumn (September to November), in °C to two decimal places
- Annual: the mean temperature of the calendar year, January to December, in °C
- Jan to Dec, in the monthly file: each month's mean temperature, to one decimal place
- Date and Value, in the daily file: the mean temperature of every day since 1 January 1772
The UK Government, coordinated by Defra, with actions from every government department. It covers England, and the matters the UK Government controls for the whole UK. Scotland, Wales and Northern Ireland write their own adaptation plans.
Link to this fileAdaptation responds to warming and does not reduce it, so do not expect to see the programme in these temperatures. Central England is inside the area the programme covers. So this file can measure, for England itself, the change the programme has to prepare for: in summers, in winters, in whole years, and in hot days from the daily file.
The tensionThe Climate Change Committee, the government's own statutory adviser, reported on 30 April 2025 that "NAP3 has been ineffective in driving a shift towards adaptation delivery". It wrote: "We do not find a single outcome with evidence of 'good' delivery on adaptation."
SourceThe UK Parliament, with the Climate Change Committee advising on five-yearly carbon budgets and reporting on progress.
Link to this fileIndirect, and say so. Central England's temperature is set by the world's emissions, and the UK's are a small share of them. So the Act cannot show in this file. Use it to explain what the UK is doing about the cause. A global series such as GISTEMP lets you set England's warming beside the world's. The same Act is the law that requires the adaptation programme.
The tensionIn October 2025 the Conservative leader, Kemi Badenoch, pledged to repeal the Act. Theresa May, whose government set the net zero target, said: "I am deeply disappointed by this retrograde step which upends 17 years of consensus between our main political parties and the scientific community."
SourceSea surface temperature at any point on the ocean, month by month since 1981
How warm the sea is, month by month, at the place your question is about: the partner file for anything living in the sea that follows the temperature, from lionfish to sharks.
- One row:
- One month at one quarter-degree point on the sea: the mean sea surface temperature in °C.
- Covers:
- Every month from September 1981 to August 2026 at any sea point on Earth, on a quarter-degree grid: 540 months for each point you ask for
Full record5 traps
- Usable rows
- 540
- Coverage
- Every month from September 1981 to August 2026 at any sea point on Earth, on a quarter-degree grid: 540 months for each point you ask for
- How it arrives
- One CSV per point, from a link you edit: change the latitude and longitude in it to your own sea point
- Language
- Page and file in English
- 1One point at a time. Each link returns one point. Copy it and change latitude and longitude for every point you need, and keep a list of the points and why you chose them: where you take the temperature is part of your method.
- 2Your point moves. NOAA uses the nearest quarter-degree cell, so asking for 46.2°N, 6.15°E returns 46.125°N, 6.125°E. Report the point in the file, not the one you typed.
- 3Land and lakes return NaN. A point on land gives NaN for every month, and so does Lake Geneva: this file is the sea only. Move the point a little out to sea.
- 4Choose the point and the month before you look. Picking the point or the month that gives the best result after seeing them all is cherry-picking. Say which you chose and why.
- 5It is the surface. A shark or a fish may live deeper, where the water is cooler. Say that your variable is surface temperature.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much warmer was February, the coldest month, south of Cyprus in 2016 to 2025 than in 1982 to 1991?
- Does the sea off Perth vary less through the year than the sea off Sydney, 1982 to 2025?
- time: the first day of the month the value is for
- latitude, longitude: the grid point NOAA actually used, which is not always the one you asked for
- sst: the monthly mean sea surface temperature in °C
Nearly every country, agreeing to hold warming well below 2 °C and to pursue efforts towards 1.5 °C, with each setting its own contribution.
Link to this fileIndirect: a global goal, and this is one point of sea. On its own this file supports a question about how the sea has changed; for the strategy, use the one on the file you pair it with (lionfish removal, shark nets), which is where the argument is.
The tensionA collective target with no enforcement, so the argument is about ambition rather than compliance, and every country's answer depends on what it thinks the others will do.
SourceIce and snow2
Swiss glacier length change, 1847 onwards
How far the tongues of Alpine glaciers have moved, year by year, across the whole period in which anybody has been measuring them.
- One row:
- One glacier over one observation interval: how far its tongue moved between two survey dates, in metres, with a minus sign for retreat.
- Covers:
- 156 glaciers, 1847 to 2023, and 10,861 observations
Full record7 traps
- Usable rows
- 10,861
- Coverage
- 156 glaciers, 1847 to 2023, and 10,861 observations
- How it arrives
- One CSV, about 725 KB, with six lines of citation and then three header rows
- Language
- Page and file in English
- 1Six lines of citation, then three header rows. The column names are on one row, a row of variable codes is under them, and a row of units under that. Import it as it is and your first two observations are the words `dL` and `m`.
- 2One row is an interval, not a year. 9,668 of them run about a year, 778 run one to two years, and 400 run longer. A five-year gap posts one large negative number, and plotted against the end date it draws five years of retreat as a single year's collapse.
- 378% of start dates and 77% of end dates are flagged as not exactly known, with an `x` in the column beside them. That is not a reason to discard them; it is a reason not to calculate rates to the nearest day.
- 418% of all observations are advances, not retreats. They are real and they cluster: 41% of the observations in the 1970s and 36% in the 1980s are positive, against 4% in the 2000s. A student who assumes retreat and filters out the positives has deleted the most interesting thing in the file.
- 5The tongue elevation column is empty in 84% of rows, so any question that needs altitude loses most of the dataset before it starts.
- 6The largest single value is −2,600 m, at Unterer Grindelwaldgletscher between September 2010 and September 2011. Check an extreme like that before deleting it: this one is real, and it is the story rather than the error.
- 7Coverage is very uneven. Rhonegletscher and Glacier du Trient carry 143 observations each; 15 of the 156 glaciers have fewer than ten. Pick your glacier for its record, not its name.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the length of one named glacier changed since records began, and when did it last advance?
- In which decades did the largest share of Swiss glaciers advance rather than retreat?
- Do two glaciers with long records, such as Rhonegletscher and Glacier du Trient, retreat at the same rate?
- Length change in metres, negative for retreat
- Glacier name and its Swiss Glacier Inventory id
- Start and end date of the interval, each with a flag saying whether the date is exactly known
- Elevation of the glacier tongue, where anyone recorded it
- The observer
The family that runs the ice grotto at the Furka pass, who lay white fleece over part of the tongue each summer. An ice grotto has been cut into that glacier every year since 1870 and it is a business as well as an attraction.
Link to this fileThe tightest fit in this file, because the glacier it acts on is the one with the longest record in it: Rhonegletscher carries 143 observations. The covers reduce melt beneath them by roughly half to two thirds, and seven years of them preserved about 35 metres of ice thickness, which is a real effect on a small patch of a glacier that kept retreating.
The tensionIt works, but it cannot be done on a large scale. Covering every Swiss glacier would cost over a billion francs a year and would still only slow the decline, so a treatment that protects one tourist attraction is not a policy for a country. That is an economic goal and an environmental one pulling in different directions on the same ice.
SourceThe Swiss Confederation. It binds the country to net zero greenhouse gas emissions by 2050, with emissions halved by 2030, and funds the replacement of heating systems and industrial processes.
Link to this fileBe honest about this one. Swiss glaciers respond to the temperature of the atmosphere above them, which Swiss emissions hardly affect, so nothing in this file can test whether the Act worked and nothing in it ever will. What the file can do is measure the thing the law was passed about, which is a different and defensible claim.
The tensionIt went to a popular vote, so the disagreement is on the public record rather than inferred: the cost of replacing heating and the pace of the change against the target.
SourceA citizens' initiative, launched by an association rather than a government, demanding that net zero be written into the Swiss constitution. It was withdrawn once parliament produced the counter-proposal that became the Act above.
Link to this fileA campaign named after the thing this dataset measures, which makes it the one strategy here you can hold directly against the numbers: the initiative said the glaciers were going, and 10,861 observations are the record of whether they were.
The tensionA constitutional demand against a parliamentary compromise. The initiative asked for a ban on fossil fuels by 2050; what passed funds a transition instead, and whether that counts as winning is exactly the argument.
SourceSnow depth and snowfall at Alpine stations, 1961 onwards
Whether it snows as often as it used to in the Alps, and what that does to a valley whose economy was built on the assumption that it would.
- One row:
- One station on one day: depth of snowfall in centimetres, and snow depth in centimetres, each in a raw version and a quality-checked one.
- Covers:
- Over 2,000 stations across Austria, France, Germany, Italy, Slovenia and Switzerland, 1961 to 2020. The French file counted here holds 279 stations and 2,979,635 daily rows.
Full record6 traps
- Usable rows
- 1,449,284
- Coverage
- Over 2,000 stations across Austria, France, Germany, Italy, Slovenia and Switzerland, 1961 to 2020. The French file counted here holds 279 stations and 2,979,635 daily rows.
- How it arrives
- One zip per country, each holding a single daily CSV; the French one is 11 MB
- Language
- Page and file in English
- 1The same measurement arrives three times. HN and HN_after_qc, HS and HS_after_qc, and then HS_after_gapfill. Analysing the raw column silently includes values the authors examined and rejected; analysing the gap-filled one includes 535,786 values in the French file that nobody measured. Pick one, and say in your method which and why.
- 2Half the file is empty, and that is normal. HN is present in 49% of rows and HS in 57%, because stations report in winter and not in July, and because a station that opened in 1990 has thirty blank years before it. Count your rows after filtering, never before.
- 3The set of stations changes over time. Pooling all French stations, one snow season rests on 15,334 station-days in 1985/86, 42,976 in 2005/06 and 30,157 in 2018/19. A percentage that moves because the network moved is the classic false finding in this file, and the fix is the same as everywhere: restrict to stations present throughout.
- 4Altitude is not in the daily file. It is in the metadata, and without joining it you are averaging a valley floor and a col together, which is the one comparison that matters most here.
- 5A day with no snowfall and a day with no observation both look like nothing if you are careless: zero is a measurement, blank is not. Days with snowfall means `HN_after_qc > 0` counted against days where `HN_after_qc` exists at all.
- 6It stops in 2020. The series was compiled for a paper, not maintained as a service, so the last five winters are not in it. If your question is about the present, this file cannot answer it and a national weather service can.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the number of days with measurable snowfall changed at French Alpine stations since the 1960s?
- Does the change differ between stations above and below 1,500 m?
- Do snow depth and snowfall tell the same story at one station, and if they diverge, when?
- HN, depth of snowfall in cm: the column that answers how often it snowed
- HS, snow depth in cm: how much was lying, which is a different question
- HN_after_qc and HS_after_qc, the same columns with failed values removed
- HS_after_gapfill, snow depth with missing days estimated
- Provider, station name and date
The Syndicat Mixte des Monts Jura, the public body that runs the resort at Lélex, Crozet and Mijoux, an hour from Geneva. The plan replaces lifts at one end of the domain and, at the other, designates certain runs piste nature non damée: no snowmaking, no grooming, left to return to nature.
Link to this fileThe best-fitting strategy on this card, because it is a plan that only makes sense if the snow is going, and this is the file that says whether it is. The Jura is not in these six countries' Alpine station list, so the honest move is to use the nearest French Alpine stations at comparable altitude and say that is what you have done.
The tensionThe same plan invests in lifts and withdraws from pistes at once, which is the argument on a small scale: a village economy of a few hundred people against a nature reserve whose objective is returning forest, with the same document funding both.
SourceThe French state, through the Agence Nationale de la Cohésion des Territoires, funding mountain areas to diversify away from a season that depends on snow: year-round tourism, renovating cold beds (holiday flats that stand empty most of the year), adapting resorts that are too low to be reliable.
Link to this fileA national response to exactly the trend this dataset measures, which makes the dataset the evidence base rather than the test: you are asking whether the premise of the policy holds at the altitudes it is aimed at.
The tensionMountain Wilderness France and others argue the money has gone on new lifts, snow cannons and hillside reservoirs, which locks in the ski model rather than moving away from it. The Cour des comptes reviewed mountain resorts and climate change in February 2024 and is the citable version of the same argument.
SourceResort operators and the communes that own them, building retenues collinaires, reservoirs high on the mountain that store water through the year to make snow in December. It is the adaptation most resorts have actually chosen.
Link to this fileThis acts directly on the quantity your file measures, and it is the reason a student must be careful: many stations sit in or near resorts, and snow depth where snow is made is not snow that fell. HN, depth of snowfall, is the column that cannot be manufactured, which is why it is the one to build on.
The tensionWater, taken and stored on a mountain, against the snow economy of the valley below. Environmental groups document the land take and the pressure on alpine water; the resorts answer that without snow there is no village left to protect.
SourceLand and forests5
Farmland in every Swiss canton, four surveys since 1979
Farmland lost to building in the lowlands, and alpine pasture going back to forest in the mountains: two different reasons a canton farms less of itself than it did.
- One row:
- One land-use category in one canton in one survey period, in hectares. Pivot it so one row is one canton in one period.
- Covers:
- 26 cantons in four national surveys (1979/85, 1992/97, 2004/09, 2013/18), plus a fifth survey (2020/25) already published for 10 cantons: 114 canton-periods in all
Full record8 traps
- Usable rows
- 114
- Coverage
- 26 cantons in four national surveys (1979/85, 1992/97, 2004/09, 2013/18), plus a fifth survey (2020/25) already published for 10 cantons: 114 canton-periods in all
- How it arrives
- One CSV, already filtered to Switzerland and the 26 cantons. The full table on stats.swiss also covers every district and commune and is 140 MB, so use this link
- Language
- Page and file in English
- Natural· Asian hornet records· by canton
- Natural· Wooded areaagainstHuman· Settlement area
- Natural· Wooded areaagainstHuman· Farmland
- Natural· Wooded areaagainstHuman· Arable land
- 1Alpine pasture is farmland here. Code 2 includes code 209, which is about 35% of all Swiss farmland. In Graubünden or Uri most of the "farmland" is summer pasture. Split 2 into 206-208 (lowland) and 209 (alpine) before you compare a mountain canton with a lowland one, or you will be comparing different things.
- 2Don't add the parent to its children. Code 2 is the sum of 206, 207, 208 and 209, and 8100 is the sum of the cantons. Add up the whole column and you count everything twice.
- 3The periods are labels, not dates. "2013-2018" means the photographs of Geneva were taken in 2012 and of Graubünden in 2015-19. Use the FJ column, and turn changes into hectares per year, because the gaps between surveys vary from canton to canton.
- 42014 falls inside a survey. The Spatial Planning Act came into force in 2014. Geneva's 2013/18 photographs are from 2012, which is before it, but Aargau's are from 2015-16, which is after. Check each canton's flight year before you call an interval before or after.
- 5Only 10 cantons have a fifth survey. 2020/25 is being published canton by canton: AG, BL, BS, FR, GE, JU, LU, NE, SO and VD so far, as of 31 August 2026. None of them is a high-mountain canton. The national row stops at 2013/18, so don't add the new cantons to it.
- 6The canton boundaries moved. On 1 January 2026 Moutier moved from Bern to Jura. The file uses today's boundaries for every period, so Jura's 1979/85 figure includes Moutier. An older download, such as the 2021 xlsx edition, will give you different numbers for 43 of the 104 canton-periods. Use one download, and cite it.
- 7It is a sample, not a census. Each hectare is judged by one point on an aerial photograph, so a small figure is uncertain. The 95% error is roughly 1.96 times the square root of the number of hectares: Basel-Stadt's 422 ha of farmland is really 422 ± 40. Jura's farmland rose between 2013/18 and 2020/25 by 103 ha, which is within that error. Don't report it as a finding.
- 8Net change, not what became what. Geneva lost 1,675 ha of farmland and gained 1,673 ha of settlement between 1979/85 and 2013/18, but a table of totals can't prove that the same hectares changed. Say "at the same time", not "was built on". Only the point data, a separate download from the same office, can show you what each hectare changed to.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Is there a correlation between the increase in settlement area and the decrease in lowland farmland across the 26 Swiss cantons between 1979/85 and 2013/18?
- In the 10 cantons surveyed in 2020/25, was the yearly loss of farmland smaller after the revised Spatial Planning Act (2014) than before it?
- Do mountain cantons (Valais, Ticino, Graubünden, Uri) lose a larger share of their alpine pasture than of their lowland farmland?
- NOAS: the land-use category. 1 settlement, 2 agricultural, 3 wooded, 4 unproductive. Inside agricultural: 206 orchards, vineyards and horticulture, 207 arable land, 208 meadows and farm pastures, 209 alpine pastures
- PERIOD: the survey, A1 (1979-1985) to A5 (2020-2025)
- REGION: the canton's two-letter code, and 8100 for Switzerland
- OBS_VALUE: the area in hectares
- FJ: the flight year, when the aerial photographs of that canton were actually taken
One table per canton, every survey (26)
In the explorer choose Downloads & Links, then Filtered data in tabular text (CSV) or Table in Excel. Not Unfiltered: that is every commune in Switzerland, 140 MB.
- Aargau (AG)Land use, 5 surveys
- Appenzell Ausserrhoden (AR)Land use, 4 surveys
- Appenzell Innerrhoden (AI)Land use, 4 surveys
- Basel-Landschaft (BL)Land use, 5 surveys
- Basel-Stadt (BS)Land use, 5 surveys
- Bern (BE)Land use, 4 surveys
- Fribourg (FR)Land use, 5 surveys
- Geneva (GE)Land use, 5 surveys
- Glarus (GL)Land use, 4 surveys
- Graubünden (GR)Land use, 4 surveys
- Jura (JU)Land use, 5 surveys
- Lucerne (LU)Land use, 5 surveys
- Neuchâtel (NE)Land use, 5 surveys
- Nidwalden (NW)Land use, 4 surveys
- Obwalden (OW)Land use, 4 surveys
- Schaffhausen (SH)Land use, 4 surveys
- Schwyz (SZ)Land use, 4 surveys
- Solothurn (SO)Land use, 5 surveys
- St. Gallen (SG)Land use, 4 surveys
- Ticino (TI)Land use, 4 surveys
- Thurgau (TG)Land use, 4 surveys
- Uri (UR)Land use, 4 surveys
- Vaud (VD)Land use, 5 surveys
- Valais (VS)Land use, 4 surveys
- Zug (ZG)Land use, 4 surveys
- Zurich (ZH)Land use, 4 surveys
Are there more Asian hornet records per km² in the Swiss cantons where settlements cover a larger share of the land, 2025?
Watch outBoth files write the canton as the same two-letter code, and Fribourg is FR in both, so all 26 cantons match: 22 have a hornet record, 20 of them in 2025, and 4 have none (AI, AR, GL, GR). Filter the hornet file to countryCode CH first. Divide by each canton's area (NOAS 1 to 4 added up), and use the 2013/18 survey, the last one every canton has. More settlement also means more people to see a hornet and report it, and the file does not say how many people looked: count the 5 km squares with a record as well as the rows, and say that neither is a count of nests. A small canton has few squares (Basel-Stadt has one), so its count of squares soon reaches the top. The hornet did not reach every canton at the same time, so check your result on the 15 cantons with a record by 2023 as well.
The Confederation sets the rules; the cantons must shrink building zones that are larger than 15 years' demand, and tax at least 20% of the gain in value when land is rezoned for building.
Link to this fileDirect: its purpose is to stop building spreading onto open land. The 10 cantons with a 2020/25 survey give an after. Every one of them lost farmland more slowly per year between 2013/18 and 2020/25 than between 2004/09 and 2013/18. Aargau fell from about 186 to 83 ha a year, Vaud from 274 to 220. Slower is not the same as caused: see the traps on flight years.
The tensionValais was the only canton to vote no, by 80.4%. It had more building land than it could use, so shrinking its zones meant taking value away from people who owned plots zoned for building. Landowners in mountain cantons against the pressure on lowland farmland is a tension you can document.
SourceThe Federal Council, through the Federal Office for Spatial Development (ARE). Each canton must keep its own quota of its best arable land.
Link to this fileIt protects arable land only, which is code 207. It says nothing about meadows or alpine pasture. So test it on arable land. Arable land in all of Switzerland fell from 434,164 ha in 1979/85 to 381,254 in 2013/18. That is below the plan's 438,460, but the two are not the same measure: the plan counts land that could be ploughed, including meadow in a crop rotation, while this survey records what the photograph shows. Compare the trend, not the totals.
The tensionProtected farmland against housing, and Geneva is where it is sharpest. Voters approved rezoning 58 ha of farmland at Les Cherpines for about 3,000 homes on 15 May 2011 (56.6% yes). On 8 March 2026, Confignon's voters rejected the plan for the district by 64.59%.
SourceThe Federal Office for Agriculture, as direct payments to farms that take livestock up to alpine pastures for the summer.
Link to this fileThe payments exist because unused alpine pasture turns into scrub and then forest, which is code 209 falling as wooded land rises. Ticino lost 20% of its alpine pasture between 1979/85 and 2013/18, while its wooded area grew by 13,410 ha. The increase came in 2014, inside the 2013/18 survey, and no mountain canton has a 2020/25 survey yet, so this data can show you the trend the payment was meant to stop, but not yet whether it worked.
SourceTree cover lost in every country, state and province, every year since 2001
Forest loss: how much tree cover each country, and each state or province within it, loses every year, and what the satellite record says removed it.
- One row:
- One country, or one state or province, at one canopy threshold: its tree cover in 2000 and 2010, and the hectares of it lost in each year from 2001 to 2025, one column per year.
- Covers:
- 226 countries and their states and provinces, 2001 to 2025, each at eight canopy thresholds from 0 to 75%. 109 countries had at least 1 million hectares of tree cover in 2000
Full record8 traps, 2 on the three lines
- Usable rows
- 226
- Coverage
- 226 countries and their states and provinces, 2001 to 2025, each at eight canopy thresholds from 0 to 75%. 109 countries had at least 1 million hectares of tree cover in 2000
- How it arrives
- One Excel workbook of 18 MB with eleven sheets, from the Download button on the global dashboard
- Language
- Page and file in English
- Human· Cattle herd· by state name and year
- Human· Protected land· by country, by name
- Human· GDP per person (PPP)· by country, by name
- 1Every country is there eight times, once for each canopy threshold (0, 10, 15, 20, 25, 30, 50 and 75%). Filter threshold to one value first, 30 unless you have a reason: at 0, every hectare of land counts as tree cover.
- 2Names, never codes. Thirteen of the 109 countries with at least 1 million hectares of tree cover are spelled differently in World Bank files (México, Russia, Vietnam, Côte d'Ivoire, both Congos, both Koreas and five more), and French Guiana and Taiwan are not in them at all. Match them by hand, and say so in your method.
- 3China, India and Pakistan appear more than once. Disputed areas get extra rows under each country that claims them: at 30%, India has six rows, China four and Pakistan two. Add a country's rows together, and say that you did.
- 4Tree cover is not forest, and loss is not deforestation. Tree cover is any vegetation over 5 metres, plantations included, and loss includes harvesting, fire, disease and storms. The drivers sheet says which: wildfire is 79% of Australia's loss since 2001, 74% of Russia's and 68% of Canada's.
- 5Fire years stand out. Wildfire was 49% of Brazil's loss in 2024 and 42% in 2025, against 12 to 25% in each year from 2019 to 2023.
- 6The method changed between 2011 and 2015. GFW warns that newer satellite data can show more loss in later years for that reason alone. Compare places over the same years, rather than early years with late ones.
- 7Hectares favour big countries. Divide the loss by the country's own tree cover (extent_2000_ha or extent_2010_ha) before you compare.
- 8The site has a new name. Global Forest Watch is now Global Nature Watch. The old address still leads there, and the file is the same.
- Across the countries with at least 1 million hectares of tree cover, is the share of land protected in 2015 related to the share of tree cover lost from 2016 to 2025?
- Across Brazil's 27 states, is the share of tree cover lost from 2001 to 2023 related to how much each state's cattle herd grew?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- In which countries does GFW put most tree cover loss since 2001 down to wildfire, and in which to permanent agriculture?
- country, and subnational1 on the state sheet: names only, never a code
- threshold: how dense the canopy has to be, in %, to count as tree cover. GFW's own default is 30
- extent_2000_ha and extent_2010_ha: tree cover, in hectares
- tc_loss_ha_2001 to tc_loss_ha_2025: hectares of tree cover lost in each year
- The Country drivers and Subnational 1 drivers sheets: the same loss split by what caused it (permanent agriculture, hard commodities, shifting cultivation, logging, wildfire, settlements and infrastructure, other natural disturbances), at 30% only
Across the countries with at least 1 million hectares of tree cover, does protecting more land in 2015 go with losing less of that tree cover from 2016 to 2025?
Watch outGFW names countries and never codes them, so match by hand: 13 of the 109 countries with 1 million hectares of tree cover are spelled differently in the World Bank file (México is Mexico, Russia is Russian Federation, Vietnam is Viet Nam, both Congos and both Koreas are written Congo, Dem. Rep. and so on), and French Guiana and Taiwan are not in it at all. That leaves 107. Filter GFW to one canopy threshold first, and add together the extra rows China, India and Pakistan carry for disputed areas.
Across the countries with at least 1 million hectares of tree cover, is GDP per person in 2015 related to the share of tree cover lost from 2016 to 2025?
Watch outMatch by hand, as for protected land: GFW names countries and never codes them, 13 of the 109 are spelled differently in the World Bank file, and French Guiana and Taiwan are not in it. Five more have no GDP in any year (Cuba, New Caledonia, North Korea, South Sudan and Venezuela), which leaves 102. Filter GFW to one canopy threshold and add together China's, India's and Pakistan's extra rows first, and divide each country's loss by its own tree cover. GDP per person does not clear forest: any link runs through what a country's land is used for and who buys what it grows, so say what you think it runs through.
Across Brazil's 27 states, is the growth of the cattle herd from 2001 to 2023 related to the share of tree cover each state lost over the same years?
Watch outThe state names are spelled identically in both files, accents included, so all 27 match. Stop at 2023: wildfire was 49% of Brazil's tree cover loss in 2024, against 12 to 25% in each year from 2019 to 2023, and IBGE's 2024 herd is preliminary. Compare shares (loss as a percentage of the state's tree cover, the herd's percentage change), not hectares and head counts, or the biggest states simply come top.
More than a hundred countries holding most of the world's forests, Brazil, Indonesia, Russia and China among them.
Link to this fileDirect: its signatories committed to halt and reverse forest loss by 2030, and this file measures forest loss country by country, year by year. It counts tree cover, plantations included, so read the drivers sheet before you call a country's loss deforestation.
The tensionTwo days after Indonesia signed, its environment minister, Siti Nurbaya Bakar, said that forcing Indonesia to zero deforestation by 2030 was "inappropriate and unfair", because the country still had to develop its economy.
SourceThe federal government. In its fifth phase, 13 ministries coordinated by the Ministry of the Environment and Climate Change.
Link to this fileDirect for the nine states of the Legal Amazon, which are on the Subnational 1 sheet. The plan measures itself by INPE's PRODES survey, which counts clearing of primary forest, not this file's tree cover, so the two will not give the same numbers.
The tensionThe 2012 Forest Code (Law 12.651) forgave fines for clearing done illegally before 22 July 2008 and cut how much land had to be restored. Conservationists said it told landowners that illegal clearing would one day be forgiven; Brazil's Supreme Court upheld the amnesty in 2018.
SourceDwellings in England, every year since 1969
Housing growth in England and the habitat it can take. The government starts from each census count of homes and adds the homes gained each year, less those demolished, so this is England's number of homes, year by year.
- One row:
- One date for the whole of England: the estimated number of dwellings on that date, all together and split by tenure (who owns or rents them). From 1969 there is one date a year.
- Covers:
- England, one count a year from 1969 to 2025, plus census counts every ten years from 1801 to 1961 (none for 1941)
Full record11 traps, 1 on the three lines
- Usable rows
- 57
- Coverage
- England, one count a year from 1969 to 2025, plus census counts every ten years from 1801 to 1961 (none for 1941)
- How it arrives
- One .ods spreadsheet (13 KB) with four sheets: Cover, Contents, Notes and the table, LT_104. Excel, LibreOffice and Google Sheets all open it.
- Language
- Page and file in English
- Natural· Butterfly abundance, one species· by year
- 1The table is on the fourth sheet. The file opens on its Cover. The numbers are on the sheet LT_104, where four lines of notes sit above the headings. [x] means no data. It is text, so delete it before you calculate.
- 2Two rows for 1971 and two for 1981. Each of these census years has an extra row dated 1 April beside the 31 December row, with the same Year. Delete the 1 April rows before you subtract, or those two years are split in two. The rows from 1801 to 1961 are census counts ten years apart: keep them out of a yearly series.
- 3The date of the count moves. Up to 1990 each count is on 31 December. From 1991 it is on 31 March. So the step from 1990 to 1991 covers three months, not twelve. Decide what to do with it, and say.
- 4A year is not always the same twelve months. Up to 1990 the step for a year runs from January to December. From 1992 it runs from April of the year before to March, so most of it falls in the year before its label. Say which months your other data covers, and check that the two line up.
- 5Added is not built. This year's count minus last year's is the net gain: new homes, plus homes made by converting a building or changing its use (an office turned into flats), minus homes demolished.
- 6Two years do not look like building. The count grew by 82,000 in 1980 and by 254,000 in 1987. MHCLG's own count of homes completed in those years, in Live Table 244 on its house-building page, is very different from both, and neither file explains why. The Cover also warns that earlier years are less reliable. Decide what to do with 1980 and 1987 before you look at your result, and say what you did.
- 7Each census corrects the years before it. After the 2001, 2011 and 2021 censuses, MHCLG spread the difference between its estimate and the census evenly over the ten years before. So every step from 1992 to 2021 includes a share of a correction, and the share is different in each ten years. The steps from 2022 have no correction yet, and may change after the next census.
- 8Rounded up to 2001, provisional from 2022. All dwellings is rounded to the nearest thousand up to 2001 and given to the single dwelling from 2002, but the Cover says the figures are not accurate to the nearest dwelling. 2022 to 2025 are marked [p] (provisional), and 2024 also [r] (revised). Record the publication date from the Cover.
- 9The tenure columns start late and change method. Private registered providers start in 1977, and other public sector dwellings in 1991. From 2003 owner occupied and private renting use a new method (note 2), so compare them before 2003 or after it, not across it. Note 3 says the 2020 council figures may be lower because fewer councils replied, not because the stock fell.
- 10A dwelling is not a hectare. A home built where a building already stood takes no new habitat. A home built on a field or in a wood takes land from nature. This file does not say which is which, so say what you are assuming. And already built does not always mean no wildlife: some brownfield land left empty for years is home to rare insects.
- 11The total rises every year. Almost anything that grew steadily since 1969 will go with the number of dwellings. Use the number added each year, never the total.
- Is the number of dwellings added in England each year related to the change in a butterfly species' England index from one year to the next, 1976 to 2023? (UK butterflies, England rows.)
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the number of dwellings added each year in England changed between the 1970s and the 2010s?
- How has the share of England's dwellings rented from local authorities changed since 1969?
- Has private renting grown faster than owner occupation in England since 2003, when the method for both changed?
- Date: the day of the count: 31 December up to 1990, 31 March from 1991, and 1 April in the census rows
- Year
- All dwellings: every dwelling in England on that date
- Owner occupied, Rented privately or with a job or business, Rented from private registered providers (housing associations), Rented from local authorities and Other public sector dwellings: the same dwellings, split by tenure
- Notes: [p] provisional, [r] revised, and a numbered note
Is the change in a butterfly species' England index from one year to the next related to the number of dwellings added in England that year, 1976 to 2023?
Watch outUse the England rows of the butterfly file and the All dwellings column on the sheet LT_104 of Live Table 104. Work out dwellings added as this year's count minus last year's, and look before you trust it: 1971 and 1981 have an extra census row (1 April) that splits a year in two, so drop those rows, and in 1991 the count moved from 31 December to 31 March, so the 1990 to 1991 step covers three months, not twelve. The move also changes which months a year means: up to 1990 a step runs January to December and includes that summer's butterfly counts, but from 1992 it runs April to March and ends before they begin. Check your result on 1992 to 2023 alone, where every step is the same twelve months: 32 changes. The steps for 1980 and 1987 are very different from the homes MHCLG counts as completed those years: decide what to do with them before you look. Stock figures are rounded to thousands before 2002, each census's correction is spread over the ten years before it, and 2022 and 2023 are provisional. Both series change steadily, so use the change against the change, not the level against the level. And a dwelling is not a hectare: homes built on land that was already built on take no new habitat, so say what you are assuming.
The UK Government, through Defra, which sets the rules and the biodiversity metric that measures habitat. Each local planning authority must approve a developer's biodiversity gain plan before building can start.
Link to this fileDirect in purpose, but not in this file's years. It is the plan that links new homes to the habitat they replace. But it applies to planning applications made from February 2024, and this file ends at 31 March 2025, so the file cannot show it working. Use it to explain what new homes now have to do for nature, not to test whether it worked.
The tensionHousebuilders and nature groups disagree about how far it should reach. On 2 April 2026 Neil Jefferson, chief executive of the Home Builders Federation, said that "The cumulative impact of Biodiversity Net Gain requirements and wider cost pressures is increasingly affecting the viability of housing developments" (whether they make enough money to go ahead). On 10 June 2026 Wildlife and Countryside Link, a group of nature charities, said it "strongly opposes the introduction of any new brownfield exemption from Biodiversity Net Gain (BNG), in any form".
SourceThe UK Government, through MHCLG. Councils have to deliver it: the National Planning Policy Framework gives each one a mandatory housing target and, where homes cannot be found elsewhere, lets it build on "grey belt" land in the Green Belt.
Link to this fileDirect on the people side. The government measures the target in net additional dwellings, the same measure as this file's change from one year to the next. But the target counts homes from 9 July 2024, and this file ends on 31 March 2025, less than nine months into it. Use it to explain why the number of homes added each year matters now, not to judge the target.
The tensionBuilding homes against keeping countryside. On 7 December 2025 CPRE, the countryside charity, reported that of 1,250 homes approved by government planning inspectors on "grey belt" land, 88% "will be built on previously undeveloped countryside". Its chief executive, Roger Mortlock, said that "local authorities are so desperate to meet revised housing targets that they are forced to approve speculative development in the Green Belt" (building that a developer proposes where the council's plan did not).
SourceThe UK Parliament, on the government's bill. Natural England writes environmental delivery plans and runs the fund. Developers pay a nature restoration levy into it, and Natural England spends it on nature at a larger scale, instead of each developer dealing with its own site.
Link to this fileIndirect. It is about protected sites and protected species, such as great crested newts and rivers harmed by nutrients from new homes, not butterflies in general. It changes how builders pay for harm to nature, not how many homes are built, and it began after the last year in this file.
The tensionThe government called the fund "a win-win for both the economy and nature", and housebuilders welcomed the Act: David Thomas, chief executive of Barratt Redrow, said it "will help unlock and accelerate the delivery of much-needed homes across the country". The government's own environmental watchdog, the Office for Environmental Protection, advised on 2 May 2025: "As drafted, the provisions are a regression." In May 2025 the RSPB's chief executive, Beccy Speight, said: "The evidence clearly shows nature isn't a blocker to growth." The watchdog's advice was on the bill as first written, so check what the final Act says.
SourceArea burnt and number of fires in 31 countries around Europe and the Mediterranean, every year since 1980
Wildfires: how much land burns in each country every year, and how many fires there are, as each country's fire service reports them to the EU.
- One row:
- One country in one year: the hectares burnt and the number of fires its fire service reported. Each country is a column, and each year is a row.
- Covers:
- 31 countries, 1980 to 2024. Portugal, Spain, France, Italy and Greece from 1980, 13 more from the 1990s, and the rest from 2000 to 2017. Algeria stops in 2022 and Lebanon in 2023
Full record8 traps
- Usable rows
- 916
- Coverage
- 31 countries, 1980 to 2024. Portugal, Spain, France, Italy and Greece from 1980, 13 more from the 1990s, and the rest from 2000 to 2017. Algeria stops in 2022 and Lebanon in 2023
- How it arrives
- One Excel workbook of about 30 KB with two sheets, from the link Country totals 2024 on the Data and services page. The page calls it .xls; it is .xlsx
- Language
- Page and file in English
- 1The five biggest columns are text. For Portugal, Spain, France, Italy and Greece, almost every year from 1980 to 2023 is stored as text, with a no-break space inside the number (44 251). Excel shows it but will not add it up, and =VALUE() on its own fails. Only 2024 is a real number. Convert first, for example with =VALUE(TRIM(SUBSTITUTE(B2,UNICHAR(160),""))). Finland's 2012 and 2013 are text too.
- 2Burnt area means different land in different countries. France's column is forest fires in mainland France only: 2,949 hectares in 2024, while France's own total for all vegetation, overseas included, was 6,869. Italy, Portugal, Spain and Greece count shrubland and other vegetation as well. Compare a country with itself over the years before you compare countries, and say what each column counts.
- 3Three fire counts for 2024 are wrong in the file. Austria reads 595, Bulgaria 106 and Slovakia 186: each is the figure from the column beside it. The report's own chapters give 130 for Austria, 595 for Bulgaria and 102 for Slovakia. Czechia's burnt area for 2006 and its fire counts for 2005 and 2006 also differ from the report's tables. Use the report's figures, and say that you did.
- 4Countries join in different years. Portugal, Spain, France, Italy and Greece start in 1980, 13 more in the 1990s, and the rest between 2000 (Cyprus, Estonia) and 2017 (the Netherlands). Algeria stops in 2022, Lebanon in 2023, and Hungary has no burnt area for 2001. An empty cell means no report, not no fire: the file has no zeros at all.
- 5More fires can mean better counting. The report says the rise in the number of fires in the 1990s was partly "due to improved recording", and Norway changed how it records fires in 2016. Small fires add many to the count but little to the area.
- 6Years run down and countries across. Turn the table round to one row per country and year before you match it with anything. Countries are codes only: DEU is Germany, CHE Switzerland, MKD North Macedonia.
- 7Each report can change the past. The report says that where two reports disagree, "the later report should be taken as the definitive version". Against the 2023 workbook, one value changed (Romania, 2023) and one was added. Greece's 2024 figures are marked provisional. Write down which edition you used.
- 8No UK, Ireland or Belgium. Only countries that send figures to EFFIS are here. For those countries, and for 2025 and 2026, see EFFIS's satellite estimates, which measure something different.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the area burnt each year in Portugal from 2018 to 2024 compare with 2007 to 2016, before its fire reforms?
- Has the average size of a fire (hectares burnt divided by the number of fires) changed in Portugal, Spain, France, Italy and Greece since 1980?
- Do Portugal, Spain, France, Italy and Greece have their worst fire years in the same years?
- What share of the 660,000 hectares in Portugal's target for 2020 to 2030 had burnt by the end of 2024?
- Year: one row for each year, 1980 to 2024
- PRT, ESP, FRA and 28 more: one column for each country, named by its three-letter ISO code
- The sheet Burnt area (ha) 1980 - 204 (the name is cut short in the file): hectares burnt
- The sheet Nr. of forest fires 1980 - 2024: the number of fires
The Government of Portugal. AGIF, the agency for integrated rural fire management, set up by Decree-Law 12/2018, coordinates the new system. ICNF, the forest and nature institute, has to approve any planting of eucalyptus.
Link to this fileDirect. The plan's own target is hectares burnt, and the Portugal column is hectares burnt, reported by Portugal's forest institute. It counts mainland Portugal only, and includes shrubland and farmland as well as forest. The file stops in 2024, so it covers less than half of the years the target is about.
The tensionIn April 2017, before the eucalyptus rule became law, CELPA, the association of Portugal's paper industry, called the ban a "huge loss to the Portuguese economy". It said the ban "encourages abandonment and growth of scrubland and unused areas and will increase the risk of fires".
SourceThe French Parliament. The law asks the interior, forest, environment and planning ministries to write a national strategy against fires in forests and other vegetation. It also extends the legal duty to clear undergrowth (the obligations légales de débroussaillement) to parts of France that did not have it before.
Link to this fileDirect for France's column, but only for part of what burns: here France is forest fires in mainland France only. Fires in other vegetation, and on Réunion, Mayotte and in French Guiana, are in France's chapter of the report but not in this file. The file ends with 2024, the first whole year after the law.
The tensionNot a clash between groups, but a gap between the law and what followed. In April 2025 three members of the National Assembly who reviewed the law regretted that the national strategy it asked for within one year had still not been adopted, so it was not yet working.
SourceThe European Commission pays for aircraft that stand ready in member states. When a country asks for help through the EU Civil Protection Mechanism, the Commission and the countries involved decide together where they go.
Link to this fileIndirect. The planes fight fires that have already started, in whichever country asks for them, so any effect is spread over many columns and has no single start date. The file cannot show where the planes went: for that you need the Commission's own list of the times the Mechanism was used.
SourceBurnt area mapped from satellites in each country around Europe and the Mediterranean, every year since 2006
Wildfires measured the same way in every country: the area burnt that satellites map each year, including the fire season that is still going on.
- One row:
- One country in one year: the hectares inside the burnt areas that EFFIS mapped from satellite images, and how many burnt areas it mapped.
- Covers:
- 65 countries and territories in the EU, the rest of Europe, the Middle East and North Africa, 2006 to 2026. 2026 is the season so far. 15 of them show no burnt area in any year
Full record7 traps
- Usable rows
- 1,000
- Coverage
- 65 countries and territories in the EU, the rest of Europe, the Middle East and North Africa, 2006 to 2026. 2026 is the season so far. 15 of them show no burnt area in any year
- How it arrives
- A CSV for one country, 21 rows, from the download menu on the country's chart (the arrow icon, then Data, then CSV). The same numbers are in a table under the chart
- Language
- Page and file in English
- 1The last year is not finished. The file ends with the current year's season so far (2026 when counted), although its name says 2006_2025. Leave that year out of any comparison of whole years.
- 2The method changed in 2018. Until 2018 EFFIS mapped burnt areas of about 30 hectares or more from MODIS images. Since 2018 it also uses Sentinel-2 images, which find smaller fires. The number of fires changes most. Compare years on the same side of 2018, or say why you did not.
- 3Every burnt patch counts. EFFIS says it makes no difference between wildfires, burning to manage land and planned burns.
- 4Not the same as what countries report. For the same country and year, this figure and the country's own (EFFIS country totals) can be far apart. Use one source for the whole of a series.
- 5The file does not say which country it is. Each export holds one country, and only its file name carries the code. Add a country column before you put two exports together.
- 6A zero is not always no fire. 712 of the 1,300 country-years up to 2025 are 0 hectares. The French overseas territories read 0 in every year, although France reports fires on Réunion and Mayotte and in French Guiana. Say whether your zeros mean no fire, or no burnt area large enough to map.
- 7The CSV is inside a menu. Hold the mouse over the arrow icon at the top right of the chart, then Data, then CSV. The overview page gives one year for every country in a zone instead.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How close is the area EFFIS mapped from satellites in each country to the area the country reported itself, 2006 to 2024?
- How much of Portugal's target of 660,000 hectares for 2020 to 2030 had burnt by the end of 2025, as mapped from satellites?
- In which years since 2006 were the largest areas mapped as burnt in the UK, and were they the same years as in Ireland?
- Year: 2006 to the current year
- Burned Area (ha): the hectares inside the burnt areas mapped that year
- Number of Fires: the number of burnt areas mapped
- On the overview page, one year for every country in a zone, with Country area (ha) beside it
The Government of Portugal. AGIF, the agency for integrated rural fire management, set up by Decree-Law 12/2018, coordinates the new system. ICNF, the forest and nature institute, has to approve any planting of eucalyptus.
Link to this fileDirect: the target is hectares burnt from 2020 to 2030, and this file reaches 2025 and the 2026 season so far. The satellite does not map the smallest fires, so its total for Portugal will not match the figure Portugal reports.
The tensionIn April 2017, before the eucalyptus rule became law, CELPA, the association of Portugal's paper industry, called the ban a "huge loss to the Portuguese economy". It said the ban "encourages abandonment and growth of scrubland and unused areas and will increase the risk of fires".
SourceThe French Parliament. The law asks the interior, forest, environment and planning ministries to write a national strategy against fires in forests and other vegetation. It also extends the legal duty to clear undergrowth (the obligations légales de débroussaillement) to parts of France that did not have it before.
Link to this fileDirect for France, and closer to the law than the national figures: the satellite maps every burnt area, forest or not, as the law covers forests and other vegetation. The French overseas territories are separate rows here.
The tensionNot a clash between groups, but a gap between the law and what followed. In April 2025 three members of the National Assembly who reviewed the law regretted that the national strategy it asked for within one year had still not been adopted, so it was not yet working.
SourceFood and farming6
US honey bee colonies lost, and what affected them, state by state every quarter since 2015
Honey bee losses and the pressures on managed pollinators. Every quarter, beekeepers with five or more colonies report how many they lost and what affected them: mites, disease, pesticides and more.
- One row:
- One state in one quarter: its colonies on the first day, the most it had, and the colonies lost, added and renovated; and, in a second table, the percentage of colonies affected by each of six stressors.
- Covers:
- Up to 45 states, 46 quarters from January 2015 to June 2026, with April to June 2019 missing
Full record10 traps, 1 on the three lines
- Usable rows
- 1,815
- Coverage
- Up to 45 states, 46 quarters from January 2015 to June 2026, with April to June 2019 missing
- How it arrives
- One zip for each year's release, with a CSV for every table, plus the same report as text and as a PDF
- Language
- Page and file in English
- Natural· Honey bee colonies lostagainstHuman· Colonies affected by pesticides
- 1One release a year, and they overlap. Each August release covers the year before and the first half of its own year; the first, in May 2016, covers 2015. For 2015 to 2026 you need all eleven. Where two releases give the same quarter, use the later one, which is revised, and record which release each quarter came from.
- 2Twenty states since July 2024. Up to June 2022, 45 states are named; up to June 2024, 44; from July 2024, only 20, and the rest are merged into Other States. For a series that runs to 2026, use the 20 states named in every quarter.
- 3April to June 2019 is missing. Every cell says (NA), not available. Leave the gap and say so; never fill it in.
- 4The symbols are values. "-" means zero and "(Z)" means less than half a percent. In the Pesticides column, 382 of 1,860 cells are one or the other. Turn them into 0 and say so. Deleting them throws away the low values and makes every average too high. "(X)" means not applicable.
- 5A state and the US divide by different numbers. A state's percent lost divides by its maximum colonies, which include every colony moved in during the quarter: California had 910,000 colonies on 1 January 2025 and a maximum of 1,260,000. The US figure divides by the first-day colonies. So averaging the states does not give you the US figure.
- 6These are livestock, not wild bees. Beekeepers replace lost colonies by splitting strong ones and by buying new bees, so the number of colonies can hold steady while losses stay high. A result about honey bee colonies says nothing on its own about wild pollinators.
- 7A stressor is what the beekeeper reported. "Affected by pesticides" is the beekeeper's judgement, not a measurement of how much pesticide was used, and one colony can be counted under several stressors.
- 8Winter is two quarters. The 2015 goal is about the whole winter. This file gives October to December and January to March separately, each as a percentage of a different starting number, so do not simply add them. Say how you combined them, and that it is not quite the goal's own measure.
- 9Losses follow the seasons. Compare the same quarter across states and years, or you will mostly be measuring the time of year.
- 10Only operations with five or more colonies. The quarterly tables leave out small beekeepers.
- Is the percentage of honey bee colonies lost from January to March related to the percentage of colonies affected by pesticides in the same quarter, across the 20 US states named every year, 2015 to 2026?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Has the percentage of US honey bee colonies lost over the winter fallen to the 2015 strategy's goal of 15% or less?
- In which quarter of the year are the most colonies lost, and is it the same quarter in California and in North Dakota?
- Is the percentage of colonies affected by varroa mites related to the percentage lost, across US states?
- Percent lost: the colonies lost in the quarter, as a percentage of the most colonies the state had in it
- Lost colonies, Added colonies and Renovated colonies (requeened, or given new bees)
- Maximum colonies: the colonies on the first day, plus all those moved into the state during the quarter
- Varroa mites, Other pests and parasites, Diseases, Pesticides, Other and Unknown: the percentage of colonies each affected at any time in the quarter, in the stressor table
The Pollinator Health Task Force, set up by President Obama in June 2014 and co-chaired by the Secretary of Agriculture and the Administrator of the Environmental Protection Agency.
Link to this fileDirect, and rare: the strategy names this survey. It says its goal is informed by the Bee Informed Partnership's surveys and by "the newly established quarterly and annual surveys by the USDA National Agricultural Statistics Service", which is this file, so you can test the target itself. But the surveys it was based on count winter as October to April, and this file splits it into two quarters: say how you put them together.
The tensionPesticides. Environmental groups said the plan did not restrict the neonicotinoid pesticides already in use: the Natural Resources Defense Council said "we must drastically cut down on today's pervasive use of neonicotinoids and other pesticides". Bayer CropScience replied that neonicotinoids, "when used according to labeled directions", can be used safely with pollinators (NPR, 20 May 2015).
SourceSheep on French farms, department by department, every year since 2010
How many sheep are kept in each French department, year by year: the flocks that wolves returning to the Alps and the Jura now meet.
- One row:
- One kind of farm animal in one department, with its head count at the end of each year from 2010 to 2025, one column per year.
- Covers:
- 100 departments (Paris is absent), 2010 to 2025, with 2025 provisional. France's sheep: 7,945,548 in 2010 and 6,449,154 in 2025
Full record8 traps, 1 on the three lines
- Usable rows
- 1,600
- Coverage
- 100 departments (Paris is absent), 2010 to 2025, with 2025 provisional. France's sheep: 7,945,548 in 2010 and 6,449,154 in 2025
- How it arrives
- A zip holding two Excel workbooks: use SAA_2010-2025_provisoires_donnees_departementales.xlsx, sheet EFA
- Language
- Page and file in French
- Natural· Confirmed wolf evidence, by department· by department and year
- 1The headings are on row 6. Five title and note rows sit above them, and the workbook has 14 sheets. Sheep are on EFA.
- 2Years run across, not down. Turn the 16 year columns into one column before you join the file to anything that has a year column.
- 3"dont" means "of which". Row 33 (dairy ewes) is already inside row 32, so adding rows 31 to 34 counts dairy ewes twice. Use row 35, all sheep.
- 4Department codes have three characters here (004), and two in the wolf file (04). Corsica is 02A and 02B.
- 5Sheep are counted at the farm, in December. A flock that spends the summer in another department's mountains, where it meets wolves, is counted at home.
- 6Some figures are estimates, and are flagged. For all sheep, 29 departments are quality 1, 57 are 2 and 14 are 3. Say how you treated the 3s.
- 7Some years just repeat. 42 department-years equal the year before (Lozère has 154,427 in both 2019 and 2020).
- 8Each edition rewrites the whole series. This one was realigned to the 2020 farm census, back to 2010. Do not mix it with older editions.
- Across French departments, is the number of confirmed wolf records related to the number of sheep kept on farms, 2013 to 2025?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of sheep in each Alpine department change between 2013 and 2025?
- LIB_DEP: the department, code and name in one cell (005 - Hautes-Alpes)
- LIB_SAA: the kind of animal, numbered 01 to 35. Row 35 is all sheep
- EFF_2010 to EFF_2025: the number of animals at the end of each year
- STATUT_QUALITE: how reliable the figure is, from 1 (reasonably reliable) to 3 (use only in totals)
Across French departments, is the number of confirmed wolf records related to the number of sheep kept on farms, 2013 to 2025?
Watch outStrip the leading zero from Agreste's three-character codes (004 becomes 04) and turn its year columns into one column first. 516 department-years appear in both, from 83 departments. The 2025 sheep figures are provisional, and 25 matched rows are in departments whose figure is quality 3. Sheep are counted at the farm in December, not on the summer pastures where wolves meet them, and a department is a coarse unit: say so.
Farmers apply through their department's DDT(M); shepherds are funded by the agriculture ministry and the EU.
Link to this fileIt is paid per flock, so the number of sheep in a department is part of what it costs. The file cannot say which flocks were protected.
The tensionThe wolf plan says the budget will not cope if the growth continues, and farmers keep asking for shepherd costs to be paid in advance, which EU farm-policy rules do not allow.
SourceThe ecology and agriculture ministries, run by the prefect of the Auvergne-Rhône-Alpes region.
Link to this fileIndirect: the plan is about wolves and farming together, and sheep farming is the side of it this file measures. It does not say which flocks the plan's measures reached.
The tensionThe national nature protection council gave the draft an unfavourable opinion, unanimously, on 19 October 2023, objecting to making lethal shooting easier.
SourceSheep on Swiss farms, canton by canton and commune by commune, since 1975
How many sheep Swiss farms keep, in each canton and each commune: the flocks that wolves returning to the Alps now meet, and that herd protection is paid to guard.
- One row:
- One place in one year: the sheep kept on its farms, counted in the commune where each farm is registered. A place can be Switzerland, a canton, a district or a commune.
- Covers:
- Switzerland, 26 cantons and 2,115 communes, in 28 years from 1975 to 2025, and every year since 2010. 355,639 sheep in Switzerland in 2025
Full record10 traps, 1 on the three lines
- Usable rows
- 728
- Coverage
- Switzerland, 26 cantons and 2,115 communes, in 28 years from 1975 to 2025, and every year since 2010. 355,639 sheep in Switzerland in 2025
- How it arrives
- An interactive table (STAT-TAB) that you download from. Choose Tiere - Schafe, Betriebssystem - Total, your places and your years, press Tabelle anzeigen, then choose Excel (xlsx) under Format wählen: the text files come without a header row. A French version of the page exists
- Language
- Page and file in German
- Natural· Wolf records, Schwyz· by commune and year
- 1The count moved from May to January in 2015. Up to 2014 sheep were counted in early May, after the lambs are born; since 2015 on 1 January. Switzerland had 402,772 sheep in 2014 and 347,025 in 2015, and the BFS says the fall is mostly the change of date ("grösstenteils auf den Wechsel des Stichtags zurückzuführen"). Start in 2015, or never calculate a change across it.
- 2Four levels in one list. Switzerland, cantons (- Schwyz), districts (>> Bezirk March) and communes (......1348 Vorderthal) are mixed together. Keep one level. The cantons add up exactly to the Switzerland row, so adding every row counts each sheep three or four times.
- 3Strip the dots and the number. A commune arrives as ......1331 Küssnacht (SZ). Take off the dots and the number before you match it with a name, and watch for the canton in brackets: the Schwyz wolf file writes just Küssnacht.
- 4Canton names in two or three languages. Bern / Berne, Fribourg / Freiburg, Graubünden / Grigioni / Grischun, Valais / Wallis. Other cards write cantons as two-letter codes or in French, so add a code yourself.
- 5Choose Total, not every kind of farm. Betriebssystem splits farms into organic, conventional and undetermined. Pick Betriebssystem - Total, or the same sheep appear on several rows.
- 6Only farms are counted. A farm needs at least 1 hectare of farmland, or another minimum. Sheep kept as a hobby on less land are not in the table, and the BFS warns that its totals can be below Switzerland's real totals.
- 7Counted at home, in winter. Sheep are counted in the commune where the farm is registered. In summer many go up to alpine pastures, often in another commune or canton, and that is where most meet wolves.
- 8Some years are missing. Before 2010 the table has gaps: 1999, 2001, 2003, 2005, 2007 and 2009 are left out. The canton-only table px-x-0702000000_101 has them, and its canton figures are the same.
- 9Today's communes for every year. The table uses the communes of 31 December 2025 ("Gemeinden per 31.12.2025") for every year, so a commune that merged appears under its new name even in 1975.
- 10Years may arrive as columns. If each year is its own column, turn them into one column before you join the table to a file with a year column.
- Across the 30 communes of the canton of Schwyz, is the number of wolf records related to the number of sheep kept on farms, 2023 to 2025?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of sheep on farms change in Graubünden, Valais, Ticino and Uri between 2015 and 2025?
- Beobachtungseinheit: what is counted. Choose Tiere - Schafe (sheep). Betriebe mit Schafen is the number of farms with sheep
- Kanton (-) / Bezirk (>>) / Gemeinde (......): the place. A canton has a dash in front (- Schwyz), a district two arrows (>> Bezirk March), and a commune six dots and its BFS number (......1348 Vorderthal)
- Betriebssystem: organic or conventional farms. Choose Betriebssystem - Total
- Jahr: the year of the survey
- LN - Landwirtschaftliche Nutzfläche Total (in ha): the same farms' farmland in hectares, if you want sheep per hectare
Across the 30 communes of the canton of Schwyz, is the number of wolf records related to the number of sheep kept on farms, 2023 to 2025?
Watch outTake the sheep from the commune rows, which start with six dots and a number (......1348 Vorderthal): strip those off, write the wolf file's Küssnacht as Küssnacht (SZ), and every commune matches. Leave out the 8 ausserkantonal rows and the one record that appears twice. A commune-year with no wolf record has no row, so give it 0: that makes 90 commune-years, and 36 of them have at least one record, from 20 communes. The 2026 records have no sheep count to meet yet. Most records are from 2025, when the canton set up more than 30 cameras to follow its first pack, so records show where people looked as well as where wolves were. Sheep are counted in January in the farm's home commune, not on the summer pastures where they meet wolves. Communes differ greatly in size, so think about sheep per hectare of farmland (the same table has it). With many zeros and only 30 communes, choose your test and your years before you look, and say why.
FOEN and the cantons. The money pays for electric fences, night pens, guard dogs and similar measures, on farms and on summer pastures.
Link to this fileDirect for sheep keepers: it pays for the fences and dogs that guard their flocks, and a kill in a flock without good enough protection is not paid for. The file counts sheep, not protected flocks, so it cannot say which flocks were protected.
The tensionThe farmers' union (3 September 2024): "Der SBV unterstützt die Herdenschutzmassnahmen. Deren Wirkung hängt mit der wirksamen Begrenzung der Wolfspopulation zusammen" (protection only works if the number of wolves is limited). Four conservation groups (23 September 2026): "Denn Abschüsse in unauffälligen Rudeln können diese auseinanderreissen und dadurch zu mehr Problemen mit Nutztieren führen statt weniger" (culling packs that cause no trouble can break them up, and lead to more problems with farm animals, not fewer).
SourceThe Federal Council, in the Hunting Ordinance (JSV). FOEN and the cantons apply it when they decide whether a wolf may be shot.
Link to this fileIndirect: it links the sheep in this file to the culling of wolves. FOEN wrote that protecting farm animals is no reason to shoot a wolf where reasonable protection was possible but not used. So what counts is how sheep are kept, not only how many there are.
The tensionThe farmers' union wants fewer kills to be enough for a wolf to be shot (3 September 2024): "Hier sind die Schadschwellen (=Anzahl gerissene Nutztiere) zu senken, die einen Abschuss der schadenstiftenden Tiere zulassen" (the damage thresholds, the number of farm animals killed, that allow the wolves causing damage to be shot should be lowered).
SourceCattle in each Brazilian state, every year since 1974
Cattle ranching, which takes most of the land cleared from the Amazon: how many cattle each Brazilian state holds, year by year.
- One row:
- One state in one year: the number of cattle on 31 December.
- Covers:
- 27 states (26 and the Federal District), 1974 to 2024, with 2024 preliminary. Brazil's herd was 176,388,726 in 2001 and 238,620,910 in 2023
Full record7 traps, 1 on the three lines
- Usable rows
- 1,356
- Coverage
- 27 states (26 and the Federal District), 1974 to 2024, with 2024 preliminary. Brazil's herd was 176,388,726 in 2001 and 238,620,910 in 2023
- How it arrives
- One Excel workbook, from a link that works in a browser but refuses other programs
- Language
- Page and file in Portuguese
- Natural· Tree cover loss, Brazil's states· by state name and year
- 1The state name appears once, on the first of its 51 rows, and the cells below it are empty. Fill it down before you sort or filter, or every year after the first belongs to no state.
- 2The number has no heading. It is the last column. The workbook's second sheet, Notas, holds the notes, in Portuguese.
- 3Use a browser. The download link refuses other programs, and on 2026-09-26 the page's CSV option returned an error while the Excel one worked.
- 42024 is preliminary, as the Notas sheet says, and can change at the next release.
- 5"..." means not available. Tocantins starts in 1989 and Mato Grosso do Sul in 1978, because neither existed before; Roraima is missing 1990 and 1993. Nothing is missing from 2001 on.
- 6Counted on 31 December, where the herd is kept at the end of the year, not where the animals grazed or were born.
- 7A cow is not a hectare. A herd can grow on the same pasture by grazing it harder, so more cattle does not have to mean more land cleared.
- Across Brazil's 27 states, is the growth of the cattle herd from 2001 to 2023 related to the share of tree cover each state lost?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Which Brazilian states' cattle herds grew fastest between 2001 and 2023, and are they the states of the Legal Amazon?
- Unidade da Federação: the state, by name, on the first of its rows only
- Ano: the year
- The number of cattle, in the last column, which has no heading. Bovino means cattle and Cabeças head
Across Brazil's 27 states, is the growth of the cattle herd from 2001 to 2023 related to the share of tree cover each state lost over the same years?
Watch outThe state names are spelled identically in both files, accents included, so all 27 match. Stop at 2023: wildfire was 49% of Brazil's tree cover loss in 2024, against 12 to 25% in each year from 2019 to 2023, and IBGE's 2024 herd is preliminary. Compare shares (loss as a percentage of the state's tree cover, the herd's percentage change), not hectares and head counts, or the biggest states simply come top.
Federal prosecutors in Pará, with ranchers and meatpackers (the TAC), and Brazil's four largest meatpackers, JBS, Marfrig, Minerva and Bertin, with Greenpeace (the G4).
Link to this fileDirect: both bar slaughterhouses from buying cattle raised on land cleared after a cut-off date, so they act on the herds this file counts. The file cannot say which animals came from which ranch.
The tensionLaundering. Cattle raised on newly cleared land are sold on through ranches that comply, and the meatpackers have not monitored these indirect suppliers as the agreements required.
SourceThe federal government. In its fifth phase, 13 ministries coordinated by the Ministry of the Environment and Climate Change.
Link to this fileIndirect: the plan is aimed at clearing, not at cattle. Most land cleared in the Amazon becomes pasture, so the herd is where the plan's success or failure should show on the farm side.
The tensionThe 2012 Forest Code (Law 12.651) forgave fines for clearing done illegally before 22 July 2008 and cut how much land had to be restored. Conservationists said it told landowners that illegal clearing would one day be forgiven; Brazil's Supreme Court upheld the amnesty in 2018.
SourceFood waste per person in 29 European countries, at each stage of the food chain, 2020 to 2023
How much food each country throws away, and where: on farms, in factories, in shops, in restaurants or at home. Food that is thrown away took land, water and energy to grow, and the EU's first binding food waste targets are counted in these numbers.
- One row:
- One country in one year, for one stage of the food chain: for example, the kilograms of food waste per person from households in Italy in 2022.
- Covers:
- 2020 to 2023: the 27 EU members and Norway in every year, Iceland in 2022 and 2023, and the EU total. Not Switzerland or the UK
Full record10 traps
- Usable rows
- 114
- Coverage
- 2020 to 2023: the 27 EU members and Norway in every year, Iceland in 2022 and 2023, and the EU total. Not Switzerland or the UK
- How it arrives
- One CSV, from the data browser's download button
- Language
- Page and file in English
- 1Six kinds of row share one number column. Each row is one stage of the food chain (nace_r2), and TOT is the other five added up. Filter to one before you calculate, and never add TOT to the stages. Kilograms per person are rounded, so the stages can differ from TOT by 1 or 2 kg.
- 2Four years, starting in 2020. The EU's common way of measuring food waste began with the 2020 data, so nothing earlier is in the file. Four points are too few for a trend you can rely on.
- 32020 may not be typical. The 2025 directive says that because of COVID-19 measures "the 2020 data are not considered representative of food waste generated in certain Member States". That is why the EU measures its targets from the average of 2021 to 2023.
- 4"Waste collected" is not what it says. Every row says COL, waste collected, but Eurostat's notes say the number is all the food waste generated in the year.
- 5Methods change from year to year. Countries must measure in depth at least once every four years and may use simpler methods in between, and for 2020 they could reuse measurements made from 2017. 158 values are estimated (e), including 30 of the 114 household values.
- 6Some definitions differ. 20 values are flagged d: Spain's households in all four years, Italy's households in 2020 and 2021, and some other stages in Italy, Greece, Latvia and the Netherlands. Leave them out, or say why you kept them.
- 7Food waste includes the parts nobody eats, such as bones and peel. It does not include crops left in the field or fish thrown back into the sea.
- 8A year, not a week. The numbers are kilograms per person per year. Divide by 52 to compare them with a weekly check of your own bin, and remember that a national average is not one household.
- 9Who is in it. The 27 EU members and Norway in every year, Iceland only in 2022 and 2023, and the EU total among them: take out EU27_2020. Switzerland and the UK are not in the file. Greece is EL, not GR: write the three-letter code beside each country before you match a World Bank file.
- 10Slovenia's primary production is 0 in all four years. Say whether you read that as none, or as not measured.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How much household food waste per person did each EU country produce, on average, in 2021 to 2023, the years the EU's 2030 target is measured from?
- In each country, which stage of the food chain wastes the most food per person, 2021 to 2023?
- Across EU countries, is household food waste per person related to GDP per person, 2021 to 2023?
- nace_r2: the stage of the food chain. HH is households, I55_I56_N-S_FOOD restaurants and food services, G46_G47_FOOD shops and other distribution, C10_C11 processing and manufacturing, A01_A03_FOOD primary production, and TOT all five together
- unit: KG_HAB, kilograms per person a year. The link keeps only this; take the filter off for T, tonnes
- geo: the country, code and name in one cell (EL: Greece). TIME_PERIOD: the year
- OBS_VALUE: the number. OBS_FLAG: e estimated, d definition differs
The European Parliament and the Council. Each of the 27 EU member states must meet the targets at national level.
Link to this fileDirect: the targets are counted in this series, and the three starting years are in the file. For the 30% target add HH, I55_I56_N-S_FOOD and G46_G47_FOOD per person. The 10% target is in tonnes, not per person, so use unit T for C10_C11. The file stops in 2023: it shows the starting point, not whether the targets work. Norway and Iceland report, but the targets bind EU members.
The tensionThe European Parliament voted on 13 March 2024 for higher targets: "at least 20% in food processing and manufacturing (instead of 10%) and to 40% per capita in retail, restaurants, food services and households (instead of 30%)". The law kept 10% and 30%. Tourism is a second argument: tourists eat and waste food but are not counted in the population, so the Commission must set a correction for it by 17 October 2027.
SourceEnergy and emissions5
CO2 emissions per person, every country, 1750 onwards
Who emits how much per head rather than in total, across the whole period of industrialisation. Nearly every climate agreement is an argument about this number.
- One row:
- One entity in one year: tonnes of CO2 per person, from fossil fuels and industry. Entity means a country for 214 of them and an aggregate for the other 17.
- Covers:
- 231 entities, 1750 to 2024, of which 214 are countries and 17 are aggregates
Full record6 traps
- Usable rows
- 22,976
- Coverage
- 231 entities, 1750 to 2024, of which 214 are countries and 17 are aggregates
- How it arrives
- One CSV, four columns, from the chart's own URL with .csv on the end
- Language
- Page and file in English
- 1The entity column is not a country column. Seventeen aggregates sit in it beside the 214 countries: World, Africa, Europe, the European Union, and the four income groups. Average the file as it downloads and you have counted the world several times.
- 2The obvious cleaning rule is worse than the problem. Dropping every row whose code starts OWID_ removes the aggregates and also deletes Kosovo, which is a country, carries the code OWID_KOS, and has 31 real rows.
- 3Per-person figures cannot be averaged across countries. In 2000 the World row reads 4.13 tonnes and the unweighted mean of the 214 countries is 5.03. In 2024 they are 4.73 and 4.63, so the error even changes direction. You need population weights, or you need the World row that is already in the file.
- 4It is fossil fuels and industry only: land use change is excluded. A country whose emissions are mostly deforestation looks clean here, and that is a property of the indicator rather than of the country.
- 5The column header is "CO₂ emissions per capita" with a subscript two, which is a different character from the 2 on your keyboard. A lookup or a formula typed as "CO2" will not match it.
- 6The series starts in 1750 for five countries and one of them is Taiwan. It does not reach 190 entities until about 1950, so a long comparison is partly a comparison of who was being counted.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Did the UK's emissions per person fall faster in the fifteen years after the Climate Change Act than in the fifteen before it?
- How do changes in emissions per person compare between EU countries inside the trading scheme and comparable countries outside it?
- How far apart are two countries at similar income, and has the gap closed since 1990?
- CO2 emissions per person, tonnes a year
- Entity, which is a country, a continent, an income group or the world
- Code, the ISO3 code, blank on some aggregates and OWID_-prefixed on others
- Year, 1750 to 2024
The UK Parliament, advised by the Climate Committee. The Act binds the government to a 2050 target and to a series of carbon budgets, each a legal cap on total emissions over five years, which must be set in law twelve years before the period begins.
Link to this fileThe tightest fit available for this file. The budgets are measured on territorial emissions, which is the same accounting basis as this series, and the UK is one of only five countries with data back to 1750. The one step you must explain is comparing per-person figures with the budgets' totals, because population changed too.
The tensionThe budgets count emissions produced in the UK, not emissions caused by it. Everything embodied in imported goods sits outside a target the country is meeting. So a fall that partly comes from moving production abroad (offshoring) still counts as a fall. From the sixth budget onwards the UK's share of international aviation and shipping is finally inside the cap, which shows that people disagreed about where the boundary should be.
SourceThe European Commission with the member states, running the world's first international carbon market: a cap on power stations and heavy industry, with allowances traded between them.
Link to this fileIt covers roughly the part of the economy this series measures, so a comparison between EU members and similar countries outside the scheme is a real design. It is not a per-person instrument, though, so you have to justify the step to per-person figures.
The tensionFree allocation. Some industries are judged at risk of carbon leakage (moving production to somewhere with no carbon price). They receive allowances free instead of buying them, so the sectors that emit most are not the sectors that pay most. The argument continues: in July 2026 the Commission proposed revising the scheme to protect industrial competitiveness alongside the 2040 climate target.
SourceNearly every country, each setting its own nationally determined contribution and revising it on a five-year cycle, with no external body able to set a country's target.
Link to this fileBe careful with this one. Everybody chooses it, and it is the hardest to test: a global agreement with no binding national numbers cannot be shown to have changed any one country's emissions. What this dataset does suit is the fairness (equity) argument the agreement rests on, which is by its nature about emissions per person. Use it for that, and say that is what you are doing.
The tensionCommon but differentiated responsibilities. A country with low emissions per person and a rising total, and a country with high emissions per person and a falling total, both claim the other should move first, and this file is where that argument gets its numbers.
SourceElectricity mix and energy use, every country, 1985 onwards
How countries actually generate their electricity, and how fast that mix moves once a government decides it should move.
- One row:
- One country in one year, with about 130 columns of energy and electricity figures beside it. Any one question uses three or four of them.
- Covers:
- 314 entities, 220 of them countries, years running 1900 to 2025. The electricity columns only start in 1985, and that is the limit that matters.
Full record6 traps
- Usable rows
- 6,551
- Coverage
- 314 entities, 220 of them countries, years running 1900 to 2025. The electricity columns only start in 1985, and that is the limit that matters.
- How it arrives
- One CSV of about 9 MB, 130 columns, with a codebook CSV beside it in the same repository
- Language
- Page and file in English
- 1A third aggregate convention, different again. 94 of the 314 entities are aggregates, and here they have a blank iso_code rather than an OWID_ one. Worse, Africa appears five times: Africa, Africa (EI), Africa (EIA), Africa (Ember) and Africa (Shift), because four compilations are stacked in one file. Group by continent without filtering and you count it five times over.
- 2113 of the 127 data columns are less than half filled. The file looks enormous and most of it is empty for any particular country-year. Check your chosen column's coverage before you design around it, not after.
- 3The years say 1900 and the electricity says 1985. Primary energy reaches back much further than generation by source, so a question about the mix quietly starts in 1985 whatever the file's range suggests. That is 6,551 usable country rows, not 23,377.
- 4A share and a quantity are different findings. A renewable share can rise because renewables grew or because demand fell, and 2020 is in this series, so at least one year moved for reasons that have nothing to do with energy policy.
- 5The change columns are already calculated. Recalculate them yourself: it is one formula, it is the processing Criterion D is asking to see, and a column you did not compute is a column you cannot explain.
- 6The codebook is a separate file in the same repository. It is the only thing that tells you which source each column came from, and the licence asks you to name that source.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How have the electricity mixes of Germany and France diverged since 2000, and what did renewables replace in each?
- Which EU member states are on course for 42.5% renewables by 2030 at their rate of change since 2015?
- Does a rising renewable share track rising renewable generation, or falling total demand?
- Share of electricity from renewables, fossil fuels and nuclear, %
- Generation by source in terawatt hours: coal, gas, oil, hydro, wind, solar, nuclear, other
- Primary energy consumption, total and per person
- Population and GDP columns, so you can normalise without going to another file
- Year-on-year change columns, already calculated, which you should recalculate rather than trust
The German federal government. The 2000 act guaranteed any renewable generator a grid connection, priority on the grid over conventional power stations, and a fixed price per kilowatt hour for twenty years. It is the policy most often credited with making solar panels cheap for everyone else.
Link to this fileA strategy is rarely this visible in a dataset. Germany's renewable share of electricity runs 6.2% in 2000 and 58.6% in 2024 in this file, while nuclear goes from 29.8% to exactly zero. The interesting question is not whether renewables grew but what they replaced, and the fossil column answers it: 64.0% to 41.4% over the same twenty-four years.
The tensionIt was paid for by a surcharge on electricity bills, with electricity-intensive industry and the railways largely exempted, so households paid for a benefit that everyone shared. Two environmental goals also collide in the same series: leaving nuclear and leaving coal could not both be done first.
SourceThe French state, through EDF. The plan announced by Prime Minister Pierre Messmer in March 1974, after the oil embargo exposed how much imported petroleum the country ran on, built 56 reactors in about fifteen years. The fleet is now governed by the programmation pluriannuelle de l'énergie.
Link to this fileThe best contrast with Germany, because the two countries decarbonised their electricity in opposite ways over the same decades and this file holds both. Comparing them is a real argument about methods, not a ranking.
The tensionAn ageing fleet can fail all at once. Stress corrosion cracking found from late 2021 put twelve reactors into inspection outages and cut EDF's output to 279 TWh in 2022, against 361 the year before and the lowest since 1992. That turned a low-carbon system into an importer in a year when electricity was scarce. Waste and the cost of life extension sit behind that.
SourceThe European Union, binding member states collectively rather than individually: the target is for the Union, and each country contributes through its own national plan.
Link to this fileIt sets the target this file lets you measure progress towards, so it suits a question about which member states are on course. Be careful with the units: the directive's target is a share of all energy, and the columns most students reach for are shares of electricity, which is a smaller and faster-moving thing.
The tensionA collective target against each country's right to choose its own energy mix. France argues that nuclear should count towards decarbonisation targets it does not currently count towards. That is a disagreement about what the word renewable is for.
SourceAtmospheric CO2 at Mauna Loa, monthly since 1958
The concentration of carbon dioxide in the atmosphere, measured continuously in one place for longer than anywhere else on Earth.
- One row:
- One month at one observatory: the mean CO2 concentration in parts per million.
- Covers:
- 822 monthly values, March 1958 to August 2026, from a single station in Hawaii
Full record6 traps
- Usable rows
- 822
- Coverage
- 822 monthly values, March 1958 to August 2026, from a single station in Hawaii
- How it arrives
- One CSV of about 40 KB, with forty comment lines before the header
- Language
- Page and file in English
- 1Forty comment lines before the header. Point a spreadsheet at the file and its first data row is a sentence about the data policy.
- 2Missing values are −9.99 and −1, not blanks. The standard deviation column reads −9.99 in 196 of the 822 rows and the days-behind-the-mean column reads −1 in 195, all in the early decades when those were not recorded. Average either column as it downloads and you get a negative standard deviation, which cannot exist.
- 3There are two CO2 columns and they answer different questions. One is the monthly mean, the other has the seasonal cycle removed. Take the deseasonalised one for a trend and the raw one for the cycle, and say which you used, because a reader cannot tell from a chart.
- 4The series starts in March 1958, not January, so the first year is incomplete and a naive annual mean for 1958 is a mean of ten months.
- 5The last few months are provisional and change. Record the date you downloaded it; two students downloading in different weeks do not have the same file.
- 6One observatory on one mountain in the Pacific. It is the cleanest CO2 record in the world and it is still one place, so a claim about the global atmosphere rests on an argument about why that site represents it, which the laboratory makes and you should cite.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has the rate of increase changed between the first twenty years of the record and the last twenty?
- How large is the seasonal cycle, and has its amplitude changed since 1958?
- What does the curve do around 2020, when global emissions fell sharply, and what does that tell you about concentration against emission?
- The monthly mean, in ppm: 315.71 at the start of the series and 427.55 at the end
- A deseasonalised version of the same month, with the annual cycle removed
- Year, month, and a decimal date for plotting
- The number of daily means behind each monthly value, and its standard deviation
Nearly every country, each setting its own nationally determined contribution and revising it every five years. No external body can set a country's target.
Link to this fileRead this one carefully, because the reasoning here is the point. This series is the global concentration: it responds to the sum of everything everyone emits, so no single agreement can be shown to have changed it, and the curve passes through 2015 without a change of direction. That is not a reason to avoid the pairing. It is the finding, and stating it precisely is worth more than a correlation.
The tensionCommon but differentiated responsibilities: who cuts first, and who pays for it, argued between countries whose emissions per person differ by a factor of fifty.
SourceThe UK Parliament, advised by the Climate Committee, with a legal cap on total emissions across each five-year period.
Link to this fileA national law against a global concentration, which is the mismatch worth writing about: the UK is about 1% of global emissions, so even a law that works perfectly is invisible here. Pair this series with the ONS industry file, which is the scale at which the law can actually be checked.
The tensionTerritorial accounting. Emissions embodied in imports sit outside a budget the country is meeting, and this curve counts them wherever they happened.
SourceEvery country in the world, controlling ozone-depleting substances rather than CO2.
Link to this fileHere on purpose, and not because it acts on CO2. It is the case where an agreement did change a concentration, and the ozone file on this page shows it happening. Comparing that series with this one is the clearest way to show what is different about carbon dioxide: a few industries made the ozone chemicals, but almost everything produces CO2.
The tensionThe price of universal agreement was a grace period for developing countries and a fund to pay for their compliance. The climate negotiations have never managed to agree a deal like that at the same scale.
SourceUK greenhouse gas emissions by industry
Which parts of an economy actually produce its greenhouse gases, and which of them have cut anything over thirty-five years of policy aimed at them.
- One row:
- One industry in one year: emissions in thousand tonnes of carbon dioxide equivalent. The sheet stores it as a matrix, so one spreadsheet row is a year and one column is an industry.
- Covers:
- 158 industry columns, 35 years from 1990 to 2024, on each of eight gas sheets
Full record6 traps
- Usable rows
- 5,460
- Coverage
- 158 industry columns, 35 years from 1990 to 2024, on each of eight gas sheets
- How it arrives
- One xlsx of about 9 MB, eleven sheets, one per gas plus a total
- Language
- Page and file in English
- 1It is a matrix, not a table. Years run down and 158 industries run across, so before you can ask anything you have to unpivot it into one row per industry-year: 35 by 156, about 5,460 rows.
- 2There is text inside the numbers. `[low]` appears 4 times and means a small figure that is not a real zero; `[u]` appears 12 times and marks estimates the ONS considers unreliable after a methodological revision. A spreadsheet treats a column containing either as text, and may leave those cells out of an average without warning.
- 3UK residence basis, not territorial. This counts emissions by UK-resident economic units, including UK-owned ships and aircraft operating abroad, which is not the basis the UK's own carbon budgets are measured on. It is the single best evaluation point in the file and most students never notice it.
- 4Two of the 158 columns are consumer expenditure, travel and non-travel. They are households, not industries, and adding them in does not count anything twice, but it does mix two different kinds of thing.
- 5Eight sheets hold eight gases, and they are not interchangeable: methane and CO2 are already converted to CO2 equivalent, so adding the gas sheets to the total sheet counts everything twice.
- 6The last year is provisional. An industry that looks like it fell sharply in 2024 may later be revised, which is the reason the release date on the cover sheet belongs in your method.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- Which industries cut emissions most between 1990 and 2024, and which have barely moved?
- Do the industries inside the UK Emissions Trading Scheme fall faster than comparable industries outside it?
- How much of the fall in total UK emissions is electricity supply alone?
- Total greenhouse gases, in thousand tonnes of CO2 equivalent
- The same for each of seven gases on its own sheet: CO2, CH4, N2O, HFC, PFC, SF6, NF3
- 158 industries, from SIC sections down to individual groups
- Two consumer-expenditure columns, travel and non-travel, which are households rather than industries
Parliament, advised by the Climate Committee. Each budget is a legal cap on total UK emissions across five years, and it must be set in law twelve years before the period starts.
Link to this fileThe natural pairing, and the mismatch is the lesson. The budgets are territorial totals; this file is by industry on a residence basis. A student who adds up the industry columns and compares the total with a carbon budget is comparing two different accounting systems, and has to say so.
The tensionA cap on the country that no single industry is responsible for. Everything embodied in imports sits outside a target the UK is meeting, and the industries that cut most are not always the ones that were asked to.
SourceThe UK government and devolved administrations, capping emissions from power stations, energy-intensive industry and aviation, and making them buy allowances to emit.
Link to this fileIt acts on a nameable subset of these columns rather than all of them, which is what makes it useful here: you can compare the industries inside the scheme with those outside it, in the same file, on the same basis. Note the start date, though. 2021 leaves four years of data after it, which is a short period to compare.
The tensionFree allowances to industries judged at risk of moving abroad, so the sectors that emit most are not the sectors that pay most, against the households who pay through their bills.
SourceThe statutory body that recommends each budget's level and reports to Parliament every year on whether the policies exist to meet it. It cannot legislate; it can only publish.
Link to this fileAn indirect one, and worth having for that reason: this is the strategy that produces the argument rather than the emissions. Its reports name the sectors that are off track, which gives you a reason to have chosen your industry column rather than a preference.
The tensionAn independent body whose advice a government may decline, publicly, with reasons. That is a political tension with an environmental consequence you can measure in these columns.
SourceRenewable energy projects in the UK planning system, one by one, since 1990
Where wind and solar farms get built, and where they are refused. Every renewable electricity project in the UK above a minimum size is listed, with where it is, how big it is, who decides and what they decided.
- One row:
- One renewable electricity or storage project in one planning application: its technology, its size in megawatts, where it is, who decides, and the date it reached each stage, from application to operation.
- Covers:
- 14,657 projects in England, Scotland, Wales and Northern Ireland, 7,460 of them solar and 2,748 onshore wind, with applications from 1990 to 2026
Full record12 traps
- Usable rows
- 10,197
- Coverage
- 14,657 projects in England, Scotland, Wales and Northern Ireland, 7,460 of them solar and 2,748 onshore wind, with applications from 1990 to 2026
- How it arrives
- One CSV (4.85 MB), 53 columns. The same table comes as an Excel file (4.95 MB) with a Definition Sheet that explains every column and every stage.
- Language
- Page and file in English
- 1It is not UTF-8. The CSV is saved as Windows-1252, so Python stops with an error at byte 5,497 unless you open it with encoding="cp1252". 183 non-breaking spaces hide in site names, addresses, planning references and coordinates. The Excel version avoids the first problem and has the Definition Sheet.
- 2Development Status is only the latest stage. A wind farm refused by the council and then allowed on appeal reads Appeal Granted, or Operational once it is built: 204 onshore wind rows have both a council refusal date and an appeal granted date. 591 onshore wind rows are refused now, but 923 carry at least one refusal date. To count refusals, count the date columns, and say which ones.
- 3One stage, three spellings, and a heading with two spaces. Planning Permission Granted appears 5,709 times, Planning permission Granted 8 times and Planning permission granted 6 times. The date column is headed Planning Permission Granted, with two spaces, so a formula or code that types one space will not find it.
- 4The minimum size changed in 2021. Until 2021 only projects of 1 MW or more were listed; since then the line is 150 kW. 3,921 rows under 1 MW were applied for from 2021 and only 277 before, so a count that crosses 2021 grows partly because smaller projects started to be listed. Keep to projects of 1 MW or more for any series that crosses 2021.
- 5Some projects have more than one row. 826 rows are Revised: the developer sent in a new application for the same project, and that has its own row. Keep both to count applications; drop the Revised rows to count projects. 13 applications also appear twice under different Ref IDs, with the same name, size and council, 8 of them wind or solar: remove the second of each. One pair, Hagshaw Hill Wind Farm, has no planning reference, so a check on the reference alone misses it. And a solar farm with a battery is often two rows with one planning reference (562 references are solar plus battery).
- 6Many solar farms are 49.9 MW. 258 solar rows are exactly 49.9 MW, 246 of them in England, and only 8 are between 50 and 51 MW. In England a solar farm over 50 MW was decided by national government, not the council, until 31 December 2025, when the line moved to 100 MW. So many projects seem to be sized to stay with the council: a question about size has to say so.
- 7The coordinates are metres on the British National Grid, not latitude and longitude, and that includes Northern Ireland, which usually uses the Irish Grid. 36 rows have none. One Northern Ireland wind farm (Ref ID 21008) is written in Irish Grid and lands near Wrexham if you read it as British. 9 Scottish battery rows have the two numbers swapped and land in the North Sea. In the Excel version 72 X and 88 Y values are stored as text, mostly because they start with a tab or a hidden space.
- 8Dates are text, and some have not happened yet. All are day/month/year except one refusal written 4/17/2026 (Ref ID 20519). Operational says NA 16 times. 7 granted dates fall after the end of June 2026, the latest in December 2026. Planning Permission Expired is, in the Definition Sheet's words, the "Date a planning permission expires (as per the planning decision)", so 2,218 of those dates are still to come.
- 9Solar Site Area is empty for more than half the solar rows. It is filled in 3,586 of 7,460, and some values cannot be a solar farm: one ground-mounted site covers 66 m². Work out hectares per MW on ground-mounted rows only, and look at the smallest and largest before you trust an average. The type of farmland, which the government said in May 2024 it would add to this database, is not in the file.
- 10Who decides is not always a council. 639 rows name Scottish Government (S36) as the planning authority, 129 The Planning Inspectorate - National Infrastructure and 90 Welsh Government (NSIP). A question about councils has to leave these out, and say so.
- 11Not every row is one UK site. Two solar farms in Jersey are filed as South West England (Ref IDs 17044 and 17120), and one row is 169 Sainsbury's stores with no country and no coordinates (Ref ID 1613). Remove them.
- 12It is compiled, and it changes every three months. Barbour ABI builds it for the government, mostly "from the web and developers", and asks people to report projects it has missed, so an older project can appear or change in a later edition. 1,345 rows have no capacity. Record which edition you used, and never mix two.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of onshore wind applications made each year in England change after 18 June 2015, and again after 8 July 2024, compared with Scotland over the same years?
- Is the share of ground-mounted solar applications refused by councils different between the regions of England, for applications made from 2015 to 2024?
- How much solar and onshore wind capacity, in MW, began operating in each region of the UK each year from 2010 to 2025?
- How many hectares does each megawatt of ground-mounted solar take up in England, and is it different for solar farms applied for before 2016 and from 2021?
- Technology Type: 26 kinds, among them Wind Onshore and Solar Photovoltaics
- Installed Capacity (MWelec): the project's size in megawatts
- Development Status: the latest stage the project has reached, such as submitted, granted, refused, withdrawn, appeal refused or operational
- Planning Application Submitted, Planning Permission Refused, Planning Permission Granted, Appeal Refused, Appeal Granted, Operational and other date columns: the day each stage was reached
- Region, Country and Planning Authority: where the project is, and who decides
- X-coordinate and Y-coordinate: the site on the British National Grid, in metres
- Solar Site Area (sqm) and Mounting Type for Solar: the land a solar project covers, and whether it is on the ground, on a roof or floating
- No. of Turbines and Height of Turbines (m): for wind farms
The UK Government, for England only. Greg Clark, Secretary of State for Communities and Local Government, set it out in a written statement to Parliament, and the National Planning Policy Framework carried it as a footnote (footnote 54 in the July 2021 version). Councils applied it to each application.
Link to this fileDirect, and the file has years on both sides of both dates. Every onshore wind row has a country and most have a date of application, so England can be compared with Scotland, Wales and Northern Ireland, where the statement did not apply. But the subsidy for renewables, the Renewables Obligation, also closed in these years: to large solar on 31 March 2015, to small solar on 31 March 2016, and to onshore wind in England, Scotland and Wales on 12 May 2016 (Ofgem). Scotland lost the subsidy too, and solar never faced the statement, so look at both before you say the statement caused any change.
The tensionLocal control against the national need for energy. Greg Clark made the statement "so that local people have the final say on wind farm applications". Announcing its removal on 8 July 2024, the Energy Secretary, Ed Miliband, said: "Every family has paid the price of the ban on onshore wind farms in higher energy bills."
SourceThe Ministry of Housing, Communities and Local Government, for England. Councils and planning inspectors must take the Framework into account in every decision. Since 31 December 2025, onshore wind and solar projects over 100 MW are decided by national government instead of the council.
Link to this fileDirect in purpose, but very recent. The July 2026 edition holds about eighteen months of decisions under the December 2024 Framework and none under the August 2026 one. Use it to describe what changed and the years before it, and say that the years after it are few.
The tensionClean energy against habitats and landscapes. Asked in 2024 whether habitats such as peat should be better protected from renewable energy development, "1,314 (90%) said they agreed with additional protections for habitats, such as those containing peat soil", but "Some developers felt that additional measures could be a barrier to development". The government chose to review its definition of irreplaceable habitats later. On 24 June 2026 CPRE, the countryside charity, warned that the draft of the new Framework dropped its "valued landscape" policy, and its Policy Lead, Emma Marrington, said that removing it "silences the everyday landscapes that matter most to local communities". The August 2026 Framework does not use the words. CPRE's warning was about all development, not wind or solar in particular.
SourceThe Department for Energy Security and Net Zero, for Great Britain (Northern Ireland is not included). It sets out how much of each kind of generation should be running by 2030. The Solar Roadmap of 30 June 2025 sets out how the solar part is to be reached.
Link to this fileDirect on the people side: the plan is about how much gets built, and this file records each project's capacity and the day it began to operate. But the file lists only projects over 150 kW (1 MW before 2021), so the panels on people's houses, which are far smaller, are not in it, and the plan counts them. Leave out Northern Ireland, and compare how fast capacity is added, not the totals.
The tensionSolar farms against farmland. In May 2024 the Energy Secretary, Claire Coutinho, told Parliament: "we are concerned that as large solar developments proceed at pace, more of our 'Best and Most Versatile' (BMV) land could be used for solar PV instead of food production." The next government's Solar Roadmap said: "we do not believe that increased solar deployment poses a threat to food security." CPRE replied on 2 July 2025: "Most people don't want to see mega solar farms swamping and transforming the character of rural places."
SourceThe Scottish Government. Councils decide most applications; the rest are decided by the Scottish Government itself, and the file names it as the planning authority: Scottish Government (S36).
Link to this fileDirect, and a useful contrast. Scotland never had England's 2015 statement, so the two countries' onshore wind rows can be compared over the same years. The years after this statement are few: 2023 to the middle of 2026 in the July 2026 edition.
The tensionWind energy against wild landscapes. The statement itself says that reaching 20 GW "will require taller and more efficient turbines. This will change the landscape." Responding to the draft on 12 January 2022, the John Muir Trust, a wild land charity, said: "The huge potential for offshore means there need be no rush towards accelerating the construction of new large scale wind developments across Scotland's Wild Land Areas." The draft had asked about 8 to 12 GW of additional capacity; the final statement set 20 GW in total.
SourcePeople and development5
GDP per person in every country, every year since 1990, at 2021 prices
Economic growth: the value of what each country produces for each person who lives there, a number often set against harm to nature.
- One row:
- One country in one year: the value of everything produced there (its GDP), divided by the number of people who live there, in international dollars at 2021 prices.
- Covers:
- 199 countries with values, 1990 to 2025, among 265 rows that include 47 aggregates such as World and Low income. In 2015, from Burundi's $992 to Luxembourg's $130,097 a person
Full record10 traps, 1 on the three lines
- Usable rows
- 6,977
- Coverage
- 199 countries with values, 1990 to 2025, among 265 rows that include 47 aggregates such as World and Low income. In 2015, from Burundi's $992 to Luxembourg's $130,097 a person
- How it arrives
- A zip of three CSVs: the data is the one whose name begins API_, and the two Metadata files say which rows are countries
- Language
- Page and file in English
- Natural· North Atlantic shortfin mako caught· by country and year
- Natural· Tree cover loss· by country, by name
- 1Four lines of notes sit above the header. The header is the fifth line and begins Country Name, and there is an empty column after the last year.
- 2Aggregates sit among the countries: 47 rows such as World, Euro area and Low income. The zip's Metadata_Country file leaves Region blank for every one of them, so use it to take them out.
- 3Years run across. One column per year: turn them into one column before you match the file to anything with a year column.
- 4Values from 1990 to 2025 only. The columns for 1960 to 1989 are empty. 14 countries start later, among them Afghanistan (2000), Kosovo (2008) and Djibouti (2013), and the last years have fewer countries: 199 have a value in every year from 2013 to 2022, then 197 in 2023, 195 in 2024 and 185 in 2025. Check your countries before you choose a year.
- 518 countries have no value in any year: Cuba, Eritrea, North Korea, South Sudan, Venezuela, Yemen and 12 smaller places such as Monaco and Guam. An empty cell is not zero: leave them out, and say so.
- 6International dollars at 2021 prices. PPP (purchasing power parity) allows for what things cost in each country, so one international dollar buys about the same everywhere. Every year is valued at 2021 prices, so a rise means more was produced, not that prices went up. You could not exchange these dollars for anything.
- 7Values change from one release to the next, old years too. In May 2024 every value changed when the World Bank moved the series to 2021 prices, and countries revise their own figures between releases. Use one download for every value, and cite its Last Updated Date.
- 8What is produced, not what people earn. GDP counts everything produced in a country, whoever takes the money. Ireland's rises by 23% from 2014 to 2015 (65,463 to 80,816), and Ireland's statistics office says much of its GDP is "profits that are generated here but then go straight out to the owners of companies abroad". Check any country that jumps before you rely on it.
- 9An average, and it may trend. One number per person says nothing about how income is shared, or about the people who fish or farm. It can also change steadily over the years, so across years it may only follow time: compare countries within one year, or correlate the change from one year to the next as well. Say what you think GDP stands for in your question.
- 10The environmental Kuznets curve is an idea, not a plan. It proposes that harm to the environment rises as a country's income grows, then falls once it is rich enough. You can use it to explain what you find, but nobody carries it out, so it cannot be your strategy.
- Across the countries with at least 1 million hectares of tree cover, is GDP per person in 2015 related to the share of tree cover lost from 2016 to 2025? (Tree cover loss.)
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How has GDP per person in your own country changed since 1990, and how does that compare with the World row in the same file?
- Country Name, and Country Code, the three-letter ISO code
- 1960 to 2025: one column per year, empty before 1990
Across the countries whose fleets catch North Atlantic shortfin mako, is the tonnage caught related to the country's GDP per person, 1990 to 2021?
Watch outICCAT names each fleet's flag in words (EU-España, Maroc, UK-Bermuda) and the World Bank uses three-letter codes, so write the code beside each flag yourself. Three flags have no GDP to match: Chinese Taipei (not in the World Bank's list), Venezuela (no value in this series) and St Pierre et Miquelon. That leaves 413 country-years with a mako catch above zero and a GDP value, from 30 countries, but only 4 of them in 1990 and 11 to 22 a year from 2006, so pool the years or compare within one. Add dead discards to landings, and stop at 2021 or say how you handled the 2022 ban. GDP per person does not catch sharks: any link runs through the size and reach of a country's fleet, and a big fleet from a middle-income country can catch more than a small one from a rich country. Say what you think the link runs through.
Across the countries with at least 1 million hectares of tree cover, is GDP per person in 2015 related to the share of tree cover lost from 2016 to 2025?
Watch outMatch by hand, as for protected land: GFW names countries and never codes them, 13 of the 109 are spelled differently in the World Bank file, and French Guiana and Taiwan are not in it. Five more have no GDP in any year (Cuba, New Caledonia, North Korea, South Sudan and Venezuela), which leaves 102. Filter GFW to one canopy threshold and add together China's, India's and Pakistan's extra rows first, and divide each country's loss by its own tree cover. GDP per person does not clear forest: any link runs through what a country's land is used for and who buys what it grows, so say what you think it runs through.
Every UN member state. The UN Statistics Division keeps the indicator for target 8.1, the yearly growth of real GDP per person. UNEP, the UN Environment Programme, keeps the two for target 8.4, which count the materials a country uses, in total, per person and for each unit of GDP.
Link to this fileDirect for target 8.1: its indicator is the change in this file's measure from one year to the next. Indirect for target 8.4: it is about growth and harm to nature together, and this file is only the growth half. Its own indicators, materials for each unit of GDP, are not in the file, so set GDP against a natural column from another card, and say how your question tests the target.
The tensionThe two targets can conflict. Jason Hickel argued in the journal Sustainable Development in 2019 that global growth of 3% a year would make any fall in the world's use of resources impossible in practice, and wrote that "Goal 8 violates the sustainability objectives of the SDGs".
SourceShip arrivals at each European country's main ports, every year since 1997
How much shipping reaches each country's ports: the traffic that carries invasive species around the world, and one side of the argument about controlling it.
- One row:
- One country in one year, for one type of vessel and one unit: for example, the number of vessels arriving in Greece's main ports in 2019.
- Covers:
- 1997 to 2024, 26 countries and 5 EU groups, with series starting in different years: Croatia 2005, Cyprus 2002, France 1998, Greece and Italy 2000, Malta 2003, Türkiye 2007
Full record7 traps, 1 on the three lines
- Usable rows
- 613
- Coverage
- 1997 to 2024, 26 countries and 5 EU groups, with series starting in different years: Croatia 2005, Cyprus 2002, France 1998, Greece and Italy 2000, Malta 2003, Türkiye 2007
- How it arrives
- One CSV, from the data browser's download button
- Language
- Page and file in English
- Natural· Lionfish records· by country, and the year of the first lionfish record
- 1Some rows are percentages, not counts. 13,330 rows of the full file are the change on the year before, in the same number column. Filter unit to NR first.
- 2An arrival is a port call, not a ship. Every ferry trip counts: Greece had 488,194 arrivals in 2024, Cyprus 2,147.
- 3Jumps with no warning. Malta goes from 2,894 (2007) to 23,248 (2008); Greece jumps in 2002 when ferries started to be counted; Spain rises in 2011 when regional ports were added. The flag column is empty in every row.
- 4A missing year has no row. It is simply absent from the CSV, not blank.
- 5EU totals sit among the countries. Five EU groups add 87 rows of total arrivals. Take them out.
- 6One number covers every coast. France includes its Atlantic ports and overseas departments, Spain includes Ceuta and Melilla, and only main ports are counted.
- 7Codes that do not match other files. Greece is EL, not GR, and Turkey is Türkiye.
- Is the year lionfish were first recorded in a Mediterranean country related to the number of vessels arriving at its main ports?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of vessels arriving in Cyprus and Malta change between 2005 and 2024?
- unit: NR is vessels arriving, THS_GT is gross tonnage in thousands, RT_PRE_NR and RT_PRE_GT are percentage changes
- vessel: the type of ship, with TOTAL for all of them
- rep_mar: the country, code and name in one cell (EL: Greece)
- TIME_PERIOD: the year. OBS_VALUE: the number
Is the year lionfish were first recorded in a Mediterranean country related to the number of vessels arriving at its main ports?
Watch outOnly 8 of the 15 lionfish countries have a traffic series: Israel, Lebanon, Egypt, Tunisia, Libya, Albania and Syria are not in Eurostat. Match by hand, because Eurostat writes Greece as EL and Turkey as Türkiye. The first records of Greece (1995) and Turkey (2005) fall before their traffic series begin, so only 6 first-record years have a traffic figure. That is a small sample: describe it rather than lean on a test, and say that first-record years rest on single early records. Lionfish swam in through the Suez Canal; shipping is not how they arrived.
The International Maritime Organization and the flag states that have joined it.
Link to this fileIt controls the ballast water of exactly the ships this file counts. It does not bear on the lionfish, which swam in through the Suez Canal rather than riding in ballast: say so if you pair them.
The tensionThirteen years passed between adoption and entry into force; the IMO says it took approved treatment technology to remove the main barriers to countries ratifying it.
SourceThe European Commission and member states.
Link to this fileMember states must act on listed species and on the pathways they travel by, shipping among them. Lionfish are not on the list, so they carry no listed-species duty.
The tensionKleitou and colleagues (2021) describe getting lionfish onto the list as lengthy and still ongoing, while the fish spread.
SourceBoats registered in every Swiss canton, every year since 1980
How many private boats each canton registers: the boats that carry invasive quagga mussels from lake to lake, and that new cleaning rules now target.
- One row:
- One canton in one year's sheet: the boats registered there on 30 September, by kind of boat.
- Covers:
- 1980 to 2025, 26 cantons, 46 sheets
Full record6 traps, 1 on the three lines
- Usable rows
- 1,193
- Coverage
- 1980 to 2025, 26 cantons, 46 sheets
- How it arrives
- One Excel workbook with one sheet per year (a German version exists too)
- Language
- Page and file in French
- Natural· Quagga mussels recorded (yes or no)· by canton
- 1One sheet per year. There are 46 sheets; stack the ones you need into one table first.
- 2The Total column moves. It is column G from 1980 to 2013, I from 2014 to 2018 and B from 2019. In the 2013 sheet, column B is motor boats only.
- 3Regions sit between the cantons. Seven regional subtotals (such as Région lémanique) and a Switzerland total are mixed in. Adding every row counts boats twice.
- 4The kinds of boat were redrawn in 2014, and Schwyz dropped that year because of a change of computer system.
- 5Two kinds of missing. Both … and ... appear, and three canton totals are not numbers (Jura 1980 and 1981, Basel-Landschaft 1983).
- 6Boats are counted where they are registered, not where they sail. Boats on Lake Geneva are split between Vaud, Geneva and Valais.
- Have quagga mussels been recorded in more of the Swiss cantons with many registered boats than of those with few, 2016 to 2025?
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the number of registered motor boats change in the cantons around Lake Geneva between 1980 and 2025?
- Total: all registered boats
- Bateaux à moteur, Bateaux à voile, Bateaux à rames, pédalos: motor boats, sailing boats, rowing boats and pedalos
- Bateaux à passagers / de marchandises / autres: passenger, goods and other boats, from 2014
- Avec moteur électrique: boats with an electric motor, from 2019 and only in some cantons
Do Swiss cantons where quagga mussels have been recorded have more registered boats than cantons where they have not?
Watch outThe mussel file writes cantons as codes (VD) and the boat file as names (Vaud): add one to the other. All 18 cantons with a record match a boat row; 8 cantons have none (AI, AR, GL, GR, JU, NW, TI, UR). Choose one year's boat sheet, take out the regional subtotals, and say which year. Boats are counted where they are registered and mussels on 5 km squares with no lake named, and the Rhine cantons' records may come from the river rather than a lake.
Each canton sets its own rules.
Link to this fileDirect: it applies to registered boats, which are exactly what this file counts. Canoes and paddleboards are only advised to clean, and they are not in the file either.
The tensionRTS reports that certified cleaning costs 300 to 1,000 francs, and Zurich itself calls its cleaning duty a relaxation of the earlier ban: boat owners' costs against keeping mussels out.
SourceAccess to clean cooking fuels, every country, 2000 onwards
Household air pollution from cooking on wood, dung, charcoal, coal and kerosene, which is among the largest environmental causes of early death, and how fast people are leaving it.
- One row:
- One country in one year: the percentage of the population whose main cooking fuel and technology are clean ones.
- Covers:
- 215 entities, 201 of them countries, 2000 to 2024, with no missing years
Full record6 traps
- Usable rows
- 5,025
- Coverage
- 215 entities, 201 of them countries, 2000 to 2024, with no missing years
- How it arrives
- One CSV, four columns, from the chart URL with .csv on the end
- Language
- Page and file in English
- 1Every value is a whole number. India reads 34, 36, 39, 42 and so on: the file is rounded to the nearest percentage point, so any change smaller than one point is invisible and a year-on-year difference is a rounded difference.
- 2643 rows sit at exactly 100. That is a ceiling, and it means a question about which countries improved most silently excludes every country that had already arrived.
- 3It is modelled, not counted. India climbs from 34% in 2010 to 81% in 2024 by three or four points every single year, with no step anywhere. These are estimates from a statistical model fitted to surveys, so smoothness is a property of the method. No policy will show up as a kink, and pretending otherwise is the most likely way to lose marks on this file.
- 4The aggregates are different here from the other Our World in Data files: fourteen of them, mostly UN regions with a blank code, plus World. A cleaning rule written for the CO2 file catches one of the fourteen.
- 5The value column is 123 characters long and ends "Residence area type: Total", which is telling you that urban and rural versions exist and you have neither.
- 6A percentage of a population is not a number of people. A country at 40% with a growing population can have more people cooking on wood every year while its figure improves, and that is a genuinely different finding.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How does India's rate of improvement since 2016 compare with Bangladesh, Nepal and Pakistan over the same years?
- Which countries have moved least since 2000, and what do they have in common?
- Does the rate of improvement track national income, using the World Bank's series for the second variable?
- Share of the population relying mainly on clean fuels for cooking, %
- Entity, code and year
- Nothing else. It is a one-variable file, which is why it wants pairing with an income or an energy series
The Government of India through the Ministry of Petroleum and Natural Gas and the state oil companies. It gives a free gas connection, in the name of a woman of the household, to families below the poverty line, replacing wood and dung at the point of use.
Link to this fileThe best-documented national strategy this indicator has, and the data will not show it as a step. India's series climbs three or four points a year on both sides of 2016. That is worth knowing before you build a before-and-after design on it: what you can honestly ask is how India's rate of change compares with its neighbours' over the same years, and that comparison gives you your evaluation.
The tensionA connection is not a fuel supply. Refills have to be bought at full price with the subsidy arriving afterwards, so reported refill rates for these households run well below the seven or eight cylinders a year that cooking on gas alone needs, and a 2026 CEEW study reported that under a quarter of rural households use it exclusively. Free firewood competes with a cylinder that costs money today.
SourceEvery UN member state, with progress tracked jointly by the WHO, the World Bank, the IEA, IRENA and the UN Statistics Division. This indicator is the one the target is scored on, which is why the series exists at all.
Link to this fileHere the strategy and the measurement are the same thing. So there is nothing to infer and no independent test: you cannot ask whether the target changed the number that defines the target. Use it to frame a question about who will and will not arrive by 2030, which is a projection question and a legitimate one.
The tensionClean cooking attracts a small fraction of the finance that goes to electricity access, despite the health burden being larger, so two halves of the same goal compete for the same money.
SourceThe World Health Organization, recommending against unprocessed coal and kerosene in the home and setting emission rate targets for household fuels and stoves.
Link to this fileIt is the definition rather than an intervention: it decides which side of the line a household falls on, so it explains what the variable means more than it explains what moved it. Good for the method section, weak as the strategy you evaluate.
The tensionA standard with no money attached does not change what a household can afford. Kerosene is ruled out by the guidelines and remains the cheapest lit fuel available in many of the places that use it.
SourceMunicipal waste per person in 37 European countries, and what happens to it, every year since 1995
Where each country's household rubbish goes: recycled, composted, burned or buried. The European Commission says the main environmental threat from food and garden waste is the methane it gives off as it rots in landfill, and the EU's recycling and landfill targets are counted in this table.
- One row:
- One country in one year, for one thing that happens to its waste: for example, the kilograms of municipal waste per person that Greece landfilled in 2022.
- Covers:
- 1995 to 2024, 37 countries and the EU total: the 27 EU members, Iceland, Norway, Switzerland, Türkiye, Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia. Not the UK. Series start later in Croatia (1997), Serbia (2006), Bosnia and Herzegovina and North Macedonia (2008), Montenegro (2011), Albania (2013) and Kosovo (2015)
Full record10 traps
- Usable rows
- 993
- Coverage
- 1995 to 2024, 37 countries and the EU total: the 27 EU members, Iceland, Norway, Switzerland, Türkiye, Albania, Bosnia and Herzegovina, Kosovo, Montenegro, North Macedonia and Serbia. Not the UK. Series start later in Croatia (1997), Serbia (2006), Bosnia and Herzegovina and North Macedonia (2008), Montenegro (2011), Albania (2013) and Kosovo (2015)
- How it arrives
- One CSV, from the data browser's download button
- Language
- Page and file in English
- 1Ten kinds of row share one number column. Each row is one thing that happened to the waste (wst_oper). Filter to GEN for waste generated, or you add landfill, recycling and incineration together. RCY already includes material recycling, composting and preparing for reuse in all 922 of its rows: add a part to it and you count that waste twice.
- 2The recycling rate is a calculation you do. Divide RCY by GEN and multiply by 100. On thousand tonnes this matches Eurostat's own rate (cei_wm011) in all 808 rows that rate has; on kilograms per person it is within 0.3 of a percentage point. Divide by GEN, not TRT: treatment runs from 59% to 116% of waste generated, because of exports, storage and waste counted at two steps.
- 3The rules changed from 2020. New EU rules changed what counts as municipal waste and where recycling is measured, and countries moved to them between 2019 and 2022. A change of method is flagged b: for waste generated, Austria, Greece, Ireland, Norway and Sweden in 2020, and 38 values since 1995. Check the flag on every year you compare.
- 4Some figures are estimates. Of the 993 values for waste generated, 102 are estimated (e), 7 provisional (p) and 15 imputed by Eurostat (i). Switzerland's recycling is estimated in every year from 2012 to 2019. Say how you treated them.
- 5A missing year has no row. It is simply absent from the CSV, not blank. 2024 has only 25 countries: Switzerland, Austria, Italy and nine others stop in 2023. Ireland has no 2013 or 2015, Denmark no 2010 and Iceland no 2019.
- 6A 0 is a real value. 52 recycling values and 20 landfill values are 0. They are not missing data.
- 7Municipal waste is not all waste. It is waste from homes, and similar waste from shops, offices and public buildings. Countries include different amounts of that similar waste, so part of a difference between countries comes from the definition. Waste from industry, farming and building work is not in it.
- 8Waste sent abroad is counted at home. Eurostat counts waste in the country that produced it, wherever it is treated. Waste from a French border town that is burned in a Swiss incinerator is in France's rows, not Switzerland's.
- 9The EU total sits among the countries, and the codes are not ISO. Take out EU27_2020 (266 rows). Greece is EL, not GR, and Kosovo is XK, with an asterisk after its name. To match a file with three-letter codes, write the code beside each country yourself: in the World Bank's files EL is GRC and XK is XKX. The UK is not in the file.
- 10Each release rewrites the series. Eurostat publishes the whole series again every year and can revise it in between. This card was counted on the release of 30 March 2026: say which release you used.
Background for sprint 1, not a research question: none of these sets what people do against what happens to nature.
- How did the share of municipal waste recycled change in Switzerland and its four neighbours between 1995 and 2023?
- How far was each EU country from the 55% recycling target for 2025 in 2023, the last year every EU country has?
- Across European countries, is municipal waste generated per person related to GDP per person in 2023?
- wst_oper: what happened to the waste. GEN is generated, RCY recycled, DSP_L_OTH landfilled (with other disposal), DSP_I_RCV_E incinerated, TRT all treatment. RCY_M, RCY_C_D and PRP_REU are the parts of RCY
- unit: KG_HAB, kilograms per person. The link keeps only this; take the filter off for THS_T, thousand tonnes
- geo: the country, code and name in one cell (EL: Greece)
- TIME_PERIOD: the year. OBS_VALUE: the number. OBS_FLAG: e estimated, p provisional, b break in series, i imputed by Eurostat
The European Parliament and the Council set the targets. Each of the 27 EU member states must meet them, and the Commission checks progress with the European Environment Agency.
Link to this fileDirect: the targets are recycling divided by waste generated, the same division you do with RCY and GEN (preparing for reuse is already inside RCY). It binds the 27 EU members: check whether it binds any other country in the file before you use that country for comparison. The first target year is 2025, and the file stops in 2024, so it shows the approach to the target, not whether it was met.
The tensionThe same targets apply to countries that start from very different places. A country that recycled less than 20% or landfilled more than 60% of its municipal waste in 2013 may put each deadline back by up to five years, if it sends the Commission a plan. The European Environment Agency adds that the 2025 target is "harder to meet because of tighter reporting rules" (June 2023).
SourceThe European Parliament and the Council. Each of the 27 EU member states must meet it.
Link to this fileDirect: the limit is landfilled waste as a share of waste generated, which is DSP_L_OTH divided by GEN. The file cannot reach 2035, so a question can only be about the path towards the limit. It binds EU members only.
SourceThe Federal Council made it. The cantons plan waste disposal and see that it is done, and the Federal Office for the Environment (FOEN) oversees it.
Link to this fileDirect for Switzerland's landfill and incineration rows. The rule is older than the ordinance: the FOEN says such waste was still landfilled "in the last century", so 2016 is not when it began. Find the date at source before you compare the years before and after it. Switzerland is not in the EU, so neither EU target binds it.
The tensionSwiss incinerators also burn waste from across the border. The European Commission says "around 200,000 tonnes" of mixed municipal waste a year go to Switzerland from border regions of Austria, France, Germany and Italy. New EU rules banned this export from 21 May 2026, which the Commission said "would disrupt long-standing, sustainable waste management practices in border regions". The EU then changed the rule again: from 30 July 2026 the waste may go to Switzerland once more, to be recycled or burned for energy.
SourceSprint 4 is where a dataset becomes a method: open it before you trust it, count what is really in it, fix an inclusion rule before you look, and write it down tightly enough that a stranger could rebuild your exact file.
Sprint 4 · Select and document your dataAbout the order: the catalogue is indexed by environmental issue rather than by publisher, because you are choosing something to find out, not somewhere to download from. Places are ordered outwards from Geneva, where this guide was written, and the order is not a ranking: a study local to Geneva does not score better than a global one.