Ten investigations you can start today
Each is a starting point, not a finished investigation. For each one I have downloaded the data, checked that the two files match, and counted enough rows for a test. Both files come straight from the people who publish them. Pick one, make it yours, and follow it.
Every one is laid out on the three lines: what people do, what happens to nature, and somebody’s plan.
- Choose the one about a place or a living thing you care about.
- Tell your teacher which, and make the choice under Make it yours, so yours is not the same as a classmate’s.
- Follow the steps. They take you as far as your result.
- Then write it up with the guide: the investigation gives you the data, not the report. Start with step 3, the strategy, and step 4, documenting your data.
Sewage and river invertebrates
The relationship between ammonium (mg N/l, the 90th percentile of each year's samples) and the invertebrate index (IBGN, a score from 1 to 20) across Geneva's river monitoring stations, 1995 to 2022.
Matched on station code (CODEMESURE) and year. 408 station-years appear in both files, from 145 stations, 1995 to 2022.
- 1Download both files from their pages below. Each is a zip: open it and use the CSV inside. The columns are separated by semicolons.
- 2In the nutrients file keep CODEMESURE, NOM_STATION, ANNEE and N_NH4 (ammonium). If any N_NH4 value is −99, delete that row: −99 means missing.
- 3In the invertebrate file keep CODEMESURE, ANNEE_MESURE and IBGN_MOYENNE (the index).
- 4In both files add a column that joins the station code and the year, for example
=A2&"-"&C2. Use XLOOKUP on it to bring each ammonium value into the invertebrate sheet, and keep only the rows where both values exist. You should have 408. - 5Average each station's rows into one ammonium value and one index value (a pivot table does this). You now have 145 stations, one row each.
- 6Run the test below and draw a scatter graph with ammonium on the x axis.
Spearman's rank correlation between ammonium and the invertebrate index, on 145 stations.
The contract's goal is cleaner water, and it names wastewater treatment, which is where most of the ammonium comes from. The invertebrates are a slower answer to whether it worked: they reflect the water over months, not only on the morning of a sample.
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.
Read the source- Station names repeat. Twenty-two stations are called Embouchure, because every river has a mouth. Match on CODEMESURE, never on the name.
- A higher invertebrate index is cleaner water. So more ammonium with a lower index shows up as a negative correlation. Say so in your write-up.
- Ammonium is each year's 90th percentile, not its mean: nine samples in ten were at or below it, so it reports the bad days. The canton's own documentation, in the zip, says so. Describe it that way.
- The same station appears in several years. Average to one value per station before you test, or you count the same station several times over.
Use the number of kinds of invertebrates (TAXON_SOMME) instead of the index, or nitrate (N_NO3, mostly from farmland) instead of ammonium. Or take the canton's renaturation of the Aire as your strategy: it protects the river's life by freeing the channel rather than cutting the ammonium, and its tension, farmers against the canton over land, is the best documented of the three on the card. Decide before you look, and tell your teacher which.
Phosphate and river algae
The relationship between dissolved phosphate (mg P/l, the 90th percentile of each year's samples) and the diatom index (DI-CH, 1.4 best to 8.0 worst) across Geneva's river monitoring stations, 1998 to 2025.
Matched on station code (CODEMESURE) and year. 420 station-years appear in both files, from 148 stations, 1998 to 2025.
- 1Download both files from their pages below. Each is a zip: open it and use the CSV inside. The columns are separated by semicolons.
- 2The diatom file repeats every row about fifty times: 24,759 rows hold only 503 real ones. Remove the duplicates first (in Google Sheets: Data, Data clean-up, Remove duplicates, on CODEMESURE and ANNEE_MESURE).
- 3In the diatom file keep CODEMESURE, ANNEE_MESURE and DI_CH. In the nutrients file keep CODEMESURE, ANNEE and P_PO4 (phosphate), and delete any row where P_PO4 is −99.
- 4In both files add a column that joins the station code and the year, for example
=A2&"-"&B2. Use XLOOKUP on it to bring each phosphate value into the diatom sheet, and keep only the rows where both exist. You should have 420. - 5Average each station's rows into one phosphate value and one diatom value (a pivot table does this). You now have 148 stations, one row each.
- 6Run the test below and draw a scatter graph with phosphate on the x axis.
Spearman's rank correlation between phosphate and the diatom index, on 148 stations.
The contract names wastewater treatment, and phosphate is one of the nutrients treatment removes. Diatoms respond to that enrichment, so they are a biological answer to the question the contract set itself.
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.
Read the source- The diatom index runs backwards. 1.4 is the cleanest water in the file and 8.0 the worst, so a positive correlation means the diatoms get worse as phosphate rises. Say so, or your reader will take it the wrong way round.
- Station names repeat. Match on CODEMESURE, never on the name.
- Most diatom scores rest on one or two sampling campaigns, so one year's score is a noisier number than it looks. Averaging each station's years helps.
Use total phosphorus (P_TOTAL) instead of phosphate, or keep one decade only. Decide before you look, and say why.
Phosphorus and Lake Geneva's algae
The relationship between the change in winter dissolved phosphate (µg P/L, January to March) and the change in growing-season phytoplankton biomass (µg/L, May to September) from one year to the next, in the middle of Lake Geneva, 1974 to 2010.
Matched on nothing: both are columns in the same file. 444 months, 1974 to 2010: 37 winters and summers, and 35 changes from one year to the next.
- 1Download the CSV from its page below. When you open it, tell your spreadsheet it is semicolon-separated and Latin-1 (Western European), or the headings turn into strange characters.
- 2Keep Year, Month, PO4 (dissolved phosphate) and PhytoTot (all the phytoplankton). NaN means missing, never zero.
- 3Fix the unit. PO4's heading says µgP/L, but the values are milligrams: 0.034 is 34 µg per litre. Multiply by 1,000 and say so.
- 4For each year, average PO4 over January to March and PhytoTot over May to September. A winter counts if at least two of its three months have a value: all 37 do.
- 5Add a column for each: this year's value minus last year's. Leave out the change from 2001 to 2002, when the phytoplankton began to be sampled from 18 m down instead of 10 m. You should have 35 pairs of changes.
- 6Run the test below, and draw a scatter graph of the changes with a line at zero on both axes.
Spearman's rank correlation between the change in winter phosphate and the change in summer phytoplankton, on 35 pairs of years.
Direct: CIPEL's phosphorus target is written for phosphorus because phosphorus drives the algae, and it is met at sewage works and by keeping phosphate out of detergents. Professional fishers now want it raised for the sake of their catch; CIPEL's scientists hold about 15 µg per litre.
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.
Read the source- Changes, not levels. Phosphate fell through most of these years, so its raw values would go with anything else that changed steadily. Comparing each year with the one before takes the steady fall out.
- Winter phosphate, summer algae. In summer the algae take phosphate out of the water they grow in, so a summer reading is partly their own doing. January to March is what is on offer before they start.
- The file stops in 2010. It tests the fishers' chain in the years the phosphorus came down, not the lake as it is now.
Use the water fleas (Daphnia) instead of the algae, the next link in the fishers' chain from phosphorus to fish, or compare the changes before and after the 1986 detergent ban. Decide before you look, and say why.
Geneva's chestnut tree against the countryside
A comparison of how far the day the horse chestnut comes into leaf moved between 1955 to 1984 and 1986 to 2015, between Geneva's official chestnut in the city centre and the horse chestnuts at 14 countryside stations of MeteoSwiss's phenology network below 600 m.
Matched on nothing to match: each site gives one number, how far its leafing moved. 15 sites: the official tree and 14 countryside stations, each with at least 20 springs in both periods.
- 1The city. Open the official chestnut's page (on Wikipedia, in French), unfold the table, copy it and paste it into Google Sheets. Delete the reference numbers in square brackets that come with five of the dates, and keep 1955 to 2015: every one of those springs is the same tree.
- 2Fix 2003. The table has two rows for 2002; the second one (29 December 2002) is the first leaf of spring 2003. Relabel it 2003. As a day of the year it is −2.
- 3Turn each French date (such as 16 mars) into a day of the year: split it into day and month, then use
=DATE(year, month, day) - DATE(year, 1, 1) + 1. - 4Average the days for 1955 to 1984 and for 1986 to 2015. The city's change is the second average minus the first: negative means earlier.
- 5The countryside. Download the 14 station files listed on the MeteoSwiss card below. In each keep reference_timestamp (the year) and maesh13d (horse chestnut, half its leaves unfolded).
- 6maesh13d is written as a number such as 19540330. 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, and leave the blanks blank. - 7For each station, average 1955 to 1984 and 1986 to 2015 and take the difference, as you did for the city. You should have 14 changes, each from at least 20 springs in each period.
- 8Run the test below, and draw a dot plot of the 14 changes with the city's change marked as a line across it.
A one-sample t-test: is the countryside stations' mean change different from the city tree's change? Fourteen values tested against one.
Direct on the city side: it plants trees to cool the city, and the city's heat is what the city centre has and the countryside does not. It passed unanimously, so find a documented disagreement before you write about tension.
The tensionNone found yet, and Criterion B needs one: two groups who want things that conflict, with a source for each. Finding it is your first research job on this title. Ask your teacher if you get stuck.
Read the source- Compare changes, never days. The official tree's register records the first bud; the stations record half the leaves unfolded. In any year the two are about six weeks apart.
- One tree against fourteen places. The city side is a single tree beside the warm stone of the Tour Baudet. Name that in your evaluation: what is true of this tree may not be true of Geneva's trees.
- Stop at 2015. The official tree was replaced in 2016, and a new tree leafs differently.
- Three low stations are in cities. Zürich / Fluntern, Zürich / Albisgüetli and Biel meet the other rules but are left out, because the question is city against countryside.
Choose your own two periods of 30 springs within 1951 to 2015, or add the three city stations as a group of their own. Decide before you look, and say why.
New homes and butterflies
The relationship between the number of dwellings added in England each year and the change in the Peacock butterfly's England index from one year to the next, 1976 to 2023.
Matched on year. 47 changes from one year to the next, 1976 to 2023.
- 1Download the butterfly zip from its page below and open the CSV. Filter COUNTRY to England and COMMON_NAME to Peacock (or your own species). Keep YEAR and COLLATED_INDEX.
- 2Work out the change in the index: this year's value minus last year's.
- 3Download Live Table 104 from the dwelling stock page below and use the All dwellings column for England.
- 4Tidy the years first. 1971 and 1981 each have an extra census row (1 April): delete those rows. In 1991 the count moved from 31 December to 31 March, so the 1990 to 1991 step covers three months, not twelve: say what you did with it. Then work out dwellings added: this year's stock minus last year's.
- 5Match the two by year, then run the test below and draw a scatter graph with dwellings added on the x axis.
Spearman's rank correlation between dwellings added and the change in the index, on about 47 years.
Direct. The target is measured by an index of species that includes butterflies, from monitoring schemes like this one. What you can test is what butterflies were doing while England kept building; whether the 2030 target is met 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.
Read the source- The index is not a count. It is on a log scale where each species' own average is 2, so compare changes, never the index values of two species.
- Change against change, not level against level. Both series move steadily over the years, so their raw values can correlate whether or not one affects the other.
- 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.
Choose your own species instead of the Peacock: 36 of the 59 species start in 1976. Choose it before you look, and say why (a garden species, a grassland one).
Protected land and forest loss
The relationship between the share of a country's land that was protected in 2015 (% of land area) and the tree cover it lost from 2016 to 2025 (% of its tree cover in 2010), across the 107 countries with at least 1 million hectares of tree cover.
Matched on country name, with 13 names matched by hand. 107 countries, each with at least 1 million hectares of tree cover in 2000.
- 1Download the tree cover workbook from its page below with the Download button, or directly: global.xlsx (18 MB). Use the sheet Country tree cover loss.
- 2Filter threshold to 30. Every country has eight rows, one for each threshold, and 30 is GFW's own default.
- 3China, India and Pakistan have extra rows for disputed areas: add each one's rows together into a single row. Then keep the countries whose extent_2000_ha is at least 1,000,000. You should have 109.
- 4For each country, add tc_loss_ha_2016 to tc_loss_ha_2025 together, divide by extent_2010_ha and multiply by 100: the share of its tree cover it lost.
- 5Download protected land from the World Bank page below as a CSV. It is a zip: use the file whose name begins API_. Delete the four lines above the header row, and keep Country Name, Country Code and 2015.
- 6Match the two by country name. Thirteen are written differently in the World Bank file: Cote d'Ivoire; Congo, Dem. Rep. (the Democratic Republic of the Congo); Congo, Rep. (the Republic of the Congo); Iran, Islamic Rep.; Lao PDR; Mexico; Korea, Dem. People's Rep. (North Korea); Korea, Rep. (South Korea); Russian Federation; Slovak Republic; Turkiye; Venezuela, RB; Viet Nam. French Guiana and Taiwan are not in it: leave them out. You should have 107.
- 7Run the test below and draw a scatter graph with protected land on the x axis.
Spearman's rank correlation between protected land and the share of tree cover lost, on 107 countries.
Direct. The target is counted in protected area, and it assumes that protecting land keeps what is on it. Your question tests that assumption on the protection that already existed in 2015. Your data can show whether the two go together; it cannot show that the protection caused it.
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.
Read the source- Loss is not deforestation. GFW counts every kind of tree cover lost, including to fire and to plantation harvests: wildfire is 79% of Australia's loss since 2001, 74% of Russia's and 68% of Canada's. Say so, or use the Country drivers sheet to count only loss to permanent agriculture.
- Protection first, loss after. Keep the order: protection measured in 2015, before the loss it might have prevented, not in the same years.
- Parks are not always where the pressure is. Protected areas are often land that is steep, remote or unwanted for farming, so a country can protect a lot and still lose forest elsewhere. Say what your result can and cannot show.
- Shares, not hectares. Russia, Brazil and Canada have the most hectares of everything. Dividing by each country's own tree cover is what makes them comparable.
Use GDP per person (World Bank, same format) instead of protected land, or keep one continent only. Decide before you look, and tell your teacher which.
Cattle and forest loss
The relationship between the change in the number of cattle in each Brazilian state from 2001 to 2023 (%) and the tree cover it lost over the same years (% of its tree cover in 2000), across Brazil's 27 states.
Matched on state name, spelled the same in both files. 27 states, every one in both files, with every year from 2001 to 2023.
- 1Download the cattle table from IBGE with this link. Use a browser: the site refuses other programs. It is an Excel file in Portuguese; Bovino means cattle.
- 2The state (Unidade da Federação, column A) is named only on the first of its rows. In F3 type
=A3, in F4 type=IF(A4="",F3,A4), and copy F4 down to the last row. Column F now names the state on every row. The number of cattle is column E, which has no heading. - 3Keep only the years 2001 and 2023 (Ano). For each state, work out the change in its herd: (2023 minus 2001) divided by 2001, times 100.
- 4Download the tree cover workbook from its page below, or directly: global.xlsx. On the sheet Subnational 1 tree cover loss, filter country to Brazil and threshold to 30. You have 27 rows.
- 5For each state, add tc_loss_ha_2001 to tc_loss_ha_2023 together, divide by extent_2000_ha and multiply by 100: the share of its tree cover it lost.
- 6Match the two by state name (subnational1 in the forest file). The names are identical, accents included, so all 27 match. Run the test below and draw a scatter graph with the change in cattle on the x axis.
Spearman's rank correlation between the change in cattle and the share of tree cover lost, on 27 states. That is a small sample: say how far your result reaches.
Direct. The agreements bar slaughterhouses from buying cattle raised on land cleared after a cut-off date, so they act on the very herds your file counts. Your data cannot say which animals came from which ranch, which is why the tension below matters to your evaluation.
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.
Read the source- Stop at 2023. Wildfire was 49% of Brazil's tree cover loss in 2024 and 42% in 2025, against 12 to 25% in each year from 2019 to 2023, and IBGE marks its 2024 herd as preliminary.
- A 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.
- Tree cover is not all of Brazil's vegetation. Much of the Cerrado is savanna, too low or too open to count as tree cover, so clearing it can go unseen in this file.
- The Federal District is tiny: 57,000 hectares of tree cover, so its share rests on very little forest. Keep it and say so, or leave it out and say why.
Use soybeans instead of cattle: IBGE's crop survey (table 1612) gives the area planted in each state. Or count only loss to permanent agriculture, from the forest file's Subnational 1 drivers sheet. Decide before you look.
Nets, drumlines and white sharks
A comparison of the share of the sharks caught that were white sharks, between nets and drumlines, in the four areas of Queensland's Shark Control Program that set both (Gold Coast, Sunshine Coast North, Sunshine Coast South and Rainbow Beach), 1996 to 2023.
Matched on nothing: both are columns in the same file. 5,228 sharks caught in the four areas, 1996 to 2023: 3,296 in nets and 1,932 on drumlines.
- 1Download the Excel file from the Shark Control Program page below. Use a browser: the site does not let scripts download it.
- 2Filter Year to 1996 to 2023, SpeciesGroup to SHARK, Gear to Net and Drum (leave out Other), and Area to Gold Coast, Sunshine Coast North, Sunshine Coast South and Rainbow Beach.
- 3Copy the rows you can see to a new tab. In column P, sort each row:
=IF(J2="WHITE SHARK", "White shark", "Other shark"), where J is CommonName. - 4Build the 2 by 2 table with four formulas, one per cell, such as
=SUMIFS(O:O, F:F, "Net", P:P, "White shark"), where O is NumberCaught and F is Gear. Add up NumberCaught; never count rows. You should have 3,296 sharks in nets and 1,932 on drumlines. - 5Work out what percentage of each gear's sharks were white sharks, and draw a bar chart of the two.
- 6Run the test below on the four numbers in your table.
A chi-squared test of independence on the 2 by 2 table: gear (net or drumline) against white shark or other shark. Every expected count is above 50.
Direct: the gear is the programme. Which to set, nets or drumlines, is its own decision, and your question asks what each one does to a protected shark.
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.
Read the source- Compare like with like. Drumlines are also set far north of where white sharks swim, which drags their share down if you use the whole of Queensland. Keep to the four areas that set both gears.
- A share is not a rate. The file does not say how many nets and drumlines were in the water, so you cannot say which gear catches more white sharks a day, only what part of each gear's catch they are.
- White sharks pass from June to October. Look at when each gear catches its other sharks before you say what a difference in share means.
- Stop at 2023. New gear and daily checks from 2024 changed the catch.
- Add up NumberCaught, never count rows. Some rows record more than one animal.
Swap white sharks for turtles and dolphins (SpeciesGroup TURTLE and MAMMAL), the catch the programme is criticised for, or test each area on its own. Decide before you look, and say why.
Marine reserves and sea urchins
A comparison of purple sea urchin (Strongylocentrotus purpuratus) density at kelp forest sites inside and outside no-take marine reserves in the northern Channel Islands, California, before (2010 to 2012) and after (2014 to 2024) sea star wasting disease removed the sunflower star (Pycnopodia helianthoides).
Matched on site code (site), between the counts file and the site table. 36 island sites surveyed in both periods: 15 inside no-take reserves and 21 outside.
- 1Open the PISCO package page below and download two of its files: PISCO_kelpforest_swath.1.11.csv (the counts, 33 MB) and PISCO_kelpforest_site_table.1.11.csv (each site's position and protection). The methods PDF on the same page explains how the divers work. If the notes column shows strange characters, open the file as Western European (Windows-1252); the columns you need are not affected.
- 2In the swath file, delete every column except campus, year, month, day, site, zone, transect, classcode and count, in that order, so they sit in columns A to I. In J2 put
=LEFT(E2,3)and fill it down. Filter J to SMI, SRI, SCI and ANA (San Miguel, Santa Rosa, Santa Cruz and Anacapa), copy what is left into a new sheet and call it Swath. You should have 108,731 rows below the headings. - 3Give each row a code for its transect: one strip of reef 30 m long and 2 m wide (60 m²), searched on one day. In Swath, replace column J with
=E2&"|"&B2&"|"&C2&"|"&D2&"|"&F2&"|"&G2and fill it down. In a new sheet called Transects, put=UNIQUE(Swath!J2:J108732)in A2. You should have 5,253 transects. - 4Give every transect its purple urchin count, zeros included. In B2,
=SUMIFS(Swath!I:I, Swath!J:J, A2, Swath!H:H, "STRPURAD"): a transect with no purple urchin row gets 0, which is correct. In C2,=B2/60gives urchins per m². In D2,=XLOOKUP(A2, Swath!J:J, Swath!E:E)brings back the site, and in E2=XLOOKUP(A2, Swath!J:J, Swath!B:B)the year. Fill all four down. - 5Label each transect's period. In F2:
=IF(AND(E2>=2010, E2<=2012), "Before", IF(E2>=2014, "After", "")). Every other year, 2013 included, stays blank and is left out. - 6Make a pivot table of Transects with site as the rows, period as the columns (untick the blank) and the average of density as the values: one mean for each site in each period. Keep only the sites with a value under both Before and After. You should have 46.
- 7In the site table, filter method to SBTL_SWATH_PISCO and delete every row whose site_status is blank. Next to each of your 46 sites, bring back its site_status and its mpa_type with XLOOKUP on the site table's site column, giving "none" for a site that is not there.
- 8Sort the sites into two groups. Inside: site_status mpa and mpa_type SMR (a state marine reserve, where all fishing is banned). Outside: site_status reference. Leave out mpa sites of type SMCA, and sites whose status is none. You should have 15 inside and 21 outside.
- 9For your introduction, the same file counts sunflower stars (classcode PYCHEL). Total them by year with a pivot table on the Swath sheet, and you can show the predator's loss in your own data.
Mann-Whitney U, run twice on the site means: inside against outside for Before, then for After, 15 sites against 21 each time. The Before test tells you whether the two groups started level; the After test is your answer. Draw a box plot with four boxes (inside and outside, before and after); in Google Sheets, a column chart of the four medians does the same job.
Direct: your human variable is the network itself. The reserves were set up to protect the whole ecosystem, not to manage urchins, so your question asks whether that protection also made these kelp forests more resilient to a shock nobody planned for, the loss of the sunflower star. The mechanism to argue is the predators the ban protects, sheephead and lobster, which both eat purple urchins.
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).
Read the source- The zeros are missing from the file. When divers see no purple urchins on a transect, it has no purple urchin row at all: 599 of the 5,253 island transects. Average only the rows that exist and every mean comes out too high. Steps 3 and 4 put the zeros back.
- Group by site_status, never by MPA_Name. Outside sites carry the name of the reserve they are paired with: SMI_CUYLER is listed under Harris Point SMR, but it is a reference site.
- Two 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, and never merge E and W.
- Conservation areas are not reserves. Painted Cave and west Anacapa are SMCAs, which allow some fishing, so they belong in neither group. Say in your method that you left them out, and why.
- Your sample is 36 sites, not 5,253 transects. Transects at one site are not independent of each other, which is why step 6 averages them first. And 2013 is left out because only 19 island sites were surveyed that year: say so in your method.
Swap purple urchins for giant kelp (MACPYRAD in the same file, the end of the chain the predators protect; a transect has a row for each number of stipes, so add them up), or for red urchins (MESFRAAD), which are fished themselves, so the reserves act on them directly. Decide before you look, and tell your teacher which.
Poison drops and forest birds
A comparison of native birds per five-minute count between forests treated with aerial 1080 poison in 2012 and untreated forests, on the East and West Coasts of New Zealand's South Island, in the summer of 2014/15.
Matched on nothing: both are columns in the same file. 586 five-minute counts in the summer of 2014/15, from 20 stations in each of four forests.
- 1Download birdsrelabund.txt from the Landcare Research page below. It ends .txt but is comma-separated: open it in Google Sheets and split it on commas.
- 2Filter Season to Summer 2014/15. Keep species, count, Site and treatment, and the columns that together make one five-minute count: Transect, Station, Date, Time and Observer.
- 3Keep the native birds: those listed as Endemic or Native on NZ Birds Online, 22 of the file's 38 names. Leave out the 14 introduced species and the two nobody identified (unknown, finch).
- 4Add up the native birds in each five-minute count, then average each station's counts: one number per station, 20 in each forest.
- 5Run the test below once for each coast: Wilberforce (poisoned) against Harper (untreated) on the East Coast, and Mt Alexander (poisoned) against Taipo (untreated) on the West.
- 6Draw one box plot per forest, poisoned beside untreated on each coast.
Mann-Whitney U on 20 stations against 20, once for each coast. Then the check: the same test on Winter 2012, before the drop.
Direct: the drop is your independent variable, and the case for aerial 1080 rests on the birds coming back. Scientists and the Parliamentary Commissioner for the Environment argue for more of it; hunters oppose a poison that also kills deer.
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.
Read the source- Stations, not counts. Each station was visited on three or four days, with two counts on most visits, so its counts are not independent of each other. Average them to one number per station.
- Compare within a coast, never across. Each poisoned forest has an untreated partner on its own coast.
- Most rows read 0, and that is real. Every count lists all 38 species, so a 0 means no bird of that kind was heard, not a gap.
- One pair of forests per coast. A difference could come from the forests rather than the poison, which is why the winter 2012 check matters.
Leave out the silvereye, the most counted bird, which flew in from Australia by itself in the 1850s, or test one native species of your own. Decide before you look, and say why.
Stuck on a spreadsheet step? Sort, filter and count in Google Sheets shows every move these steps use. Want something different? Every dataset is on the datasets page.