Ecological Sampling — Quadrats, Transects and Mark-Recapture
In 2017, CSIRO researchers Sarah Legge and colleagues published the most comprehensive survey of Australian feral cats ever attempted, estimating 2.1–6.3 million animals across the continent. They could not count a single cat directly — instead, they combined camera-trap data, spotlight surveys, and mark-recapture models across 91 sites. Their estimate revealed each feral cat kills 5–30 animals per night, producing a national toll of 1.4 billion birds and 1.7 billion reptiles per year — data that is only possible because of the population sampling techniques this lesson teaches.
Practise this lesson
Four printable worksheets that build from the foundations up to exam-style questions — start at whatever level suits you.
Q1. A national park covers 5,000 hectares. Rangers need to know how many brushtail possums live there. The possums are nocturnal, tree-dwelling, and hide in hollows during the day. Describe two methods you could use to estimate the population, and explain why you cannot simply walk through the park and count every individual.
Q2. Scientists place a 1 m×1 m quadrat randomly in a grassland and count 8 kangaroo grass plants inside it. There are 200 such quadrats that could fit across the entire field. Predict whether simply multiplying 8 × 200 gives an accurate population estimate. What could go wrong?
Core Content
Best for sessile and slow-moving organisms — plants, barnacles, coral, soil invertebrates
CSIRO's 2017 feral cat survey needed to estimate 2.1–6.3 million animals spread across 7.7 million square kilometres of Australia. It is physically impossible to count every individual. Instead, Legge and colleagues deployed camera traps at 91 sites, counted animals in defined areas, and applied mark-recapture statistics to scale up. The same principle underlies quadrat sampling: count organisms in small, representative areas, then extrapolate to the whole habitat. The key word is representative — a biased sample produces a biased estimate.
Why sampling is necessary: Most populations are too large, too mobile or too widely distributed to count entirely. Sampling provides estimates of distribution and abundance that are statistically reliable if the method is sound.
How quadrat sampling works:
- Define the habitat boundary and total area.
- Use a random number generator or grid coordinates to place quadrats — never place them where they look convenient.
- In each quadrat: record presence/absence, percentage cover, or individual counts.
- Calculate mean density = total individuals counted ÷ total quadrat area sampled.
- Extrapolate: population estimate = mean density × total habitat area.
Mean density = total individuals counted ÷ total quadrat area. Population estimate = mean density × total habitat area. Quadrats must be placed at random coordinates to avoid observer bias.
Pause — copy the highlighted formula and the five steps into your book before moving on.
Ten 1 m×1 m quadrats placed randomly across a 5,000 m² paddock give counts of: 7, 12, 9, 15, 8, 11, 6, 13, 10, 9.
Total = 100 individuals · Total quadrat area = 10 m²
Mean density = 100 ÷ 10 = 10 plants per m²
Population estimate = 10 × 5,000 = 50,000 plants
Appropriate for: sessile organisms — plants, barnacles, coral colonies, lichens. Unsuitable for: highly mobile animals — birds, insects, fast mammals (they leave before counting is complete).
A researcher places 5 quadrats of 1 m² each in a 2,000 m² meadow and counts 10, 14, 8, 12, 6 clover plants. What is the best estimate of total clover plants in the meadow?
Reveals how distribution changes along environmental gradients — tide height, altitude, distance from disturbance
We just saw that quadrats give density estimates for a whole habitat. That raises a question: what if organisms are not evenly spread — what if their distribution changes across the habitat? This card answers it → transects reveal how distribution shifts along a gradient.
Quadrats tell you how many organisms are in an area. Transects tell you where they are — revealing how distribution changes along an environmental gradient.
Line transect: A straight line is laid across the habitat. Every species that touches the line is recorded. Produces qualitative data — presence/absence and relative abundance. Quick, ideal for rapid zonation surveys, but does not give actual density.
Belt transect: A strip of defined width (e.g., 1 m) is surveyed on either side of the line. All individuals within the belt are counted or percentage cover is estimated. Produces quantitative data comparable to quadrat sampling, arranged spatially.
Line transect: qualitative (presence/absence); quick but no density data. Belt transect: quantitative (counts/percentage cover); used when actual density comparison along a gradient is needed.
Copy the highlighted comparison before moving on.
Belt transects from high-tide to low-tide mark reveal consistent zonation:
- High zone: Limpets and small snails — tolerate long air exposure and desiccation.
- Mid zone: Barnacles (Tesseropora rosea) — filter-feed when submerged, close tightly when exposed.
- Low zone: Mussels, sea anemones and algae — thrive where submersion is longest.
This zonation is driven primarily by abiotic tolerance (desiccation, temperature, wave action).
When to use transects: intertidal zonation; altitudinal gradients (treeline studies); recovery after fire, logging or mining; edge effects (forest into farmland).
What is the main advantage of a belt transect over a line transect?
For mobile animals — catch, mark, release, recapture; the maths does the rest
We just saw that quadrats and transects work for plants and sessile organisms. That raises a problem: mobile animals move out of any fixed sampling area. This card answers it → mark-recapture uses probability to estimate populations that cannot be directly counted.
For animals that move too fast to count in quadrats, ecologists use a statistical trick: catch some, mark them, release them, then see what fraction of marked animals appear in a second catch.
Steps:
- Capture a sample; count and mark all individuals (M). Release and allow mixing.
- After sufficient mixing time, capture a second sample; count total caught (C) and those marked (R).
- Calculate: N = (M × C) / R
N = (M × C) / R where M = marked and released, C = total in second sample, R = marked recaptures. Assumptions: closed population; marks don't affect survival; marks retained; second sample is random.
Copy the formula and four assumptions into your book before the check below.
M = 45 tagged and released. Second sample: C = 60, of which R = 15 carry tags.
N = (45 × 60) / 15 = 2,700 / 15 = 180 wallabies
Critical assumptions:
- Closed population — no births, deaths, immigration or emigration between captures.
- Marks do not affect survival or behaviour — marked animals must not be more likely to die or become trap-shy/trap-happy.
- Marks are retained — tags must not fall off or fade before recapture.
- Second sample is random — every individual has an equal chance of capture.
Researchers mark 40 fish, release them, then catch 50 fish and find 8 are marked. What is the estimated population size?
Every sampling method introduces uncertainty — understanding these errors is essential for evaluation questions
We just saw that mark-recapture gives a population estimate, not an exact count. That raises a question: what exactly makes these estimates unreliable? This card answers it → each method has specific sources of error that need to be recognised and minimised.
Quadrat errors:
- Observer bias: Placing quadrats where organisms look abundant. Fix: random coordinates.
- Non-random distribution: Clumped organisms missed or over-sampled. Fix: more quadrats; stratified sampling.
- Edge effects: Boundary organisms counted inconsistently. Fix: count only individuals whose centre is inside.
- Size mismatch: Quadrat too large for mosses or too small for trees. Fix: match quadrat size to organism.
Mark-recapture errors:
- Open population: Births, deaths or migration between samples. Fix: short study duration; closed-season studies.
- Trap shyness/happiness: Marked animals avoid or seek traps. Fix: vary trap locations; use different bait.
- Mark loss: Tags fall off or marks fade. Fix: durable marks; double-mark a subset.
- Heterogeneous capture probability: Some individuals harder to catch. Fix: spatially explicit capture-recapture models.
Quadrat errors: observer bias, non-random distribution, edge effects, size mismatch. Mark-recapture errors: open population, trap shyness/happiness, mark loss, heterogeneous capture probability. Mark-recapture gives a statistical estimate — not an exact count.
Copy the two error categories into your book, then check below.
In 2017, CSIRO scientists estimated 2.1–6.3 million feral cats across Australia. Cats are nocturnal, territorial and avoid humans — direct counting is impossible. Instead, scientists used remote camera trapping across hundreds of sites. Individual cats were identified by their unique coat patterns. By treating each camera location as a "capture occasion" and tracking which cats appeared at which cameras, scientists built detection histories and applied spatially explicit capture-recapture models to estimate both population size and the area each cat roams.
The wide range (2.1–6.3 million) reflects genuine uncertainty: varying detection rates across habitats, high population turnover, and the need to extrapolate from sampled to unsampled regions. No single traditional method could produce a national estimate — multiple data sources and models were combined.
For each ecological investigation below, select the most appropriate sampling method and justify your choice.
- Estimating the population of koalas in a eucalypt woodland where individuals are visible in the canopy.
- Measuring how grass cover changes from the edge to the centre of a dried salt lake.
- Estimating the population of eastern grey kangaroos that move across a 500 ha grassland.
- Comparing the abundance of two competing shrub species across a rocky hillside.
Part A — Calculation: Researchers studying eastern grey kangaroos in a Victorian reserve caught and ear-tagged 38 individuals. One week later, they caught 52 kangaroos, of which 11 had ear tags.
- Calculate estimated population size N. Show working. (2 marks)
- Describe one biological reason why the actual population might be higher than your estimate. (1 mark)
- Describe one biological reason why the actual population might be lower than your estimate. (1 mark)
Part B — Quadrat data: A student surveyed a 2,000 m² meadow using 1 m×1 m quadrats. Counts of white clover were:
| Quadrat | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Count | 14 | 9 | 22 | 7 | 16 | 11 | 19 | 10 |
- Calculate the mean density of white clover per m². (1 mark)
- Estimate the total population in the meadow. (1 mark)
- Quadrats near a stream had higher counts than those on drier ground. Explain why simply averaging all eight quadrats might underestimate the true population, and suggest one improvement. (3 marks)
Which assumption of mark-recapture is violated if many tagged fish emigrate from the study lake between the first and second capture events?
A fresh set drawn from this lesson's question bank — feedback shown immediately. +5 XP per correct · +25 XP all correct
ApplyBand 4(4 marks) 1. Compare line transects and belt transects as methods for studying species distribution along an environmental gradient. In your answer, describe what data each method produces and state one situation where each would be the preferred choice.
AnalyseBand 5(5 marks) 2. Describe how you would use quadrat sampling to estimate the population of kangaroo grass (Themeda triandra) in a 10-hectare paddock. Your answer should include: how you would ensure random placement, how many quadrats you would use and why, what you would measure in each quadrat, and how you would calculate the final population estimate.
EvaluateBand 5–6(6 marks) 3. In 2017, CSIRO scientists estimated that Australia holds 2.1–6.3 million feral cats. (a) Explain two reasons why the estimate range is so large. (b) Discuss whether traditional mark-recapture alone could produce this national-scale estimate, or whether additional methods and data sources were needed.
Show all answers
Activity 1 — Sampling Method Selection
1. Koalas (visible in canopy): Mark-recapture — koalas are mobile mammals that cannot be reliably counted in quadrats. Individual identification from coat patterns or ear-tags allows Lincoln-Petersen estimation.
2. Grass cover along a salt lake gradient: Belt transect — reveals how grass cover changes across an environmental gradient (salt concentration, soil moisture) in quantitative terms.
3. Eastern grey kangaroos (500 ha): Mark-recapture — kangaroos are highly mobile and would leave any fixed quadrat area before counting is complete.
4. Competing shrub species on a hillside: Quadrat sampling with random placement — shrubs are sessile, allowing direct counting. Placing quadrats randomly across the hillside gives density estimates for both species.
Activity 2 — Calculation and Error Analysis
Part A: N = (38 × 52) / 11 = 1,976 / 11 ≈ 180 kangaroos. Higher if some tagged kangaroos died before recapture (open population — fewer in denominator inflates N). Lower if tagged kangaroos became trap-happy and were disproportionately captured (R too high, deflates N).
Part B: Mean density = (14+9+22+7+16+11+19+10) / 8 = 108/8 = 13.5 plants/m². Population = 13.5 × 2,000 = 27,000 plants. The stream creates a habitat gradient: high moisture near stream inflates some quadrat counts. If fewer quadrats are placed near the stream than the proportion of the meadow it represents, the mean is skewed downward. Improvement: stratified random sampling — divide meadow into dry and riparian zones, sample each zone proportionally to area.
Short Answer Model Answers
Q1 (4 marks): Line transects record every species that touches a straight line, giving qualitative presence/absence and relative abundance data — fast and suitable for rapid reconnaissance when only zonation patterns are needed (1 mark). Belt transects survey a strip of defined width, counting all individuals or estimating percentage cover, giving quantitative density data (1 mark). Line transect preferred when time is limited and only species zonation is needed (1 mark). Belt transect preferred when actual density comparison between zones is required, e.g., barnacle density across intertidal zones (1 mark).
Q2 (5 marks): Random placement: use a random number generator to produce grid coordinates — prevents observer bias (1 mark). Use at least 10–20 quadrats of 1 m×1 m to reduce the effect of random variation and clumping (1 mark). In each quadrat, count the number of kangaroo grass individuals (1 mark). Mean density = total counted ÷ total quadrat area; population = mean density × 100,000 m² (10 ha) (1 mark). If the paddock has obvious zones (wet/dry), use stratified random sampling — weight estimates by zone area (1 mark).
Q3 (6 marks): (a) Wide range reflects: varying capture probability across habitats (cats harder to detect in forest than open desert) (1 mark); high population turnover — cats breed rapidly and die from control programs and starvation, so the population is never truly closed (1 mark). (b) Traditional mark-recapture alone could not produce a national estimate — it is designed for local, closed populations over short periods (1 mark). Scientists needed: remote camera trapping across hundreds of sites (1 mark); spatially explicit capture-recapture models accounting for varying detection probability; occupancy modelling and extrapolation from habitat suitability maps (1 mark). The wide range acknowledges the uncertainty from combining heterogeneous data sources (1 mark).
Five timed questions on quadrat sampling, transects, and mark-recapture. Beat the boss to bank a tier.
Enter the arenaCSIRO's 2017 Legge et al. study estimated 2.1–6.3 million feral cats — a 3-million-animal margin of error — because mark-recapture estimates are only as good as their assumptions. The wide range reflects violations of the closed population assumption (cats moved between sites), unequal catchability (trap-shy individuals), and mark loss. These are the same limitations that apply to any mark-recapture study, including the fish population calculations in the Think First section.
Return to your Think First response. Apply the Lincoln-Petersen formula to the numbers in the hook: N = (M × C) / R. Why does the 3-million uncertainty in the cat estimate reflect the assumptions this formula requires?