Lesson 5 of 10 · 15 min
Capture–recapture and the Lincoln index
Estimate the size of an animal population with the Lincoln index, , rearrange it, and judge how its assumptions and small samples affect the estimate.
What you will be able to do
- Describe the capture–recapture technique.
- Use the Lincoln index to estimate population size, and rearrange it to find , or .
- State the assumptions of the Lincoln index and predict whether breaking each one makes the estimate too high or too low.
- Explain why small recapture numbers make an estimate uncertain.
You can't put quadrats over animals that run, swim or fly away. To estimate how many there are, ecologists catch some, mark them harmlessly, let them go, and come back later for a second catch. If a quarter of the second catch carries a mark, then about a quarter of the whole population was marked the first time. Since you know how many you marked, you can work out the whole.
Capture–recapture
Capture a first sample
Trap or net animals across the site, using the same method everywhere.
Mark and release them
Use a mark that lasts, doesn't harm the animal and doesn't change its behaviour or its chance of being eaten: a dab of non-toxic paint on a beetle's wing case, a fin clip, a microchip, a numbered ear tag. Record the number marked, .
Allow time to mix
Leave long enough for marked animals to mix back through the population, but not so long that many are born, die or move in or out.
Capture a second sample
Use the same method. Count the total caught, , and the number of those that are marked, .
Key idea
The Lincoln index
The fraction marked in the second sample estimates the fraction marked in the whole population:
- : estimated population size
- : number caught and marked in the first sample
- : total number caught in the second sample
- : number of marked animals in the second sample
Because , the estimate is never smaller than , and never smaller than .
Worked example
Estimate a skink population with the Lincoln index
Students catch 35 garden skinks in a schoolyard garden, mark each with a small dot of non-toxic paint and release them. Three days later they catch 42 skinks, 7 of which have paint dots. Estimate the population of skinks in the garden.
Step 1: Identify the values
(marked in the first sample), (total in the second sample), (marked in the second sample).
Labelling first avoids swapping and .
Step 2: Write the formula and substitute
Step 3: Calculate
Step 4: Check it makes sense
of the second sample was marked, so about one-sixth of the population was marked: . The estimate is larger than both samples, as it must be.
Try it yourself first, then check the working one step at a time.
Answer
About 210 skinks.
Method
Label , and from the question.
Substitute into .
Check: and .
Watch out for
Calculating : an estimate smaller than the samples is impossible.
In a farm dam, 48 freshwater crayfish (yabbies) are caught, tagged and released. A week later 60 are caught, and 15 of them are tagged. Estimate the population of yabbies in the dam.
Run capture–recapture on a population you can't see. Repeat it a few times with small and then large samples: how much do the estimates vary?
The assumptions
The Lincoln index only gives a fair estimate if the second sample really is a fair mix of marked and unmarked animals. That needs several assumptions.
| Assumption | If it is broken… | Effect on |
|---|---|---|
| The population is closed: no births, deaths, immigration or emigration between the samples. | Unmarked animals born or moving in dilute the marked ones, so is smaller than it should be. | Too high (relative to the population when marking was done) |
| Marks are not lost (paint not worn off, tags not shed). | Marked animals are counted as unmarked, so is too small. | Too high |
| Marking does not affect survival (bright paint doesn't attract predators). | Marked animals die faster, so fewer are recaptured. | Too high |
| Marked animals are no more or less likely to be caught again. | If marked animals avoid traps ("trap-shy"), is too small; if they return for the bait ("trap-happy"), is too large. | Too high if trap-shy; too low if trap-happy |
| Marked animals have had time to mix evenly through the population. | If they are still near the release point and the second sample is taken there, is too large. | Usually too low |
Some of the paint marks on a sample of beetles wear off before the second capture. What effect does this have on the Lincoln index estimate?
Key idea
Small samples, big uncertainty
The estimate is very sensitive to when is small. With and :
| 2 | 3 | 4 | 5 | |
|---|---|---|---|---|
| 200 | 133 | 100 | 80 |
Going from to , one extra marked animal cuts the estimate by a third. Large samples (and so a larger ) and repeated estimates give a more reliable result. Report the mean of repeated estimates and their range.
Worked example
Evaluate repeated capture–recapture estimates
Bush rats (Rattus fuscipes) in a 4 ha patch of forest were sampled with Elliott traps three times, a month apart. Each time, 30 rats were caught and marked with a fur clip in the first night, and the traps were reopened two nights later.
| Trial | |||
|---|---|---|---|
| 1 | 30 | 28 | 6 |
| 2 | 30 | 32 | 8 |
| 3 | 30 | 25 | 3 |
(a) Calculate the population estimate for each trial and the mean. (b) Identify the least reliable estimate and explain why. (c) Fur clips grow out over several weeks. Explain whether this affects these results. (d) Suggest one improvement.
Step 1: Calculate each estimate
Trial 1:
Trial 2:
Trial 3:
Step 2: Calculate the mean
rats.
Step 3: Identify the least reliable estimate
Trial 3 is anomalous: it is about twice the others. It rests on only marked recaptures, so one more or one fewer marked rat would change it to or . Small values of give very uncertain estimates.
Reliability is about how consistent repeated results are, and how much a small change in the data changes the answer.
Step 4: Judge the effect of fur clips growing out
Within one trial, the recapture is only two nights after marking, so clips will not have grown out and is not affected. The risk is between trials: a month later, a clip from the previous trial may not have grown out fully, so a rat clipped last month could be counted as one marked this time, making too large and too low. Clipping a different position in each trial removes the risk.
Step 5: Suggest an improvement
Increase the number of traps and nights so that and (and so ) are larger, reducing the effect of chance; or use a mark unique to each trial (different clip positions).
Try it yourself first, then check the working one step at a time.
Answer
(a) 140, 120 and 250; mean 170 rats (or about 130 without the anomalous trial 3, if that is justified). (b) Trial 3: is so small that the estimate is very uncertain. (c) Not within a trial: recapture was two nights after marking, too soon for clips to grow out. But a clip left from the previous month could be mistaken for a new one, so each trial needs its own mark. (d) Larger samples (more traps or nights) so that is larger; distinct marks for each trial.
Method
Calculate each estimate before averaging.
Look for an anomaly and explain it from the data (small , an assumption broken).
Judge each assumption against the timing and method given.
Suggest an improvement that targets the specific weakness.
Watch out for
Dropping an anomalous result without giving a reason.
Saying "do more trials" without saying how it improves reliability.
Rangers estimate that a creek holds about 400 Mary River turtles. They plan to mark 80 turtles, and then catch 50 in a second sample. If their estimate is right, how many of the 50 should be marked?
Common mistake
Mixing up the letters
(marked first time) and (marked in the second sample) differ only in case. Write them out in words before you substitute. A check: must be at least as large as both and . If it isn't, you have and swapped.
Summary
Capture, mark and release animals; later catch , of which are marked.
, because the fraction marked in the second sample estimates the fraction marked in the population.
Assumptions: a closed population, marks not lost, marks don't affect survival or catchability, and time to mix.
Anything that makes too small (lost marks, deaths of marked animals, unmarked immigrants, trap-shy animals) makes too high.
Small makes the estimate very uncertain; larger samples and repeats improve reliability.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.