Mark-recapture Sampling 1

Lab 4. MARK-RECAPTURE SAMPLING

I. Introduction

One of the most basic field problems in ecology is the determination of how many individuals of
a species there are in a given area. The task of enumeration is relatively straightforward for non-
mobile or sessile organisms. However, mobile organisms present a suite of problems for a
researcher interested in ascertaining the abundance of these organisms.

One method used to estimate population sizes is the capture-recapture or mark-recapture

technique. The premise of this technique is that one catches a random sample of a population,
marks individuals, releases them so that they remix with the rest of the population, and then
catches a second random sample of the population. Population size is then estimated from the
proportion of the second sample that bears a mark from the first capture period. Generally
speaking, if the population is large, the marked individuals will have become diluted within it,
and only a few would be expected to appear in the second sample. If assumptions about the
sampling and animals’ distribution are correct, then the proportion of marked individuals in the
second sample is the same as that in the entire population.

II. Background

Assumptions
Like all estimation procedures, a number of assumptions must be met to validly use this method:
1. The two samples taken from the population must be random samples. That is, all
individuals in the population have an equal and independent chance of being captured
during the time of sampling. An adequate interval of time between captures must be
given to allow dispersal of the marked animals throughout the population.
2. There is no change in the ratio of marked to unmarked animals. This means that during
the time from initial capture to recapture, there must be no significant addition of
unmarked animals to the population through births or immigration.
3. Similarly, population losses from mortality and emigration must remove the same
proportion of marked and unmarked individuals.
4. Marking of individuals does not affect their mortality.
5. Individuals do not lose marks.

It is called Peterson’s method (Peterson 1896) by fisheries biologists, Lincoln’s method (Lincoln
1930) by mammalogists and ornithologists, and the Lincoln-Peterson index by ecologists who
are appreciative of the efforts of both contributors to this methodology. Surprisingly though,
Peterson did not mean for it to be used for population-size estimation, and Lincoln was scooped
by 13 years by another researcher, Dahl (LeCren 1965).

III. Calculations

Great, but how is this index calculated? Glad you asked. The total population may be estimated
as follows:
 Mark-recapture Sampling 2

Assume the total population size to be estimated contains N individuals. From this population,
take a sample of M individuals, mark these animals, and return them to the population. At a later
time, take a second sample of n individuals from the population; this sample contains R
recaptured animals (i.e., individuals captured and marked in the first sampling). Then the
population size, N, may be estimated by the following considerations:

M R
= (Equation 1)
N n
or
N n
= (Equation 2)
M R

Equation 1 says that the proportion of marked animals in the entire population is equal to the
proportion of marked animals in a random sample taken from the population. Equivalently,
Equation 2 says that the ratio of the total population to the number of animals marked on the first
date is equal to the ratio of the number caught on the second date to the number that were
recaptured on the second date. Rearrangement of either equation yields the following;

Mn
N= (Equation 3)
R

Example calculation
Are you like me, would an example help solidify your understanding of the theory behind this
method of population-size estimation? OK, suppose you take 100 white balls out of a pot having
an unknown number of white balls, paint them black, return them to the pot, and mix all the balls
in the pot thoroughly. If you then take 150 balls from the pot and find 25 of them to be black,
then M = 100, n = 150, R = 25, and the unknown total number of balls could be estimated using
Equation 3 from above:

Mn (100 )(150 )
N= = = 600 balls
R 25

However, Equation 3 may overestimate the population size (i.e., it is biased) when samples are
relatively small (Chapman 1951). NC is a nearly unbiased estimate of population size if the
number of recaptured animals, R, is at least 8 (Krebs 1989). This bias can be reduced by
computing

 (M + 1)(n + 1) 
Nc =   −1 (Equation 4)
 R +1 
 Mark-recapture Sampling 3

From the example above with the white balls,

 (100 + 1)(150 + 1) 
Nc =   −1
 25 + 1 
15,251
NC = –1
26
NC = 586.6 – 1
NC = 585.6 balls

The approximate variance, s2, of this estimate is

s2 =
(M + 1)(n + 1)(M − R )(n − R ) (Equation 5)
(R + 1)2 (R + 2)
From the example above with the white balls,

s2 =
(100 + 1)(150 + 1)(100 − 25)(150 − 25)
(25 + 1)2 (25 + 2)
142,978,125
s2 =
18,252
s2 = 7833.6
s = 88.5

With the standard deviation, s, (simply calculating the square root of the variance, s2) 95% and
99% confidence limits on the population estimate are given by

N (or NC) + 1.96(s) (95% confidence limits) (Equation 6)

and N (or NC) + 2.58(s) (99% confidence limits) (Equation 7)

For the above example, the 95% confidence interval (CI) for NC could be calculated as

95% CI = NC + 1.96(s)
= 585.6 + 1.96(88.5)
= 585.6 + 173.5
The 95% confidence interval = (412.1, 759.1)

Thus, we could say with 95% confidence that the true population size (total number of balls in
the pot) is between 412.1 and 759.1 balls.

As you might assume, one generally can compare the similarity in population estimates between
two populations (or two time periods for the same population) by simply calculating the
confidence limits of an estimate for both populations and determining whether these limits
overlap. If the confidence limits do not overlap, we generally consider them statistically
different.
 Mark-recapture Sampling 4

The precision with which the capture-recapture technique estimates population size is inversely
dependent on the number of marked animals recaptured. Thus, attempt to obtain a reasonably
large R by making n large.

Several newer, alternative methods use theory and sampling procedures based on multiple
markings and/or multiple recaptures for more complex study designs. These include the
Schnabel method (relaxes single marking and single recapture assumption) and the Jolly-Seber
method (relaxes closed population and survival assumptions).

Keep in mind that a population estimate is different from a density estimate. How so? Density
estimates refer to a number of individuals per unit area, so one must know the exact amount of
area from which animals were trapped. This is a tricky proposition for live-trapping studies, as
much research has shown that animals of different species, ages, or sexes may be drawn to traps
to differing degrees. Thus, it is difficult to determine exactly how much area is being trapped.
Population estimates simply yield a relative abundance or number of individuals per sampling
effort (i.e., number/trapline or number/watershed).

The data you will analyze for this lab are real data collected from an ongoing study on Konza
Prairie. You will use raw data sheets from live-trapping of small mammals to calculate
population estimates and confidence intervals, compare effects of a treatment (fire frequency) on
abundance of small mammals, and determine whether this real-world example is valid in light of
the inherent assumptions of the Lincoln-Peterson index.

Brief explanation of the data sheets

The six data sheets at the end of this lab are the results from live trapping done on the “Switch
Lines” on Konza Prairie. The study site is two watershed treatment units (R01A and R20A) that
bound each other in the southwestern corner of Konza. Unit R20A had been burned every spring
from 1981-2000. Unit R01A went unburned during that same time period (except for a wildfire
in 1994). Beginning in 2001, the treatments were switched on the two watersheds, with R01A
being burned annually, and R20A no longer being burned. The data you will be using were
collected on June 26 and 27, 2001; exactly two months after the first planned burn on R01A in
20 years. Six traplines are located on the site; three in watershed unit R20A, and three in
treatment unit R01A. Two of the traplines in each watershed have 20 stations each, whereas the
3rd trapline in each has only 10 stations. Two large (7.6 x 8.9 x 22.9 cm) Sherman live traps
were set at each station, for a total of 200 traps set across the entire site each night. Traps were
baited with a “kiss” of peanut butter and rolled oats, a delicious concoction which rodents (and
hungry graduate students) cannot resist. Most of the notation on the data sheet is self-
explanatory, but some parts require explanation. About 15 species of small mammals can be
captured on Konza, but the following abbreviations will be most useful for this exercise: PL =
Peromyscus leucopus (white-footed mouse), PM = Peromyscus maniculatus (deer mouse), SH =
Sigmodon hispidus (hispid cotton rat), and NF = Neotoma floridana (Eastern woodrat). Mass is
in grams. Toe clipping is a marking technique used to uniquely identify individuals, so each toe
clip number identifies a different individual in the population.
 Mark-recapture Sampling 5

Mark-Recapture Sampling Questions (20 pts. total)

Due March 4, 2003

1. Give a population estimate and 95% confidence interval for all rodents (all species combined)
on: a) the entire study site; b) each of the two treatment watershed units separately.
2. Give a population estimate and 95% confidence interval for only white-footed mice
(Peromyscus leucopus) on the entire study site.
3. Give a population estimate and 95% confidence interval for only deer mice (Peromyscus
maniculatus) on the entire study site. Note: you should not need to calculate anything new to
answer this question.
4. Compare the total rodent abundance (all species combined) on the two watershed units. Does
fire treatment appear to affect abundance of rodents in tallgrass prairie? How did you decide if
differences were significant?
5. Were the assumptions of the Lincoln-Peterson index satisfied for this study?

Related References

Brower, J. E., J. H. Zar, and C. N. von Ende. 1997. Field and laboratory methods for general
ecology, 4th Ed. W.C. Brown Publishing, Boston.

Chapman, D. G. 1951. Some properties of the hypergeometric distribution with applications to

zoological censuses. University of California Publications in Statistics 1:131-160.

Krebs, C. J. 1989. Ecological Methodology. Harper Collins Publishers, New York.

LeCren, E. D. 1965. A note on the history of mark-recapture population estimates. Journal of

Animal Ecology 34:453-454.

Lincoln, F. C. 1930. Calculating waterfowl abundance on the basis of banding returns. U.S.
Department of Agriculture Circular No. 118:1-4.

Peterson, C. G. J. 1896. The yearly immigration of young plaice into the Limfjord from the
German Sea. Report of the Danish Biological Station (1895) 6:5-84.