Written and maintained by CASRAI Editorial Board
Published
On this page
Mark-recapture is a quantitative ecological census methodology used to estimate the absolute abundance, density, survival rates, and demographic parameters of wild animal populations where a complete direct count is physically impossible. Operating on the principle of proportional sampling, researchers capture a known number of individuals, mark them with harmless identifiers, release them to mix freely within the parent population, and conduct subsequent recapture events to calculate total population size from the ratio of marked to unmarked individuals.
Definition and scope
Mark-recapture, also termed capture-mark-recapture (CMR) or capture-recapture, provides the mathematical framework for estimating unobservable biological population parameters across terrestrial, freshwater, and marine habitats. In wild populations, animals are mobile, cryptic, and non-uniformly distributed. Direct total counts fail because detection probability is rarely 1.0; unadjusted counts provide only an index of relative abundance confounded by variable catchability. Mark-recapture overcomes this barrier by treating encounter histories as probabilistic sampling events where catchability is explicitly modeled to derive unbiased absolute abundance.
The conceptual foundation originated with Danish biologist C.G.J. Petersen in 1896, who tagged European plaice (*Pleuronectes platessa*) to quantify marine fishery exploitation, and Frederick C. Lincoln in 1930, who published waterfowl population estimates from banding returns for the U.S. Department of Agriculture. Contemporary quantitative frameworks have expanded these historical indices into multi-state likelihood models accounting for temporary emigration, transition dynamics, and individual detection heterogeneity.
In scientific research and environmental management, mark-recapture protocols operate under guidelines established by federal and international agencies, including the U.S. Geological Survey (USGS), NOAA Fisheries, the U.S. Forest Service (USFS), and the UN Food and Agriculture Organization (FAO). These bodies employ capture-recapture to set marine commercial harvest limits, track recovery of endangered species under the Endangered Species Act, and monitor zoonotic disease transmission. In structured research programs, establishing clear boundaries through formal protocols for defining the population is mandatory before field sampling commences, ensuring data integrity across broader ecosystem ecology monitoring networks.
How it works or how it is used in practice
Field implementation of mark-recapture proceeds through structured stages, progressing from initial capture and tagging to post-release mixing and recapture. Rigorous field programs follow a six-stage standard protocol:
- Boundary Delineation and Closure Assessment: Biologists establish geographic boundaries and ensure demographic closure holds throughout the sampling interval. In stream studies, teams install upstream and downstream block nets to prevent movement, while terrestrial mammal surveys target brief intervals between reproductive pulses.
- Initial Capture (M): A representative cohort is caught using standardized gear, such as electrofishing units, mist nets, or box traps. Every individual is inspected for health, developmental stage, and physical condition.
- Mark Application: Animals receive permanent, biocompatible marks that do not impede movement or increase predation risk, ranging from fin clips and visible elastomer to passive integrated transponder (PIT) tags or acoustic transmitters.
- Recovery and Release: Marked individuals are held in aerated tanks until recovered from handling stress. Any injured or impaired specimens are excluded from the marked pool prior to release.
- Homogeneous Mixing Interval: The marked cohort is released across initial capture sites. Sufficient time is allowed for marked animals to distribute randomly among unmarked residents before resampling.
- Recapture Census (C): A second sampling event collects a total sample (C), recording newly caught unmarked individuals alongside recaptured marked animals (R).
The foundational mathematical relationship is the Lincoln-Petersen ratio. Assuming the proportion of marked animals in the second sample equals the proportion of marked animals in the total population, the formula is:
M / N = R / C
Solving for total population abundance (N) yields:
N̂ = (M × C) / R
Where N̂ is estimated population size, M is the number of marked animals released, C is total catch in the second sample, and R is the number of recaptured marks.
While mathematically direct, statistical evaluations by D.G. Chapman in 1951 demonstrated that the Lincoln-Petersen estimator is biased when sample sizes are small or recapture counts (R) fall below 10. Because R appears in the denominator, small values create positive bias, overestimating population size, and become undefined if R = 0. Chapman derived a nearly unbiased estimator:
N̂_c = [ (M + 1)(C + 1) / (R + 1) ] – 1
Chapman proved that N̂_c is exactly unbiased whenever total captures across both events equal or exceed population abundance (M + C ≥ N). White et al. (1982) in their Los Alamos National Laboratory and federal wildlife reference manual (Report LA-8787-NERP) established Chapman’s formulation as the standard for closed populations. For uncertainty analysis, G.A.F. Seber (1970, 1982) derived an exact, unbiased variance estimator:
Var(N̂_c) = [ (M + 1)(C + 1)(M – R)(C – R) ] / [ (R + 1)² (R + 2) ]
The standard error equals the square root of the variance, SE(N̂_c) = √Var(N̂_c). Researchers then construct an approximate 95% confidence interval using N̂_c ± 1.96 × SE(N̂_c), providing defensible bounds for wildlife management.
Types or related concepts
Capture-recapture models divide into two broad statistical classes: closed-population models that assume fixed abundance during the survey, and open-population models that incorporate demographic turnover.
In Wildlife Monographs No. 62, Otis et al. (1978) established the classic framework for closed populations, categorizing models by sources of variation in capture probability (p):
- Model M₀ (Null Model): Assumes constant capture probability across all individuals and sampling sessions.
- Model M_t (Temporal Variation): Accounts for capture probabilities varying over time due to weather, discharge, or changing trapping effort.
- Model M_b (Behavioral Response): Models learned responses where first capture alters subsequent catchability, distinguishing trap-shy behavior (recapture probability c falls below initial capture probability p) from trap-happy behavior (c exceeds p).
- Model M_h (Individual Heterogeneity): Models innate individual differences in catchability resulting from age, sex, social rank, or territory quality, using jackknife and Chao non-parametric estimators.
- Combination Models (M_tb, M_th, M_bh, M_tbh): Integrate multiple concurrent sources of variation using likelihood estimators implemented in Program CAPTURE and Program MARK.
Multi-sample closed surveys spanning three or more visits employ the Schnabel (1938) method, which treats successive sessions as cumulative Lincoln-Petersen censuses to narrow confidence bounds, or the Schumacher-Eschmeyer regression technique.
Over extended monitoring periods, demographic closure fails as births, deaths, and migration occur. Biologists then transition to open models formalized by Kenneth Pollock et al. (1990) in Wildlife Monographs No. 107:
- Cormack-Jolly-Seber (CJS) Model: Evaluates marked cohorts across time to estimate apparent survival rates (φ) and capture probabilities (p), isolating mortality and permanent dispersal.
- Full Jolly-Seber (JS) Model: Estimates absolute population size (N_t) at each sampling interval alongside apparent survival and new recruitments (B_t).
- Pollock’s Robust Design: Combines open and closed methodologies. Primary periods (e.g., annual intervals) allow demographic turnover, while nested secondary periods (e.g., consecutive daily visits) assume closure. This permits robust closed-model abundance estimation while tracking open-model survival and temporary emigration.
- Spatial Capture-Recapture (SCR): Incorporates spatial coordinates of capture devices (camera stations, hair snares) to estimate animal home ranges and absolute density (individuals per square kilometer) without arbitrary study area boundaries.
These dynamic methods differ fundamentally from static forest mensuration techniques such as the timber cruise, which measures fixed spatial plots without encounter probabilities. The interactive data table below compares these statistical capture models across population states, parameter outputs, and federal literature sources.
Worked example: Chapman estimator and variance calculation
The following worked example illustrates the complete calculation workflow for a closed two-sample census of rainbow trout (*Oncorhynchus mykiss*) in an isolated mountain stream reach, following the field methods evaluated by Rosenberger and Dunham (2005) in USDA Forest Service research.
Field Sampling Data: A 200-meter stream section is closed using upper and lower block nets to prevent fish movement.
- Day 1 (Marking Event): Backpack electrofishing yields M = 140 trout. Each fish is marked with an upper caudal clip and released after full recovery.
- Day 3 (Recapture Event): After a 48-hour mixing window, a second electrofishing pass collects C = 110 trout. Inspection reveals R = 35 marked recaptures and 75 unmarked fish.
Step 1: Raw Lincoln-Petersen Estimate:
N̂ = (M × C) / R = (140 × 110) / 35 = 15,400 / 35 = 440.0 individuals
Step 2: Chapman Bias-Corrected Estimate:
N̂_c = [ (M + 1)(C + 1) / (R + 1) ] – 1
N̂_c = [ (141 × 111) / 36 ] – 1 = [ 15,651 / 36 ] – 1 = 434.75 – 1 = 433.75 ≈ 434 individuals
The Chapman correction lowers the estimate by 6 individuals, counteracting small-sample positive bias.
Step 3: Seber Unbiased Variance:
Var(N̂_c) = [ (M + 1)(C + 1)(M – R)(C – R) ] / [ (R + 1)² (R + 2) ]
Var(N̂_c) = [ (141)(111)(140 – 35)(110 – 35) ] / [ (36)² (37) ]
Var(N̂_c) = [ 141 × 111 × 105 × 75 ] / [ 1,296 × 37 ]
Var(N̂_c) = 123,251,625 / 47,952 = 2,570.31
Step 4: Standard Error and 95% Confidence Interval:
SE(N̂_c) = √2,570.31 = 50.70 individuals
CI₉₅ = 433.75 ± (1.96 × 50.70) = 433.75 ± 99.37
Lower 95% Bound = 433.75 – 99.37 = 334.38 ≈ 334 individuals
Upper 95% Bound = 433.75 + 99.37 = 533.12 ≈ 533 individuals
Step 5: Sampling Precision (CV):
CV = [ SE(N̂_c) / N̂_c ] × 100% = (50.70 / 433.75) × 100% = 11.69%
Because the CV is below 20%, the survey meets the precision threshold established by White et al. (1982) for management planning.
Where it matters for labs, field stations and supply decisions
Implementing mark-recapture programs requires careful hardware procurement and laboratory coordination. Selecting appropriate marking equipment directly impacts animal welfare, mark retention, and statistical accuracy:
- Streamer and Dart Tags: In marine pelagic programs, including NOAA Fisheries’ Cooperative Tagging Center, dart tags provide durable tracking for billfish and tuna. Stations must stock stainless steel applicator needles, dart poles, and corrosion-resistant anchors, ensuring tags seat properly into dorsal musculature without inducing trauma.
- Passive Integrated Transponders (PIT Tags): For freshwater fish, amphibians, and small mammals, 134.2 kHz ISO-compliant PIT tags offer permanent electronic identification. Because they operate without internal batteries, tags last the animal’s lifetime. Stations require sterile implanters, disinfectant baths, and portable RFID readers or stationary pass-through loop antennas.
- Visible Implant Elastomer (VIE): For larval organisms and small fish, medical-grade fluorescent elastomer injected under transparent tissue provides an unobtrusive batch mark, requiring micro-syringes, specialized needles, and LED excitation flashlights.
- Non-Invasive Genetic Sampling: Modern wildlife surveys increasingly substitute physical tagging with DNA extracted from hair snares, scat, or environmental DNA (eDNA). Laboratories stock DNA extraction kits, PCR primers, and capillary sequencers to genotype microsatellite or SNP loci.
Operational compliance demands strict biosecurity and chemical hygiene under institutional standard operating procedure (SOP) mandates. Field crews require pharmaceutical-grade anesthetics such as buffered MS-222, aerated recovery vessels, and broad-spectrum disinfectants (such as Virkon S) to sanitize waders, nets, and holding tanks between watersheds, preventing pathogen transmission.
Common misunderstandings
Mark-recapture analyses are subject to critical practitioner misconceptions that introduce severe parameter bias:
- Assuming Population Closure Without Testing: Applying closed models over extended periods violates core assumptions. If unrecorded births or immigration occur, newly arrived unmarked individuals dilute the marked ratio, causing positive bias that overestimates population size. Equal mortality among marked and unmarked individuals does not bias closed abundance at marking, but distorts open vital rate estimations.
- Confusing Apparent Survival with True Survival: In open Cormack-Jolly-Seber models, apparent survival (φ) represents the combined probability that an individual survives and remains in the study area. Apparent survival conflates mortality with permanent emigration; individuals emigrating outside the sampling zone appear statistically deceased unless auxiliary telemetry is used.
- Ignoring Trap-Response Bias: Unmodelled behavioral responses invalidate Lincoln-Petersen estimates. If handling induces trap-shyness (recapture probability falls below initial capture probability), recapture counts are suppressed, heavily inflating population estimates. If food rewards induce trap-happiness (recapture probability exceeds initial capture probability), recapture counts are inflated, severely underestimating true abundance. Goodness-of-fit testing in Program CAPTURE identifies these behavioral shifts to fit Model M_b.
- Overlooking Tag Loss: Shed tags cause recaptured individuals to be recorded as unmarked. This artificially depresses R and inflates population estimates. Monitoring programs use double-tagging protocols to empirically measure tag loss and adjust estimates accordingly.
- Equating Removal Counts with Total Abundance: Uncalibrated catch-per-unit-effort or simple removal counts fail to account for declining catchability. As shown by Rosenberger and Dunham (2005), fish exhibit behavioral avoidance during successive electrofishing passes, causing removal models to produce severe negative bias compared to mark-recapture standards.
Frequently asked questions
What is the primary difference between closed and open mark-recapture models?
Closed mark-recapture models assume the population experiences zero demographic turnover—no births, deaths, immigration, or emigration—during the study period, focusing on estimating absolute abundance (N). Open models, such as Cormack-Jolly-Seber and Jolly-Seber, accommodate ongoing demographic processes across multiple seasons, estimating apparent survival, recruitment, and population growth alongside abundance.
When should researchers use the Chapman estimator instead of the Lincoln-Petersen formula?
The Chapman estimator should serve as the standard default for two-sample closed-population studies. The classical Lincoln-Petersen formula produces positive bias when sample sizes are small or recapture numbers fall below 10, and it is undefined if zero marked animals are recaptured. Chapman’s modification eliminates this bias and is exactly unbiased when total captures across both sessions equal or exceed population abundance.
How does tag loss affect population estimates in mark-recapture studies?
Tag loss causes marked individuals to be misclassified as unmarked upon recapture, reducing the recorded number of recaptures (R) relative to total catch. This introduces positive bias, leading to significant overestimation of the true population. Field programs employ double-tagging trials to quantify shedding rates and mathematically correct abundance estimates.
What is Pollock’s robust design and why is it used?
Pollock’s robust design is a hierarchical sampling structure combining open and closed models. Widely spaced primary periods (such as annual seasons) permit demographic turnover, while nested secondary periods (such as consecutive daily surveys) assume closure. This allows researchers to generate robust closed-model abundance estimates while simultaneously tracking open-model survival and temporary emigration.
Sources
- Otis, D.L., Burnham, K.P., White, G.C., and Anderson, D.R. (1978). Statistical inference from capture data on closed animal populations. Wildlife Monographs, No. 62, pp. 1–135. U.S. Geological Survey Publications Warehouse. https://pubs.usgs.gov/publication/70119899. Checked 2026-11-14.
- Pollock, K.H., Nichols, J.D., Brownie, C., and Hines, J.E. (1990). Statistical inference for capture-recapture experiments. Wildlife Monographs, No. 107, pp. 1–97. U.S. Geological Survey Publications Warehouse. https://pubs.usgs.gov/publication/5230200. Checked 2026-11-14.
- White, G.C., Anderson, D.R., Burnham, K.P., and Otis, D.L. (1982). Capture-recapture and removal methods for sampling closed populations. Los Alamos National Laboratory Report LA-8787-NERP, U.S. Department of Energy Office of Scientific and Technical Information. https://www.osti.gov/biblio/6729610. Checked 2026-11-14.
- NOAA Fisheries, Southeast Fisheries Science Center (2024/2026). Cooperative Tagging Center: Marine Tagging and Mark-Recapture Data for Highly Migratory Species. National Oceanic and Atmospheric Administration. https://www.fisheries.noaa.gov/southeast/atlantic-highly-migratory-species/cooperative-tagging-center. Checked 2026-11-14.
- Rosenberger, A.E., and Dunham, J.B. (2005). Validation of abundance estimates from mark-recapture and removal techniques for rainbow trout captured by electrofishing in small streams. North American Journal of Fisheries Management, 25(4): 1395–1410. U.S. Forest Service Research and Development Treesearch. https://research.fs.usda.gov/treesearch/23869. Checked 2026-11-14.
- Pauly, D. (1983). Some simple methods for the assessment of tropical fish stocks. FAO Fisheries Technical Paper / Circular No. 729, Food and Agriculture Organization of the United Nations. https://www.fao.org/4/x6845e/X6845E03.HTM. Checked 2026-11-14.
- Thorpe, J.E. (1975). Estimation of the exploitable stock of brown trout Salmo trutta L. in Loch Leven, Kinross, Scotland. Symposium on the Methodology for the Survey, Monitoring and Appraisal of Fishery Resources in Lakes and Large Rivers, EIFAC Technical Paper T23, FAO. https://www.fao.org/4/aa043b/AA043B20.htm. Checked 2026-11-14.
Data: Capture-Recapture Statistical Models and Parameter Specifications
Original methodological compilation comparing closed, open, and robust design capture-recapture models, detailing population state assumptions, minimum sampling occasions, estimable demographic parameters, small-sample bias corrections, analytical software frameworks, and primary federal research references from USGS, DOE OSTI, and NOAA Fisheries.
Period covered: 1978/2026. Coverage: United States and Global Marine/Terrestrial Ecosystems.
| Model / Estimator | Population State | Minimum Sampling Events (occasions) | Primary Estimated Parameters | Small-Sample Correction | Primary Analytical Software | Primary Source Reference |
|---|---|---|---|---|---|---|
| Lincoln-Petersen (1930) | Demographically Closed | 2 | Population Abundance (N) | Chapman (1951) modification | Program CAPTURE, R (Rcapture) | Otis et al. (1978); White et al. (1982) |
| Schnabel (1938) Multi-Census | Demographically Closed | 3 | Population Abundance (N) | Weighted cumulative least-squares | Program CAPTURE, Program MARK | Otis et al. (1978); White et al. (1982) |
| Model M_b (Behavioral Response) | Demographically Closed | 3 | Abundance (N), Initial catch (p), Recapture (c) | Generalized removal estimator | Program CAPTURE, Program MARK | Otis et al. (1978) |
| Model M_h (Individual Heterogeneity) | Demographically Closed | 3 | Abundance (N), Mean detection probability (p_bar) | Jackknife and Chao estimators | Program CAPTURE, Program MARK | Otis et al. (1978) |
| Cormack-Jolly-Seber (CJS) | Demographically Open | 3 | Apparent Survival (phi), Capture Probability (p) | Profile likelihood and bootstrap | Program MARK, R (RMark) | Pollock et al. (1990) |
| Full Jolly-Seber (JS) | Demographically Open | 3 | Abundance (N_t), Survival (phi_t), Recruitment (B_t) | POPAN parameterization / logit link | Program MARK, R (mra) | Pollock et al. (1990) |
| Pollock's Robust Design | Hierarchical (Open Primary / Closed Secondary) | 4 | Abundance (N), Survival (phi), Temporary Emigration (gamma) | Full-rank joint likelihood estimation | Program MARK, R (marked) | Pollock et al. (1990) |
Sources
- Statistical inference from capture data on closed animal populations (accessed 2026-11-14)
- Statistical inference for capture-recapture experiments (accessed 2026-11-14)
- Capture-recapture and removal methods for sampling closed populations (accessed 2026-11-14)
- Cooperative Tagging Center: Marine Tagging and Mark-Recapture Data for Highly Migratory Species (accessed 2026-11-14)
- Validation of abundance estimates from mark-recapture and removal techniques for rainbow trout captured by electrofishing in small streams (accessed 2026-11-14)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







