Population Size Estimation Using Capture-Recapture Methods

Capture–recapture methods estimate the size of a population by repeatedly sampling individuals, marking or identifying them, and examining the overlap among capture occasions. Originally developed in ecology to assess wildlife abundance, these methods have been extended to human populations, epidemiology and criminology, where units of study may be animals, case reports or individuals in hidden communities. Classical estimators assume closed populations and uniform detectability, yet real‐world applications must address open dynamics, heterogeneous detection probabilities, behavioural responses to capture and spatial structure. Advances over the past decade include spatially explicit capture–recapture (SECR) models that incorporate movement and landscape features, Bayesian hierarchical frameworks that borrow strength across strata, and the integration of molecular methods—such as environmental DNA—to detect elusive species. The global significance of these approaches spans conservation planning for endangered taxa, evaluation of disease‐control measures through contact‐tracing data, and assessment of clandestine activities. Robust design protocols, non‐parametric lower‐bound estimators and model‐averaging techniques further enhance precision and account for uncertainty in detection processes.

Research from Nature Portfolio

Recent studies have demonstrated the fusion of environmental DNA metabarcoding with traditional capture–recapture frameworks to estimate aquatic and semi‐aquatic species. By treating eDNA detections as non‐invasive capture events, researchers have achieved more reliable abundance estimates for cryptic fish and amphibian populations, especially in remote river systems. Another line of work has refined Bayesian hierarchical capture–recapture models to accommodate spatial heterogeneity in detection probability across fragmented landscapes. These models employ random effects to represent site‐specific covariates such as habitat quality or observer effort, yielding improved inference on total population size and enabling scenario‐based predictions under alternative management strategies.

Research from all publishers

A novel continuous‐time interaction process has been introduced to model self‐exciting and self‐correcting events in capture–recapture data, with applications to estimating the number of individuals involved in illicit networks. By formulating a conditional likelihood that includes only subjects with at least one capture, this approach captures temporal dependence and covariate effects in a unified framework. Separately, work on zero‐truncated and one‐inflated count data has produced estimators that correct severe biases arising when single captures occur more often than expected. Modified lower‐bound estimators based on counts of two and three avoid overestimation in one‐inflated settings, and bootstrap methods provide variance estimates. Finally, a Bayesian treatment of one‐inflated count distributions for elusive populations has employed Gibbs sampling to obtain posterior distributions for population size under Poisson, geometric and negative‐binomial baselines, illustrating practical gains in official statistics and wildlife assessments.

Population Size Estimation Using Capture-Recapture Methods publication trend

The graph below shows the total number of articles in population size estimation using capture-recapture methods across all publications each year (not limited to Nature Index journals).

Technical terms

Detection probability: The likelihood that an individual present in the population is observed or recorded during a capture occasion.

Zero-truncation: A feature of count data in which unobserved (zero‐count) individuals are absent from the sample, necessitating special modelling to account for unseen units.

One-inflation: An excess of individuals observed exactly once relative to a baseline count distribution, which can bias population size estimates if unaddressed.

Bayesian hierarchical model: A statistical framework that layers parameter estimation across multiple levels (for example, sites or time periods), incorporating prior information and quantifying uncertainty via posterior distributions.

Environmental DNA (eDNA): Genetic material shed into the environment by organisms, used as a non‐invasive proxy for presence or “capture” in population surveys.

References

1. The VGAM Package for Capture-Recapture Data Using the Conditional Likelihood. Journal of Statistical Software (2015).
2. A modification of Chao’s lower bound estimator in the case of one-inflation. Metrika (2018).
3. Population size estimation based upon zero-truncated, one-inflated and sparse count data. Statistical Methods & Applications (2021).
4. Continuous Time-Interaction Processes for Population Size Estimation, with an Application to Drug Dealing in Italy. Biometrics (2022).
5. Bayesian analysis of one‐inflated models for elusive population size estimation. Biometrical Journal (2022).
6. Modeling COVID-19 Contact-Tracing Using the Ratio Regression Capture–Recapture Approach. Biometrics (2023).