Capture-Recapture Methods: Estimating Population Size
Introduction
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
Estimating the size of wildlife populations is crucial for effective conservation and
management. However, directly counting every individual in a population is usually
impossible, especially for elusive species that reside over large areas. Capture-recapture
methods provide a scientifically robust way to estimate population size indirectly through
capturing, marking, releasing, and recapturing a sample of individuals from the target
population. Mark-recapture studies have been widely used across different taxonomic groups
including animals and plants. This assignment will discuss the theory, assumptions, and
different types of capture-recapture models used to estimate population size using this
approach.
The Basic Capture-Recapture Model
The most basic capture-recapture model involves capturing a sample of individuals from the
target population, uniquely marking them, releasing them back into the population, and then
recapturing another sample on a subsequent occasion. The key parameters estimated through
this model are N, the estimated total population size, and n1 and n2, the number of
individuals captured on the first and second sampling occasions respectively. From these, we
can derive the following probabilities:
- p1 = the probability of capture on the first occasion
- p2 = the probability of capture on the second occasion
- p12 = the probability of being captured on both occasions
Based on binomial probability theory, the relationship between these parameters is:
p12 = p1 * p2
And the probability of not being captured on either occasion is:
q1 * q2 = (1 - p1) * (1 - p2)
We can then estimate population size N as:
N = n1/p1
This assumes that:
1) The population is closed between sampling occasions - no births, deaths, immigration, or
emigration.
2) All animals have the same probability of capture on each sampling occasion.
3) Marks are not lost or missed during both sampling occasions.
4) Releases do not affect the probability of recapture.
Violating these assumptions will bias population estimates from the basic model. More
advanced models relax some of these assumptions.
Variations on the Basic Model
Many variations and extensions of the basic capture-recapture model have been developed to
relax its assumptions and provide more realistic population estimates:
Heterogeneous Catchability Model:
This model accounts for heterogeneity in individual capture probabilities by allowing p1 and
p2 to vary across the population. It fits a mixture model where a proportion π of the
population has capture probability θ1 and the remaining 1- π have θ2.
Open Population Models:
Jolly-Seber and related models relax the closed population assumption to allow for births,
deaths, immigration and emigration between sampling occasions. They provide estimates of
apparent survival (φ) and recruitment (pent) probabilities in addition to population size N.
Robust Design Model:
This combines information from closed-population sampling occasions within primary
periods that are open to demographic changes between them. It allows estimation of within-
and between-season demographic parameters.
Multistate Capture-Recapture Models:
These characterize individuals according to state (e.g. territory, group) and allow transition
probabilities between states. They provide more information on movement, dispersal and
survival related to different states.
Capture-Recapture Sampling Designs
There are various sampling designs that can be used in capture-recapture studies depending
on project objectives, resources, and characteristics of the study species/system. The main
designs include:
- Instantaneous sampling: Captures occur simultaneously over a very short time period to
minimize demographic changes between samples.
- Sequential sampling: Captures occur in discrete sampling sessions separated by some time
interval to allow for changes in the population.
- Mixture of methods: Combining instantaneous and sequential sampling into a robust design
to estimate within and between occasion parameters.
- Single vs. multiple occasions: Using two or more discrete capture periods to increase data
and relax assumptions.
- Single vs. multiple areas/traps: Sampling across an array of spatially separated capture
devices/locations to improve estimates.
- Trappability covariates: Collecting individual data (e.g. sex, size, trap behavior) to model
heterogeneity.
The most appropriate design depends on questions asked and maximizing number of marked
and recaptured individuals for robust estimates. Pilot studies help evaluate different designs.
Capture-Recapture Analysis and Estimation
Data from capture-recapture studies are typically analyzed using closed capture-recapture
models in Program MARK or open population models in program PRESENCE. The general
approach involves:
1. Defining candidate models based on biological hypotheses and objectives.
2. Fitting models and estimating parameters (N, survival, recruitment etc)
3. Model selection based on information criteria like AICc to identify most parsimonious
model.
4. Assessing model fit and assumptions using tools like Chi-square, bootstrap or simulation.
5. Validating estimates against independent data where possible.
6. Calculating precision of estimates usually reported as 95% confidence intervals.
7. Testing sensitivity of ‘best’ model to violations of assumptions through simulation.
8. Providing technical details on design, methods, analysis for transparency and replication.
With increasing computing power, Bayesian approaches are also gaining popularity as they
allow incorporating uncertainty and external information as priors.
Examples of Capture-Recapture Studies
Some well-known examples of capture-recapture studies estimated population size across a
range of taxa:
- Tigers in Nepal - Estimated 165 tigers using camera traps across habitat patches.
- Monarch butterflies overwintering sites - Estimated winter population across sites in
Mexico at approximately 300 million using mark-recapture of tagged individuals.
- Whale sharks off Seychelles - Aerial and boat based photo-ID mark-recapture estimated
population at 400-450 individuals.
- Elephants insavannas of Kenya - A combination of physical marking and naturalfeatures
like ear tears, tusk breaks used to estimate population declining from 12,000 to 5,000 over 25
years.
- Loggerhead sea turtles nesting beaches - Open robust design used tagging data across
seasons to estimate 34,000 females using Georgia nesting beaches annually.
- Polar bears in Southern Beaufort Sea - Capture-recapture of photo-identifiedindividuals
estimated population declining from 1,526 to 857 from 2001-2006.
Challenges and Limitations
While capture-recapture is a powerful approach, challenges and limitations exist:
- Reliance on adequate numbers captured for robust estimates. Low capture rates biasresults.
- Assumptions rarely completely met, biasing estimates. Model selection helps quantify
departures from assumptions.
- Individual heterogeneity in capture probabilities increases variance in estimates.
- Labor-intensive field work over long time scales often required for demographic studies.
- Capture stress can affect survival and behavior across repeated sampling.
- Limited by species characteristics like elusiveness, habitats, trapability, turnover rates.
- Analysis requires sophisticated statistical tools that may be inaccessible to some.
- Independent validation data not always available to check estimates.
- Practical constraints of cost, access, permits limit study scope and design options.
Despite limitations, when designed well, capture-recapture provides a robust technique for
quantifying otherwise elusive ecological parameters like population size over scales relevant
to management and policy.
Conclusion
In conclusion, capture-recapture methods have greatly advanced our ability to robustly
estimate wildlife population sizes that were previously unknown or poorly quantified.
Decades of development have relaxed unrealistic assumptions and improved ability to
estimate additional demographic parameters beyond just population size. New statistical tools
allow increasingly complex multi-state, multi-season, and Bayesian analysis approaches.
Though challenging to implement, capture-recapture remains one of the most scientifically
defensible approaches for monitoring wildlife populations and evaluating trends over time -
crucial information for effective conservation decision making. Future opportunities lie in
integrating capture-recapture data with other monitoring tools like remote cameras,
environment DNA and telemetry to exploit new statistical modeling capabilities.
