Thursday, August 7, 2008: 1:30 PM
101 B, Midwest Airlines Center
Background/Question/Methods
Capture-mark-recapture methods for estimating growth and survivorship of individual organisms are well-established in ecology, but many species cannot be effectively marked because of handling or logistical constraints. However, researchers can often approximately identify individuals by size or other characteristics, and can often make good informal guesses about identity based on continuity --- i.e., "an individual of this approximate size was seen here last week, so this is probably the same individual". We use a Bayesian hierarchical model to formalize this approach. Given prior information about measurement error and resight probability as a function of size, we use a Gibbs sampler that alternates between estimating the true identities and fates (settlement, growth, mortality) of individuals based on size-dependent demographic rates and estimating the demographic rates based on the "known" identities and fates. In principle, this approach can be further extended to estimate variation in demographic rates in time (e.g. settlement pulses) and space (e.g. variation in habitat quality).
Results/Conclusions
We apply the method to a data set of repeated surveys of Thalassoma hardwicke in Moorea, French Polynesia. Given sufficiently strong prior information about the measurement error, and if necessary providing some prior constraints on survivorship and growth, we can successfully estimate demographic rates of unmarked (but approximately identifiable) individuals.
Capture-mark-recapture methods for estimating growth and survivorship of individual organisms are well-established in ecology, but many species cannot be effectively marked because of handling or logistical constraints. However, researchers can often approximately identify individuals by size or other characteristics, and can often make good informal guesses about identity based on continuity --- i.e., "an individual of this approximate size was seen here last week, so this is probably the same individual". We use a Bayesian hierarchical model to formalize this approach. Given prior information about measurement error and resight probability as a function of size, we use a Gibbs sampler that alternates between estimating the true identities and fates (settlement, growth, mortality) of individuals based on size-dependent demographic rates and estimating the demographic rates based on the "known" identities and fates. In principle, this approach can be further extended to estimate variation in demographic rates in time (e.g. settlement pulses) and space (e.g. variation in habitat quality).
Results/Conclusions
We apply the method to a data set of repeated surveys of Thalassoma hardwicke in Moorea, French Polynesia. Given sufficiently strong prior information about the measurement error, and if necessary providing some prior constraints on survivorship and growth, we can successfully estimate demographic rates of unmarked (but approximately identifiable) individuals.
