Estimating nest density with detection probability

Forum for discussion of general questions related to study design and/or analysis of existing data - software neutral.

Estimating nest density with detection probability

Postby mattstep » Wed Jun 02, 2021 12:07 pm

Hello,

I have a dataset that I would like to use to estimate abundance/density of grassland bird nests corrected for detection probability, but it is unclear which analysis framework is most appropriate. The data was collected by searching 300 0.1 ha plots once per week with alternating pairs of searchers. Plots were randomly placed within 11 opportunistically selected study sites and stratified by conservation practice with incomplete replication among sites and years. The searchers did not share information about nests they found so that when the next pair of searchers came the following week, they would have an opportunity to discover (or not) a nest that was known to be present. Nests were confirmed to still be active by the pair that originally discovered them after plots were searched for the day and nest fate was followed through time for a related nest survival analysis. The setup was generally similar to Probability of detection of nests and implications for survey design by Smith et al 2009, but those authors focused on modeling the detection probability and did not model the density.

These “detection trials” could be analyzed separately in a binomial GLMM framework to predict detection probability for each plot search, but then it is not clear how to pass the uncertainty of the detection probability estimate through to a prediction of nest density modeled in a separate Poisson or negative binomial GLMM analysis. My study based on a double-observer approach, but because of the sessile nature of nests, once nest locations are known by a pair, subsequent ‘re-discoveries’ by that pair are not independent because searchers can remember the locations of nests they have found previously, violating assumptions of equal catchability for marked and un-marked individuals. We also know that new nests are added each week and some nests fail each week, violating closed population assumptions for many approaches. I am interested in plot-level estimates, which means “site” violates the independence assumption for many approaches as well, although this may be mitigated with carefully chosen spatial covariates.

Does MARK have a method that could be adapted for this study? Our data violates the Jolly-Seber and Pradel assumption of equal probability of capture for marked and unmarked individuals. Robust design does not seem to work because there is no replication when the secondary period is short enough to ensure demographic closure and the fidelity and availability parameters don’t make much sense in this context. I have not been able to find a method in MARK where my data does not violate one or more assumptions.

If possible, I would like to find a method that leverages my double-observer or removal data, but I could set that part of the dataset aside if there isn’t an existing method that makes use of it. For example, outside of MARK I have considered using an occupancy framework to estimate abundance following Royle and Nichols (2003), or using the N-mixture modeling extension from Dail and Madsen (2011) that allows estimation of abundance from repeated counts in a demographically open metapopulation. Both of these methods are implemented in the R package ‘Unmarked’.

I would appreciate any comments, suggestions, hints, or directions for where to look in the literature.

Much Thanks,

Matt
mattstep
 
Posts: 1
Joined: Tue Jun 01, 2021 3:53 pm

Return to analysis & design questions

Who is online

Users browsing this forum: No registered users and 15 guests