Estimating species richness with capture-recapture models: Choice of model when sampling in heterogeneous conditions

Abstract

Capsule: The capture-recapture heterogeneity model M(h) can be used confidently to estimate species richness when using Breeding Bird Survey (BBS) data. Aims: To test the relevance of various probabilistic capture-recapture models for estimating species richness using data from large-scale, multi-observer monitoring schemes (e.g. BBS) sampling heterogeneous habitats. Methods: The program CAPTURE and its model selection algorithm were used to investigate the relevance of different capture-recapture models for estimating species richness within 601 BBS sampling units. We investigated whether the habitat in each sampling unit had any influence upon the model selected. We tested the efficiency of models to account for spatial variation in detection probability in sampling units exhibiting heterogeneous habitat cover and structure. We tested the potential efficiency of some models to account for behavioural variations in detection probabilities of data collected by under-qualified observers. Results: The most frequently selected model, among all sampling units, was the heterogeneity model, M(h), which assumes variation among species in detection probability. Spatial variations of the probability of detection among sampling points within the unit were not related to heterogeneity in habitat cover or structure. A lower probability of detection in the first sampling points than in the following ones was not linked to 'detection-happiness' of under-qualified observers. Conclusion: The heterogeneity capture-recapture model M(h) and its associated jackknife estimator are reliable in estimating local species richness from data obtained through large-scale, multi-observer breeding bird surveys, when habitats are heterogeneous within sampling units. © 2005 British Trust for Ornithology.