Reliability of multinomial N-mixture models for estimating abundance of small terrestrial vertebrates

Abstract

Information on population abundance is important to correctly plan conservation and management of animal populations. In general, capture-mark-recapture (CMR) is considered the most robust technique to estimate population abundance, but it is costly in terms of time and effort. Recently, binomial N-mixture models, based on counts of unmarked individuals, have been widely employed to estimate abundance. These models have limits and their reliability has been criticized. In the majority of cases, multinomial N-mixture models based on multiple observer protocols, that are hierarchical extensions of simple CMR, are applied in estimating abundance of animals with large body size, conspicuous behavior or high detection probabilities. We applied and evaluated the reliability of a multinomial N-mixture modelling approach with multiple observer data to a small and cryptic terrestrial salamander, found in different habitats where populations possess different level of detectability. Estimates obtained with multinomial N-mixture models were compared to estimates obtained with classical methods, such as removal sampling, and their reliability has also been evaluated by simulations scenarios. Our results show that multinomial N-mixture models, applied within a multiple observer framework, give reliable and robust estimates of population abundance even when detection and density are relatively low. Therefore, multinomial N-mixture models appear efficient and cost-effective when planning and identifying management actions and conservation programs of small terrestrial animals such as amphibians and reptiles.