Unraveling abundance from occurrence: Modeling an endangered rodent population with low capture probability

TL;DR

This paper introduces a new method to estimate population abundance for elusive species like the Pacific pocket mouse using occupancy and detection data, which is more practical than traditional methods.

Contribution

The study presents a novel Bayesian approach to predict abundance from occupancy and detection data for species with low capture probability.

Findings

Occupancy and detection were positively correlated with density but could not fully explain all density variation.

The Bayesian method effectively integrates uncertainty in site-level abundance for population-level predictions.

The approach is applicable to other species with low capture probability and limited resources for direct density estimation.

Abstract

Predicting population abundance while accounting for uncertainty is an essential task for managers of endangered species but is often hindered by the challenge and expense of comprehensive data collection. Many traditional methods for estimating abundance of rare or elusive species are costly and logistically difficult, with occupancy‐based methods being a popular alternative. While the theoretical relationship between occupancy and abundance is well studied, there are few examples of methodological approaches for predicting abundance from occupancy. This study presents a novel approach to bridge the gap between abundance and occurrence for species with low capture probability, using the Pacific pocket mouse (Perognathus longimembris pacificus; PPM) in Southern California, USA, as a model system. PPM have been monitored across three subpopulations in this region using track tubes to inform occupancy over space and time and live captures to inform PPM demography and phenology. Paired capture–recapture data and presence–absence data collected between 2012 and 2022 were used to estimate density, occupancy, and detection, respectively. Density was modeled as a function of both occupancy and detection, and abundance at monthly and annual scales was predicted from estimates of density for all subpopulations. Our methodology leverages all available data in an integrated Bayesian analysis where uncertainty in site‐level abundance is naturally accounted for when scaling abundance estimates to the population level. While occupancy and detection were both predictive of and positively correlated with density, a meaningful amount of variation in density was not explained by our model, revealing avenues for future study as well as providing a realistic assessment of uncertainty in population‐level abundance predictions. In addition to advancing the current understanding of Pacific pocket mouse population dynamics, this approach is applicable to a wide array of species and ecosystems where population management is necessary, but individuals have low capture probability and available resources may preclude direct estimation of density at relevant spatial scales. From a design perspective, our results demonstrate the utility of strategically deploying density‐based monitoring methods within long‐term occupancy monitoring programs. More generally, our findings underscore the potential of this approach to inform methods to include abundance estimation in spatial occupancy monitoring programs for endangered species.

Figures