Monitoring and habitat inferences change with population metric: a case study with mesocarnivores

Context Accurate and precise estimates of wildlife abundance and distribution are critical for robust ecological inference and effective management. However, obtaining this information for mesocarnivores is challenging because they are elusive and highly mobile. Aims To compare four common population metrics (occupancy, local abundance, relative abundance, and density) for monitoring unmarked populations and the influence of three habitat covariates on these population metrics. Methods For five mesocarnivores species we used data collected at 74 camera traps deployed in the northeastern USA in summer 2021 to fit (1) models that estimated probabilistic occupancy, (2) Royle–Nichols models that estimated local abundance, (3) Poisson distributed general linear models that estimated relative abundance, and (4) random encounter and staying time (REST) models that estimated density. We also quantified habitat relationships across these four different models and compared the resultant inferences. Key results Density and relative abundance had the highest correlation (Pearson correlation (r) = 0.91), whereas occupancy and density had the lowest correlation (r = 0.19). Density estimates for all species were consistent with expectations and similar to those reported in previous studies. The effects of habitat covariates changed across metrics, such that a significant effect of a covariate on one metric was not indicative of a significant influence on the other metrics. There were only two instances of a significant effect of a covariate on all metrics, and two instances where the influence of a covariate had opposite, albeit insignificant, effects on two metrics. Conclusions Estimates of occupancy and local abundance for mesocarnivores derived from camera traps may not be reliable proxies for density. However, relative abundance, as derived from detection rates, could be a promising means of monitoring density with less intensive data processing. Mesocarnivore habitat relationships changed across these metrics. Implications When designing monitoring or research programs, practitioners should be cautious about assuming that inferences derived from camera trap estimates of these four population metrics are interchangeable. Further, we highlight how the REST model offers a promising new means for monitoring multiple mesocarnivores simultaneously, and likely other unmarked species, via density estimates.