Does movement bias abundance estimates?: Improving unmarked abundance estimators by accounting for animal activity patterns

Estimating demographic rates is critical for informed wildlife management, and non-invasive methods such as camera traps are an important tool for estimating abundance. Unmarked abundance models for species without individually identifiable marks can provide accurate estimates under ideal conditions, but the imperfect availability of animals may bias estimates. Cameras sample twenty-four hours a day, but animals may not be active for the entire diel period, creating the potential for individuals to be unavailable for detection during some sampling periods. Additionally, small study areas are vulnerable to animals leaving and violating geographic closure. Our objectives were to quantify how animal activity patterns and movement out of a study area influence abundance estimates from unmarked models and to develop potential correction factors to reduce bias in estimates. First, we simulated animal populations with varying activity patterns and with and without geographic closure. We quantified the bias imperfect availability and geographic closure had on Space to Event estimates and tested three methods to correct it. We determined that when animals violated availability in one way (i.e., unavailable for the full diel period or lack of geographic closure), estimates without corrections were negatively biased by 50–75% while corrections-adjusted estimates captured true abundance. However, when animals violated availability in both ways, applying one correction still underestimated by 50–62% while combining corrections that addressed both facets of availability yielded underestimates of only 14% or captured true abundance. Next, we estimated elk abundance in the Ruby Mountain Range in southwest Montana, leveraging concurrent camera trap and GPS-collar data to improve estimates with our correction factors. We determined that all three corrections increased estimates; the largest were unreasonable for our study area compared to similar areas, and the smallest provided reasonable estimates. We also employed two additional corrections specific to the Ruby Range elk, which either decreased or had little effect on estimates. By combining our simulations and real-world application to elk, we show that accounting for species’ activity patterns and space use is essential for accurately estimating population parameters, and our tested correction factors will improve abundance estimation for management and conservation of many taxa.