Using drones to estimate moose densities at a landscape scale

Accurate estimates of wildlife densities are important for understanding population dynamics and informing management decisions, but obtaining such estimates can be logistically challenging and costly. Drones and associated sensors provide a novel and potentially efficient pathway to estimate wildlife densities but require critical evaluation of efficacy across large spatial scales. We evaluated the capability of drones to estimate moose ( Alces alces ) density across a large landscape (~12,000 km 2 ) in northern New Hampshire, USA, using infrared imagery collected via local drone surveys. Our objective was to model moose population densities using drones while accounting for imperfect detection (sightability) in forested landscapes. We also compared drone‐based density estimates to existing estimates derived from hunter surveys. Using a stratified random study design, we conducted 62 drone surveys that covered 64.6 km 2 (0.5% of the study area) during January and February 2024. We fit 2 Bayesian binomial mixture models that accounted for sightability to 33 adult moose detections. A simple survey model resulted in a density estimate of 0.64 moose/km 2 across the study area (90% credible interval [CrI] = 0.56–0.77 moose/km 2 ). A second spatial model incorporated covariates to make extrapolated predictions of moose densities at the wildlife management unit (WMU) scale across the study area. Overall, both drone‐based densities were approximately twice the moose density estimated from deer hunter surveys. The precision of drone estimates from the survey model was within the benchmark range for moose management (90% CrI within ±25% of mean), but extrapolated predictions from the spatial model based on environmental covariates were less precise and exhibited somewhat weak correlation (Spearman's correlation = 0.42) with hunter survey estimates at the WMU scale. Our study is among the first to employ a thermal sensor via drone to estimate the density of an unmarked ungulate across a large spatial scale (thousands of square kilometers) using local sampling within a stratified random study design. Our work suggests that drones offer a promising method for estimating density over broad spatial scales without the need for tagged or global positioning system (GPS)‐collared individuals.