Sample size guidelines for mapping migration corridors and population distributions using tracking data

Abstract Animal distribution maps are a key tool for wildlife conservation, guiding high‐profile decisions, such as legally designating priority habitat or building highway crossing structures. GPS tracking data enhances these efforts but requires balancing statistically robust sample sizes with minimizing researcher impacts on wildlife and costs. Nevertheless, rigorous guidelines that leverage a priori information are still lacking on how to determine the optimal number of tracked animals (i.e. sample size) for accurately mapping migration corridors and seasonal ranges at the population level, particularly in the context of ungulate conservation. We used a cumulative curve resampling approach to evaluate the consequences of reduced animal sample size, assessed sample size sufficiency and extrapolated where sample size sufficiency might occur outside of the observed data. We illustrate our approach with simulated data. We then compiled GPS data from 77 ungulate populations and aggregated individuals' spatial distributions in each study area to create population‐level migration and seasonal distributions and examined whether known explanatory variables (e.g. population abundance, environmental metrics) could predict sufficient sample sizes to map population distribution. Our simulated and empirical analyses to assess and model sample size sufficiency demonstrated that sample size varies depending on the species, season, population‐level percent volume contour of interest and population abundance. For example, for migration distributions at the 95% volume contour, the interquartile range for number of individuals needed to reach an adequate sample was 10–23 for bighorn sheep, 51–93 for elk and 58–164 for mule deer. Practical implication . For existing datasets, the resampling approach quantifies the sensitivity of population distribution maps to sample size. To guide study design for future GPS tracking projects aimed at mapping population distributions, our models provide specific sample size recommendations incorporating known population covariates. If adequate model training data are available, our approach can be extended across a wide range of taxa and populations to inform sample size requirements for estimating robust distribution patterns.