How to estimate overall population trends when sites show varying population counts over time?

Population trends can be used to assess the conservation status of a species. Current methods used for estimating population trends have drawbacks when it comes to estimating a robust supra-regional trend from available population monitoring data. Because most monitoring data are collected from many subpopulations rather than a single homogeneous population, the population-level data can be highly heterogeneous with respect to the dynamics of each subpopulation over time. Thus, a supra-regional population trend should represent the common trend across all subpopulations like a global smoother in a hierarchical generalized additive model (HGAM) does. We show and compare how commonly used population trend regression models and HGAMs perform when applied to (i) simulated data consisting of subpopulations with homogeneous versus heterogeneous population dynamics and (ii) real-world data of the Greater mouse-eared bat Myotis myotis monitored in hibernacula throughout Germany over a 34-year period. The true population trend is significantly overestimated in models that ignore subpopulation structure and the inflated overdispersion inherent in monitoring data.