A comparison of hierarchical models to estimate species-environment relationships using spatially misaligned data

Spatial misalignment occurs in regression analysis when a response variable is sampled at spatial locations that differ from those where predictors are available. If not address during modelling, spatial misalignment can bias estimates and conduct to misleading conclusions. Area-to-point misalignment is a particular case when response variable is sampled at point-level while predictors are observed at area-level. Such spatial misalignment is frequent in species distribution modelling (SDM). However, it is still commonly overlooked by ecologists while known to bias estimates of species-environment relationships.

In this work, we compare several hierarchical models to estimate abundance-environment relationships in presence of area-to-point misalignment. We perform a sim ulation study to investigate the effect of the size of scale mismatch between ecological response and observed environment on the accuracy of tested models. Our work aims to allow new insights on when to worry about point-to-area spatial misalignment in SDM, and help ecologists to select appropriate model in those cases. We also present the results of a case study.

Keywords. Spatial misalignment; Spatial scales; Species distribution;
Bayesian inference; Measurement error.