Discrepancies between point process models and environmental envelopes identify the niche centroid – geography configuration

Contrasting outcomes of statistical modelling methods for the same data may represent critical pieces of information about the biological process. Ecological niche and species distribution modelling are notorious for their keenness on algorithm testing although differences between method outcomes are seldom used to gain biological insight. Here we use the differences and similarities between point process models (PPMs) and minimum volume ellipsoids (MVEs) to help identify the configuration of species’ climatic niches in relation to geographic space and species’ interactions. Poisson PPMs represent the abundance of points in an euclidean plane as a function of spatially defined covariates, while MVEs are used to estimate the (niche) centroid and represent similarity with its centroid via Mahalanobis distance. The niche centroid is receiving increasing attention due to its simplicity and ability to capture complex demographic processes such as species’ abundance. MVEs’ simplicity makes them sensitive to deviations from symmetry in the statistical distribution of environmental axes, or the lack of a defined centroid in geographic space. Using synthetic and real species we test the ability of PPMs and MVEs to characterise niche centroids in relation to each other and to the statistical properties of the environment. We also study a scenario in which non-linear responses and biological interactions define a species’ fundamental niche and realised distribution respectively. MVEs were less precise but estimated useful centroids more frequently than PPMs. When centroids clearly existed in geographic and environmental spaces, PPMs’ estimated centroids that were closer to the truth. MVEs’ ability to estimate a similarity surface, unlike PPMs, depends on the correct estimation of the centroid. We suggest then that contrasting similarity surfaces estimated by both methods indicate the absence of the centroid in geographic space.