Unpacking the cross-scale niche: A multi-grain modeling framework to evaluate scale-dependence in species-environment relationships

Background/Question/Methods

Multiple lines of ecological theory and research indicate that species respond to environmental conditions differently at different spatial scales. Quantifying the spatial grains at which species respond to specific environmental variables is a key goal in ecology, and represents a critical step toward selecting appropriate scales for analysis and making accurate inferences about ecological relationships. We used a multi-grain modeling framework to investigate how a species' spatial distribution is shaped by two interacting, scale-dependent factors: the spatial structure of environmental conditions, and the species' selection of environmental variables. We analyzed artificial landscapes with varying spatial structure and presence/absence data generated under different selection modes to develop structure-based expectations for multi-grain models. Biologically relevant environmental variables were derived from high-resolution remote sensing imagery and field data, and standardized at six grains from 10-1000m. During 2011-2014, we conducted timed surveys for the Von der Decken’s hornbill (Tockus deckeni), a medium-sized (150-250g) omnivorous resident species, at sites distributed across key environmental gradients in the study area. We modeled the relationship between T. deckeni occurrence and environmental conditions using generalized linear mixed models with spatial random effects to account for spatial autocorrelation, and assessed multi- and univariate model performance across spatial grains.

Results/Conclusions

Multivariate mixed models did not indicate one or a few grain(s) at which probability of occurrence was most accurately predicted; instead, model performance was relatively high across all study grains (AUC 0.85-0.96). In both multi- and univariate models, T. deckeni occurrence was influenced in a scale-dependent hierarchy according to the spatial structure of key environmental variables, demonstrating that spatial structure is a critical factor in structuring species responses to the environment. This finding contrasts with the current view that habitat-related variables shape occurrence patterns at fine spatial grains, and climatic variables at coarse grains, and highlights the need to account for spatial structure when predicting species' spatial distributions and exploring species-environment relationships. Through characterizing cross-scale variation and correlation in environment spatial structure and multi-grain distribution models, we describe the cross-scale niche of the focal species.