OOS 39-6
Developing spatially explicit habitat models for vertebrate distribution studies using fine-scale vegetation maps and the National Vegetation Classification
Most species distribution modeling is carried out at coarse resolutions (>800 m) dictated by the spatial resolution of available predictor variables; however, coarse resolution data does not capture finer-scale variation of habitats and may not match the scale at which species perceive the landscape. Availability of fine-scale vegetation maps for the northern Sierra Nevada foothills using the National Vegetation Classification System (NVCS) and downscaled climate data at 270 m resolution gave us the opportunity to model habitat suitability at a scale that better matches the variability of habitats across the landscape. The detailed vegetation data allowed us to explore three different ways to incorporate vegetation data into the models: continuous and 15 class categorical vegetation data for Maxent models, and 65 class categorical vegetation data for expert opinion models. We selected 30 vertebrate focal species representing a range of taxonomic groups and habitat specialization. We generated four Maxent modeling scenarios for each species to compare categorical versus continuous vegetation data, both with and without climate variables, as well as an expert opinion model based solely on vegetation data. Close collaboration with species experts to review and refine these models helped us identify the best habitat suitability model for each species.
Results/Conclusions
In models selected as the best representation of suitable habitat by species experts, vegetation variables were among the top three predictors of habitat suitability 73% of the time, and vegetation was the top predictor of habitat suitability for 14 of 30 (47%) species. Expert opinion models based solely on vegetation data were selected for 7 of 30 (23%) species. This underscores the importance of using spatially accurate vegetation maps at a scale relevant to the species’ use of the landscape when developing spatially explicit habitat suitability models. The NVCS provided a hierarchical classification, based on field data, for systematically grouping vegetation types into categorical and continuous habitat units for modeling. Both categorical and continuous vegetation variables produced successful models with slightly different views of the landscape. Models using categorical vegetation data were more often selected as the best models for habitat generalist species, while models using continuous vegetation data were more often selected as best for habitat specialist species.