Evaluating the importance of scale and variable selection when modeling rare and endangered plants in regions with rugged topography

Background/Question/Methods

In the past decade, species distribution modeling (SDM) for rare and endangered species has rapidly progressed and become an informative tool for identifying key areas for restoration and habitat conservation.  Despite its widespread use, there remains great uncertainty regarding the appropriate scales for SDMs, which are often constrained by coarse-resolution climate data.  Although 1 km may be fine enough to capture topographic variation for large, relatively flat regions, this scale may not be ideal for areas with a small area and highly variable topography (e.g. Hawaiian Islands).  This research uses the SDM algorithm, Maxent, to model habitat suitability for 11 federally endangered dry forest plant species on Oahu, Hawaii at a landscape (1km), local (250 m) and site specific (10 m) spatial scale.  The objective of this research is to determine which combination of variables and spatial scale yields the best results for modeling rare and endangered plants on islands with extreme topographic variation for the purposes of restoration site selection.

Results/Conclusions

Using a wide range of environmental variables (e.g. bioclimatic, topography, soil) we found that the most accurate models for all species were at a 10 m spatial resolution with a reduced variable set including summer precipitation, elevation, slope, aspect, hillshade, curvature, and soil great group.  PCA analysis showed high correlations (> 0.75) between all temperature metrics and elevation, and between all precipitation variables. Very low correlations were found between topographical, soil, and precipitation variables.  For models at all spatial scales, statistical analysis suggests overall good model performance and thus, usefulness of the models for conservation management.  Our results suggest that for areas with rugged topography (e.g. Hawaiian Islands), SDMs at the site scale may best capture plant micro-habitats and prove most useful for guiding endangered species restoration.  Furthermore, SDMs using high-resolution topography variables rather than coarse-scale climate data may be more useful for informing endangered plant reintroductions.