Abstract

Estimation of biodiversity patterns in poorly known areas is hampered by data availability and biased collecting efforts. To overcome the former, patterns can be estimated at higher taxonomic levels and larger spatial units. To deal with the latter, species distribution models (SDMs) can be employed. We explored the ability of higher-rank taxonomic units to surrogate patterns of species diversity at different aggregation levels and the use of SDMs to correct collection bias. We used Chile as a study case and employed three biodiversity measures (taxon richness, weighted endemism and turnover), four spatial aggregation levels or resolutions (100, 75, 50 and 25 km grid cells) and three taxonomic levels (species, genera and operational taxonomic units (OTUs)) to evaluate the spatial agreement of biodiversity measures. OTUs are monophyletic groups at the finest taxonomic resolution given the available phylogenetic information. We used a specimen database of 3684 species (84%) of the Chilean vascular flora and evaluated its redundancy. Agreement in spatial patterns was calculated using the fuzzy Kappa index. SDMs were generated for the three taxonomic levels to estimate taxon richness. For each spatial aggregation level, we calculated agreement between specimen-based and SDM-based richness and surrogacy among taxonomic levels with and without SDMs. Density of sampling for specimen-based data allowed for a resolution of 25 km before reaching a critical low redundancy value for all taxonomic levels. Genera and OTUs are good surrogates of species for all biodiversity measures, but their predictive power decreases with spatial scale. Agreement in richness patterns between taxonomic levels is greatest for SDMs at 100 and 75 km resolution, suggesting that biodiversity patterns are best estimated at 75 km resolution using SDMs for this data set. While these results cannot be extrapolated beyond the study area, this framework can be implemented in other data-deficient regions to describe biodiversity patterns and to choose the appropriate aggregation level for downstream biodiversity studies, such as spatial phylogenetics, where species-level data availability is a more generalized problem, since sequence data are normally available for only few species.