An evaluation of the validity of fish species richness estimates from species distribution models

Stream biological communities in tropical developing countries tend to have high levels of biodiversity, but are generally poorly sampled and thus present challenges for the estimation of general community characteristics.  Species distribution modeling is an approach that can leverage information available at sampling localities to create continuous predictions of species ranges and areas of occupancy.  This study examines the potential to use species distribution models generated for multiple fish species to estimate fish species richness.  I used correlation analysis to compare richness estimates gained from summation of distribution models with those from observed (e.g., sampled) richness, and a jackknife estimate of richness derived from repeated observation data in headwater streams.  Pair-wise correlation coefficients between all three richness estimates were statistically significant, but observed richness values tended to be lower than those derived from the jackknife estimator, and summed distribution models.  Results suggest that the summed species distribution model approach provides a valid, though potentially inflated, first estimate of richness.