IMPACTS OF WEIGHTING SCHEMES AND TRANSFORMED ENVIRONMENTAL VARIABLES ON BIODIVERSITY MODELING WITH PRESENCE-ONLY DATA

Abstract

Biodiversity modeling techniques at the community- and species-level can be used to address questions in ecology, management, and conservation. I addressed aspects of community-level and specie-level models using virtual and inventoried species in North and South America. Firstly, I assessed the effectiveness of two weighting schemes in reducing impacts (if any) of five sampling routines (simulating unrepresentative sampling in presence-only data) on the model performance of Generalized dissimilarity model (GDM). Unrepresentative sampling lowers model performance, but weighting species can reduce this negative impact to a certain extent. However, PO data severely impacts GDM’s ability to detect the relative contribution of environmental gradients. Secondly, I examined the potential of (GDM) transformed environmental variables in improving the performance of Maxent models (presence-only) along with the influence of range size, sample size, and species dependence type. Transformed environmental variables improved model performance, especially when used with small-ranged species and/or low sample sizes.