Project Report

Natural areas worldwide are challenged to protect their natural resources in the face of rural development and unsustainable land use practices. As these areas become islands, the species distributed in their boundaries become increasingly isolated. Understanding how these species are dispersed ecologically as well as geographically is critical to the conservation of natural resources. Using algorithms to model species' ecological niches and predict geographic distributions is an important tool in the management of protected areas; nevertheless, few studies have compared the performance of different algorithms. Thus, resource managers have no particular guide for making decisions about the accuracy of such models. This study aims to develop a general framework to provide conservation planners with a decision making tool that allows them to accurately identify critical habitat for restricted species. This study will use different algorithms (Bioclim, GLM, GAM, DOMAIN, GARP, and BioMapper) to model and evaluate the potential distribution of birds, trees, and beetles in a tropical dry forest. Evaluation of those models will be based on field work through the Balsas Basin tropical dry forest, Mexico. Model performance will be evaluated using ROC analysis. Model errors (omission, commission) will be evaluated using maximum Kappa and confusion matrix. The results of this work will empower natural resource managers for conservation of natural areas in the Balsas Basin, as well as natural areas worldwide.

Until recently, I have been using four algorithms (Bioclim, DOMAIN, GARP and MaxEnt*) to model and evaluate the potential distribution of birds and Bursera trees in the Balsas Basin tropical dry forest region of southern Mexico, where data collection involved fieldwork in 60 1km x 1km plots. Preliminary results indicate that performance of MaxEnt is higher for narrowly distributed species, whiles GARP and DOMAIN achieve greater success in predicting the occurrence of widely distributed species. Additionally, the inclusion of both direct (e.g., temperature) and indirect (e.g., elevation) variables showed better results in predicting the distribution of target species. The results of this work will empower natural resource managers in their priority assessment efforts and contribute to developing strategies for biodiversity conservation in the different national parks sampled in this project, as well as for natural areas worldwide. Researchers at the Natural History Museum, Smithsonian Institution and Facultad de Estudios Superiores Zaragoza, UNAM, Mexico, are currently collaborating with this project. Carabid beetle identification is still in progress, along with the evaluation of four additional species distribution models (GLIM, Biommaper, MARS* and CART*). A final report that includes the comprehensive outcome of this project will be sent to the Cleveland Metroparks Zoo in June of 2006.