Close Window

Application of Remote Sensing Data in Predictive Models of Species’ Distribution

Wolfgang Buermann, Center for Tropical Research, Institute of the Environment, UCLA, California, buermann@ucla.edu (Presenting)
Sassan Saatchi, Jet Propulsion Laboratory, California Institute of Technology,Pasadena, CA, saatchi@congo.jpl.nasa.gov
Brian R. Zutta, Center for Tropical Research, Institute of the Environment, UCLA, California, bzutta@ucla.edu
Jaime Chaves, Center for Tropical Research, Institute of the Environment, UCLA, California, jachaves@ucla.edu
Borja Mila, Center for Tropical Research, Institute of the Environment, UCLA, California, bmila@ucla.edu bmila@ucla.edu
Cathrine H. Graham, Dept. of Ecology and Evolution, Stony Brook, New York, cgraham@life.bio.sunysb.edu
Tom B. Smith, Center for Tropical Research, Institute of the Environment, UCLA, California, tbsmith@ucla.edu

Predicting geographical distribution of species requires data that can describe species' environmental requirements accurately. With the recent launch of a number of earth orbiting satellites, a vast array of direct physical measurements of ecological variables at high spatial and temporal resolution have become available. This study explores the relative merit of using these newly available remote sensing data in species distribution modelling. We applied a recently developed algorithm, based on the maximum entropy approach (Maxent), to model the distribution of two South American bird species, the wedge-billed woodcreeper and the speckled hummingbird, at 1 km spatial resolution. The models were developed separately for three scenarios of input data layers with bioclimatic and remote sensing layers in isolation and combined. Results from quantitative performance measures and visual inspections showed that Maxent scenarios with remote sensing and bioclimatic layers in isolation performed almost equally in predicting general patterns of species distributions, suggesting each of these data sets contain useful information. However, Maxent model runs with a combination of remote sensing and bioclimatic layers resulted in the best model fits and generally higher spatial accuracy including less overprediction, suggesting a more constrained characterization of the two bird’s ecological niche. In Ecuador, the inclusion of high-resolution remote sensing data were critical in resolving known geographically isolated populations of these species. Further, due to their sensitivity to vegetation and landscape patterns, the remote sensing data were also essential in excluding areas in the Maxent predictions that had lost their natural forests, leading to much more detailed range maps. The findings suggest that remote sensing data can play a major role in modelling geographical ranges of species and in predicting and monitoring any changes due to human-induced fragmentation or climate-related stresses on their habitat.

Presentation Type:  Poster

Abstract ID: 86

Close Window