Spatial Modeling - Tree Species Distribution

Research at the Iwokrama Forest in central Guyana has shown the potential for using Geographic Information Systems (GIS) to predict the distribution of tree species. A predictive spatial model of Greenheart (Ocotea rodiei) distribution in the Forest was constructed using a Digital Elevation Model (DEM) and georeferenced tree survey data.

A loosely-coupled unbiased spatial model was constructed that combined well-founded components of GIS and multivariate logistic regression with the known ecology of a species. It demonstrated the integration capability of GIS to derive a value-added product from meso-scale topographic and GPS - tree inventory data.

The output was a map of the study area showing the most probable areas for finding Greenheart. The results were tested against independent field survey data and achieved about 60% classification accuracy, which is comparable to other landscape analyses. The model also demonstrated increased classification accuracy with model-calculated probability, and this behaviour is encouraging. The model shows potential for guiding surveys of forest resources and for assisting strategic planning in forest management.

Future efforts will be on improving the statistical performance of the model and incorporating other influences on species distribution such as geology and climate.

This study built on a Masters thesis at Edinburgh University Geography Department and was presented at the URISA 2001 Caribbean GIS Conference. A full report may be found in: Datadin, V.K. (2001). Predictive modeling of tree distribution in tropical forest using GIS in Proceedings of the Urban and Regional Information Systems Association (URISA) Caribbean Conference, Montego Bay, Jamaica, pgs. 234-247.