Document: ERI-3-90-2

Identifying optimal habitat for multiple species with a combination of kernel estimation and odds ratio statistics.

PETERSON, E.B.*

Oregon State University, Corvallis, OR 97331 USA ¹

Abstract:
Predicting species occurrence over landscapes with GIS is useful for identifying critical habitat for rare species. There are several ways to combine maps of different species to identify optimal habitat for multiple species. However, when the species occurrence maps plot probabilities of occurrence, choices must be made regarding how to combine the maps. Simply adding the maps together will bias toward the more frequent species because they tend to have higher probabilities than infrequent species. Various relativizations are available, but each comes with its own bias. One solution is to develop probability maps with a kernel estimation technique combined with odds ratio statistics. I use a simple form of kernel estimation to assess the probability of a species occurring across a landscape from a database of previously sampled sites. Probabilities are estimated from the proportion of known occurrences within the environmental neighborhoods of sites being predicted for. I then apply odds ratio statistics to compare the odds of the species occurring within the environmental neighborhoods, with the odds of it occurring outside the neighborhoods. Both methods are non-parametric and well suited to typical species occurrence data. The resulting map shows habitats where a species has a statistically significant odds ratio. Maps calculated for different species may be added together to identify habitats that provide high odds of occurrence for multiple species without bias toward the more frequent species. The final maps of optimum habitats are useful for land management, species conservation, and GAP analysis.

Keywords: Species occurrence GIS GAP model landscape

This abstract is being presented at: 10:30 AM in session:
STATISTICAL ECOLOGY