Mapping the Risk of Sudden Oak Death for California
Ross Meentemeyer1, Liz Lotz2, Dave Rizzo3, Wally Mark4, and Maggi Kelly5
Information on Sudden Oak Death risk is needed to target distribution surveys and predict potential impacts on California's natural resources. We present a geographical model of disease risk for California. The model is based on host species distribution data and several critical environmental factors. These factors include GIS-derived maps of proximity to infested locations, weather and climate conditions, proximity to commercial nurseries and human features, and landscape structure of host connectivity and edge effects. Two modeling approaches are evaluated. One is statistically based while the other uses human input from preliminary field and laboratory studies of environmental conditions conducive for pathogen reproduction and dispersal. In the statistical approach, observations of disease incidence (the response variable) are obtained from FIA field plots and COMTF ground surveys, and linked with the predictor variables in the GIS database. Multivariable relationships are statistically analyzed to develop a model that quantifies the degree to which predictor variables are related to disease presence. The model's equation is applied to mapped predictor variables in the GIS to calculate the risk that an unsampled location could develop infestation, or already has. Twenty-five percent of the observations are held back during model development to evaluate model performance.
We also develop a standard overlay model, which uses human input, rather than statistical inference. In collaboration with Dave Rizzo (UC Davis), each of the mapped predictor variables (described above) are scored and ranked based on potential infestation risk. The maps are overlayed and summed to produce a composite map of risk for the whole state. Risk values range continuously from 0 (low risk) to 100 (high risk). Proximity to infected sites gives the model a dynamic component. As additional sites are infected, the model targets new areas at risk and tracks patterns of our knowledge through time. Model performance is evaluated by comparing the resultant risk maps to field observations of disease incidence.
1Sonoma State University, Department of Geography, 1801 E. Cotati Avenue, Rohnert Park, CA 94928; (707) 664-2558; meenteme@sonoma.edu
2Sonoma State University, Department of Geography, 1801 E. Cotati Avenue, Rohnert Park, CA 94928
3University of California, Department of Plant Pathology, One Shields Avenue, Davis, CA 95616
4California Polytechnic State University, San Luis Obispo
5University of California, Environmental Sciences, Policy and Management Department, 151 Hilgard Hall, Berkeley, CA 94720
|
