Developing a methodology to predict oak wilt distribution using classification tree analysis

Abstract

Oak wilt (Ceratocystis fagacearum), a fungal disease that causes some species of oak trees to wilt and die rapidly, is a threat to oak forested resources in 22 states in the United States. We developed a methodology for predicting the Potential Distribution of Oak Wilt (PDOW) using Anoka County, Minnesota as our study area. The PDOW utilizes GIS; the classification tree statistical test; field sample data; commonly acquired, coarse-resolution auxiliary data; and a unique application of data from the Landsat Thematic Mapper (TM) satellite to predict the spatial distribution of oak wilt. Two accuracy assessments, one being a ten-fold cross validation, the other using verified oak wilt data from a later date, indicate that, at the landscape scale, PDOW correctly models the presence of oak wilt, and accurately predicts oak wilt distribution in Anoka County. Important variables in predicting oak wilt distribution in Anoka County included: Landsat TM Bands 3, 4, 5, and 7; distances between sample sites and lakes (Distance to Lakes); density of streams within a 400 x 400 meter grid surface (Stream Density); slope; aspect; and density of roads within a 400 x 400 meter grid surface (Road Density).

Parent Publication

• Monitoring Science and Technology Symposium: Unifying Knowledge for Sustainability in the Western Hemisphere

Citation

Downing, Marla C.; Thomas, Vernon L.; Reich, Robin M. 2006. Developing a methodology to predict oak wilt distribution using classification tree analysis. In: Aguirre-Bravo, C.; Pellicane, Patrick J.; Burns, Denver P.; and Draggan, Sidney, Eds. 2006. Monitoring Science and Technology Symposium: Unifying Knowledge for Sustainability in the Western Hemisphere Proceedings RMRS-P-42CD. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. p. 771-778