Abstract:The Dendrolimus punctatus Walker is the most serious forest defoliator in China. The study of pest remote sensing prediction is helpful to the comprehensive and timely identification of possible damage regions, and can provide the technical support for the direct and indirect losses reduction. Taking Fujian Province as the study area, the Dendrolimus punctatus Walker overwintering generation was investigated in 13 monitoring stations, Sanming City, Jiangle County, Shaxian County, Nanping City, Huaan County, Yunxiao County, Nan’an City, Anxi County, Putian City, Changting County, Jianyang City, Ningde City, and Fuqing City, successively from February to May 2012. The investigation indices contained GPS coordinates, pine forests canopy spectrum, climate, terrain, stand, pest source, human environment and so on. After the analysis and extraction of these factors above, the sub-factors used to predict the Dendrolimus punctatus Walker pest levels were obtained with correlation analysis. With the help of Fisher discriminant analysis, the discriminant functions of 4 pest levels, non-damage, mild damage, moderate damage and severe damage, were constructed, and the self-test result with samples of modeling group showed that the discriminant accuracies were 71.0%, 84.9%, 87.2%, 100% respectively, with total discriminant accuracy of 80.8%. The test result with samples of validation group showed that the discriminant accuracy and accurate rate were 73.00% and 75.99% respectively, which indicated that there was good feasibility to use Fisher discriminant analysis to predict the Dendrolimus punctatus Walker pest levels.
