Monitoring exotic forest pest based on high-resolution remote sensing image and CART model

Zijin Mountain National Forest Park in Nanjing City was selected as the study area and pine wilt disease was chosen as the monitoring object. High resolution remote sensing images of QuickBird in 2004 and 2007 were collected as the primary information source, while level two forest inventory data in 2007 were used as the auxiliary information. After image procession of change detection between 2004 and 2007, occurrence points of pine wilt disease with geographical coordinates were extracted. Five environmental variables of land categories, NDVI, elevation, slope and aspect were gathered, followed by generation of Classification and Regression Trees (CART) model to monitoring occurrence area of pine wilt disease in 2007. Results showed that in 2007, occurrence area of pine wilt disease was 9.461 ha, which were mainly distributed on the southern slope, northwestern slope and east slope of Zijin Mountain; at the landscape scale, the occurrence of pine wilt disease was not only related with the host pine trees, but also with the terrain, forest growth condition; pine wilt disease tend to occur on gentle southeastern slope (<;6.67°) with low altitude (<;54.081m). As for stand site selection, pine wilt disease likes to infect pine forest with low to medium growth condition.