Accuracy Assessment of Planted Area Proportion Using Landsat TM Imagery

In this Paper, the accuracy of planted area proportion using Landsat TM is assessed with two pilot sites at Kaifeng of Henan province and Taigu of Shanxi province. In Kaifeng, one Landsat TM imagery on April 1 2001 was acquired for summer crop. In Taigu, Landsat TM image on Oct. 14 2003 was acquired for autumn crop. Landsat TM images were processed for extracting the planted area proportion with 6 steps: atmospheric correction, geometric correction, SAVI calculation, non-arable land masking, unsupervised classification,and labeling. Planted pixels were counted and the planted area proportion can be calculated by the divided the number of arable land pixels extracted from the land use map at a scale of 1∶100 000. At the same time, in order to assess the accuracy of planted area proportion from Landsat TM, an IKONOS imagery (11km×11km Coverage) on Mar. 21, 2001.was acquired for Kaifeng. After an unsupervised classification based on ground survey, the planted area proportion are obtained. A QuickBird imagery was applied to Taigu site, and the planted area proportion are obtained by field works, which fills every parcel with crop or landuse type. These two planted area proportions for two sites can be compared to assess the accuracy of extracting the planted area proportion with TM data. It shows that the accuracy is 99% in Kaifeng and 97% in Taigu. And this is acceptable for the operational purpose of China Crop Watch System with remote sensing.