Cultivated area estimation combining moderate spatial resolution remote sensing data and MODIS time series data

Numerous mixed pixels from coarse resolution data and missing date from moderate resolution data are major concern in cropland identification. Therefore, estimation using combining the coarse time series data and moderate resolution data has become the popular tendency for cultivated area identification. We proposed a novel method which integrated the coarse time series and moderate resolution data whose estimation results was in moderate resolution rather than lower resolution such as most of former studies. Linkou, Heilongjing Province, China was used as test site, and Landsat5 TM, and MODIS-NDVI time series (250m) were the major materials. High resolution imageries at 2.5 m from Google Earth were sampled as reference and validated data. The result had an overall accuracy of 94.57%, Kappa of 0.92, producer accuracy of cropland was from 71.24 % to 88.07% and user accuracy was from 86.10 % to 94.74%.