Irrigated area mapping in heterogeneous landscapes with MODIS time series, ground truth and census data, Krishna Basin, India

摘要: Diverse irrigated areas were mapped in the Krishna River Basin (258,912 km 2 ), southern India, using an fraction approach and multiple ancillary data sources. Unsupervised classification of a monthly time series net difference vegetation index (NDVI) images from Moderate Resolution Imaging Spectrometer (MODIS) over January–December 2002 generated 40 classes. Nine generalized classes included five with distinct NDVI signatures: continuous irrigation, double-cropped, low biomass, minor groundwater irrigation. Areas by surface water began greening 45 days after groundwater-irrigated areas, which allowed separation irrigation classification. The each class area was determined three different methods: ground truth data, linear regression model calibrated to agricultural census visual interpretation Landsat TM imagery. Irrigated fractions methods varied least for double-cropped (0.62–0.79) rangeland (0.00–0.02), most (0.06–0.43). Small patches (,0.1 ) accounted more than all major combined. differed widely method, suggesting that mapping patchy small remains challenging, but comparison sources improves confidence highlights requiring intensive fieldwork.