A Global Map of Rainfed Cropland Areas at the end of last millennium using Remote Sensing and Geospatial Techniques

摘要

Rainfed agriculture plays a critical role in most part of the tropics and subtropics of the world. Eighty percent of the agricultural land worldwide is under rainfed agriculture; and significant proportion of rural economy still depends on rainfed agriculture with characteristically low yield levels. In this context the International Water Management Institute (IWMI) produced the first satellite sensor based Global map of rainfed cropland areas at lOKm resolution (GMRCAlOKm). The study used a mega-file of 159 global data layers involving the AVHRR and SPOT time-series, GTOPO30 DEM, mean monthly rainfall, and forest cover. A suite of innovative techniques were developed that begins with the image segmentation, quantitative spectral matching techniques (SMTs) and spectral correlation similarity (SCS R2). The SCS was found to be the most useful technique in grouping identical classes. Mixed classes were resolved using a decision trees, time series plots, and principal component analysis algorithms. A wide array of groundtruth data, and high-resolution images were used to identify and label classes. The outcome was the GMRCAlOKm estimated to be 1.75 billion hectares for the main cropping period. The sub-pixel areas (SPAs) of GMRCAlOKm provide more realistic estimates of the actual area cultivated unlike the full pixel areas (FPAs) often calculated from the raster datasets. Three distinct GMRCAlOKm maps have been produced: viz., Aggregated 7-class, Dis-aggregated 18-class and Generic 255-class. The aggregated classes will suffice for broad range of users at global level. The GMRCAlOKm product line consists of maps, images, area calculations, snap-shots, class characteristics, and animations.