Major Land Use (key crops), 2010

Abstract

Land use maps show the spatial distribution of dominant crop types (at the per-parcel level) in the major part of the Fergana Valley in 2010, 2011, 2012, and 2014. Major crop types like cotton or rice were classified as separately, whilst other minor crop classes, e.g. sorghum, maize, or melons were aggregated to the class “other”. Such data can be used as input for crop acreage estimations, crop type distribution analysis, modeling land cover and land use change, yield, water demand calculations etc. Crop classification was done by means of an object (=per-parcel) based supervised classification procedure, based on ground reference samples that were recorded in the Fergana Valley in 2010, 2011, 2012, and 2014. The Random Forest algorithm was used for the classification of one Landsat-5 TM and six RapidEye images (30 m and 6.5 m, respectively pixel size). The overall accuracy of this maps is 86%, 86%, and 85% (2010, 2011, and 2012, respectively).