Object-based identification of vegetation cover decline in irrigated agro-ecosystems in Uzbekistan

摘要

The sustainability of irrigated croplands in Central Asia is threatened by degradation of their productive function. To quantify the extent of the cropland degradation, this paper combines object-based change detection and spectral mixture analysis for vegetation cover decline mapping in irrigated agro-ecosystems in Uzbekistan based on multitemporal Landsat TM images from 1998 to 2009. The results of the change detection reveal that the cadastral field parcels, characterized by vegetation cover decrease, occupied 18% (52,938 ha) of the cropland area. Spatial distribution of fields with decreased vegetation cover was mainly associated with abandoned cropland and land with inherently low-fertility soils located on the outreaches of the irrigation system and bordering natural sandy deserts. The comparison with the land degradation map based on trend analyses of 250 m MODIS NDVI time series 2000-2010 yielded an overall agreement of 93%. The proposed approach is a useful supplement to the commonly applied trend analysis for detecting land degradation in cases when plot-specific data are needed but satellite time series of high spatial resolution are not available. The derived parcel-specific, spatial information supports better informed decisions on cropland rehabilitation measures.