Abstract:

In this study， the ESA-LUC and MODIS-NDVI data were adopted for pixel comparisons to calibrate matching values （thresholds） between the two datasets. Based on which， ESA-LUC types were clustered into three types including oasis vegetation （OV）， desert vegetation （DV） and bareland （BL） by application of the calibrated NDVI thresholds. Correspondingly， it is the water outside the arid desert-oasis system that supports the OV growth， while local precipitation provides water for physiological consumption of the DV. The third type of the BL includes desert， hamada， gravel desert， and so on， featuring very low greenness or non-vegetative land surface. It is concluded that the referenced LUC is essential for NDVI thresholds determination for Clustering. NDVI thresholds were remarkably different due to hydrometeorology， matrix characteristics and oasis development in the middle and lower reaches where dominate the desert-oasis system， in the selected four representative basins. Classification by clustering results into strong effectiveness of oasis extension， local precipitation and desert matrix on vegetation dynamics. It is found that the stable OA become greener while the stable BL， barer during the time period from 2001 to 2015. In the unstable area， more increase is found near the stable OA， while less of that is located mainly in the interior desert where local precipitation feeds vegetative prosperity. Overall， a simple filtering technique provides opportunities for people to classify the arid desert-oasis system for better understanding of relationship between ecological water consumption and vegetation growth， then benefits soil-water planning and desert ecology protection in the resource-scarce and water-shortage area in Northwestern China

Key words: oasis vegetation, desert vegetation, bareland, ESA-LUC, MODIS-NDVI