Using Multiple Remote Sensing Perspectives to Identify and Attribute Land Surface Dynamics in Central Asia 2001-2013

Document Type


Publication Date



MODIS; vegetation indices; land surface temperature; trends; Central Asia


To understand the land surface changes that Central Asia experienced between 2001 and 2013, we applied a non-parametric change analysis method to the standard vegetation indices (NDVI and EVI), as well as to the MODIS Tasseled Cap indices Brightness, Greenness and Wetness. In addition, we evaluated the MODIS nighttime and daytime land surface temperature products and the MODIS evapotranspiration product. We compared the change results by country, land cover type, and anthropogenic biome, and we also evaluated the results according to an index of human influence (HII). We found that EVI, NDVI and Tasseled Cap Greenness reveal very similar changes (r > 0.8), while there was a much lower correlation between the vegetation indices and results based on other portions of the electromagnetic spectrum. Thus, we found it informative to expand the analysis beyond the optical and near infrared portions of the electromagnetic spectrum, into the thermal regions. We found that the majority of the changes occurred in Kazakhstan and Uzbekistan, while Turkmenistan, Kyrgyzstan and Turkmenistan appeared more stable during this period. The observed changes were attributable to a combination of anthropogenic changes and weather effects. For example, changes in crop type south of the Aral Sea were revealed as increases in vegetation indices but declines in evapotranspiration, resulting from a shift from cotton to wheat. Across Kazakhstan large patches of negative vegetation changes, combined with increasing temperatures and declines in evapotranspiration were attributable to persistent droughts. Generally, we found that most browning of vegetation occurred in areas with lower human influence (except for areas with very high human influence) and most greening of the vegetation occurred in the areas with intermediate human influence. The use of multiple indicators of significant trends improved trend interpretation and attribution of proximate causes.

Publication Title

Remote Sensing of Environment



First Page


Last Page


DOI of Published Version





Copyright © 2015 Elsevier Inc.