Vol. 7, No. 1, 30-45, 2008

Abstract
Spatial relationship between vegetation patterns and precipitation in drylands of Kazakhstan was investigated with Normalised Difference Vegetation Index (NDVI) derived from the Advanced Very High Resolution Radiometer (AVHRR) and climate records from weather stations. A local regression technique known as geographically weighted regression (GWR) was used to model the growing season relationship between these variables at the local scale. GWR models were established for each pixel and every analysis year during the period of 1985-2000. The models exposed spatial and temporal non-stationarity of the NDVI-precipitation relationships. This non-stationarity was estimated for each pixel and each year and mapped. We traced at the per-pixel basis the vegetation response to precipitation over the period of 1985-2000 and compared it with land use/land cover change that happened in the study region. For the four cases investigated the study proved that any temporal drift of this response is a result of any land use/land cover change. The areas with clear signs of land degradation displayed either extraordinary low response of NDVI to rainfall or a permanent decrease of this response throughout the observation time. On the contrary, the areas associated with improvement of vegetation cover are characterised by an increase of the vegetation response to precipitation. The results of this study improve our understanding of the interaction between the vegetation cover, the climatic components and human activities of dryland's ecosystem in Central Asia.

ISSN 1729-3782