Spatial Non-stationarity and Time-dependency in the Relationship between NOAA-AVHRR NDVI and Climatic Determinants: Sensitivity to Landuse/Landcover Change

Abstract

The spatial relationship between vegetation patterns and climatic variables and its trend over the period 1985-2001 in the shrubland, grassland, and cropland of Middle Kazakhstan was investigated with normalized difference vegetation index (NDVI) images (1985-2000) derived from the Advanced Very High Resolution Radiometer (AVHRR), and climate records from weather stations located in the study area. A local regression technique known as geographically weighted regression (GWR) was used to examine growing season relationship between NDVI and a set of climatic determinants, comprising total rainfall and mean air temperature over April-October. Regression models between NDVI and precipitation, NDVI and temperature, as well as multiple regression models including NDVI and the both climatic predictors for each pixel and every analysis year (1985-2000) were calculated, and temporal drifts of regression parameters and drifts of determination coefficients, R2, for every pixel were estimated and mapped. The relationships between NDVI and the explanatory variables were found to vary spatially and temporally. The determination coefficients of multiple regressions generally increase in order from cropland, to shrubs vegetation, and to grassland. Further, the results indicate that the strength of the relationships and the parameters of the regression models for every pixel varied from year to year over the study period. For the most pixels in shrubs vegetation and grassland, we found a general increase of NDVI determination by climatic factors during the period 1985-2001. On the contrary, the determination coefficients associated with cropland have been decreasing over 1985-1992 increased rapidly after 1992. The drifts of the determination coefficients between NDVI and climatic predictors over the period 1985-2000 provided evidence to be depending on the land use/land cover change in the study area. The areas with clear signs of land degradation displayed whether extraordinary low response of NDVI to climatic predictors or a permanent decrease of this response over the time. Four types of drift behaviour were determined and analysed.