Determining land surface fractional cover from NDVI and rainfall time series for a savanna ecosystem

摘要: Savanna ecosystems are water limited and responsive to rainfall on short time scales, characteristics that can be exploited estimate fractional cover of trees, grass, bare soil over large-scale areas from synthesis remote sensing measurements. A method is presented based upon the differing ways in which grasses trees respond rainfall, implementation this demonstrated along Kalahari Transect (KT), an aridity gradient southern Africa. Seasonally averaged normalized difference vegetation index (NDVI) sensitivity NDVI interannual variations wet season used as state-space variables a linear unmixing model. End members for analysis were determined basis best fit observed data. The realized end consistent with qualitative (high NDVI, low rainfall), (low sensitivity), transient grass/ area (moderate high sensitivity). Observed was measured relationship between 16-year period (1983–1998). model yields north-to-south decrease tree corresponds mean precipitation 1600 300 mm KT (R 2 =.87). results compare favorably available ground-based observations. potential extent grass by dominance northern transect, peaks at location approximately 450 arid portion transect. With inferred data, yearly estimates tree, derived. These annual estimates, difficult obtain traditional procedures, important parameters fuel load land–atmosphere exchange models. No calibration or training sets required analysis, has additional capability predict fractional-cover components future scenarios. D 2002 Published Elsevier Science Inc.