干旱区生态环境知识资源中心

A study involving empirical relationships between average growing season NDVI and total annual evapotranspiration (ET) from eddy covariance (EC) examined two approaches to extract average growing season NDVI values for regression analysis. Analysis and accuracy assessment included a single pixel (SP), average within the footprint boundary (FB), and a weighted average footprint (FW) approach to better account for the areas with the most influence on tower ET measurements. The SP approach produced a 17% average error in year 1, 20% in year 2, and a combined two-year regression relationship resulted in 14% error. The FW approach was only slightly less accurate with average errors of 27% and 26%, in years 1 and 2 respectively, and a two-year regression relationship yielded 12% error. The FB approach yielded poor results (51% error). These results indicate that a multi-year regression analysis produces more accurate estimates of annual ET than a single year regression relationship and that a SP approach performed as well as the FW approach.