An unmixing-based spatial downscaling fusion approach for the MODIS evapotranspiration product

Abstract

Although the Moderate Resolution Imaging Spectroradiometer global evapotranspiration product (MOD16 ET) has been successfully applied to mesoscale evapotranspiration (ET) monitoring, its coarse spatial resolution (500-m) cannot meet the needs of locally scaled water management and environmental change research. Based on the correlation between surface ET differences and corresponding land cover types, and spectral mixture analysis theory, we propose an unmixing-based spatial downscaling approach for the MOD16 ET (500-m) product to obtain temporally continuous high spatial resolution 16-m ET data corresponding to the spatial resolution of the Chinese GaoFen Satellite No. 1 Wide Field-of-View camera (GF-1 WFV) images. Three different landscape areas in China are selected for an experimental study. The experimental results indicate that the downscaled 16-m ET has higher spatial details in spatiotemporal patterns than the MOD16 ET (500-m) product. The MOD16 ET product’s assessment of agreement indicates that the downscaled ET has high correlation and less error with the MOD16 ET product. The assessment of agreement with 29 in situ ET measurements shows that the downscaled ET has high correlation and less error with in situ ET measurements. Our downscaling method is reliable and applicable in downscaling MOD16 ET product to meet the needs of local-scale water management and environmental change research.