Region- and pixel-based image fusion for disaggregation of actual evapotranspiration

This paper compares a region-based and a pixel-based disaggregation method used to improve obtaining actual evapotranspiration (aET) data from MODIS images. Using these methods and the relationship between different vegetation indices form Landsat-5 and aET from MODIS, a 1 km resolution aET image was disaggregated to 250 and 30 m resolutions in two steps. Disaggregated aET images were compared with aET data obtained from a Landsat-5 TM image. A sensitivity analysis using synthetic data showed the impacts of land-cover homogeneity and registration error of the input images at the three scale levels. Accuracy assessment illustrated that the region-based disaggregation method using the Normalized Difference Vegetation Index (NDVI) has a good agreement with the Landsat-5 aET, having a mean absolute error equal to 0.93 mm. This method can be powerful for improving irrigation management, as it allows to increase the spatial resolution of aET derived from remote sensing images. The study concluded that a region-based method with NDVI data performs best to disaggregate MODIS aET data.