Biological Systems Engineering


Date of this Version



Sharma, V., A. Kilic, and S. Irmak (2016), Impact of scale/resolution on evapotranspiration from Landsat and MODIS images, Water Resour. Res., 52, 1800–1819, doi:10.1002/ 2015WR017772.


Copyright 2016. American Geophysical Union.


Understanding the role of landscape heterogeneity and its influence on the scaling behavior of surface fluxes as observed by satellite sensors with different spatial resolutions is a critical need to investigate. In this study, the effects of pixel scales on ETc estimation and other parameters that are used to calculate ETc were investigated over different vegetation surfaces in south central Nebraska, USA. Surface Energy Balance System (SEBS) was used to estimate spatially distributed ETc by combining ground-based meteorological data for Landsat and MODIS imagery. The estimated surface energy fluxes were compared and validated to the measured Bowen Ratio Energy Balance System (BREBS) ETc fluxes. Validation results showed that Landsat has more preferable spatial resolution (30 m) to map and analyze ETc; regression models explained 91% of the variability in the observed data (RMSD50.064 mm/h; MBE50.04 mm/h). However, for MODIS-based ETc, the regression model explained only 59% of the variability in observed ETc with a larger error (RMSD50.17 mm/h; MBE50.15 mm/h). MODIS-based ETc was about 31% higher than the measured ETc. Imperfect assessment in MODIS-based retrievals is due to the underlying assumption of spatial heterogeneity and coarser sensor pixel scale (500 m), which was summarized by up-scaling the Landsat images to MODIS images using output flux aggregation and input up-scaling procedure using simple average and nearest neighbor aggregation techniques and comparisons were made on both image and pixel scales. Aggregation results illustrate that simple average with output flux aggregation provides close interpretation in aggregating fluxes to coarser resolution than other aggregation approaches. Pixel-by-pixel comparison using output aggregation with simple average resulted in close agreement (error range 5%–35%) between measured and up-scaled fluxes, compared to input up-scaling using simple average (error range 25%–60%). Larger error in input up-scaling is due to the changes in the surface roughness parameters due to aggregation in SEBS model. In addition, the magnitude of errors in ETc estimation was observed to be a function of the heterogeneity of the land surface and evaporative elements over the study region. Comparison between up-scaled ETc at 480 m spatial resolution with original MODIS image at 500 m showed that the output aggregation using simple average aggregation method provided closer representation of ETc at 500 m MODIS pixel resolution than the nearest neighbor resampling method.