See more from this Session: Remote Sensing and Regional Scale Modeling
Wednesday, November 3, 2010: 1:00 PM
Long Beach Convention Center, Room 102A, First Floor
Remotely sensed imagery of the Earth’s surface via satellite sensors provides information to estimate the spatial distribution of evapotranspiration (ET). The spatial resolution of ET predictions depends on the sensor type and varies from the 30 – 60 m Landsat scale to the 250 – 1000 m MODIS scale. Therefore, for an accurate characterization of the regional distribution of ET, scaling transfer between images of different resolutions is important. Scaling transfer includes both up-scaling (aggregation) and down-scaling (disaggregation). In this paper, we address the up-scaling problem. The Surface Energy Balance Algorithm for Land (SEBAL) was used to derive ET maps from Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Moderate Resolution Imaging Spectroradiometer (MODIS) images. Landsat 7 bands have spatial resolutions of 30 to 60 m, while MODIS bands have resolutions of 250, 500 and 1000 m. Evaluations were conducted for both “output” and “input” up-scaling procedures, with aggregation accomplished by both simple averaging and nearest neighboring resampling techniques. Output up-scaling consisted of first applying SEBAL and then aggregating the output variable (daily ET). Input up-scaling consisted of aggregating 30 m Landsat pixels of the input variable (radiance) to obtain pixels at 60, 120, 250, 500 and 1000 m before SEBAL was applied. The objectives of this study were first to test the consistency of SEBAL algorithm for Landsat and MODIS satellite images and second to investigate the effect of the four different up-scaling processes on the spatial distribution of ET. We conclude that good agreement exists between SEBAL estimated ET maps directly derived from Landsat 7 and MODIS images. Among the four up-scaling methods compared, the output simple averaging method produced aggregated data and aggregated differences with the most statistically and spatially predictable behavior. The input nearest neighbor method was the least predictable but was still acceptable. Overall, the daily ET maps over the Middle Rio Grande Basin aggregated from Landsat images were in good agreement with ET maps directly derived from MODIS images.