SCALING OF SURFACE ENERGY FLUXES USING REMOTELY SENSED DATA

Accurate estimates of evapotranspiration (ET) across multiple terrains would greatly ease challenges faced by hydrologists, climate modelers , and agronomists as they attempt to apply theoretical models to real-world situations. One ET estimation approach uses an energy balance model to interpret a combination of meteorological observations taken at the surface and data captured by remote sensors. However, results of this approach have not been accurate because of poor understanding of the relationship between surface energy flux and land cover heterogeneity, combined with limits in available resolution of remote sensors. The purpose of this study was to determine how land cover and image resolution affect ET estimates. Using remotely sensed data collected over El Reno, Oklahoma, during four days in June and July 1997, scale effects on the estimation of spatially distributed ET were investigated. Instantaneous estimates of latent and sensible heat flux were calculated using a two-source surface energy balance model driven by thermal infrared, visible- near infrared, and meteorological data. The heat flux estimates were verified by comparison to independent eddy-covariance observations. Outcomes of observations taken at coarser resolutions were simulated by aggregating remote sensor data and estimated surface energy balance components from the finest sensor resolution (12 meter) to hypothetical resolutions as coarse as one kilometer. Estimated surface energy flux components were found to be significantly dependent on observation scale. For example, average evaporative fraction varied from 0.79, using 12-m resolution data, to 0.93 , using 1-km resolution data. Resolution effects upon flux estimates were related to a measure of landscape heterogeneity known as operational scale, reflecting the size of dominant landscape features. Energy flux estimates based on data at resolutions less than 100 m and much greater than 400 m showed a scale dependent bias. But estimates derived from data taken at about 400-m resolution (the operational scale at E l Reno) were susceptible to large error due to mixing of surface types. The El Reno experiments show that accurate instantaneous estimates of ET require precise image alignment and image resolutions finer than landscape operational scale. These findings are valuable for the design of sensors and experiments to quantify spatially-varying hydrologic processes