Characterizing the Effects of Temporal Upscaling on Remote Sensing-based Estimates of Evapotranspiration at Field Scales

Thermal remote sensing-based techniques exploit the linkage between land surface temperature and evapotranspiration (ET) to provide spatially distributed estimates of the moisture flux. While these techniques have demonstrated broad utility for estimating ET at large spatial scales, their applicability at field scales is limited by the infrequent acquisition of high-resolution thermal imagery. To overcome this limitation, a number of temporal upscaling methods have been developed to fill the gaps in ET retrievals. However, these approaches are predicated on the assumption that the scaled metric, e.g. evaporative fraction, used is self-preserving (persistent) and varies smoothly over time. As a result, they may not be able to adequately capture the effects of changing surface and atmospheric conditions that rapidly alter the partitioning of the surface energy fluxes. In order to quantify the impacts of temporal upscaling, in-situ measurements of ET collected over a range of land use type was used as proxy for remote estimates of ET. This study considered return intervals of up to 32 days along with a number of differing scaled metrics and interpolation methods. The results of the study suggest that the degree of persistence, thus the accuracy of the ET estimate, varied significantly both with the scaled metric used and the land use type. For example, it was found that agricultural ecosystems exhibit a higher persistence as measured via autocorrelation than natural landscapes. Similarly, the accuracy of the ET estimates tends to be greater at the agricultural sites. Assuming a 10-day return interval – this is necessary for the comparison because the error in the ET estimates increases logarithmically with return interval – the percent error of the daily ET estimates for agricultural sites averaged 20% while it averaged nearly 27% for the natural ecosystems. Overall, the study indicates that the impact of temporal upscaling is strongly influenced by both surface and atmospheric conditions.