Temporal upscaling of instantaneous evapotranspiration: An intercomparison of four methods using eddy covariance measurements and MODIS data

Abstract

The quantification of land surface evapotranspiration (ET) at daily or longer time scales is of great significance in modeling the global hydrological cycle, studying climate change, and managing water resources. However, current remote sensing-based ET models can generally only provide snapshots of ET at the time of a satellite overpass and do not satisfy the expectations of hydrologists, irrigation engineers, or water resource managers concerned with practical applications. Four commonly used ET upscaling schemes, namely, the constant reference evaporative fraction (EFr) method, the constant evaporative fraction (EF) method, the constant extraterrestrial solar radiation ratio (R-p) method, and the constant observed global solar radiation ratio (R-g) method, were evaluated in this study using ground-based eddy covariance (EC) measurements and the Moderate Resolution Imaging Spectroradiometer (MODIS) derived estimates from a two-source energy balance model. Analysis was made of both closed and un-closed surface fluxes and of different assumed satellite overpass times from mid-morning to mid-afternoon. Data for the analysis were collected at the Yucheng comprehensive experimental station in Northern China, spanning the period from late April 2009 to late October 2011. The results show that all four upscaling factors from noon to mid-afternoon had a better agreement with their corresponding daily averages. Overall, the EFr (R-g) method had the best (second best) performance of the four upscaling methods. The EF method was found to significantly underestimate the daily latent heat flux (LE) and performed the worst of the four upscaling methods. With the correction of the energy imbalance of EC measurements, the R-g method's performance was improved and outperformed the EF, method in the morning. The presence of clouds either increased or decreased the BIAS but generally increased the root mean square error (RMSE) for all four upscaling methods. When the upscaling methods were applied to convert the instantaneous MODIS remote sensing estimates of ET to daily values, the accuracy of the extrapolated daily ET was controlled by the accuracy of both the remote sensing ET estimates and the upscaling methods themselves. With the insignificantly biased estimates of the instantaneous LE from the N95 model at 101 MODIS overpass times in this study, using the EF method underestimated the daily ET by 11%, while all the remaining three factors overestimated the daily ET by a range of 5%-18%. This study was conducted to provide a scientific basis for developing an operational and more accurate ET-upscaling method with easy access to data in future studies. (C) 2013 Elsevier Inc. All rights reserved.