Impact of insolation data source on remote sensing retrievals of evapotranspiration over the California Delta

Authors: Anderson, Martha; DIAK, G. - University Of Wisconsin; Gao, Feng; Knipper, Kyle; HAIN, C. - Goddard Space Flight Center; EICHELMANN, E. - University Of California; HEMES, K. - University Of California; BALDOCCHI, D. - University Of California; Kustas, William - Bill; Alfieri, Joseph

Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/4/2019
Publication Date: 1/22/2019
Citation: Anderson, M.C., Diak, G., Gao, F.N., Knipper, K.R., Hain, C., Eichelmann, E., Hemes, K., Baldocchi, D., Kustas, W.P., Alfieri, J.G. 2019. Impact of insolation data source on remote sensing retrievals of evapotranspiration over the California Delta. Remote Sensing. 11:216. https://doi.org/10.3390/rs11030216.
DOI: https://doi.org/10.3390/rs11030216

Interpretive Summary: Evapotranspiration (ET) describes the rate of consumptive water use by crops and other landcover types. The ability to accurately estimate ET spatially and temporally over large areas is critical to a broad range of applications, for example in managing water resources, in assessing agricultural water use, and in monitoring ecosystem health and impacts of drought. The primary factor governing ET in many cases is the solar radiation load (insolation), which largely determines the energy available at the land surface to evaporate water. Therefore, the accuracy of ET modeling systems is tied to the accuracy of the insolation datasets used as input to the models. This paper assesses the accuracy of three different gridded insolation datasets available over the United States in comparison with ground-based measurements collected in central California. A new near real-time satellite-based product from the University of Wisconsin shows excellent agreement with observations, and shows excellent promise for operational water use assessments within the U.S. When used in remotely sensed ET retrievals over central California, this insolation dataset provided improved water use estimates in comparison with standard model-based dataset. However, the improvement in ET accuracy was significantly smaller than that found for the insolation inputs, indicating that other inputs, modeling assumptions and site conditions are limiting performance over these sites. This is likely due in part to the predominantly clear sky conditions that are prevalent in California during the peak growing season. Larger improvements in water use estimates are expected over more humid sites, as in the U.S. Corn Belt and eastern U.S. crop growing regions.

Technical Abstract: The energy delivered to the land surface via insolation is a primary driver of evapotranspiration (ET) – the exchange of water vapor between the land and atmosphere. Spatially distributed ET products, available in near real time, are in great demand in water resource management community for real-time operations and sustainable water use planning. The accuracy and deliverability of these products will be determined in part by the characteristics and quality of the insolation data sources used as input to ET models. This paper investigates the practical utility of three different insolation datasets within the context of a satellite-based remote sensing framework for mapping ET at high spatiotemporal resolution, in an application over Sacramento-San Joaquin Delta region in California. The datasets tested include one reanalysis product: the Climate System Forecast Reanalysis (CFSR) at 0.25O spatial resolution, and two remote sensing insolation products generated with geostationary satellite imagery: a product for the continental United States at 0.2O developed by the University of Wisconsin Space Sciences and Engineering Center (SSEC) and a coarser resolution (1O) global Clouds and the Earth’s Radiant Energy System (CERES) product. The three insolation data sources are compared to pyranometer data collected at flux towers with the Delta domain to establish relative accuracy. The satellite products significantly outperformed CFSR, with root-mean square errors (RMSE) of 2.7, 1.52 and 1.43 MJ m-2 d-1for CFSR, CERES and SSEC, respectively, at daily timesteps. The satellite-based products provided more accurate estimates of cloud cover occurrence and radiation transmission, while the reanalysis tended to underestimate solar radiation under clouds. This difference in insolation performance, however, did not translate into comparable improvement in ET retrieval accuracy. RMSE in daily ET was 0.98 and 0.94 mm d-1 using CFSR and SSEC insolation data sources, respectively, although some individual sites improved in RMSE by up to 0.1 mm d-1. The lack of notable impact on aggregate site performance may be due in part to the predominantly clear-sky conditions prevalent in central California, under which the reanalysis and satellite-based insolation data sources have more comparable accuracy. Residual errors are largest at complex sites where observed ET is less strongly correlated with insolation due to intensive agricultural management or other controls on vegetation growth and heat storage, which may not be fully captured in the energy balance framework. While satellite-based insolation data could improve ET retrieval in more humid regions with greater cloud-cover frequency, over the California Delta and climatologically similar regions in the western U.S., the CFSR data may suffice for real-time ET modeling efforts.