CALFRET: A high-performing, CALibration-FRee 1-km EvapoTranspiration product over the conterminous USA
Year: 2025
Type: preprint
Abstract: Obtaining accurate spatially explicit estimates of evapotranspiration (ET) is challenging. This study utilizes a calibration-free approach based on Surface Flux Equilibrium Theory to develop a 1 km daily gridded ET product, CALFRET, over the conterminous United States (CONUS) from 2000 to 2023. When evaluated at 129 inland eddy covariance sites, CALFRET achieved KGE, NSE, r, PBIAS, and MARE values of 0.56, 0.47, 0.72, 21.27, and 0.47, respectively. CALFRET had a relatively better performance in wet and forested sites. The product outperforms MODIS, GLEAM, and ECOSTRESS across all six considered evaluation metrics, SSEBop on five, and OpenET on four. The product outperforms all considered products across all six metrics when evaluated against raw (i.e., uncorrected) evapotranspiration data. It demonstrates better performance than OpenET in subhumid, semiarid, and arid regions, as well as in grasslands, evergreen forests, mixed forests, and shrublands. Importantly, it outperforms the compared products during high evapotranspirative stress periods. Furthermore, CALFRET can enhance existing ensemble ET platforms such as the OpenET framework. Overall, CALFRET provides accurate, gap-free, daily, long-term, fine-resolution ET estimates across CONUS, making it valuable for both research and applied decision-making. INTRODUCTIONEvapotranspiration (ET) data is widely used in water budget analysis (Lv et al., 2017; Pan et al., 2012; Rodell et al., 2011; Trenberth et al., 2007) and hydrological modeling (Chen et al., 2015; Herman et al., 2018; Kumar and Duffy, 2016; López López et al., 2017; Rajib et al., 2018; Zhao et al., 2013). It also plays a critical role in analyzing components of water, carbon, and energy cycles (Fisher et al., 2017; Katul et al., 2012; Yang et al., 2023). Beyond these applications, ET data supports the understanding of diverse ecohydrological processes and land-atmosphere feedback (Berg et al., 2016; Dirmeyer et al., 2012; S. Kim et al., 2023; Meng et al., 2014; Salvi and Kumar, 2024), plant stress responses (Anderson et al., 2007; Lorenz et al., 2021; Raghav and Kumar, 2024), tree mortality (Anderegg et al., 2016; Das et al., 2013; Liu et al., 2017), crop water use and productivity (Djaman et al., 2013; Katerji et al., 2013; Ko and Piccinni, 2009; Liu et al., 2002), and drought dynamics (Beguería et al., 2014; Kim and Rhee, 2016; Teuling et al., 2013). Despite its importance, obtaining a precise estimate of ET in a spatially-explicit fashion, on a daily basis, and without gaps remains a significant challenge. Several gridded products of ET has been developed over the years to address aforementioned needs (Miralles et al., 2025; Senay et al., 2022). A recent study by (Thakur et al., 2025) showed that ET estimates using the Surface Flux equilibrium theory (SFET) (McColl et al., 2019; McColl and Rigden, 2020) outperformed continental remote sensing products such as ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS), Simplified Surface Energy Balance (SSEBop), and Moderate Resolution Imaging Spectroradiometer (MODIS), at 33 eddy covariance sites. The SFET method assumes an approximate balance between surface heating and surface moistening terms in the near-surface relative humidity budget for inland areas over daily to longer time scales (McColl et al., 2019; McColl and Rigden, 2020; Salvucci and Gentine, 2013). This translates to atmospheric states having enough information to estimate ET. Given that the atmospheric states respond to changes in land use and land cover, the method is suitable to track ET under changing environment. The method holds a promise for providing ET estimate in a spatially-explicit fashion over large spatial scales. Already, SFET has been applied over 221 watersheds over the CONUS, where it showed comparable performance with Coupled Model Intercomparison Project Phase 6 (CMIP6) models, and captured the multi-year mean annual latent heat obtained from catchment water balance estimates (Chen et al., 2021). Another study used SFET to estimate evaporation at 0.25 o spatial resolution (Zhu et al., 2024) globally for 1991-2014 using inputs from ERA (9km) and Global Land Data Assimilation System (GLDAS) (0.25o) models. In their study, for comparison at finer resolutions, SFET based evaporation fraction values were resampled to 0.08o. Overall, most applications of SFET across continental or larger spatial extents has primarily been performed at coarse resolutions. This study (a) implements SFET at a 1 km scale thus offering a more spatially detailed ET estimate over the conterminous United States (CONUS); (b) thoroughly evaluates the method’s capabilities to quantify ET using a recent exhaustive compilation of quality-checked eddy covariance data (Volk et al., 2023); (c) intercompares the method’s performance against sister continental or larger scale gridded ET products such as OpenET, SSEBop, GLEAM, MODIS, ECOSTRESS and X-BASE; and finally but importantly (d) open-sources the data for use by the scientific community and the general public for the period spanning 2000 to 2023.MATERIALS AND METHODSThe following sections describe the methodology outlined in Figure 1 to estimate daily ET at 1 km spatial resolution over the CONUS using the SFET-based ET.
Authors: Mukesh Kumar, Fitsume T Wolkeba, Pushpendra Raghav, Hitesh Thakur
Institution: University of Alabama
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: gold
OpenAlex ID: W4414629415
Cites:
Cited by:
Related to: