Continental-scale net radiation and evapotranspiration estimated using MODIS satellite observations

Evapotranspiration (ET) is a major pathway for water loss from many ecosystems, and its seasonal variation affects soil moisture and net ecosystem CO2 exchange. We developed an algorithm to estimate ET using a semi-empirical Priestley–Taylor (PT) approach, which can be applied at a range of spatial scales. We estimated regional net radiation (Rnet) at monthly time scales using MODerate resolution Imaging Spectroradiometer (MODIS) albedo and land surface temperature. Good agreement was found between satellite-based estimates of monthly Rnet and field-measured Rnet, with a RMSE of less than 30 W m− 2. An adjustable PT coefficient was parameterized as a function of leaf area index and soil moisture based on observations from 27 AmeriFlux eddy covariance sites. The biome specific optimization using tower-based observations performed well, with a RMSE of 17 W m− 2 and a correlation of 0.90 for predicted monthly latent heat. We implemented the approach within the hydrology module of the CASA biogeochemical model, and used it to estimate ET at a 1 km spatial resolution for the conterminous United States (CONUS). The RMSE of modeled ET was reduced to 21.1 mm mon− 1, compared to 27.1 mm mon− 1 in the original CASA model. The monthly ET rates averaged over the Mississippi River basin were similar to those derived using GRACE satellite measurements and river discharge data. ET varied substantially over the CONUS, with annual mean values of 110 ± 76 mm yr− 1 in deserts, 391 ± 176 mm yr− 1 in savannas and grasslands, and 840 ± 234 mm yr− 1 in broadleaf forests. The PT coefficient was the main driver for the spatial variation of ET in arid areas, whereas Rnet controlled ET when mean annual precipitation was higher than approximately 400 mm yr− 1.

Research highlights
► We estimate surface Rnet and ET at a high spatial resolution using satellite data.
► The Priestley-Taylor ET algorithm optimized with AmeriFlux data performs well.
► The algorithm improves ET and soil moisture estimate in CASA model.
► Basin-scale runoff estimate agrees well with other water budget analysis.
► Spatial distribution of evapotranspiration reveals different controls on ET.