Estimates of evapotranspiration from MODIS and AMSR-E land surface temperature and moisture over the Southern Great Plains

We apply a new stationarity-based method for estimating the parameters of a simplified land surface model over a 10° × 10° study area in the Southern Great Plains. The model is then used to calculate evapotranspiration (ET) using the estimated parameters, atmospheric forcing data (from NLDAS-2) and remotely sensed surface soil moisture (from AMSR-E) and temperature (from AMSR-E and MODIS). By exploiting the statistical relationship between the forcing (e.g., precipitation and radiation) and surface states (soil moisture and temperature), the stationarity-based parameter estimation avoids the restrictive requirements of direct calibration (i.e., the need for ground-truth ET), enabling broad applicability. Modeled annual mean ET rates are compared with benchmark values based on runoff data from Global Runoff Data Center (GRDC), and the root mean square error (RMSE) between them is less than 10 Wm− 2. Spatial patterns of model estimated ET and of key model parameters indicate that radiation and moisture availability are the controlling factors on ET in the study area. The spatial distributions of the estimated parameters appear reasonable: they are smooth and not overly patchy or random; and they are consistent with the land cover and soil moisture distribution through the area. An important feature of the stationarity-based method demonstrated here is that forcing and surface states need not be continuous. In this application, approximately 30% of the daily surface temperature and moisture data were missing.

► Parameters estimated from covariance of meteorological and surface data, not fluxes. ► Evaporation is calculated with estimated parameters and satellite surface states. ► Widely applicable method since evaporation not needed for calibration of parameters. ► Meteorological and remotely sensed data can be sparse and need not be continuous.