Modeling soil evaporation efficiency in a range of soil and atmospheric conditions using a meta-analysis approach

A meta-analysis data-driven approach is developed to represent the soil evaporative efficiency (SEE) defined as the ratio of actual to potential soil evaporation. The new model is tested across a bare soil database composed of more than 30 sites around the world, a clay fraction range of 0.02–0.56, a sand fraction range of 0.05–0.92, and about 30,000 acquisition times. SEE is modeled using a soil resistance (rss) formulation based on surface soil moisture (h) and two resistance parameters rss;ref and hefolding. The datadriven approach aims to express both parameters as a function of observable data including meteorological forcing, cut-off soil moisture value h1=2 at which SEE50.5, and first derivative of SEE at h1=2, named Dh 21 1=2. An analytical relationship between ðrss;ref ; hefoldingÞ and ðh1=2;Dh 1=2Þ is first built by running a soil energy balance model for two extreme conditions with rss5 0 and rss 1 using meteorological forcing solely, and by approaching the middle point from the two (wet and dry) reference points. Two different methods are then investigated to estimate the pair ðh1=2;Dh 1=2Þ either from the time series of SEE and h observations for a given site, or using the soil texture information for all sites. The first method is based on an algorithm specifically designed to accomodate for strongly nonlinear SEEðhÞ relationships and potentially large random deviations of observed SEE from the mean observed SEEðhÞ. The second method parameterizes h1=2 as a multi-linear regression of clay and sand percentages, and sets Dh 1=2 to a constant mean value for all sites. The new model significantly outperformed the evaporation modules of ISBA (Interaction Sol-BiosphèreAtmosphère), H-TESSEL (Hydrology-Tiled ECMWF Scheme for Surface Exchange over Land), and CLM (Community Land Model). It has potential for integration in various land-surface schemes, and real calibration capabilities using combined thermal and microwave remote sensing data.