Evapotranspiration model selection for estimation of actual evaporation from bare soil, as required in annual potential groundwater recharge studies of a semi-arid foothill region

Improved evapotranspiration estimation is instrumental in annual potential recharge (Re) evaluations of semi-arid regions, in the context of the water balance approach. Complementary relationship areal evapotranspiration (CRAE) and advection-aridity (AA) and a general combination equation proposed by Granger and Gray (GG) were evaluated for the estimation of actual evaporation (E-a) from bare soil under occurrence of Re. Model capability was evaluated by utilizing data from lysimeter measurements during the hydrological years [(2011-2012) and (2012-2013)]. Measured parameters included soil moisture (SM), rainfall and related meteorological data and Re (lysimeter output). All models with original parameters performed poorly during 2011-2012 and 2012-2013, underestimating annual Re values. The 2011-2012 data were used for model calibration, considering annual E-a [criterion (I)] and cumulative E-a during the occurrence period of Re [criterion(II)]. The calibrated CRAE model [criterion (II)] produced the best simulation for E-a (NRMSE<30%). The 2012-2013 data were used to validate the models, and the calibrated CRAE model produced acceptable results (NRMSE<10%) in SM simulation (both criteria). But calibrated CRAE under criterion (II) provided the best estimation of annual Re (Delta Q<6%). According to the results, CR models can produce acceptable estimation of annual Re, if initially calibrated utilizing criterion (II).