Abstract
Soil is the largest reservoir of water in a catchment and evapotranspiration is the largest yearly output flux of water from the soil in semiarid environments at the catchment scale. Hydrologic stress is one of the typical characteristics of semiarid environments. The shortage of water affects all the biophysical processes and increases the nonlinearity in the rainfall–runoff relationship. A proper description of semiarid watersheds should include an accurate description of the wetting–drying cycle of the soil, which is critical to properly describe important processes such as the connection between the surface and the soil saturated zone, the production of runoff or the growth of plants. Evapotranspiration is typically modelled in comprehensive distributed hydrologic models (e.g. MIKE‐SHE or MODHMS) using functions that relate evapotranspiration to soil moisture and plant biomass through a set of empirical coefficients, which typically adopt some default values regardless of the conditions of the area where the model is applied. In this work, using a global and a local search algorithm, a three‐dimensional soil water flow model coupled to an evapotranspiration model is calibrated in a soil profile covered with grass in a semiarid environment. Four parameters for the water flow model and three parameters for the evapotranspiration model are calibrated. An existing trend between daily average soil moisture and daily maximum temperature is used as extra information during the calibration process. The results showed that the known relationship between maximum temperature and average soil moisture includes information on the seasonality of the atmospheric demand. Although total evapotranspiration is similar in all calibrations, the trade‐off between evaporation and transpiration is sensitive to this extra information mainly through one model parameter. Further research is necessary to identify the value of the parameter that produces the correct trade‐off. Copyright © 2008 John Wiley & Sons, Ltd.