Six crop models differ in their simulation of water uptake

Root water uptake is an essential component of crop models since it affects plant growth and, through its effect on the soil water balance, multiple soil and nutrient cycling processes. Several methods to simulate water uptake exist; however, the differences among them have not been evaluated. We compared the water uptake methods implemented in six crop models: APSIM, CropSyst, DSSAT, EPIC, SWAP and WOFOST. These methods range from simple empiric approaches (WOFOST) to mechanistic approaches based on the water potential gradient and root distribution in the soil-plant system (CropSyst). We compared the six models' water uptake algorithms in scenarios with different evaporative demand, soil texture, and water distribution with depth. The main difference among methods derived from the degree to which each model enabled the use of water in the subsoil (below 0.5 m). In a rooted, 1-m deep silt loam soil in which the root density decreased geometrically with depth and which was subjected to an evaporative demand of 5 mm d(-1) for 60 days, APSIM, EPIC, DSSAT and SWAP transpired about 83% of the total plant available water while SWAP and CropSyst transpired about 65% of it. When methods were compared with initially dry bottom soil layers, cumulative transpiration became similar for all methods, while the opposite initial condition exacerbated differences. All methods, except CropSyst, increased transpiration as the evaporative demand rose to relatively high rates (10 mm d(-1)) because they lack a feedback mechanism that reduces transpiration when the demand exceeds the plant's ability to conduct water. CropSyst, DSSAT, EPIC and SWAP developed a drying front, as usually observed in field conditions, while APSIM and WOFOST showed relatively uniform water depletion with depth in the soil profile. In conclusion, the models differ meaningfully in their simulation of water uptake and careful consideration of these differences is needed to properly use and interpret the outcome of model simulations. (C) 2016 Elsevier B.V. All rights reserved.