Root water uptake under non-uniform transient salinity and water stress

All existing reduction functions and those newly developed in this study for separate and combined salinity and water stress are used in the macroscopic root water uptake model and tested against experimental data. The experimentally obtained data are used to derive the parameter values needed for the simulation model HYSWASOR. Under salinity stress, both experimental and simulated results indicate that the well-known linear crop response function can be used as a reduction function. When experimentally derived parameters are used in the simulation model calibration is still needed. The parameter values obtained by calibration differ slightly from the experiments because the experimentally derived parameter values are based upon mean soil solution salinity. Simulation results indicate that different salinity reduction functions can provide almost the same results if the parameter values are well specified. Under water stress, since the experimental data shows nonlinear trend, the linear reduction function cannot fit the data. The existing nonlinear reduction functions can fit only half of the data range satisfactorily. The best agreement is obtained with the newly developed nonlinear two-threshold reduction function. Soil water pressure head heterogeneity over the root zone does not play an important role in water uptake. The roots appear to take up water from the relatively wetter parts of the root zone to compensate for the water deficit in the drier parts. While the simulated transpiration agrees closely with the experimental data, the main reason for the discrepancy between the simulated and actual water contents appears to be water uptake during the night. Under combined water and salinity stress, the additive and multiplicative reduction functions are first tested against the experimental data and then inserted in the simulation model. A new combination reduction function is introduced that differs conceptually from the additive and multiplicative functions. Both the experimental and simulated results show that the newly proposed model fits the data best, while the worst results are obtained with the additive model.