On the identification of macroscopic root water uptake parameters from soil water content observations

Abstract

In this paper, we analyze the identification problem of macroscopic root water uptake parameters from soil water content observations. For this study, the macroscopic root water uptake is considered to be linearly decreasing with depth with A and B parameters conditioning, respectively, the maximum root water uptake at the surface and the decreasing rate with depth. For identification of parameter A and B, two different identification approaches are tested using a detailed soil moisture data set consisting of vertical profiles measured with time domain reflectometry (TDR) and neutron probe on 28 locations within a small maize cropped field during a dry period of 26 days. The first approach is based on a simplified water balance, while the second one uses an integrated soil‐vegetation‐atmosphere transfer (SVAT) model in an inverse mode. Results of the simplified water balance show first that the root water uptake measured within the field is quite variable with CV ranging from 22 to 34% for the root uptake parameters A and B, respectively. Furthermore, positive correlation between A and B suggests that low superficial root water uptake could be compensated by high deep root water uptake. Second, numerical simulations used to test the validity of this simplified approach show that the method is quite robust, at least for a certain range of A parameter values (0.008 < A < 0.013), for B, and for all investigated differently textured soils, except the coarse sand. To investigate the feasibility and the robustness of the second approach, we address the problems of insensitivity, instability, and nonuniqueness by means of numerical simulations. Results show, for soils with different textures, that the soil water content and, to a lesser extent, the time derivative of soil water content are quite insensitive to root water uptake parameters, at least compared to soil hydraulic parameters. Furthermore, instability analysis clearly illustrates that root water uptake parameters are quite and even very instable with respect to small uncertainties on the apparently known parameters (e.g., soil parameters). Finally, results show that if root water uptake plus additional other parameters (e.g., soil parameters) are simultaneously optimized, nonuniqueness of parameter sets must be expected. Therefore a robust identification of root water uptake parameters by means of a full SVAT inversion is unlikely to be achieved during a dry period with soil water content observations at least when some other parameters driving the system, like soil hydraulic properties, are not error free. In this context, further in depth research is needed to investigate if the use of longer periods characterized by both drying and redistribution events may increase the success of a full SVAT inversion. Finally, to extend the results of this study, we recommend applying a similar analysis to other macroscopic conceptual root water uptake models.