Inverse Estimation of Soil Hydraulic and Root Distribution Parameters from Lysimeter Data.

Simultaneous determination of soil hydraulic properties and a rooting depth parameter was studied by inverse modeling of unsaturated water flow with root water uptake in weighable lysimeters under transient atmospheric boundary conditions. Inclusion of pressure head data in the objective function yielded unique parameter estimates even for layered profiles. We investigated the feasibility of simultaneous identification of soil hydraulic and root-distribution parameters by inverse simulation of soil water flow in monolithic lysimeters under atmospheric boundary conditions using the Richards equation and a macroscopic root water uptake model. Weighable lysimeters are powerful test systems for this purpose because the boundary fluxes (precipitation, actual evapotranspiration, and seepage across the bottom) can be determined very precisely. We analyzed the amount of information needed for the unique identification of parameters and investigated the magnitude of their uncertainties. First, we examined synthetic data sets for different scenarios and instrumentation campaigns that differed in their information content and complexity of soil properties. Atmospheric boundary conditions as measured at the lysimeter station in Wagna, Austria, were used as forcing data. The results show that for homogeneous profiles, cumulative outflow and profile-averaged water content data contain enough system information to allow the simultaneous estimation of soil hydraulic properties and root-distribution parameters. In contrast, for soil profiles consisting of two layers, unique soil hydraulic parameters and the correct rooting depth could only be estimated if matric potential measurements from both layers were included in the objective function. Finally, soil hydraulic properties of the grass-reference lysimeter in Wagna were estimated using real measurements. Water dynamics in the lysimeter could be described well by an effective parameterization assuming a homogeneous soil profile. Furthermore, the system behavior under different boundary conditions could be predicted adequately with the estimated parameters. This demonstrates the usefulness of the identified system properties for predictive modeling. [ABSTRACT FROM AUTHOR]