Parameter estimation for a physics‐based distributed hydrologic model using measured outflow fluxes and internal moisture states

Abstract

We use an inverse simulation strategy to estimate soil hydraulic parameter values for an extensively measured planar hillslope plot in Seattle, Washington, United States. Both the integrated (subsurface outflow) and internal (piezometer water levels, volumetric water contents) hydrologic responses are measured at the plot. Inverse simulation scenarios are configured in the physics‐based variably saturated hydrologic model, HYDRUS‐2D, for a nonhysteretic drainage scenario starting from saturated initial conditions. Multiple inverse simulations calibrate the model either to single‐measurement time series or to combinations of multiple types of measurements. Inverse simulations calibrated to different types of measurements give a wide range of parameter combinations, including over 2 orders of magnitude in predicted saturated hydraulic conductivity (Ks), in part because the calibrations to a single measurement type are poorly constrained and biased. Parameter values are better constrained with multiobjective inverse simulations (Ks from 30 to 55 cm h−1). All parameter combinations from inverse simulations were tested in 2‐month‐long continuous simulations of the plot flow response to natural precipitation and evapotranspiration. The long‐term outflow response was predicted best (Nash‐Sutcliffe E = 0.94) by the parameters from a multiobjective inverse simulation calibrated to both the outflow and the piezometer water levels. Overall results show that for an assumed nonhysteretic soil a physics‐based hydrologic response model can be calibrated using one short‐duration drainage‐from‐saturation event if both integrated (outflow) and internal (saturated water level) measurements are used as calibration objectives.