PNNL

Abstract

A series of numerical experiments are used to evaluate the accuracy of two approaches for modeling topographically-driven subsurface flow suitable for inclusion in spatially distributed hydrologic models. The first is an explicit grid cell-by-grid cell approach, and the second is a statistical-dynamic method widely used in Topmodel. Both approaches were compared with analytical solutions to the kinematic wave equation for flow down an inclined plane of constant slope using a power law transmissivity function. The hillslope discharge and water table profiles simulated by the explicit method were in good agreement with the analytic solution in all test cases. The statistical-dynamic method converged on the correct steady-state solution, but failed to reproduce accurately transient conditions. The two algorithms were then compared using a 30-m DEM representation of the Mahantango, PA basin and hourly precipitation measured at the site. The agreement in discharge between the two methods improved with increasing soil thickness, surface conductivity, and power law exponent. However, the percent root-mean-squared-difference (RMSD) in hourly discharge was above 20 percent in all cases. The statistical-dynamic model was then "calibrated" to discharge simulated by the explicit method. Calibration was most effective for soils with low power law exponents, generally producing discharge hydrographs in close agreement for soils with linear transmissivity profiles. However, even under these conditions, there are large discrepancies between local water table depths simulated by the two models. The level of agreement is best during dry conditions and decreases during wetter conditions as the number of cells with surface soil saturation increases.