Evaluation of the Soil Model of the Hydro–Thermodynamic Soil–Vegetation Scheme by Observations and a Theoretically Advanced Numerical Scheme

Abstract

The soil module of the Hydro?Thermodynamic Soil?Vegetation Scheme is evaluated by soil temperature observations and independent theoretical numerical results. To gain the latter, a Galerkin weak finite-element (GWFE) scheme is implemented for solving the heat and water balance equations that are originally solved by a Crank?Nicholson finite-difference (CNFD) scheme. The GWFE scheme captures discontinuities well and has a high phase fidelity. When/where frozen ground thaws and under moderate advection-dominated regimes, peak temperatures simulated with the CNFD scheme are up to seven days off compared with observations and the results of the GWFE scheme. If freeze?thaw cycles repeat for more than a month, CNFD predictions will oscillate ±1 K around the observations but will converge to the observations and results of the GWFE scheme afterward. Under diffusion-dominated regimes, CNFD runs perform well with similar quality to the GWFE predictions. Comparisons of the results of both numerical schemes substantiate that the long spinup time of CNFD simulations results from the numerical scheme and not from the initialization procedure and that the diffusive nature of the CNFD scheme and not parameterized physical processes causes phase shifts. GWFE requires 1.6?2.8 more CPU time than CNFD in this study. Unless CPU time is an issue, the GWFE scheme is recommended because of its high phase fidelity and short spinup.