Water infiltration and redistribution in Land Surface Models

In this report, we examine the potential of integrated Land Surface Models that are part of Numerical Weather Prediction (NWP) systems to produce surface runoff in case of intense rainfall events. We chose to run two infiltration experiments using the SWAP, JULES and CHTESSEL models. In the first experiment, intensive rainfall was represented as a step function and in the second constant intensive rainfall was considered. These types of problems (with discontinuities) are challenging for the numerical solvers. We performed a mesh dependence study changing the vertical resolution of the soil column. This revealed non-physical oscillations in predicted soil moisture profiles, hydraulic conductivity profiles and surface/subsurface runoff in JULES and SWAP, while CHTESSEL does not present this behaviour due to its approach in computing the hydraulic conductivity. It has been also identified that infiltration representation in JULES does not allow for the gradual increase of surface runoff expected in the experiments and predicted by the other two codes. We recommend future work should consider correcting the maximum infiltration in JULES and investigating the numerical schemes in order to make it high resolution ready.