Resolution Matters: Numerical Analysis of the Effect of Subgrid Heterogeneities on Soil Moisture Distribution with a Physically Based Hydrological Model

Studies of hillslope hydrology reveal complex process thresholds associated with soil layering, heterogeneity, flow convergence, and unsaturated storage dynamics; critical for simulating runoff generation and slope stability. Such nuances are often neglected in coarser scale models, but have an important impact on the hydrologic response dynamics in space and time. To compensate for the use of a coarse resolution hydrological model, downscaling methods have been developed. However, the errors associated with such downscaling and their sources at the subgrid scale are not well known or understood. The choice of coarser resolutions is often linked with time constraints in operational setups or stochastic simulations, but also to the lack of high-resolution inputs. Reducing the resolution leads to a loss of information (e.g. topography smoothing), and to a lower accuracy in the representation of flow. Thanks to ongoing improvements in computing capabilities, in particular parallel computing, the use of regional, national, or continental-scale physically based hydrological models has spread in several fields. However, these have often focused on increasing spatial coverage of model domains rather than spatial resolution. In this work we provide a systematic numerical analysis of the effect of subgrid heterogeneities on hydrological fluxes, by choosing a domain of the size of a typical large-scale hydrological model (1.6km²) and comparing the baseline simulation, with no subgrid heterogeneity, to others in which topographic and soil heterogeneities are introduced. We use the fully coupled physically based hydrological model ParFlow.CLM to (a) quantify the effect of subgrid heterogeneities (slope, soil layering) on hydrological variables; (b) compare how these change in typical semiarid and wet temperate climates and two different soil types; and (c) to verify if downscaling techniques can compensate for a coarser resolution of the hydrological model, and where and why they fail. We show that while all the aspects considered affect the hydrological variables, the smoothing of concave topography has the strongest effect on the spatial and temporal distribution of soil moisture, leading to underestimation of runoff and, marginally, evapotranspiration.