Fully coupled subsurface-land surface hydrological models: A scaling approach to improve subsurface storage predictions

The flow conditions (e.g., river network) for rivers and open channels are often forced into hydrogeological models that use a constant horizontal grid resolution without correction for grid mismatching. As a result, the flow velocity will be significantly underestimated if the width of rivers is substantially narrower than the grid size of these models. Furthermore, the exchange between the river and the subsurface is overestimated, resulting in an erroneously large vertical exchange.

In response to this challenge, in this work, the subscale channel flow is approximated in the kinematic wave equation by a scaled roughness coefficient. A relationship between grid cell size and river width is used for this purpose, which follows a simplified modification of the Manning-Strickler equation. In addition, the exchange between the subsurface and the river, as well as the rate of ex- and in-filtration, are scaled across river beds based on grid resolution. As a result, even though the grid size is relatively large, the exchange rates are corrected across river beds. The effectiveness of the scaling of river parametrization is validated against groundwater gauges and remote sensing-based surface soil moisture in a fully coupled subsurface-land surface ParFlow-CLM at a spatial resolution of 0.055° (~6 km) over the Upper Rhine Basin. The validity of the results is examined through an innovative application of the First Order Reliability Method (FORM) for the time period 2012-2014. Results indicate that the scaling approach improves the estimates of soil moisture, particularly in the summer and autumn seasons when cross-validated with independent CCI-SM observations. This improvement is achieved (SM RMSE reduction from 0.03 to 0.005) due to the effective impacts of the scaling river parametrization on SM estimation. FORM results show that the accuracy of ParFlow-CLM soil moisture simulations by using scaling approach is more than 95, 89, 85 and 92 percent for Autumn, Winter, April and Summer, respectively. The scaling river parametrization also shows overall improvements in groundwater level estimation, particularly over the central and northern regions where the groundwater level is shallow.