The sensitivity of a critical zone model to the representation of hydraulic conductivity heterogeneity in a deeply weathered hard rock aquifer in West Africa

Currently 40 % of Africa's population still lacks access to clean water. Twice as many rural people live in hard rock areas as compared to sedimentary areas. In these hard rock areas a thick weathered regolith layer covers the crystalline basement, where groundwater (GW) circulates. In the Sudanian area of West Africa (WA) ,groundwater levels are shallow enough to interact directly with the surface water. Therefore, constructing coupled surface-groundwater models helps to estimate quantities of both, GW and surface flows, and their evolution over time to facilitate integrated water management. However, the sensitivity of such models to aquifer properties (saturated hydraulic conductivity (Ks), porosity, geometry), which are difficult to obtain in heterogeneous crystalline contexts, is still poorly constrained. The heterogeneity of aquifer properties at the scale at which most water management decisions are taken, is twofold: 1) bimodal vertical heterogeneity with an unconsolidated weathered zone (high porosity, low Ks) overlying a fissured zone (low porosity, high Ks) and 2) lateral heterogeneity controlled by substratum features and weathering history. We assessed the sensitivity of a coupled surface-groundwater model (PARFLOW-CLM) to vertical and lateral heterogeneity of Ks. The sensitivity to the lateral heterogeneity was explored either using simulations with homogeneous or distributed Ks following random field approaches with a range of spatial correlation lengths. The representation of a vertically uniform aquifer layer was compared to a two-layer scenario for each of the lateral heterogeneity cases. Here, we focused our analysis on the Northern Oueme catchment in Benin (14 000 km²) and we constructed a model with a spatial resolution of 1 km², preventing the analysis of smaller-scale features, such as macropores or clay accumulations. Hydraulic conductivity and aquifer geometry data to constrain the sensitivity experiments were derived from the literature specific to the target area, but also from regional hard rock aquifers in West Africa. As an output of the model, we obtained streamflow, water table head and evapotranspiration time series (in a monthly and daily resolution). The results that we gained with our model configuration (and resolution) point towards a low sensitivity of the model to lateral and vertical heterogeneity. However, we observed a significant impact of the magnitude of Ks on water table head and particularly on the streamflow amplitude. Regarding the water balance our results show that further exploration of the subsurface is crucial to improve critical zone modeling in the context of WA.