Improving river-floodplain relationship in PARFLOW-CLM in continental West Africa as an early step to build an multi source hydrologic reanalysis

Advances in hydrology science are opening new opportunities for using hydrology models as part of larger information systems in order to produce hydrologic reanalysis datasets. These products or services are especially important in areas with limited field-based information. These have to be consistent in terms of available data and in terms of water budget. In this regard, models that integrate surface and groundwater hydrology at fine resolution, based on physical principles, such as the PARFLOW-CLM model, offer advantages as they are fed by measurable parameters, they can be run at large scale, and provide results at high spatial and temporal resolution. Finally these models are able to be upgraded with missing or poorly represented processes.

One example of this continuous improvement is the representation of floodplains: these wetlands correspond to areas that are regularly flooded by large rivers and maintain a complex relationship between surface water, groundwater, and land surface fluxes. Furthermore, due to the flooding conditions, floodplains present highly damageable hydrological risk and high-frequency saturated soil conditions, which favour a significant hotspot of biodiversity, sustain key ecosystem services for human communities, and regulate hydrological flows. But floodplain dynamics are difficult to represent in large-scale hydrologic models because of the control that small-scale topography exerts on water flow and storage. Furthermore in the case of a fine resolution, the simulation must involve the use of explicit relationships and physics-based equations with a high computational cost.

A zone where these difficulties are clearly depicted is the Continental West Africa (CONWA). The CONWA domain covers an area of 3.5 million km² that contains some of the world's largest floodplains, such as the inland Niger River delta. It also covers other smaller intermittent endorheic ponds, with not measured data, and where the combination of wetlands, rivers and aquifers controls both low water levels in dry seasons, downstream high water levels, and induces preferential recharge pathways.

In these perspectives PARFLOW-CLM is implemented in the CONWA domain at 1 km² resolution using the ERA-5 reanalysis, IMERG precipitation dataset, and Copernicus Leaf Area Index data. The methodology representing floodplains prescribe an anisotropic layer near the surface in areas that are “regularly flooded” to allow up-slope flows driven by water head gradient. This anisotropic layer is defined by a depth and a tensor factor affecting horizontal permeability, and allows river grids to connect to neighboring floodplain grids when the water level is high enough to flood them.

We will focus our analysis in disentangle and evaluate the effects of floodplains in three main variables: river discharge seasonality, evapotranspiration, and groundwater storage. These results are important to improve the representation of key hydrologic elements in large-scale hydrology models and Earth system models, and constitute a step toward creating a multi source hydrologic reanalysis system.