Process refinements improve a hydrological model concept applied to the Niger River basin

This study examines a method to improve a process-oriented hydrological model concept applied to another region than it was first developed for. In principle, we propose to analyse and refine each major hydrological process separately, sequentially, and iteratively. To test the method, the HYPE model concept (HYdrological Predictions for the Environment, originally developed for Sweden) was here applied to the data-sparse Niger River basin in West Africa. Errors in the baseline Niger-HYPE model were analysed to identify inadequately described processes. These process descriptions were subsequently isolated and refined through a set of experiments focusing on concept development, input data enhancement, and multivariable calibration. The refinements were guided by in situ discharge observations, earth observations, local expert knowledge, and previous studies. The results show that the original model concept could simulate the annual cycle of discharge, but not the magnitudes or daily dynamics (56-station average Nash-Sutcliffe Efficiency=-1). The main processes requiring improved descriptions were precipitation, evaporation, surface runoff, infiltration, soil storage, reservoir regulations, aquifer recharge, and flooding and river-atmosphere exchange in the Inner Niger Delta. Of these, evaporation, flooding and river-atmosphere exchange differ so much between Sweden and the Niger River that the model concept had to be refined. All refinements were synthesized in a new model version (Niger-HYPE2.0) performing significantly better across the basin (56-station average Nash-Sutcliffe Efficiency=0.4). This study demonstrates the danger of applying a model off the shelf, and the obligation to carefully evaluate and revise process descriptions when applying a model concept to a new region. Moreover, the results indicate that our approach to separately, sequentially, and iteratively refine processes together with local experts can substantially improve process-oriented hydrological models.