Evaluating the Feasibility of Hydrologic Parameter Transferability to Ungauged Domains Under Data-Sparse Conditions in Nigeria

Abstract

Hydrologic modelling has become a vital tool for formulating policies to plan and manage water resources sustainably. However, inadequate or non-available hydro-meteorological datasets significantly hinder its application in ungauged basins. This issue has highlighted critical research gaps and deprived such regions of the advantages of continuous hydrologic modelling for understanding and mitigating hydrologic extremes and developing and managing water resources infrastructure. The major objectives of this study are to: (1) evaluate the ability of remotely-sensed precipitation products to capture rainfall dynamics across different locations in Nigeria, (2) assess the mesoscale hydrologic model (mHM) streamflow simulation within a multi/uni-variable calibration framework, driven by gridded precipitation datasets, (3) evaluate the regionalization of mHM parameters from donor to ungauged basins for streamflow predictions, and (4) model actual evapotranspiration and soil moisture across Nigeria using mHM parameters acquired when constrained using only streamflow data. <br/> The ability of several gridded precipitation products (CHIRPS, PERSIANN-CDR and TAMSAT) to replicate rainfall characteristics at 24 climatic stations distributed across Nigeria was evaluated against in-situ measurements. The results indicate that all products well captured the observed annual cycle and spatial trends across selected locations. Statistical assessment reveals that the CHIRPS dataset was consistent with observations across most climatic stations, accurately reproducing local rainfall characteristics. Next, various gridded precipitation products within a uni- and multi-variable calibration framework were employed to evaluate the performance of the mHM across four different data-scarce basins in Nigeria. This model utilizing CHIRPS and ERA5 rainfall datasets as input, consistently generated acceptable Kling-Gupta efficiency (KGE) values (0.5 < KGE < 0.75) for streamflow simulation during model validation under both calibration frameworks. However, constraining model parameters in both calibration schemes did not significantly improve model simulations in all selected study domains. Furthermore, the transferability of optimized mHM parameter sets from gauged to ungauged domains was assessed under a multi-domain modelling configuration. Optimized mHM streamflow simulations, driven by CHIRPS, ERA5 and MSWEP precipitation datasets, demonstrated significant improvement (KGE > 0.5) across all modelling domains compared to using mHM default parameters. Subsequently, the optimized model parameters were transferred to three independent basins for streamflow prediction. Acceptable streamflow simulation using regionalized mHM parameter sets was shown only in one basin, presenting a KGE of 0.54. Lastly, optimized mHM parameter sets derived from distinct basin-precipitation configurations were utilized to simulate actual evapotranspiration (aET) and soil moisture across three agro-climatic zones in Nigeria. Spatial patterns of mean annual aET for all mHM configurations exhibited similar trends with observations (GLEAM and FLUXNET). CHIRPS-driven aET simulations demonstrated satisfactory correlation scores (r > 0.5) with the GLEAM datasets. Similarly, all mHM setups showcased comparable trends in the annual aET cycle, with acceptable model fits (KGE > 0.7) observed in the Sahel region. The monthly temporal variation of soil moisture anomaly exhibited acceptable agreements (r > 0.8) across all agro-climatic zones. This study represents the first evaluation of mHM under sparse input data constraints in Nigeria. The results of this study not only align with the objectives of the International Association of Hydrological Sciences initiative on prediction in ungauged basins but also address the challenges of hydrologic modelling in Nigeria (and in regions with similar climatic conditions).