A proper estimation of future water availability is vital information for water planners. However, most contemporary quantitative hydrological predictions are mostly streamflow-centred which makes them fundamentally uncertain. Since, total streamflow represents only 35% of the total incoming rainfall at the global scale and even less than that in regions like West Africa, an alternative of using soil moisture and evapotranspiration is a possibility. This study explored these alternative avenues for more accurate hydrological prediction. From a theoretical perspective, water resources were treated as blue water (BW, sum of streamflow and deep aquifer recharge) and green water (GW, sum of actual evapotranspiration and soil moisture). The theoretical aspect also addressed the uncertainties associated with the use of a single hydrological model and adopted a multi-model evaluation instead. This theoretical framework was applied to the Benin Portion of the Niger River Basin, a conglomerate of four understudied and poorly gauged basins, Coubéri, Gbassè, Yankin, and Kompongou. This area provides a number of ecosystem services whose sustenance requires better hydrological knowledge. To this purpose, four objectives were addressed in order to assess the impact of climate change on future BW and GW availability in the study area. The first objective was to identify within a set of hydrological models the most suitable ones to better simulate streamflow and soil moisture. The performance of four hydrological models (HBV-light, UHP-HRU, SWAT and WaSiM) to simulate daily streamflow and the ability of three of these models (UHP-HRU, SWAT and WaSiM) to reproduce daily remotely-sensed soil moisture dynamic were compared. The results showed that none of the hydrological models clearly outperformed the others in the simulation of streamflow in all the basins. While WaSiM was the most suitable to mulate streamflow in the Yankin and Kompongou basins, HBV-light was the best for the Coubéri and Gbassè basins. However, regardless of the basin, UHP-HRU was the most adequate model to simulate soil moisture. The second objective dealt with the downscaling of three regional climate models outputs (HIRHAM5, RCSM, and RCA4) under RCP4.5 and RCP8.5 for the historical period (1976-2005) and the future (2021-2050). To this end, the Statistical DownScaling Model (SDSM) was used. The results suggested that rainfall will increase (1.7 to 23.4%) for HIRHAM5 and RCSM under both RCPs but will show mixed-trends (-8.5 to 17.3%) for RCA4. Mean temperature will also increase (-0.1 to 0.48°C) for HIRHAM5 and RCSM but decrease for RCA4 (-0.37 to 0.1°C). On the basis of the results of the two previous objectives, the third objective was to quantify the future BW only with the models found robust to simulate streamflow and future GW solely with the models found suitable to simulate soil moisture. It was found that GW will increase in all the four investigated basins while BW will only increase in the Kompongou basin.
The last objective was about the quantification of the overall uncertainty associated with the evaluation of future BW and GW. This was done by computing the interviuartile range of the total number of model realizations for ach basin. The results show that BW evaluation is associated with larger uncertainty than GW quantification. To cope with the projected decrease in BW that could adversely impact livelihoods and food security of the local population, some recommendations for the development of adequate adaptation strategies including the rational use of BW are briefly discussed.