*2.1.1 The PCRaster global water balance model*

The PCR-GLOBWB is a large scale hydrological model, which incorporates human activities into hydrology [31]. Global simulation of hydrology is daily and at a spatial resolution of 5 arcmin (0.08o at the equator). In this study, our focus is on river basins in West Africa, specifically for gauging stations located within Nigeria. The PCR-GLOBWB simulates water stored in two top soil layers S1 and S2; one bottom groundwater reservoir S3 for each grid cell and time step. The model also simulates; water movement between the atmosphere and layer of topsoil (precipitation, evaporation, transpiration, and snowmelt); among the soil; in between the soil and the active layer of groundwater and estimates interception by canopy and snow storage. Distinct land cover types (forest, grassland, irrigated paddy field, irrigated non-paddy field, and open water), soil types, and elevation are considered to determine sub-grid variability. The Improved ARNO scheme is used in the model to estimate the fraction of the area of saturated soil [32]. Precipitation can be intercepted, evaporated, or infiltrated into the soil layers. The excess surface runoff (Qdr), second soil layer (interflow) runoff (Qsf), and groundwater (baseflow) runoff (Qbf) make up the runoff from each cell. To obtain the river discharge (Qchannel), Specific runoff from each cell is gathered and then routed through the drainage network following the travel-time solution of [33]. At each time step, the model simulates (i) livestock, household irrigation and industrial water demand, (ii) water withdrawn from surface water, groundwater and desalinisation. For this work, we adopted the guideline of standard parameterisation of [31] of PCR-GLOBWB, which is based on available global datasets.

#### **2.2 Data**

#### *2.2.1 Forcing datasets*

Datasets, as described in [34], is followed and repeated in this study. The meteorological datasets required to drive the model are precipitation, temperature, and reference potential evapotranspiration. We obtained these data from the CRU TS 3.2 [34]. These data were processed by interpolating station observation past time-series to a global grid resolution of 0.50 . Due to the daily resolution of PCR-GLOBWB, the monthly CRU TS 3.2 data were downscaled to daily resolution with ERA 40 (1958–1978, [35]) and ERA-Interim (1979–2015, [36]). ERA 40 and ERA-I had been spatially downscaled from their initial spatial resolutions of 1.2o and 0.7° to 0.5° in the resampling scheme of the European Centre for Medium-Range Weather Forecasts (ECMWF). This downscaling was done by first allotting the larger values ERA40 and ERA-I to the middle of the cells and then interpolating spatially to the higher resolution of 0.5°. Firstly, downscaling of precipitation was done by temporarily assigning

a threshold of 0.1 mm day−1 to the daily time series of ERA, thereby estimating the number of days with rain and eliminating the drizzle effect. The rainfall quantity below this threshold was proportionally allocated to the rainy days. Thereafter, CRU monthly precipitation was reproduced by multiplicative scaling of the daily rainfall totals. Also, monthly reference potential evaporation, estimated from the CRU dataset with Penman-Monteith, was scaled using multiplicative scaling and downscaled to daily data using a daily temperature-based ET product derived from daily ERA temperatures. An additive scaling method was used for air temperature (see [31] for more details). For this work, standard parameterisation guideline, as provided in [31], was adopted. We used available global datasets, including vegetation, geological information, and soil properties, to parameterise the model and simulate discharge at daily time steps over the selected river basins (from 1958 to 2015). Monthly averages were used to report the output from the Model.
