**2.4 Method of analyses**

For the analyses, we used discharge climatologies, i.e. long-term average monthly river discharge of each river. Projected daily discharge from five climate models was averaged to monthly climatologies and analysed for two periods, the reference (1971–2000) and the far-future period (2070–2099). Absolute changes between Reference and far future conditions was computed for each climate model historical (1971–2000).

*Future Climate Change Impacts on River Discharge Seasonality for Selected West African River… DOI: http://dx.doi.org/10.5772/intechopen.99426*


#### **Table 2.**

*Overview of monthly performance ratings [45].*

#### **2.5 Seasonality index (SI)**

The seasonality of river discharge volumes shows the extent of variation in monthly discharge magnitude throughout the year. In this study, we used SI, developed by [46]. SI is the summation of the absolute changes of the monthly river discharge volumes from the mean monthly discharge, divided by the total annual river discharge of a given year. It is given as:

$$SI = \frac{\sum\_{n=1}^{12} X\_n - \frac{R}{12}}{R} \tag{1}$$

where Xn is the mean discharge of the nth month, and R is the mean annual discharge. SI varies from zero (all months having equal discharge distribution) to 1.83 (the discharge occurs in one month). A seasonal pattern in the discharge regime is established when SI is 0.6 and above. A region with high SI would be susceptible to drought because a high SI value in an area translates to the high variability of water resources and the shortage with respect to time. The classification of the degree of SI values is presented in **Table 3**.

#### **2.6 Results**

#### *2.6.1 PCR-GLOBWB validation*

PCR-GLOBWB models' validation results in terms of five metrics are presented in **Table 4** and **Figure 2** for the gauges. Overall, the performance of the PCR-GLOBWB on the monthly time step for validation periods was satisfactory. As seen in **Figure 2**, the hydrographs monthly flow pattern was well reproduced, close agreement in discharge distribution is seen in accord with performance statistics. The PCR-GLOBWB model performance was adequate in all basins, displaying very good to satisfactory values, established on ratings detailed in [44, 45]. **Table 5** sums up the five performance metrics values obtained for validation periods in each study basin. Concerning KGE values, the best model fit was found for the Hadejia (0.88) and Yobe (0.87) basins. The Niger, Hadejia, and Yobe river have very good NSE (0.8, 0.79, and 0.79), RSR (0.4, 0.45, and 0.4), and r2 (0.88, 0.79, and 0.82) values, while Jamaare has satisfactory values (NSE = 0.6, RSR = 0.62, r2 = 0.62). The highest PBIAS values reaching −18 to −25% were obtained in the Niger and Jamaare river. The PCR-GLOBWB model mostly reproduced well the flow dynamics of the


#### **Table 3.**

*Classification of seasonality index (SI).*


#### **Table 4.**

*Model's validation performance for the river basins.*

#### **Figure 2.**

*Hydrographs for the different validation period of the four rivers (a) Niger (b) Yobe (c) Jamaare (d) Hadejia. Red line is PCR-GLOBWB and Black line is Observed.*

*Future Climate Change Impacts on River Discharge Seasonality for Selected West African River… DOI: http://dx.doi.org/10.5772/intechopen.99426*


**Table 5.**

*Seasonality index (SI) of climate models median discharge climatologies for reference and projected far-future periods under RC4.6 and RCP8.5. Minimum and maximum values of the climate models' combinations are shown in brackets.*

observed but displayed some disparity. Peak values overestimation was also seen in most cases for the river basins. The overestimation of flow can be attributed to the CRU TS 32.1, forcing input data. CRU Precipitation across Africa is of low quality due to sparse CRU stations and limited data assimilated during the reanalysis of ERA-40 [47]. As a result, of the temporal and spatial disparity in station density, the datasets are subjected to uncertainties; this uncertainty explains the overestimation of the stations' hydrographs. However, CRU TS 3.2 is a preferred data as it based on observation. Using another meteorological dataset may reduce the overestimation, but at the cost of the temporal variability, because no other datasets cover a long period of 1958–2015.
