https://orcid.org/0000-0002-4276-6093
Abstract
Water resources in the West African Sudan Savannah are concentrated in fractured crystalline and sedimentary rock aquifers. Due to their low storage, these aquifers are vulnerable to changes in climate. This study assesses future temperature and precipitation trends in the Koumfab watershed of northern Togo. The impact of these future climate trends on surface water and the two types of fractured rock aquifers is also investigated. A fully integrated hydrologic model of the Koumfab watershed in Northern Togo is used to study the impact of climate change on water resources. Bias-corrected temperature and precipitation from four regional climate models (RCMs) and two greenhouse emissions scenarios are used as climate forcings for the integrated model. Climate projections from the four RCMs indicate a rise (1.4 −5.5 °C) in average temperature, from the control (1971–2000) to the future period (2031–2060). Projected precipitation follows two contrasting trends with some RCMs predicting an increase (by up to 14%) in mean annual precipitation, while others predict a decline (by up to −14%). These changes in precipitation and temperature will have a noticeable effect on groundwater levels and stream flow. The hydrologic model indicates a decline in groundwater levels for all future climate projections by 1.4−5.7 m. The decline in groundwater levels is not uniformly distributed in the two types of aquifers. The decline in the fractured sandstone aquifer will be four times greater than that of the fractured granodiorite/migmatite aquifer. Projected stream flow at the watershed outlet shows increasing (59 -332%) and decreasing trends (−39 −100%).
HIGHLIGHTS
Future climate projections are likely to reduce the availability of water resources in Northern Togo.
Groundwater storage in the weathered horizon of the West African craton can act as a buffer to mitigate the impact of climate change.
Average temperatures are projected to increase by 1.4–5.5 °C for the period 2031–2060 relative to the control period (1971–2000).
Precipitation trends for the period 2031–2060 show significant fluctuations (±14%).
INTRODUCTION
The West African Sudan Savannah (WASS) is considered one of the most vulnerable regions to climate change (Callo-Concha et al. 2013; IPCC 2014). The WASS spans several West African countries including Togo. From a water resources perspective, the WASS geology and recent climate trends explain its high vulnerability. Precipitation trends in the region are highly variable (Nicholson,1980, 2013; Mann & Gupta 2022). For example, the average annual precipitation between 1901 and 2002 in the Volta Basin varied from 730 to 1,314 mm/year, with a coefficient of interannual variation of 10.4% (Oguntunde et al. 2006). The region has experienced extreme drought events in the 1970s, with rainfall deficits ranging from 180 to 200 mm (le Barbé et al. 2002; L'Hote et al. 2002). Since the 1970s, precipitation has declined due to droughts and shifts in precipitation regimes (Oguntunde et al. 2006; Owusu et al. 2008; Ekwueme & Agunwamba 2021). The average annual temperature has steadily increased, since 1970, by 0.5–0.8 °C (Jenkins et al. 2002; Collins 2011). Studies show that these historical trends have impacted groundwater levels and surface runoff (Leblanc et al. 2008; Favreau et al. 2009; Mahe 2009). Climate projections for the WASS suggest warming trends, with frequent occurrences of extreme flood/drought events (Serdeczny et al. 2017), which will further intensify the hydrologic cycle.
The hydrogeology of the WASS is another contributing factor to the vulnerability of the region. Most of sub-Saharan Africa is underlain by crystalline rocks (40%) and consolidated sedimentary rocks (32%) (MacDonald et al. 2008). Due to the low primary porosity of crystalline and consolidated sedimentary rocks, interconnected networks of fractures constitute the main pathways for fluid flow and groundwater storage. Consequently, these fractured rock aquifers are particularly vulnerable to changes in precipitation (Parashar & Reeves 2017). The two types of fractured rock aquifers (crystalline and consolidated sedimentary), however, have different hydrostratigraphic structures. The crystalline rock aquifers consist of a weathered horizon (also known as saprolite) underlain by a fissured/fractured layer (Chilton & Foster 1995; Dewandel et al. 2006). The saprolite has a high porosity and can be relatively thick. When saturated, saprolite contributes to groundwater storage. The fissured/fractured layer consists predominantly of sub-horizontal fractures where groundwater flow is concentrated. The saprolite layer is not well developed in sedimentary fractured rock aquifers, and thus, groundwater storage and flow are concentrated in open fractures.
The population of sub-Saharan Africa is projected to double by 2050 (United Nations 2019). Urbanization is also projected to increase in sub-Saharan Africa (United Nations 2022). These two trends indicate that water demand for domestic and agricultural usage will increase, while water supply will likely decrease due to climate change. To mitigate the potential impact of future climate on water resources in the WASS, sustainable adaptation strategies are needed. Numerical simulations of groundwater/surface water can serve as a reliable tool to anticipate the effect of different climatic and anthropogenic forcings on water resources. The results of such numerical simulations can be used to inform sustainable water planning strategies.
Future climate trends and their potential effect on water resources in the WASS have been studied. For example, different bias-correction techniques have been used to downscale the Coupled Model Intercomparison Project (CMIP5) data for the Volta Basin in Ghana (e.g., Siabi et al. 2021; Yeboah et al. 2022). Overall, these studies project increasing temperatures and decreasing precipitation in the Volta Basin. The WASS exhibits a high degree of climate variability, and it is unclear whether the magnitude and direction of trends for the Volta Basin will hold elsewhere in the WASS. Only a few studies have used numerical models (e.g., MODFLOW, SWAT) to investigate how changes in future climate will affect groundwater resources (Toure et al. 2016) or surface water resources (Angelina et al. 2015; Abubakari et al. 2019). Although the results of these studies are useful, they suffer from not explicitly simulating the interactions between surface water and groundwater. Simulation of the interactions between surface water and groundwater in numerical models is particularly important because studies have found field evidence of how groundwater level variations affect streamflow in the WASS (e.g., Tirogo et al. 2016; Belemtougri et al. 2021). Very few studies (e.g., Boko et al. 2020) have investigated the impact of future climate trends on surface and groundwater resources in the WASS using a fully integrated modeling approach. Boko et al. (2020), however, use only a small number of future climate scenarios in their simulations. Climate forecasts, based on RCMs, contain uncertainties. These uncertainties arise from the complexity of earth climate and the inability of numerical models to perfectly represent such complex climate processes. Thus, any RCM output represents one out of many possible future climate states. Numerical simulations used to assess the impacts of climate change on water resources are most useful if climate uncertainty considers several RCMs. Moreover, the difference in how the two types of fractured (crystalline rock and consolidated sedimentary rock) aquifers respond to climate change has not been explored in previous studies.
This study aims at assessing future climate trends in the Koumfab watershed of Northern Togo, and how these future trends will affect surface and groundwater resources. The study also investigates the difference in how fractured crystalline and consolidated sedimentary rock aquifers respond to changes in climate. To achieve the study objectives, climate forecasts from the CMIP5 are downscaled to the Koumfab watershed scale. An integrated hydrologic model of the Koumfab watershed is developed and calibrated. The calibrated hydrologic model is then used to analyze trends in groundwater levels and streamflow in response to climate forecasts. Section 2 describes the hydrogeology and climate of the Koumfab watershed. The hydrologic modeling approach and the analysis of the RCMs’ data are presented in Section 3. In Section 4, future climate trends and numerical modeling results are presented and discussed. Concluding remarks in Section 5 provide a synthesis of the study findings.
STUDY SITE
(a) Location of the Koumfab watershed shown with the red square. (b) Map of the Koumfab watershed showing the geologic units as well as hydrologic features. Well C1 is represented with the green square. Please refer to the online version of this paper to see this figure in colour: https://dx.doi.org/10.2166/wcc.2022.264.
(a) Location of the Koumfab watershed shown with the red square. (b) Map of the Koumfab watershed showing the geologic units as well as hydrologic features. Well C1 is represented with the green square. Please refer to the online version of this paper to see this figure in colour: https://dx.doi.org/10.2166/wcc.2022.264.
The Koumfab watershed is comprised of two geologic domains: the West African craton (60%) to the north and the Volta basin (40%) to the south. The West African craton consists of metavolcanic rocks (amphibolite) and granitic (gneiss, granodiorite, and migmatite) rocks that were last deformed ∼2,000 Ma ago during the Eburnean orogeny (Kalsbeek et al. 2008). The Volta basin lies unconformably on the southeastern margin of the West African craton and gently dips (1–10°) to the southeast. The Volta basin strata are divided into three supergroups, in order of youngest to oldest: Tamale, Oti, and Bombouaka supergroups. Of these, only the Bombouaka supergroup outcrops in the Koumfab watershed. The Bombouaka supergroup is comprised of coarse-grained quartzitic and feldspathic sandstones, interbedded with clayey-silty layers. It is overlain by the Oti supergroup, which consists of shales and greywackes. The Tamale supergoup is largely composed of conglomerates.
Piper diagram showing the distribution of major cations and anions. The data suggest a bicarbonate water type.
Piper diagram showing the distribution of major cations and anions. The data suggest a bicarbonate water type.
Oxygen and hydrogen isotope compositions of groundwater samples. The blue and red lines represent the local and global water meteoric line (GWML), respectively. The regression line of the study site (black dashed line) has a similar slope to that of the local water meteoric line, suggesting rapid recharge. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wcc.2022.264.
Oxygen and hydrogen isotope compositions of groundwater samples. The blue and red lines represent the local and global water meteoric line (GWML), respectively. The regression line of the study site (black dashed line) has a similar slope to that of the local water meteoric line, suggesting rapid recharge. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wcc.2022.264.
Land cover types in the Koumfab watershed.
METHODS
The effect of climate change on groundwater and surface water resources is investigated using an integrated numerical modeling approach. Geologic logs and pumping test data are used to assign geologic units and parameterize the Koumfab watershed integrated model. Historical groundwater level and streamflow data are then used to calibrate the model. Future climate data from the CMIP5 are used to force the calibrated model. Two greenhouse emission scenarios (RCP4.5 and RCP8.5) are considered; for each emission scenario, an ensemble of four regional climate models is selected. This leads to a total of eight future climate scenarios. Simulations are subdivided into a control period (1971–2000) and a future period (2031–2060). Changes in groundwater level and streamflow relative to the control period are used to elucidate the effect of climate change.
HydroGeoSphere numerical code
The HydroGeoSphere (HGS) software is used to simulate all dominant hydrologic processes (Therrien et al. 2012). HGS is a 3D fully integrated, physically based, and controlled volume finite element code capable of simulating surface–subsurface flow and reactive/non-reactive transport processes (Sudicky et al. 2008; Brunner & Simmons 2012). In HGS, surface flow is described by the 2D diffusion-wave approximation to the Saint Venant equation. The 3D form of the Richards equation is used to simulate variably saturated subsurface flow. The coupling of flow between the surface and subsurface domains is achieved by a flux exchange term. The flux exchange is a function of the head difference between the two domains, and thus, surface and subsurface flow equations can be solved simultaneously at each timestep. HGS uses an adaptative procedure to adjust timestep values based on changes in the hydraulic head.
Model setup
The surface domain of the Koumfab numerical model is discretized using triangular finite elements. In total, the surface mesh contains 4,720 triangular elements and 2,481 nodes. A 30 × 30 m digital elevation model is used to define the surface mesh topography and extract the Koumfab river network. The surface domain is subdivided into three land-use types (cropland, urban, and open water) with different evapotranspiration parameters. Two soil zones overlying the sandstone and granodiorite/migmatite units are represented with different unsaturated hydraulic parameters. Similarly, the subsurface parameters are assigned based on the sandstone and granodiorite/migmatite units. The spatial extent of the land-use types and soil zones are derived from the Copernicus dataset and the geologic map, respectively (Figures 1 and 4). The Copernicus dataset provides high-resolution global maps of land cover and land use (Buchhorn et al. 2020). In HGS, several parameters are required to parameterize surface flow, subsurface flow, and evapotranspiration. In this study, these parameters are prescribed using values from previous studies with similar soil composition and vegetation (Goderniaux et al. 2009; Boubacar et al. 2020; Tagnon et al. 2020). Details of these parameters are presented in Table 1.
HGS model parameters
| Parameters | Symbol and unit | Value | Reference |
|---|---|---|---|
| Subsurface | |||
| Total porosity | n (–) | 0.001–0.39 | Fontodji et al. (2011), Tossou et al. (2021) |
| Specific storage | Ss (m−1) | <0.01 | Clark (1985), Vouillamoz et al. (2005) |
| Saturated hydraulic conductivity | K (m/s) | 4.6 × 10−7–10−3 | Acheampong & Hess (1998), Bruand et al. (2005), Akara et al. (2020) |
| Residual water saturation | Swr (–) | 0.034–0.269 | Hodnett & Tomasella (2002) |
| van Genuchten parameter | α (–) | 0.14–0.67 | |
| van Genuchten parameter | β (m−1) | 1.38–2.44 | |
| Surface | |||
| Manning roughness coefficients | nx, ny (m−1/3s) | 0.02–2 | Unami et al. (2009); Quenum et al. (2022) |
| Coupling length | Lc (m) | 0.05–0.1 | Tagnon et al. (2020) |
| Obstruction storage | Os (m) | 10−3–4 × 10−2 | Tagnon et al. (2020) |
| Rill storage | Rs (m) | 0.07–0.15 | Tagnon et al. (2020) |
| Evapotranspiration | |||
| Evaporation limiting saturations | θe1, θe2 (–) | 0.32, 0.2 | Panday & Huyakorn (2004) |
| Evaporation depth | Le (m) | 3 | Tagnon et al. (2020) |
| Leaf area index | LAI (–) | 1–5 | Schroeder et al. (1994) |
| Transpiration fitting parameters | C1, C2, C3 (–) | 0.28–0.35, 0.07–0.3, 10–259 | Li et al. (2008), Vazquez (2003) |
| Transpiration limiting saturations | θt1, θt2, θt3, θt4 (–) | 0.2–0.9 | Tagnon et al. (2020) |
| Root depth | Lr (m) | <4 | Li et al. (2008) |
| Parameters | Symbol and unit | Value | Reference |
|---|---|---|---|
| Subsurface | |||
| Total porosity | n (–) | 0.001–0.39 | Fontodji et al. (2011), Tossou et al. (2021) |
| Specific storage | Ss (m−1) | <0.01 | Clark (1985), Vouillamoz et al. (2005) |
| Saturated hydraulic conductivity | K (m/s) | 4.6 × 10−7–10−3 | Acheampong & Hess (1998), Bruand et al. (2005), Akara et al. (2020) |
| Residual water saturation | Swr (–) | 0.034–0.269 | Hodnett & Tomasella (2002) |
| van Genuchten parameter | α (–) | 0.14–0.67 | |
| van Genuchten parameter | β (m−1) | 1.38–2.44 | |
| Surface | |||
| Manning roughness coefficients | nx, ny (m−1/3s) | 0.02–2 | Unami et al. (2009); Quenum et al. (2022) |
| Coupling length | Lc (m) | 0.05–0.1 | Tagnon et al. (2020) |
| Obstruction storage | Os (m) | 10−3–4 × 10−2 | Tagnon et al. (2020) |
| Rill storage | Rs (m) | 0.07–0.15 | Tagnon et al. (2020) |
| Evapotranspiration | |||
| Evaporation limiting saturations | θe1, θe2 (–) | 0.32, 0.2 | Panday & Huyakorn (2004) |
| Evaporation depth | Le (m) | 3 | Tagnon et al. (2020) |
| Leaf area index | LAI (–) | 1–5 | Schroeder et al. (1994) |
| Transpiration fitting parameters | C1, C2, C3 (–) | 0.28–0.35, 0.07–0.3, 10–259 | Li et al. (2008), Vazquez (2003) |
| Transpiration limiting saturations | θt1, θt2, θt3, θt4 (–) | 0.2–0.9 | Tagnon et al. (2020) |
| Root depth | Lr (m) | <4 | Li et al. (2008) |
A 3D mesh consisting of four layers is used to simulate subsurface flow. The four layers reflect the lithological units observed in well logs. The four layers and corresponding thicknesses are as follows: the soil horizon (2 m), the weathered horizon (6 m), the saprock horizon (25 m), and the fractured bedrock (67 m). A finer level (20–60 cm) of vertical discretization is used in the first two layers to better represent interactions between surface and subsurface domains. For the bottom two layers, a coarse vertical resolution (3–7 m) is used to optimize model runtime. Saturated hydraulic conductivity values for the different layers are derived from 27 pumping tests (Assouma 1988; Akara et al. 2020).
Boundary conditions
Two flux boundary conditions are assigned to the surface domain: precipitation and potential evapotranspiration. These two fluxes are based on monthly temperature and precipitation measured at a single weather station. Potential evapotranspiration is estimated using the Thornthwaite method (Thornthwaite 1948). Average temperature and latitude are required by the Thornthwaite method. No flow boundary conditions are assigned to the bottom and outer boundaries of the numerical model. Water is allowed to exit the surface domain only at the Koumfab river outlet. Nodes at the outlet are assigned a critical depth boundary condition, which allows for the flow rate to vary naturally as a function of the water depth (Sudicky et al. 2008). A sink flux is also used to represent water withdrawal from the Koumfab reservoir.
Model calibration and sensitivity analysis
The Koumfab numerical model is calibrated in three phases. The first step of calibration (also known as spin-up) involves running the model until minimal variation in storage is achieved. The mean annual precipitation and potential evapotranspiration from 1980 to 2010 are used during the spin-up phase. The solution at the end of the spin-up phase is used as an initial condition for the second phase of calibration. In the second phase, the average monthly precipitation and potential evapotranspiration from 1980 to 2010 are used to force the model. The model is run iteratively until a quasi-steady state is achieved. The quasi-steady state is reached when the interannual variations in hydraulic head and streamflow become negligible. The best fit between simulated and observed hydraulic head and streamflow values is achieved by manually adjusting surface (nx and ny) and subsurface (K, Ss, α, and β) parameters. Observed static levels from 19 wells and average streamflow at the watershed outlet are used to assess the model performance (Figure 1). Statistical parameters such as the Normalized Root Mean Square Error (NRMSE) and the coefficient of determination (r2) are used for performance assessment. Next, the quasi-steady-state solution is used to perform the transient-state calibration. Model parameters from the quasi-steady state are further adjusted to better match transient hydraulic head (1984–1986) and streamflow rate (1960–1961). The mean absolute error (MAE) and the Nash–Sutcliffe efficiency (NSE) are used as goodness-of-fit indicators during the transient-state calibration (Hamzah et al. 2021). Evapotranspiration parameters are kept constant for steady and transient calibrations. Tables 2 and 3 show the calibrated hydraulic parameters.
Calibrated surface flow and evapotranspiration parameters
| Parameters | Cropland | Urban | Open Water |
|---|---|---|---|
| Surface parameters | |||
| Manning friction (m−1/3s) | 0.18 | 0.18 | 0.18 |
| Rill storage (m) | 0.11 | 0.11 | – |
| Obstruction storage (m) | 0.03 | 0.03 | – |
| Coupling length (m) | 0.01 | 0.01 | 0.01 |
| Evapotranspiration | |||
| LAI | 1.25 | 0.25 | 1.25 |
| Root depth (m) | 2 | 0.1 | 1 |
| Parameters | Cropland | Urban | Open Water |
|---|---|---|---|
| Surface parameters | |||
| Manning friction (m−1/3s) | 0.18 | 0.18 | 0.18 |
| Rill storage (m) | 0.11 | 0.11 | – |
| Obstruction storage (m) | 0.03 | 0.03 | – |
| Coupling length (m) | 0.01 | 0.01 | 0.01 |
| Evapotranspiration | |||
| LAI | 1.25 | 0.25 | 1.25 |
| Root depth (m) | 2 | 0.1 | 1 |
Calibrated subsurface parameters
| Domains | Layers | Porosity | Hydraulic conductivity (m/s) | Specific storage (m−1) | van Genutchen parameters | Residual water saturation | |
|---|---|---|---|---|---|---|---|
| α (m−1) | Β | ||||||
| Sandstone | Layer 1 [0–2 m] | 0.3 | 1.7 × 10−3 | 2.0 × 10−4 | 0.60 | 1.42 | 0.1 |
| Layer 2 [2–8 m] | 0.1 | 4.3 × 10−5 | 1.8 × 10−3 | 0.166 | 1.13 | 0.06 | |
| Layer 3 [8–33 m] | 0.1 | 6.1 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.06 | |
| Layer 4 [33–100 m] | 0.1 | 6.5 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.06 | |
| Granodiorite/migmatite | Layer 1 [0–2 m] | 0.25 | 9.7 × 10−4 | 1.8 × 10−4 | 1.00 | 1.40 | 0.10 |
| Layer 2 [2–8 m] | 0.15 | 1.7 × 10−5 | 1.8 × 10−4 | 0.166 | 1.28 | 0.06 | |
| Layer 3 [8–33 m] | 0.01 | 5.9 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.08 | |
| Layer 4 [33–100 m] | 0.01 | 5.2 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.08 | |
| Domains | Layers | Porosity | Hydraulic conductivity (m/s) | Specific storage (m−1) | van Genutchen parameters | Residual water saturation | |
|---|---|---|---|---|---|---|---|
| α (m−1) | Β | ||||||
| Sandstone | Layer 1 [0–2 m] | 0.3 | 1.7 × 10−3 | 2.0 × 10−4 | 0.60 | 1.42 | 0.1 |
| Layer 2 [2–8 m] | 0.1 | 4.3 × 10−5 | 1.8 × 10−3 | 0.166 | 1.13 | 0.06 | |
| Layer 3 [8–33 m] | 0.1 | 6.1 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.06 | |
| Layer 4 [33–100 m] | 0.1 | 6.5 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.06 | |
| Granodiorite/migmatite | Layer 1 [0–2 m] | 0.25 | 9.7 × 10−4 | 1.8 × 10−4 | 1.00 | 1.40 | 0.10 |
| Layer 2 [2–8 m] | 0.15 | 1.7 × 10−5 | 1.8 × 10−4 | 0.166 | 1.28 | 0.06 | |
| Layer 3 [8–33 m] | 0.01 | 5.9 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.08 | |
| Layer 4 [33–100 m] | 0.01 | 5.2 × 10−6 | 1.8 × 10−4 | 0.134 | 1.34 | 0.08 | |
Future climate data
Climate data from the Coordinated Regional Downscaling Experiment (CORDEX-Africa) are used to force the calibrated model (http://www.csag.uct.ac.za/cordex-africa/). These climate data are based on the CIMP5, because the CIMP6-derived RCMs are currently unavailable. The effect of climate change is investigated by comparing the hydraulic head and streamflow rate of the control period (1971–2000) with that of the future period (2030–2060). Four RCMs and two representative concentration pathways (RCP4.5 and RCP8.5) are considered in this study. The RCP4.5 is an intermediate emission scenario characterized by a stabilization of the radiative forcing to 4.5 W/m2 by 2100. The RCP8.5 is often considered the worst-case scenario because radiative forcing is expected to reach 8.5 W/m2 by 2100. The choice of the four RCMs is based on the availability of climate data from 1971 to 2060.
The coarse grid resolution of RCMs requires climate data to be bias-corrected prior to climate impact assessment. Several methods (e.g., delta change, quantile mapping, local intensity scaling, and multiple linear regression) can be used to perform bias correction. Studies have compared these methods (e.g., Gudmundsson et al. 2012; Themeßl et al. 2011) and found that quantile mapping shows the best performance. Hence, the quantile-based bias correction method is used to adjust monthly temperature and precipitation data. In the quantile-based bias correction approach, the RCMs’ outputs are statistically transformed such that their new distribution is identical to the distribution of observed precipitation and temperature. In practice, the empirical cumulative distribution function (CDF) of the modeled and observed data for a control period is first determined. The CDF of the modeled data is then reshaped and shifted to match the CDF of the observed. The same shape and shift parameters used during the control period phase are used to correct future climate data. The ‘qmap’ package in R is used for bias correction (Gudmundsson et al. 2012). Monthly precipitation and temperature data from the period 1971 to 2000 is used to derive the historical cumulative distribution function, which in turn is used to bias correct the four RCMs.
is the average annual precipitation over the control period (1971–2000), and Sref is the standard deviation of the annual precipitation over the control period. A dry (respectively, wet) year is defined as a year with a rainfall anomaly index less than −1 (respectively, higher than 1).RESULTS AND DISCUSSION
Calibration and sensitivity analysis
Simulated vs. observed hydraulic head for the 19 wells.
Relative sensitivity of (a) hydraulic head and (b) streamflow. Sensitivities are computed for n (porosity), K (hydraulic conductivity), Ss (specific storage), α (van Genuchten parameter), β (van Genuchten parameter), Swr (residual water saturation), nx,y (Manning's coefficient), Rs (rill storage), Os (obstruction storage), and LAI (leaf area index).
Relative sensitivity of (a) hydraulic head and (b) streamflow. Sensitivities are computed for n (porosity), K (hydraulic conductivity), Ss (specific storage), α (van Genuchten parameter), β (van Genuchten parameter), Swr (residual water saturation), nx,y (Manning's coefficient), Rs (rill storage), Os (obstruction storage), and LAI (leaf area index).
Transient-state calibration of hydraulic head at well C1.
Transient-state calibration of streamflow rate at the watershed outlet.
Projected changes in climate
Changes in the average annual temperature from the control period (1971–2000) to the future period (2031–2060)
| Change (RCP4.5) | Change (RCP8.5) | |
|---|---|---|
| CCLM4–8–17 | +1.8 | +2.3 |
| RACMO22T | +3.1 | +3.6 |
| RCA4 | +3.6 | +5.5 |
| REMO2009 | −0.3 | +1.4 |
| Change (RCP4.5) | Change (RCP8.5) | |
|---|---|---|
| CCLM4–8–17 | +1.8 | +2.3 |
| RACMO22T | +3.1 | +3.6 |
| RCA4 | +3.6 | +5.5 |
| REMO2009 | −0.3 | +1.4 |
Changes in monthly temperature from the control period (1971–2000) to the future period (2031–2060) for (a) RCP4.5 and (b) RCP8.5 greenhouse emission scenarios. The four RCMs are indicated with different colors.
Changes in monthly temperature from the control period (1971–2000) to the future period (2031–2060) for (a) RCP4.5 and (b) RCP8.5 greenhouse emission scenarios. The four RCMs are indicated with different colors.
The future temperature trends are similar to previous studies; however, differences in the magnitude are noted. Badou et al. (2018) investigated future temperature trends in the Niger River Basin and reported a temperature change of −0.37 to +0.48 °C. Yira et al. (2017) reported only an increase in temperature (0.1–2.6 °C) in south-western Burkina-Faso. Note that all studies do not predict a decrease in temperature, and the maximum temperature increase is significantly different from one study to the other. These differences are explained, in part, by the RCMs used. Most importantly, these differences suggest that the spatial variability in the current WASS climate will persist, if not be exacerbated, under future climate. A comparison of our results with Yira et al. (2017) suggests that the effect of climate change will be more pronounced moving southward in the WASS.
Changes in mean annual precipitation from the control period (1971–2000) to the future period (2031–2060)
| % Change (RCP4.5) | % Change (RCP8.5) | |
|---|---|---|
| CCLM4–8–17 | −13.4 | −14.3 |
| RACMO22T | +4.3 | +11.5 |
| RCA4 | +5.4 | +13.9 |
| REMO2009 | −6.9 | −0.7 |
| % Change (RCP4.5) | % Change (RCP8.5) | |
|---|---|---|
| CCLM4–8–17 | −13.4 | −14.3 |
| RACMO22T | +4.3 | +11.5 |
| RCA4 | +5.4 | +13.9 |
| REMO2009 | −6.9 | −0.7 |
Average monthly precipitation and relative change under the RCP4.5 (a and c) and RCP8.5 (b and d) scenarios. The reference period corresponds to 1971–2000.
Average monthly precipitation and relative change under the RCP4.5 (a and c) and RCP8.5 (b and d) scenarios. The reference period corresponds to 1971–2000.
Rainfall anomaly for the eight climate scenarios. The black dashed lines correspond to rainfall anomaly indexes of 1 and −1. Dry (respectively, wet) years correspond to rainfall anomaly indexes less than −1 (respectively, greater than 1).
Rainfall anomaly for the eight climate scenarios. The black dashed lines correspond to rainfall anomaly indexes of 1 and −1. Dry (respectively, wet) years correspond to rainfall anomaly indexes less than −1 (respectively, greater than 1).
Effect of climate scenarios on the hydrologic cycle
Hydraulic head variations from the historical to the future period for (a) the Granodiorite/migmatite unit and (b) the Sandstone unit.
Hydraulic head variations from the historical to the future period for (a) the Granodiorite/migmatite unit and (b) the Sandstone unit.
Overall, the decrease in precipitation is the main contributor to the groundwater level decline (Table 5). Surprisingly, the increase in precipitation projected for RACMO22T–RCP4.5 and RACMO22T–RCP8.5 scenarios is not sufficient to offset the increase in evapotranspiration induced by higher temperatures. The maximum groundwater decline is 1.4 m for the granodiorite/migmatite unit and 5.7 m for the sandstone unit. Toure et al. (2016) investigated the impact of climate change on groundwater levels in a fractured sandstone aquifer and reported a groundwater decline of 15 m. These results indicate that groundwater resources in fractured sedimentary and crystalline rocks will be affected differently by decreasing precipitation. The small groundwater decline observed in the granodiorite/migmatite unit can be explained by the presence of the saprolite layer. Because of its high storativity, the saprolite layer acts as a buffer to lessen the impact of climate change (Dewandel et al. 2006). The saprolite layer is not well developed in the sandstone unit, which explains the greater decline in groundwater levels.
Streamflow variations from the historical to the future period.
The results presented above have implications for access to potable water and the safety of inhabitants in the Koumfab watershed. Access to potable water in the Koumfab watershed depends primarily on boreholes equipped with manual pumps and, in some cases, solar pumps. Typically, the static groundwater level and maximum drawdown during pumping tests are used to determine the depth at which pumps are installed. Hence, a groundwater level decline of 1.4–5.7 m will likely increase borehole costs and hinder access to potable water because the pumps will need to be installed at greater depths. From a water budget perspective, a groundwater level decline of 1.4–5.7 m corresponds to approximately 20 million m3 of water loss. Streamflow is expected to increase by up to 332%. Large increases in streamflow pose a serious flooding risk to the local population. Some models, however, predict a total drying up of the Koumfab River, which will adversely impact water supply and fishery activities in the region.
Study limitations
Due to the complex geology of the Koumfab watershed and the scarcity of hydrologic data, this study has a few limitations. Precipitation is assumed uniformly distributed over the watershed because data are available for only one weather station. While this assumption is justified given the small size and low topographic relief (Amponsah et al. 2022) of our study site, it is worth noting that spatial variability can strongly influence hydrologic responses. Data regarding the temporal and spatial fluctuations of groundwater level and streamflow are very limited. Consequently, the scope of the transient-state calibration is limited. Although extensive calibration data are often desired, studies have shown that the length of the calibration period has little effect on the predictive capability of hydrologic models (Duethmann et al. 2020). Nonetheless, future hydrologic investigations will benefit from the installation of permanent groundwater/surface water monitoring stations. Given that this study focuses on the impact of climate change on water resources, vegetation and land use are assumed stationary throughout the simulations. Li et al. (2007) investigated the impact of land use on the hydrologic cycle in West Africa and found that the removal of vegetation leads to a significant increase in streamflow. As shown in Figure 4, vegetation is very limited in the Koumfab watershed. Hence, vegetation removal will not significantly change the results of this study. Regardless of the model limitations, this study provides a valuable insight into how groundwater and surface water will respond to different climate scenarios.
CONCLUSION
Future climate (2031–2060) trends in northern Togo are investigated using an ensemble of four regional climate models. Results of this analysis indicate that average temperature will increase by 1.4–5.5 °C, and average precipitation will change by −14 to +14%. Seasonal variations in temperature are expected to be similar to that of the current climate, with the highest temperature occurring in March–April. For precipitation, however, a slight shift in the duration of the rainy season is noted. The shift is characterized by unusually high rainfall during the month of November. The analysis also indicates the frequent occurrence of extremely wet and dry years during the 2031–2060 period. The impact of these changes on the hydrologic cycle was investigated using a fully integrated hydrologic model and indicated that groundwater levels will decline by 1.4–5.7 m in the Koumfab watershed. The fractured sandstone and granodiorite/migmatite aquifers are found to respond differently to changes in climate. Groundwater level decline in the sandstone aquifer is four times greater than that of the granodiorite/migmatite aquifer. Projected streamflow rates show two contrasting trends. Some climate scenarios indicate an increase (59–332%) in streamflow, while other scenarios predict a decline (−39 to 100%). These results suggest that fractured rock aquifers in the WASS are at risk of depletion due to projected changes in climate. The effect of climate change on groundwater levels is less pronounced in the granodiorite/migmatite unit, which suggests that the thick weathered horizon often observed in the West African craton has enough storage to lessen the impact of climate change. This is an important finding because of the widespread occurrence of the weathered horizon in the West Africa Craton. These results provide a valuable insight regarding future water supply in the watershed and thus improve water resource planning.
ACKNOWLEDGEMENTS
This work was supported by Western Michigan University's Climate Change Research Graduate Scholarship.
DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.
%20cropped.png?versionId=7143)