Introduction

Climate change, its impacts on water resources, as well as adaptation measures constitute a major concern across the world (Huntington 2006; Mendez et al. 2022). Under climate change and anthropogenic stress, the future of surface water resources and groundwater is becoming a major concern across the world (Thompson et al. 2017). In this context of climate change, the management of water resources which have become increasingly vulnerable remains problematic (Donevska and Panov 2019). Hydrological models are frequently utilized to assess the potential effects of climate change on hydrological systems. These models are driven by data from general circulation models (GCM) and regional climate models (RCM) under different scenarios. This approach is adopted to evaluate the consequences of climate change on hydrological systems in a logical manner (Saha and Quinn 2020). Global circulation model (GCM) is a model that simulates general circulation of planetary atmosphere or oceans. The term general circulation is used to indicate large-scale atmospheric or oceanic motions with its persistent as well as transient features on various scales. GCM employs a combination of mathematical expressions that represent governing physics of circulations processes and empirical calculations which replicate processes based on data. RCMs are commonly used for dynamic downscaling and are much more granular than GCMs, which have mesh sizes of 10–50 km.

To construct future projections of climate change, the most used tools are general circulation models (GCMs) (Quesada-Chacón et al. 2021). However, the limited spatial resolution of 100–300 km offers challenges in utilizing them for impact assessment studies at the local level (Zubler et al. 2016). For this reason, regional climate models (RCMs) which have a higher resolution are more suitable for such studies (Teutschbein and Seibert 2012). Furthermore, climate projections are associated with uncertainties which cannot be neglected in the use made of them (Piani et al. 2010). They result from uncertainties in greenhouse gas evolution scenarios, uncertainties associated with climate models and impact models (hydrology, biodiversity, etc.) and natural climate variability, which makes the direct use of these results is unreliable in impact evaluation studies (Ayugi et al. 2020). Furthermore, compared to regions with temperate climatic conditions, GCM-RCM projections show poorer performance compared to tropical regions, due to the nature of precipitation (Kotlarski et al. 2014).

Climate change would also bring multiple benefits for health, crops and ecosystems, and would limit the risks to development due to flooding and water shortages. An increase in temperature and atmospheric CO2 concentration promotes plant anabolism and photosynthesis. Even if the other resources necessary for plant growth are present, plant productivity would really increase. On the other hand, the thawing of permafrost, thanks to the increase in temperature, will open up new areas for agriculture, while thermal conditions will allow different plants to grow at higher altitudes. These modifications should therefore increase the agricultural productivity of the land around the poles.

Water resources vary greatly in space and time. Their estimation at the watershed scale requires hydrological models whose use is constrained by the complexity of the basins and the difficulty of measurement (basin characteristics and meteorological forcing) (Guan et al. 2019; Sidle 2021). For the management of water resources in a context of climate change, hydrological models have become standard tools to answer many questions such as the prediction and assessment of floods and drought (Mendez et al. 2022). Drought is an abnormal deficit, over a prolonged period, of one (at least) of the components of the terrestrial hydrological cycle. If meteorological drought which corresponds to a prolonged deficit of precipitation, hydrological drought corresponds to a deficit in flow of watercourses, low levels of water tables or reservoirs, over a period or a year during which the flows are very lower to the average. Hydrological drought can be characterized by its consequences on water as a living environment and as a resource. The ecological and economic consequences of hydrological drought can be aggravated by wastewater discharges, due to less dilution of polluting flows.

One of the challenges for hydrological models used to simulate the effects of climate change on river regimes is to be able to adequately represent hydrological processes under wide ranges of climatic conditions (Beven, 2001; Moradkhani and Sorooshian, 2018; Fowler et al. 2018). However, current models have difficulty performing well in this regard. Numerous works show a general difficulty in simulating the flow rates of watercourses outside the conditions encountered when calibrating their parameters (Acuña et al. 2019; Ahmed et al. 2023). Physics-based models encounter limitations as a result of the complex nature of the rainfall–runoff transformation process (De Vos and Rientjes 2005). Conversely, conceptual models offer a closer approximation to the fundamental physical mechanisms that govern hydrological processes by employing simplified equations (Parajka et al. 2007). Beside conceptual model with a lot of parameters limitation, it is better to mention other methods such as AI and of remote sensing hydrology estimation. To address these uncertainties and challenges, the CWatM and HMF-WA models simulate spatially coherent river flows on a spatial resolutions respectively of 0.5° × 5′ and 0.1° × 0.1° grid (approximately 10 km × 10 km) continuously over the entire domain and includes estimates of anthropogenic water use, wetland flooding, and local hydrological features such as endorheic regions.

The Community Water Model (CWatM) is a comprehensive hydrological model that incorporates several surface and groundwater hydrological processes, as well as human activities related to water usage and reservoir management (Burek et al. 2020). On the other hand, the hydrological modeling framework for West Africa (HMF-WA) is a distributed hydrological model that operates on physical principles (Rameshwaran et al 2022). The model in question is a regional model that incorporates a hydrological modeling framework known as HMF (Crooks et al. 2014). This framework consists of a geographically distributed hydrological model that operates on a grid structure, which bears resemblance to the G2G model (the G2G is a national-scale hydrological model for Great Britain that runs on a 1 × 1 km grid) (Bell et al. 2007). Its application in West Africa (Rameshwaran et al. 2021) takes into account a certain number of parameters such as anthropogenic use of water, flooding of wetlands and local hydrological characteristics such of endorheic regions in order to carry out spatially coherent simulations of river flows throughout the West African region (Rameshwaran et al 2022).

The Aga-Foua-Djilas watershed is located in the northern and northwest part of the Sine Saloum delta. In this basin, agriculture remains the main means of subsistence of the populations and shows the importance of rainfall and its spatiotemporal distribution in agricultural activities (Dione 2011). Therefore, it is necessary to study the evolution of the flow rate and the availability of water in the said basin, especially in a context of climate variability and change. The general objective of this study is to analyze the impacts of climate change on current and future water resources in the Aga-Foua-Djilas watershed, based on a comparative study of the CWatM and HMF-WA hydrological models. As a specific objective, the work aims to (i) characterize the flows of the Aga-Foua-Djilas watershed over the historical period and (ii) evaluate the impact of climate change on future flows of the Aga-Foua-Djilas watershed. Aga-Foua-Djilas, under different scenarios.

Study area

Aga-Foua-Djilas watershed is located to the north and northwest of the Sine Saloum delta, one of the main hydrological basins of Senegal. This basin is crossed by two major rias named Sine and Saloum. Despite the persistent drought (which occurred during the 1970s, 80s and part of the 90s), it is crossed by numerous small rivers which collect significant runoff each rainy season. Geographically, it extends between latitudes 14°15′N and 14°25′N, and between longitudes 16°37′W and 16°53′W, covering an area of 317.5 km2 on a perimeter of 115.7 km. From an administrative point of view, the basin straddles the municipalities of Malicounda, Sandiara, Séssène, and Nguéniène (Mbour department), as well as Tattaguine, Loul Séssène, and Djilas (Fatick department) (Fig. 1).

Fig. 1
figure 1

Source: Dione et al. (2023)

Presentation of the Aga-Foua-Djilas basin

Full size image

Materials and methods

ISIMIP data

The Inter-Sectoral Impact Model Intercomparison (ISIMIP) project, accessible at https://www.isimip.org, is a systematic approach for evaluating the consequences of climate change on various sectors, employing diverse scales (Lange 2019). ISIMIP encompasses a variety of global impact models, including hydrological models. The ISIMIP3b protocol was specifically developed to assist the IPCC special report on the 1.5 °C target (Frieler et al. 2017). It offers bias-corrected climate forecasts (Lange 2019) with a spatial resolution of 0.5° × 0.5°. These projections are based on data from five climate models that contributed to the IPCC sixth assessment report. Following the ISIMIP3b simulation protocol (www.isimip.org/protocol/#isimip3b), simulations are carried out on all general circulation models at a spatial resolution of 0.5° × 0.5°.

CWatM data

The Community Water Model (CWatM) is a comprehensive hydrological model that incorporates the fundamental processes of surface and groundwater hydrology (Burek et al. 2020). The model, created by the International Institute for Applied Systems Analysis (IIASA), measures the amount of water that is accessible, the amount of water used by humans, and the impact of water infrastructure, such as reservoirs, groundwater pumping, and irrigation (Burek et al. 2020). CWatM is a model that is built on a grid system. The most recent version of the model includes spatial resolutions of 0.5 degrees and 5 min at a daily temporal resolution. Figure 2 provides a schematic representation of the processes in the CWatM model.

Fig. 2
figure 2

Schematic figure of the processes included in the CWatM model

Full size image

We used precipitation P and flow rates Qm3/s. If for precipitation, a set of five global climate models (CESM2-WACCM, CNRM-CM6-1-HR f2, GFDL-ESM4, CNRM-CM6-1 f2 and MRI-ESM2-0) are used, to characterize the flow rates flowed into the basin, only the CWatM outputs from the MRI-ESM2-0 model (one of only two models available among the five on the ISIMIP platform) are used. The MRI-ESM2-0 model represents a subset of those participating in Phase 6 of the Coupled Model Intercomparison Project (CMIP6). Note that a limited number of models may not cover the full range of CMIP6 model propagation.

In this study, ISIMIP 3a data from 1981–2019 and ISIMIP 3b data from 2015–2100 in SSP 126, SSP 370 and SSP 585 scenarios were used. SSP 126 (low adaptation problem, low mitigation challenge) defines a world characterized by strong international collaboration and a focus on sustainable development, which is closely related to the “Proaction” family. SSP 370 (high adaptation issue, high mitigation issue) depicts a world affected by challenge among countries, limited economic growth, policies centered toward security and industrial production, and little regard for the environment, extremely close to the “Reply” family. SSP 585 (low adaptation issue, high mitigation issue) explains a world that focuses on traditional, rapid growth of developing countries, based on high-energy utilization and carbon-emitting technologies; an increase in standard of living allows for an increase in adaptation capacity, particularly as extreme poverty is reduced. This future is described by the “Growth at All Costs” family.

From the available trend-preserving statistical bias correction algorithm bias-corrected simulation data (Lange 2018), the MRI-ESM2-0 model was selected. Through this bias correction algorithm, the average monthly temperature and precipitation were corrected by adjusting the long-term differences between simulations and observations (Hempel et al. 2013; Lange 2018). For this study, the multi-model ensemble output data from the hydrological impact models at the outlet of the Aga-Foua-Djilas basin are available free of charge from ISIMIP (www.isimip.org), are used.

The choice of the CWatM model compared to other available hydrological models is justified by the fact that in this model the water balances are calculated for different land use classes and that the CWatM model is open source and community-driven to promote a works with the wider global water community and is flexible enough to connect with other planned developments such as water quality and hydro-economic modules. Additionally, the multi-model ensemble data input used in the case study is provided in the supplementary excel file. CWatM can use different daily weather forcing datasets, which are used as inputs to calculate potential evaporation. The model calculates the preferential bypass flow, that is, the flow that bypasses the soil matrix and flows directly to groundwater. The redistribution of soil moisture in three soil layers is calculated using the van Genuchten simplification of the Richards equation.

HMF-WA data

Hydrological modeling framework for West Africa (HMF-WA) is a model configured to simulate continuous flows across West Africa and the Sahel and to provide ensemble hydrological model estimates of average river flows monthly and annual maximums (m s3-1) on a 0.1° × 0.1° grid (approximate 10 km × 10 km grid) across West Africa for historical periods (1950–2014) and projected future periods (2015–2100) (Rameshwaran et al 2022). HMF-WA is also an open-source model, as is the case with CWatM. The data used in the modeling are those of CMIP6 (Fig. 3). These consisted of precipitation and temperature variables post-processed and bias-corrected on a 0.5° × 0.5° (~ 50 km × 50 km) grid over Africa (Rameshwaran et al 2022).

Fig. 3
figure 3

Schematic of the modeling methodology

Full size image

Before being used to study the impact of climate change on river flows in West Africa (Rameshwaran et al 2022), the performance of the HMF-WA model was first evaluated using daily flows measured by monitoring stations in the region (Rameshwaran et al. 2021). Model training data over the evaluation period (1981–2010) are daily precipitation from CHIRPS (Climate Hazards Group Infra-Red Precipitation with station data (Funk et al. 2014) and daily potential evaporation (PE) from the Global Land Evaporation Amsterdam Model (Martens et al., 2017), as mentioned by Rameshwaran et al. (2022).

Although this study carefully examined the variable features of major water cycle elements in the Aga-Foua-Djilas basin and anticipated the effect of hydrological processes under climate change scenarios, it is still constrained by a number of uncertainties. These may have an impact on the strength of the conclusion research. The uncertainty that exists in multi-source remote sensing satellite products directly affects the trustworthiness of the data. Previous research, for example, has shown that the accuracy of GPM (Global Precipitation Measurement) precipitation products on a daily basis and throughout the dry season is poor, and their ability to capture downpour occurrences remains limited. This could impact the basin’s precipitation inversion performance during extreme hydrological events (Huang et al. 2021). In addition to the deviation of satellite products, post-processing of products at different time and space scales causes some differences, such as spatial averaging of raster datasets and time scale conversions. The selection and optimization of hydrological model parameters can play a critical role in hydrological forecasting based on CMIP6 climate scenarios, and their impact will be amplified in the context of multiple climate-related uncertainties, which can have a significant negative impact on terrestrial waters, runoff, and evaporation simulation results (He et al. 2023).

Given the aforementioned constraints, future studies ought to concentrate more on examining the possible influence of multi-source uncertainty on the estimation of different hydrological variables in the Aga-Foua-Djilas basin. Along with artificial intelligence, it should incorporate other forms of hydrometeorological data, such as radar measurement, into the comparison. The technology computes the best answer using numerical techniques (such Bayesian weighted average) while also illuminating the possible range of variables. In order to provide scientific references, it is also important to further quantitatively study the role that climate change and human activity play in the disruption of water cycle processes in the basin (He et al. 2023). New remote sensing satellite missions, like the Surface Water Ocean Topography (SWOT) satellites, and hydrological modeling techniques should be used to clarify the impact mechanisms and interactions of these factors for better utilization of local hydrological resources.

The CWatM and HMF-WA models were not calibrated or validated in this study because flow rate data from the basin under investigation was not available. As a result, we were unable to assess the models’ dependability and performance in simulating hydrological processes.

Results

Simulation and estimation of flow rates over the historical period

Over the historical period (1981–2019), different hydrological parameters are analyzed. These are rainfall, average temperature, potential evapotranspiration and flow rates simulated with the CWatM and HMF-WA models. Table 1 presents the results of the Mann–Kendall and Pettitt tests which are applied to the values.

Table 1 Mann–Kendall test and Pettitt test on hydro-climatic parameters over the historical period (1981–2019) in the Aga-Foua-Djilas basin
Full size table

There are several uncertainties related to the results of the trend analysis and model simulations used to simulate flows in the Aga-Foua-Djilas basin. These uncertainties constitute reasons which explain the differences between the model simulations, particularly with regard to the divergent trends between the models. These may have some impact on the robustness of the conclusion research. The uncertainty inherent in multi-source remote sensing satellite products directly determines the reliability of the results. For hydrological forecasting based on CMIP6 climate scenarios, the selection and optimization of hydrological model parameters can play a key role, and its impact will be further amplified in the context of multiple climate-related uncertainties, which can have a negative impact significantly on terrestrial waters and runoff and evaporation simulation results (He et al. 2023). Despite these inevitable uncertainties, this study can still be a valuable resource to better understand water resource regimes during historical and future periods in the Aga-Foua-Djilas basin. Our study highlights the role of remote sensing in regional hydrological monitoring in ungauged basins as well as the great potential for early warning of hydrological extremes such as floods and droughts. These valuable results can help policy makers and water managers increase the regulation and withdrawal of local water resources, particularly in the context of climate change.

For the sources of uncertainty, the most frequently mentioned are the uncertainties in the input data and the boundary conditions, the uncertainty linked to the estimation of the parameters and the uncertainty linked to the structure of the model. Uncertainties in observational data, which include both forcing data such as precipitation and potential evapotranspiration, and response data such as flow, result both from errors in measurements but also from processing of this data. Precipitation is particularly uncertain due to the spatial sampling of the precipitation field.

Beyond the difference in scales of precipitation data (CWatM using daily precipitation data and HMF-WA using monthly precipitation data), the potential differences in the structure of the two models, the parameterization and the representation of hydrological processes in each model, could explain the divergent results obtained. Conceptual hydrological models use parameters that allow the response of the model to be adjusted to that of the rainfall–runoff transformation of a watershed. These parameters provide the flexibility needed to adapt a fixed mathematical structure to different watersheds. Faced with the difficulties encountered in specifying the values of the parameters from other information, the approach commonly adopted consists, when flow measurements are available (which is not the case in the basin studied), to adjust the parameters so as to reproduce as best as possible the flow rates observed at the outlet of the catchment basin considered.

Anthropization brings together all human activities that transform environments to exploit resources with the aim of establishing and/or developing human societies (Elissalde 2014). Anthropogenic activities can cause disruption of ecosystems. Urbanization, for example, leads to processes of erosion and waterproofing of surfaces, as for anthropogenic erosion, it plays an important role in soil morphogenesis (Tchotsoua and Bonvallot 1994). It is therefore necessary to take into account changes in land use for the transfer of pollutants due to the modification of flows, but also the intensification of discharges from activities and domestic wastewater depending on the housing densities which are the reflection of the anthropization of environments. Their impacts are generally illustrated by the modification of habitats (destruction, fragmentation), the loss of associated biodiversity (dos Santos et al. 2021) and could influence hydrological trends and flow dynamics in the basin.

Except for flows simulated with HMF-WA, all other parameters indicate positive trends. These are statistically more significant with flows simulated with CWatM which have a Kendall Tau of 0.433 m3/s/year. This Tau is 0.101 mm/year for rainfall, 0.134 °C/year for average temperature and 0.126 mm/year for potential evapotranspiration. Regarding the flow rates simulated with HMF-WA, the Tau de Kendall highlights a drop in the volume of water flowing at the basin outlet of −0.009 m3/s/year.

Pettitt test applied to the different hydrological parameters detects years of rupture. This occurred earlier with average temperature and evapotranspiration (in 1994) and was later with the other parameters. It occurred 6 years later for HMF-WA (in 2000), 10 years later for CWatM (in 2004) and 13 years later for rainfall (in 2007). The rate of change indicates an increase after breaking the parameters, except the flow rates under HMF-WA. This increase is 19% for rainfall, relatively low with 0.19% for average temperature, moderate with 2.6% for potential evapotranspiration and very significant for flow rates simulated with CWatM. For the flows simulated with HMF-WA, the rate of variation of −6.55% indicates their decrease in flows after the rupture, although this is not very strong.

Figure 4 shows the ten-year evolution of the hydrological parameters studied, with the exception of average temperatures.

Fig. 4
figure 4

Annual evolution of hydro-climatic parameters over the historical period (1981–2019) in the Aga-Foua-Djilas basin

Full size image

For an average rainfall of 415 mm recorded during the 1981–1990 decade, the average flow simulated by CWatM is 3 m3/s compared to 3.2 m3/s for the HMF-WA model (Table 2). During the following decades, rain continued to increase. It went from 426.1 mm in 1991–2000, to 455 mm in 2001–2010 then to 457.1 mm in 2011–2019. The simulated flows do not follow the same trend as the rain. With CWatM, they fell to 1.5 m3/s in 1991–2000 before gradually increasing to reach 5.1 m3/s in 2001–2010, then 11 m3/s in 2011–2019. During these last two periods, the flow rates simulated with this model therefore followed the rate of change in rainfall. The reproduction of flow rates simulated with the HMF-WA model only followed that of rainfall during the decade 1991–2000, with a slight increase of 3.6 m3/s. A drop was subsequently recorded with a flow rate of 3.1 m3/s in 2001–2010 and 2.8 m3/s in 2011–2019. In fact, general flow trends do not always follow those of rainfall. This shows that there are other parameters or factors that can influence the results of the models. Moreover, the peak of the 2011–2019 period with the CWatM model despite relatively stable precipitation proves this instability.

Table 2 Annual average values of hydro-climatic parameters over the historical period (1981–2019) in the Aga-Foua-Djilas basin
Full size table

The watershed is not gauged, so there are no measured data to validate those simulated by the two models which present variable results. However, the analysis of the hydrographs represented in Fig. 5 also shows that on the inter-monthly and inter-annual scales, the simulated flow rates follow the rhythm of rainfall.

Fig. 5
figure 5

Hydrographs simulated by the CWatM model under ISIMIP and the HMF-WA model over the historical period (1981–2019) at the outlet of the Aga-Foua-Djilas basin

Full size image

Using the monthly flow coefficient (CMD), the hydrological regime of the watershed is studied. The CMD is obtained by making the ratio between the flow of a month considered and the average annual flow. It allows you to determine the months of high and low water. According to the classification of Pardé (1968), if the CMD is greater than or equal to 1, the month corresponds to a period of high water, on the other hand, if the CMD is less than 1, the month coincides with a period of low water. Based on this categorization, the CWatM model shows 4 months of high water (August–November) compared to 3 months for the HMF-WA model (August–October). For the CWatM model, the water rise goes from May to September. The flow rate increases from 3.29 m3/s in May to 50.4 m3/s in September, with a very sudden descent of the waters from September to April. With HMF-WA, the rise begins in August, peaks in September, then declines in October, and the descent continues gradually until July. Its flow rate increases from 4.55 m3/s in August to 14.11 m3/s in September, one month after the peak of the rains. This last model more correctly reproduces the hydrological regime according to the quantity of rain received. In fact, wintering begins in June–July and ends in October.

Figure 6 shows the correlation between rainfall and flow duration curves simulated by the CWatM model under ISIMIP and by and the HMF-WA model over the historical period.

Fig. 6
figure 6

Correlation between rainfall and flow duration curves simulated by the CWatM model under ISIMIP and by the HMF-WA model over the historical period (1981–2019) at the outlet of the Aga-Foua-Djilas basin

Full size image

To address the relatively weak relationship between simulated precipitation and flow, duration curves of simulated precipitation and flow are calculated and from which correlations are determined, resulting in very high coefficients of determination. These flow duration curves comparing the flow rates simulated by the two models at the basin level show a better correlation (Fig. 6) than their monthly hydrograph (Fig. 5). The correlation equations show that there is a positive linear relationship between the precipitation duration curves and the simulated flow duration curves. The coefficient of determination R2 of 0.9371 indicates that 93% of the variation in flow rates simulated with CWatM can be explained by rain. This coefficient is, however, higher with the HMF-WA flow duration curves (0.9886). This suggests that this model has a better ability to explain the variation of simulated flows as a function of precipitation (Karamage et al. 2018; Souley Tangam et al. 2024). There is also a positive correlation between the flow duration curves simulated by the two models (0.9371).

Simulation and estimation of flow rates over the future period

Evolution on an annual scale

In order to characterize climate change and its impacts on water resources in the Aga-Foua-Djilas basin, CMIP6 climate models are used coupled with CWatM hydrological models under ISIMIP and HMF-WA. This prediction is made under different climate scenarios, which are SSP 126, SSP 370 and SSP 585 for all parameters except for HMF-WA where there are SSP 126, SSP 245 and SSP 585. The Mann–Kendall and Pettitt tests are applied to the basin data for a better understanding of future flow conditions in the Aga-Foua-Djilas basin, but also for a verification of trends and possible breaks in the series of the future period (2021–2100). The results are recorded in Table 3, showing significant fluctuations, as well as statistically significant trends.

Table 3 Mann–Kendall test and Pettitt test on hydro-climatic parameters from the flow rates simulated by the CWatM model under ISIMIP and the HMF-WA model over the future period (2015–2100) in the Aga-Foua-Djilas basin
Full size table

Given that climate change will have potentially significant impacts on the availability and management of water resources (Parry et al. 2007), it has become necessary to assess this impact on hydrological systems. This assessment involves taking into account uncertainty, which is an essential and indispensable step in climate change impact studies (Clark et al., 2016). The five main sources of uncertainty that must be explored throughout the hydrological modeling process in the context of climate change are: 1) the emission scenarios (SSP), 2) the global climate model (GCM), 3) the method downscaling, 4) the structure of the hydrological model, and 5) the parameters of the hydrological model (Xu and Singh 2004). Added to these is natural climate variability (Chen et al. 2011; Prudhomme and Davies 2009). Like any hierarchical system, uncertainty propagates through components. The uncertainties of the SSP scenario and that of the GCM will be amplified by the reduction of scale and the correction of biases. Finally, the structural uncertainty of the hydrological model puts an end to the cascade which leads to the total uncertainty of the hydrological impact assessment (Honti et al. 2013).

Average temperature and potential evapotranspiration record positive trends, synonymous with their increase, according to Tau Kendall, unlike the other hydrological parameters (precipitation and simulated flow rates) which show a decreasing trend. This drop is more significant with SSP 585, the most pessimistic scenario. For precipitation, the decrease will be − 0.157 mm/year according to SSP 126, − 0.321 mm/year with SSP 370 and − 0.472 mm/year, if the World lives at the pace predicted by SSP 585. Unlike precipitation, the average temperature and the ETP will increase according to the Kendall Tau of 0.405 °C/year (SSP 126), 0.808 °C/year (SSP 370), 0.87 °C/year (SSP 585) for temperatures and 0.43 mm/year (SSP 126), 0.802 mm/year (SSP 370) and 0.88 mm/year (SSP 585) for ETP. This increase in average temperature and water loss through the ETP, as well as the drop in the volume of rainwater that will fall on the basin, will naturally encourage a reduction in the amount of water to drain. To this end, with the CWatM model, the volume of water that will flow will experience a drop of around − 0.207 m3/s/year under SSP 126, of − 0.44 m3/s/year under SSP 370 and − 0.565 m3/s/year under SSP 585. Under HMF-WA, the drop will be − 0.248 m3/s/year, − 0.389 m3/s/year and − 0.579 m3/s/year, respectively, under SSP 126, SSP 245 and SSP 585. Under both models, the volume of water required to decrease is relatively the same.

The Pettitt test indicates ruptures, which will occur between 2036 and 2070. For precipitation, the rupture will be earlier under SSP 585 (2055) than under the other SSPs (in 2058 under 126, in 2064 under SSP 370). For the average temperature and water losses by evapotranspiration, the rupture dates will occur during the same years, in 2048, 2054 and 2055, respectively, for SSP 126, SSP 370 and SSP 585. Under SSP 126, the rupture will be earlier with the flow rates simulated by the CWatM model (in 2052) than under the HMF-WA model (in 2070). On the other hand, under SSP 585, it will intervene earlier with HMF-WA in 2046 than with CWatM in 2058, i.e., 12 years later. The rupture planned for 2036 and under SSP 245 with HMF-WA will be the earliest. This predicted under SSP 370 for CWatM is expected in 2049. For all scenarios, humidity and those after by drought will mark the periods before rupture. Temperatures and water losses through evapotranspiration will increase at the same time as precipitation and volumes of water flow will decrease. The more pessimistic the scenario, the more pronounced the decline will become. It will be around − 22.0%, − 38.4%, − 56.1%, respectively, under SSP 126, SSP 370 and SSP 585 for precipitation. With the flow rates simulated using CWatM, the drop will be further accentuated with values of − 66.3%, − 79.1% and − 98.3%, respectively, under SSP 126, SSP 370 and SSP 585. On the other hand, it will be less intense with the flow rates of the HMF-WA model and will be of the order of − 22.6%, − 27.0%, − 49.7%, respectively, under SSP 126, SSP 245 and SSP 585. This decreasing trend of HMF-WA flow follows that of precipitation with a relatively small difference, unlike that noted with CWatM, which presents a significant difference, even if the evolution is similar.

Figure 7 and Table 4 show the parameter trends in the future period, compared to the historical baseline.

Fig. 7
figure 7

Compared annual variation in flow rates simulated by the CWatM model under ISIMIP and the HMF-WA model over the historical period (1985–2014) and the future period (2021–2100) at the outlet of the Aga- Foua basin—Djilas, according to climate change scenarios SSP 126, SSP 370 and SSP 585

Full size image
Table 4 Flow trend simulated by the CWatM model under ISIMIP and the HMF-WA model and other parameters over the future period (2021–2100) compared to the historical one (1985–2014) at the outlet of the basin Aga-Foua-Djilas
Full size table

Both temperature and ETP show overall positive trends over the years. The more the scenario is provided, the more the increase will become more pronounced. Conversely, simulated rainfall and flow rates display variable overall trends from one sub-period to another, with increases in volumes of precipitated or runoff water during the first half of the century and decreases during the second.

The simulated flows show variable trends with an increase in flow compared to the reference period until the decades 2041–2060, followed by a more marked decline with CwatM than with HMF-WA. With an increase of 131% (SSP 126), 224% (SSP 370) and 320% (SSP 585), in the near future, the flow rates simulated with CwatM will experience an increase of 152% (SSP 126), 14.1% (SSP 370) and 137% and (SSP 585) at future mean 1, before recording a drastic drop of − 29.0% (SSP 370), -94.8% (SSP 585) at future mean 2. However, under SSP 126, during these decades 2061–2080, the flow will always be in excess (22.7%) compared to the reference period (1985–2014), even if this excess is far less significant than that of the future average 1. In the last decades of the century (2081–2100), the drop in CwatM flow rates will be excessive with rates of − 32.8% (SSP 126), − 80.8% (SSP 370) and − 94.6% (SSP 585). The variations predicted under HMF-WA are less significant than those under CwatM. With HMF-WA, SSP 126 forecasts a deficit of − 1.4% in the near future, followed by a surplus of 0.8% in the mean future 1. On the other hand, under SSP 245 and SSP 585, the near future will be marked by an excess volume of water to runoff of 4.2% (SSP 245) and 9.7% (SSP 585). From future mean 1, the volume of flows will start to decrease under SSP 245 (− 11.2%) and SSP 585 (− 29.9%). This downward trend will continue and become more accentuated in the mid future 2 (− 17.3% SSP 126; − 21.8% SSP 245; − 45.6% SSP 585) and at the end of the twenty-first century (− 22.3% SSP 126; − 19.6% SSP 245; − 50.9 SSP 585).

Simulated flow rates, particularly with CWatM, respond significantly to changes in precipitated water volumes. They show varying trends over the decades. The variation in flow trends obtained using the two models (CWatM and HMF-WA) indicates differences in their sensitivity to climate change. The results of the models show hydrological dynamics in the basin, through the increase in evaporation with the increase in temperature noted, particularly at the end of this century, and the decrease in precipitation. Thus, the reduction in runoff is very consistent with that of precipitation, which confirms the role played by precipitation, which constitutes the dominant factor of the water cycle in the basin.

As the studied basin is ungauged, it remains difficult to carry out a sensitivity analysis to assess the understanding of the model results in relation to the main input parameters and assumptions. Furthermore, it is impossible to discuss the reliability of the model simulations and potential biases, taking into account factors such as model performance in reproducing historical trends and limitations in representing complex hydrological processes. Although limitations and inconsistencies in model projections between the CWatM and HMF-WA models exist, it remains difficult to discuss the reasons behind them. Unlike the historical period where divergent results are noted by the two models, over the future period, the results obtained, although different, recorded concordant trends. It would then be appropriate to identify areas of future research, such as the installation of a water height measurement station at the outlet of the basin to determine the flow rates and use them to validate the results of the modeling, to respond to the remaining uncertainties and improve understanding of future hydrological changes in the basin.

Evolution on a monthly scale

Over the historical reference period (1985–2014) and the future period (2021–2100), on a monthly scale, the variations in monthly flow rates are indicated in Table 5 and Figs. 8 and 9. In light of these data, the hydrological regime of the Aga-Foua-Djilas watershed is monomodal. It is in fact characterized by a period of high water which lasts 4 months according to the CwatM model (August–November) and 3 months according to the HMF-WA model (August–October); and a low water period varying from 9 (HMF-WA) to 8 months (CwatM) depending on the models. The month of September corresponds to the hydrological maximum, which varies greatly from one model to another.

Table 5 Monthly evolution of the flow simulated by the CWatM models under ISIMIP and HMF-WA over the historical period (1985–2014) and the future period (2021–2100) at the outlet of the Aga- Foua—Djilas basin
Full size table
Fig. 8
figure 8

Monthly variation of future flow simulated by the CWatM model under ISIMIP over the future period (2021–2100) at the outlet of the Aga-Foua-Djilas basin, according to the climate change scenarios SSP 126, SSP 370 and SSP 585 for the period 2021–2040, 2041–2060, 2061–2080 and 2081–2100

Full size image
Fig. 9
figure 9

Monthly variation of future flow simulated by the HMF-WA model over the future period (2021–2100) at the outlet of the Aga-Foua-Djilas basin, according to the climate change scenarios SSP 126, SSP 370 and SSP 585 for the period 2021–2040, 2041–2060, 2061–2080 and 2081–2100

Full size image

Under CwatM over the reference period, the maximum monthly flow in September is 13.4 m3/s. It will increase over the future period (2021–2100) to 23.9 m3/s under SSP 126; at 22.4 m3/s under SSP 370 and at 26.65 m3/s under SSP 585.

Under HMF-WA the maximum flow in September over the reference period is approximately equal to that under CwatM. It is in fact 13.8 m3/s. Over the future period (2021–2100), it will experience a decline, unlike the flow rate under CwatM, going from 13.1–12.9 m3/s before reaching 9.64 m3/s, respectively, under SSP 126, SSP 245 and SSP 585.

Under both models, future flows will follow the same seasonal trends, as is the case over the reference period, with peak periods during the rainy season and no flow during the dry season. The HMF-WA model best illustrates this absence of flow during the dry season marked by the absence of rain.

The analysis of the rates of variation under the different scenarios and according to the models reveals different situations, especially during the rainy season.

Under the CwatM model

Under SSP 126: In the near future, flow rates will increase compared to the reference period. This increase will increase in the average future (146%). In future mean 2, there will be a slight drop in flow rates compared to the previous period, but the latter will still be higher than the flow rates of the reference period. In the last two decades of this century (2081–2100), a flow deficit of − 32.9% is expected.

Under SSP 370: The flow rates follow the same trend as those under SSP 126. However, from future mean 2, the volume of water flow will drop considerably by − 23.2% and reach an alarming deficit of − 77.3% at the end of the twenty-first century.

Under SSP 585: A big contrast is expected between the flow rates of the first half of the century and those of the second half. Indeed, during the near and medium future 1, a strong increase in runoff during the rainy season is expected. This will be of the order of 372.9% and 183.1%, respectively, in 2021–2040 and 2041–2060. From 2061, the decline will be drastic and will reach and exceed − 94%.

Under the HMF-WA model

The month of August, despite the fact that it is that of maximum rainfall, will be deficit in terms of runoff throughout the future period (2021–2100).

Under SSP 126: The near future (with 0.50%) and medium 1 (with 2.51%) will be slightly in surplus compared to the reference period. In future mean 2, flows will suddenly drop by − 15.72%, a decline which will increase and reach − 20.74% at the end of the century.

Under SSP 245 and SSP 585: The expected flows during the rainy season will follow the same trend over the decades. Compared to the reference period, flows will first increase during the near future, at the start of the century, an increase which will be 5.5% under SSP 245 and more sustained under SSP 585 with 10.9%. From mean future 2, a drop in flows during the rainy season is also expected. The more the years go by, the more the deficit will increase. And the more intense the scenario, the greater the deficit will be. This expected drop in flows will in fact be − 9.7% in 2041–2060, − 20.6% in 2061–2080 under SSP 245 and − 17.9% in 2081–2100. For these same decades, the flow loss will be − 29.3%, − 45.7% and − 51.3% respectively under SSP 585.

In short, for the different models and the different scenarios, an initial increase in flow rates is expected during the first half of the century followed by a decrease during the second. This drop will be earlier with the HMF-WA model and will occur from 2041, while with the CwatM model, it is expected at least in 2061. It should also be emphasized that the predicted flow deficits will be less accentuated with the flow rates simulated by the HMF-WA model.

Compared to the reference period, the biennial evolution of flow rates, according to the models and under the different scenarios, has several nuances.

Under the CwatM model

The elapsed flow rate for the reference period is 4.6 m3/s. Under SSP 126, the average flow rate will be 10.6 m3/s, 11.6 m3/s, 5.6 m3/s and 3.1 m3/s, respectively, over 2021–2040, 2041–2060, 2061–2080 and 2081–2100. During these same biennials, the flow will vary under SSP 370, respectively, by 14.9 m3/s, 5.2 m3/s, 3.2 m3/s and 0.9 m3/s. Finally, under SSP585 and over the same biennials, the average volume of water to flow will be 19.2 m3/s, 10.9 m3/s, 0.23 m3/s and 0.24 m3/s, respectively.

Under the HMF-WA model

The average flow rate for the reference period is 3.22 m3/s. This volume of water flow will pass under SSP 126 during the first two-year period 2021–2040 at 3.21 m3/s, before reaching 3.28 m3/s, 2.69 m3/s and 2.53 m3/s, respectively, on 2041–2060, 2061–2080 and 2081–2100. Under SSP 245, the volume of water flow will be approximately equal to that under SSP 126 and over all vice-annuals (3.39 m3/s in 2021–2040, 2.89 m3/s in 2041–2060, 2.54 m3/s in 2061–2080 and 2.62 m3/s in 2081–2100). Under SSP 585, the first two-year period will record a flow rate of 3.57 m3/s. A drop in flow rates is expected from the period 2041–2060 (2.28 m3/s) and will increase in 2061–2080 (1.77 m3/s) and in the last two years of the century (1.6 m3/s).

The variation in flow rates will be much more pronounced under the CwatM model than under HMF-WA. Indeed, under SSP126 and during all biennials, the flow rates generated with CwatM will be almost double those generated with HMF-WA. It will be the same under SSP 585 during the near and medium future 1. Moreover, the flow projected in 2021–2040 under CwatM will be triple that with HMF-WA. However, from the 2061–280 biennial, the situations are reversed. The speeds under CwatM will drop drastically and will even be doubled by those under HMF-WA. Changes in the contribution of different hydrological components of the hydrological system to climate change in the hydrographic basin, despite the nuances between models, are of great importance. This study attempted to simulate the changes in the main hydrological components for the future periods of the scenario compared to the reference simulated flow components (1985–2014) obtained from the two models. In the near and distant future, Fig. 10 shows that the flow generated by the two models and under the different scenarios will decrease compared to the control period (1985–2014). The flow variations as well as their magnitude are very different, especially between models, but they remain consistent in the different scenarios with rainfall.

Fig. 10
figure 10

Quantile–quantile plot of residuals and correlation between flow duration curves simulated by the CWatM model under ISIMIP and by the HMF-WA model over the future period (2021–2100) at the outlet of the Aga- Foua-Djilas basin

Full size image

In statistics, a Q-Q (quantile–quantile) plot, which is a graphical method for comparing two probability distributions by displaying their quantiles versus quantiles, is used (Fig. 10). These normal curves of the Q-Q plot indicate that our sample is a bit skewed, with the points not aligning along the line. For a better relationship between the flow rates simulated on the two models, duration curves of the flow rates simulated over the future period are calculated and from which correlations are determined, which gives rise to very high determination coefficients (Fig. 10). These flow duration curves comparing the simulated flows at the basin level show a better relationship (Fig. 6) than the monthly hydrograph (Fig. 5). The correlation equations show that there is a positive linear relationship between the flow duration curves simulated by the two models. The coefficient of determination R2 of 0.9573 under SSP 126 and 0.8989 under SSP 585 between the simulated flow duration curves of the CWatM model under ISIMIP and HMF-WA are very high.

The Aga-Foua-Djilas basin is located in the West African region and the evidence of this hydro-climatic variability there is consistent with the projections of the Intergovernmental Panel on Climate Change. Indeed, at the outlet of the Aga-Foua-Djilas basin, the models predict a reduction in average monthly flow rates in the future over these four horizons under the two scenarios (SSP 126, SSP 245, SSP 370 and SSP 585). These results indicate that the availability of surface water resources in the watersheds should continue to decrease by the end of the twenty-first century, especially with SSP 585. The Aga-Foua-Djilas basin remains very vulnerable to climate change, given a greater reduction in precipitation.

Given the climate change conditions in the Aga-Foua-Djilas basin, the downward trend in water resources in a very unfavorable scenario, such as SSP 585, compared to SSP 126, is worrying. The above results are similar to those reported in other regions but with the extreme scenarios of CMIP6. In the basin, precipitation is the main source of reconstitution of the flows. This means that the latter are affected by the variability and seasonal distribution of precipitation in climate models. Thus, patterns of seasonal change in runoff follow those of precipitation.

Just as on the annual scale, on the monthly scale, the simulated future flows are greater on the CWatM model than on the HMF-WA model. Nonetheless, the most significant changes or trends are noted under the SSP 585 scenario. Provide further discussion on why there are differences in the hydrologic regimes predicted by the CWatM and HMF-WA models. Modeling hydrological processes involves errors due to two main sources of uncertainty. These sources are those related to the parameters and structure of the model (Liu and Gupta 2007). Parameter uncertainty may exist due to the dependence of model parameters on the characteristics of the calibration period, which may differ from the simulation period (Brigode et al. 2013).

Beyond the difference in configuration of the model structures, numerous uncertainties intervene and can explain the differences. Some work has shown that although the uncertainty linked to the GCM is dominant for average flows (Aryal et al. 2019; Gao et al. 2020), the contribution of other sources of uncertainty can sometimes exceed that of the GCMs and especially for low flow rates. Therefore, consideration of various sources of uncertainty is particularly important in this case (Wang et al. 2020). These model differences may have potential implications for water resources management and decision-making in the basin, which could be replicated in other basins. This is why studies which are based on a single GCM must be handled with great caution (Minville et al. 2008), and it is strongly recommended to use a multi-model analysis and to exploit several GCMs in order to overcome this challenge (Trzaska and Schnarr 2014). The use of less than five (5) GCM leads to an underestimate of the uncertainty linked to the latter and subsequently it is strongly recommended to use at least ten (10) GCM or more in order to properly take into account consideration this issue (Wang et al. 2020).

These changes noted between the two models can have an impact on water availability, ecological habitats and socio-economic activities. Uncertainty related to future projections can have potential implications for decision-making under different scenarios. Uncertainty refers to the fact that the consequences of a decision are not known in advance, which exposes the decision maker to risk. The risk incurred is objectively quantifiable when the probability of occurrence of the different possible consequences is objectively known in advance.

Discussion

Senegal being very vulnerable to climate change, it is important for sustainable use of water resources, to be able to predict future hydrology, the impacts of climate change, to quantify at the watershed scale, the components of the balance sheet water. CMIP6 climate models and two additional hydrological models, namely CWatM and HMF-WA, are used to predict the climate change process and hydrological feedback dominated by “warming” in the Aga-Foua-Djilas basin in the framework of future emissions scenarios.

In the West African basins, many studies have predicted more rapid temperature changes. During the dry season, an increase in temperature is alarming for the hydrological system because it has implications for water loss. This increase in temperature during the dry season is consistent with previous studies (Azmat et al. 2018; Hasson et al. 2014). The increase in temperature will ultimately lead to a reduction in water availability. Using the CwatM and HMF-WA hydrological models to predict the hydrological feedback effect of the basin, we see that the evaporation process will be reinforced with the increase in temperature, especially at the end of the twenty-first century.

The results of this study from precipitation projections in the Aga-Foua-Djilas basin showed a large variation, which makes future forecasts of water availability very uncertain in the long term, and a downward trend. The low average annual rainfall in the Aga-Foua-Djilas basin as well as the downward trend in rainfall could lead to difficulties in water use, despite its strong agricultural potential. The climate trends calculated based on the SSP scenarios in this research are aligned with previous research around Senegal (Mbaye et al. 2019; Sadio et al. 2020; Diakhate et al. 2022) as well as exterior (Almazroui et al. 2021; Díaz et al. 2021; Ortega et al. 2021; Phan and Nguyen 2023). The results are different from the results where a statistically insignificant increasing trend in precipitation was recorded in all seasons (including annual time scale) (Girma et al. 2016) and statistically significant in July and November (Arragaw and Woldeamlak, 2017) in the central highlands of Ethiopia have been reported.

On a regional and global scale, climate change has had a significant impact on precipitation, temperature and flows in river basins. To facilitate the prediction of changes and management of water resources for a variety of sectors and uses, trend analysis of these meteorological and hydrological variables has received much attention (Ahmad et al. 2015). In previous studies, it is well documented that increasing temperatures generally promote an increase in actual evapotranspiration, which results in a decrease in water supply and soil moisture (Sun et al. 2016; Guo et al. 2017). These GCM outputs were used to simulate the flow of the river from the watershed. Runoff showed a consistent change with that of precipitation in the analysis, which is consistent with the conclusions drawn from the analysis of historical data; thus, precipitation is the dominant factor in the water cycle in the basin. More precisely, the degrees of change in runoff at the end of this century (2081–2100) are expected to be for the CWatM model in the order of − 32.8% (SSP 126), − 80.8% (SSP 370) and − 94.6% (SSP 585) and under HMF-WA of the order of − 22.3% (SSP 126), − 19.6% (SSP 245) and − 50.9% (SSP 585). These results are in concert with those of Diedhiou et al. (2018), Panthou et al. (2018), Taylor et al. (2017), and Todzo et al. (2020) who studied the past and future impact of global warming on flow in West Africa.

In general, it was observed that the river flow, in the future, for different scenarios decreases. Overall, the trend test results demonstrated that the basin has a statistically significant trend of increasing temperature and a statistically significant trend of decreasing precipitation and flow over the future period. Indeed, climate change will have an impact on the future evolution of the trend change in the hydro-climatic variables of the watershed. It is expected to have an effect on future water supplies for irrigation, rain-fed agriculture and other uses. This requires better management plans for the future. The analysis of the impact of climate change in the present study was limited to a particular region of the Aga-Foua-Djilas basin. This can also be extended to other river basins.

Given that climate change will have potentially significant impacts on the availability and management of water resources (Parry et al., 2007), it has become necessary to assess this impact on hydrological systems. This assessment involves taking into account uncertainty, which is an essential and indispensable step in climate change impact studies (Clark et al., 2016). The five main sources of uncertainty that must be explored throughout the hydrological modeling process in the context of climate change are: 1) the emission scenarios (SSP), 2) the global climate model (GCM), 3) the method downscaling, 4) the structure of the hydrological model, and 5) the parameters of the hydrological model (Xu and Singh 2004). Added to these is natural climate variability (Chen et al. 2011; Prudhomme and Davies 2009). Like any hierarchical system, uncertainty propagates through components. The uncertainties of the SSP scenario and that of the GCM will be amplified by the reduction of scale and the correction of biases. Finally, the structural uncertainty of the hydrological model puts an end to the cascade, which leads to the total uncertainty of the hydrological impact assessment (Honti et al. 2013).

In general, there will be a reduction in the water resources of the basin, which will have implications for the management of water resources in Senegal. Likewise, we highlight the need for continued research efforts to monitor and assess the evolving impacts of climate change on water resources. To advance readers understanding of the impacts of climate change on hydrology and water resources management in Senegal, possible areas of future research identified are the analysis of the impacts of climate change and land use and soil on the hydrological dynamics of watersheds.

Conclusion

In the context of decreasing water resources and increasing water demand, hydrological models are useful for understanding and projecting future trends at different scales. This study assesses the impacts of climate change on flow rates in the Aga-Foua-Djilas basin using CMIP6 data using CWatM hydrological models under ISIMIP and HMF-WA, a first in this basin. Despite the absence of measured flows to validate them, the results showed that the models worked well and reproduced the observed flow of the basin. However, the HMF-WA model has been validated on the West African scale with flow rates measured over large river basins. Over the historical period (1981–2019) in the Aga-Foua-Djilas basin, the analysis of all hydrological parameters indicates positive trends, although statistically not significant (with the exception of flow rates). Over the future period, unlike temperatures and PET, which have an upward trend in all scenarios, precipitation and flow rates recorded downward trends, more significant under SSP 585.

Changes in precipitation and temperatures have a significant impact on the components of the water balance with runoff declining increasingly toward the end of the century. These results help stakeholders, policy makers and farmers to take precautionary measures such as proper planning of water management and agricultural practices. However, this study does not take into account future land use change. Therefore, the future scope of the study is to assess the combined effects of future climate and land use changes in the Aga-Foua-Djilas basin.

The simulations analyzed as part of this study would need to be reproduced with forcings climatic from of models different, and with ideally A together of simulations in order to of better take in account the uncertainties linked to models climatic used as well as those linked to inter-annual variability. In addition, they deserve to be carried out with more hydrological models to be able to better estimate the uncertainties associated with our understanding of the response of the water cycle to climate forcing.

In the context of water resources management, hydrological risk prevention and water policy decision-making, the comparison between the global CwatM model and the regional HMF-WA model is relevant. It demonstrated how on a spatial scale, differences in scale influence the results and that the global model CwatM is capable of capturing hydrological characteristics on a regional scale and predicting regional changes. Global models provide trends on a global scale. Regional models are calibrated and validated for specific local conditions. The comparison between the two allows local decision-makers to have information on global and local trends to improve the understanding of hydrological processes at different scales in order to develop better strategies for adaptation to climate change and sustainable management of water resources water.

When the results of models are used for decision-making, the estimated magnitude of the model uncertainty due to uncertainty on input parameters should be considered as the lowest or minimum bound for the potential variability. The question about practical approaches to quantify the upper bound, which should include effects of structural errors and amount and quality of observed data, remains an open question. The fact that hydrological models cannot provide absolute answers but only approximations to the behavior of natural systems, may be seen as a main limitation for their use in practice. On the contrary, models should be seen as tools to produce uncertain forecasts that by using appropriate methods to integrate data and analyze their performance are based on interpreted or measured information. Through this, models can help to make better-informed decisions.

Ultimately, the results highlighted in this study make it possible to study the biophysical component of the vulnerability of water resources in the Aga-Foua-Djilas watershed. The lack of water resources would increase the strain already placed on them in this pool, which is already small. It is also necessary to consider the rise in water demand associated with predictions of Senegal's rapid population development. It is also necessary to consider the rise in water demand associated with predictions of Senegal's rapid population development. It therefore appears necessary to put in place adequate strategies to deal with the risks highlighted in this study, as well as to continue to insist on an increase in global efforts to reduce greenhouse gas emissions in international fora.