Assessment of climate change impacts on streamflow in the Inner Niger Delta under CMIP6 scenarios

Since the 1980s, climate change has been a prominent topic of discussion by scientists. Questions such as, Why is the climate changing? What are the effects of greenhouse gases? How much of the increase in the atmosphere's CO₂ concentration and temperature is human-induced? What are the possible impacts and the extent of these changes? Etc. have been the topic of numerous studies (IPCC 2001; Hitz & Smith 2004; Blöschl & Montanari 2010; Singh 2013; Roudier et al. 2014; Schramek & Harmeling 2017; Hagg et al. 2018; Vozinaki et al. 2018; Iltas et al. 2024). Vozinaki et al. (2018) investigated the hydrometeorological impacts of climate change in two Mediterranean river basins. They reported a significant increase in temperature and a decline in river discharge due to reduced precipitation. Iltas et al. (2024) reported similar projections for the Mediterranean Edremit Eybek Creek. In their assessment of climate change impact, Hagg et al. (2018) reported that a significant increase in annual runoff sums is expected under future climate conditions.

Climate change has become a reality aggravating the situation, specifically in parts of the world already suffering from water shortages and extreme events such as floods, hurricanes, droughts, and heat waves (Okpara et al. 2013). Africa, especially the Sahel region, is highly affected, with a rise in temperature of 0.2 °C per decade in the 1980s, increasing to 0.6 °C per decade at the end of the 20th century (Zwarts 2010). The observed increase in surface temperature has generally been more rapid in Africa than the global average. The IPCC Sixth Assessment Report (AR6) projects significant changes in the region, including continued climate warming, an increase in rainfall extremes, and reduced water availability driven by increased evaporation rates. By 2100, average temperatures in the region are expected to rise by 2–4 °C. Moreover, an increase in meteorological droughts, heavy precipitations, and river flooding is anticipated with high confidence (IPCC 2021). Climate affects the water resource systems of a river basin since it is part of the hydrological cycle. There is, therefore, a necessity to research, comprehend, and evaluate the possible effects of climate change on hydrologic regimes and climate-sensitive sectors such as water, energy, agriculture, fisheries, health, forestry, transport, tourism, and disaster risk management, inter alia.

Knowledge about future climate can be obtained from general circulation models (GCMs) combined with regional climate models (RCMs). However, different combinations of GCMs and RCMs can provide a large range of climate predictions, with specific large uncertainties over the Sahel region (Held et al. 2005; Thompson et al. 2016). This is affected by multiple competing mechanisms affecting rainfall. It is thus difficult to predict whether there will be more or less rainfall in the Sahel. Hulme (2001), in a comparative study, concluded that annual rainfall in the western Sahel would possibly remain at the same level, but that a decrease of 10%–40%, is more likely. Held et al. (2005) presented a climate model that captured several aspects of the 20th-century rainfall record in the Sahel. The model predicted that rainfall until 2020–2040 will remain at about the same low level as the last 20 years of the 20th century but will then gradually decrease by about 20% in the next 50–100 years, thus projecting a drier Sahel in the long-term future, primarily due to increasing greenhouse gases. Thompson et al. (2016) assessed the impact of climate change on a 2 °C increase in global mean temperature in the Upper Niger Basin (UNB) and the Inner Niger Delta (IND). They reported that, notwithstanding the considerable inter-GCM variations, the mean annual precipitation is predominantly projected to decrease (−24%) across the catchment, while all models agree on an increase (+4%) of the potential evapotranspiration (PET) for all sub-catchments. However, precipitation is projected to increase over the Delta in every month except January. In a follow-up study (Thompson et al. 2021), the authors assessed the impact of climate change on environmental flow using 12 GCM groups and the Ecological Risk due to the Flow Alteration method (ERFA). They reported a roughly equal number of GCM groups projecting an increase in river discharge as those projecting a decrease, and also an important inter-GCM variability in the ERFA-indicators change. Angelina et al. (2015) found that the mean annual flow at Koulikoro station would increase by about 7% (2026–2050), 1% (2051–2075), and 6% (2076–2100). Aich et al. (2014) investigated the future trends in average flows and indicators of extremes over 30-year periods during the first and second halves of the 21st century. They observed a trend toward increasing flows in three of the four basins studied, including the Niger River Basin (NRB). In summary, in this debate, it is important to note that however uncertain the future might be for the Sahel in general and for the NRB in particular, climate change is undoubtedly expected to have a significant impact not only on water resources but also on human society at large. The divergence in the findings emphasizes the need for more studies that may help to elucidate the climate change signal for the region's hydrology. This is important because the region is vulnerable to major hydro-climatic events and depends heavily on water supplies for economic development.

The IND has been the subject of various studies. However, the climate variables as well as the discharge time-series used in most published studies were limited to the period before 2000 (Oguntunde & Abiodun 2013; Angelina et al. 2015; Thompson et al. 2016, 2017). Yet, as argued by Ozer et al. (2003), the drought in the Sahel may have ended during the 1990s. Indeed, many researchers reported an increase in rainfall in the Sahel since the 1990s, i.e., recent rainfall amounts are found to be well above the 1961–1990 average (Olsson et al. 2005; Lebel & Ali 2009; Mahe et al. 2013; Sanogo et al. 2015; Nicholson et al. 2018). This wetting of the Sahel may skew any climate projection analysis if not considered. Therefore, as suggested by Ozer et al. (2003), including recent years such as the 2000–2010 decade would yield better precision. Moreover, little to no investigation of climate change impacts has been made in the NRB considering the recently developed Shared Socioeconomic Pathways (SSPs), which examine how global society, demographics, and economics, in short, socioeconomic factors, might change over the next century (Hausfather 2018). This is important, as population growth in the area will put pressure on land use and land cover and subsequently intensify the hydrological cycle.

This paper aims to contribute to the growing area of research by providing scientific evidence to substantiate the future impacts of climate change on the hydrology of the IND. This will help to improve understanding of potential future changes, assisting policymakers in designing adaptation and mitigation strategies. The paper is structured as follows: the information about the river basin, as well as, the various datasets and models used in the study are described in Section 2. The results are presented and discussed in Section 3. Some conclusions and recommendations are provided in the last section.

During the rainy season, the different types of natural wetlands in the basin (i.e., rivers, lakes, pools, and floodplains) join to form one body of water of 20,000–30,000 km², creating a complex ecosystem in the IND. This vast seasonal floodplain, which is influenced by the Niger River, is of paramount importance to the rural economy of Mali. It is the driver of flood recession agriculture (especially rice, cotton, and wheat cultivation), fishing, and cattle herding in the area. However, it exerts a significant influence on the river flow. In fact, evaporation and seepage from this area cause a significant loss in the river discharge.Zwarts et al. (2005) reported that on average, 45% of the combined inflows to the IND at Ke-Macina and Beney-Kegny are lost. The magnitude of losses increases with larger inundation extents. The IND is also a biological hotspot. It is the largest reservoir of biological diversity in Mali and Africa more broadly. It was classified by UNESCO in February 2004 as a Ramsar World Heritage Site for Humanity (Gonet & Stausee 2004).

There are a couple of important water resource schemes in the basin that have different impacts on the river flow. There is the Sotuba hydropower dam (constructed in 1929 just below Bamako), the Sélingué dam (completed in 1982 on the Sakarani River), and the Markala dam (completed in 1945), which diverts water to irrigate the Office du Niger (ON), one of the biggest and most intensive irrigation schemes in West Africa. There are also plans for future dams. Future changes in the streamflow can thus be expected due to two main factors. First, the upstream water resources scheme, and second, climate-induced changes to precipitation and evapotranspiration, will likely impact river flows. The current study aims to assess the extent of climate-related changes in river flows. Throughout both the baseline and each scenario, the existing dams were simulated as being operational. Therefore, the differences between the scenarios and the baseline could be attributed to climate change.

In situ weather data from 11 pluviometric stations located in the basin were collected from the National Meteorological Agency of Mali (MALI-METEO) for a data consistency check. They are comprised of monthly precipitation (rainfall) and temperature for the period 1981–2010. In addition, daily discharge data were also collected for the same period for the stations Mandiana, Pankourou, Koulikoro, Douna, Ké-Macina, and Diré. The hydrometric data were provided by the National Directorate of Hydraulics of Mali (Direction Nationale de l'Hydraulique DNH). The missing data were filled before any processing, using the Multivariate Imputation by Chained Equations (MICE) with the Python Sklearn library. MICE is a machine-learning method that inputs the multivariate missing data in a dataset through an iterative series of predictive models (Azur et al. 2011). Due to the very limited spatial allocation of meteorological stations, gridded daily rainfall and temperature data for all 149 subbasins (Figure 1) were sourced from CHIRPS (Funk et al. 2015) and WFDEI (Weedon et al. 2014), respectively. Numerous research studies have validated these datasets in many parts of the world (Raimonet et al. 2017; Wu et al. 2019; Aksu & Akgül 2020; Gupta et al. 2020; Shen et al. 2020; Bamweyana et al. 2021; López-Bermeo et al. 2022). These gridded datasets have also been used in the West African region and have been proven reliable (Andersson et al. 2017; Dembélé et al. 2020; Kwawuvi et al. 2022).

In climate research, socioeconomic and emission scenarios are used to provide plausible descriptions of how the future may evolve in terms of a variety of variables, such as socioeconomic change, technological change, energy, land use, greenhouse gas, and air pollutant emissions (Van Vuuren et al. 2011). The research community has developed many climate scenarios over the years. The most internationally used scenarios are the RCPs, which superseded the Special Report on Emissions Scenarios, and the recently developed SSPs. This research used climate simulation data taken from a combination of the sixth phase of the Coupled Model Intercomparison Project (CMIP6) (O'Neill et al. 2016). The models were statistically downscaled from 2° to 0.5° spatial resolution by the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP3b), an initiative that provides cross-sectorally consistent projections of the impacts of different levels of global warming in the 21st century (Lange 2019). The models were selected to capture all the ranges of equilibrium climate sensitivity (ECS) (Sherwood et al. 2020). ECS defines the sensitivity of the global climate to radiative perturbation and has been used for four decades to estimate the sensitivity of global mean surface temperature to a doubling of atmospheric CO₂. While the likely range of the ECS has been 1.5–4.5 °C, a significant number of CMIP6 models exhibit greater values. Table 1 provides a detailed description of the ten climate models incorporated into the multi-model ensemble. These selected climate models were also used in previous research studies in West Africa (Yangouliba et al. 2022; Kwawuvi et al. 2023). Because CMIP6 historical simulations stopped in 2014, for each model, daily minimum, mean, and maximum temperature and daily precipitation were collected for the historical period (1981–2010) and the near future period (2021–2050). The near future was selected as this is more relevant for policymakers, whose focus is primarily on the near-time climate change implications.

The efficiency of the models (as compared with the observed data) was first assessed for the baseline (historical) period (1981–2010). For this purpose, the Taylor diagram was used to compare the simulated historical rainfall with the observed (reference) rainfall on a monthly scale. Some basic statistical indicators of fit, such as the correlation and the root mean square error ⁠, were used to evaluate the models' performance.

On the need for bias correction of climate models, Christensen et al. (2008) note that an application of a simple bias correction to each model of an ensemble would possibly reduce the spread among the models' projected climate-change signals. In this research, the individual model outputs (temperature and precipitation) were corrected using the distribution method of quantile mapping (QM). This method has been proven in previous studies to be effective and even outperform other methods, especially for the analysis of precipitation data (Luo et al. 2018).

Many models have been developed for hydrological studies, and very limited consensus exists on which model is best or most applicable. The choice of a specific model is, therefore, based on the purpose for modeling, the priorities, the limitations of data availability, time, and budget for model help. For the present study, the semi-distributed catchment model Hydrological Predictions for the Environment (HYPE) was used for the hydrological modeling. Developed by the Swedish Meteorological and Hydrological Institute (SMHI), HYPE is a small-scale and large-scale hydrological model capable of simulating daily fluxes and turnover of water, nitrogen and phosphorus (Lindström et al. 1997). The model divides the landscape into classes based on soil type, land use, and elevation (i.e., soil type–land use combinations, SLCs). HYPE was developed in 2003, applied to the entirety of Sweden in 2008, then to other continents, and finally to the entire world (Arheimer et al. 2020). In hydrological studies, usually, a single model is developed and applied to different regions, assuming the model is generic enough to handle the differences. However, there is a danger in applying a model off the shelf because the model may not be able to represent the dominant hydrological processes of the new region, a region for which the model was not first developed. To tackle this challenge for the NRB, SMHI and the Regional Centre AGRHYMET proposed a refinement of the HYPE model concept for the Niger basin. The model was set up using readily available large-scale datasets such as HydroSHEDS (Lehner et al. 2008), GlobCOVER (Arino et al. 2008), Harmonized World Soil Database (FAO et al. 2012), and WISE (Batjes 2012). More details about the model setup can be found in Andersson et al. (2017). Andersson et al. (2017) reported that the resulting model (Niger-HYPE) performed significantly better across the Niger basin. In this research, the Niger-HYPE was, therefore, applied. The model was calibrated, validated, and then forced with future rainfall, temperature, and PET to simulate the future daily streamflow. The daily discharge data were used as the basis for calibration and validation. Previous research studies suggest that long calibration periods of hydrological models could be beneficial for large datasets, such as future scenarios (Al-Safi & Sarukkalige 2017). Therefore, all the datasets for the six hydrometric stations were used in the process. The first five years of the dataset (1981–1985) have been used for model initiation (warm-up period), the next 15 years (1986–2000) for calibration, and the last ten years (2001–2010) for validation. The calibration was performed using the Differential Evolution Markov Chain (DEMC) automatic optimization method. DEMC is often used for HYPE calibration due to its speed and convergence. Detailed information on the DEMC method can be found in Braak (2006). The calibrated parameters consisted of general parameters (damp, epotdist, gldepi, krs, limqprod, lp, pcelevadd, pcelevth, pcelevmax, rivvel, sswcorr, tcobselev, and tcalt), soil-dependent parameters (wcfc, wcep, wcwp, srrate, rrcs, and mperc) and land-use-dependent parameters (kc and srrcs). More details about the parameters can be found in the HYPE file reference.

A statistical trend analysis was performed on the different climate variables (temperature, rainfall, and PET) for the historical period. The PET was estimated by the hydrological model using the modified Hargreaves–Samani method. It is a temperature-based method and a simplified alternative to the FAO Penman–Monteith method. More details about the method are available at Hargreaves & Samani (1985). For the purpose of trend analysis, a modified version of the Mann–Kendall (MK) test and the Discrete Wavelet Transform (DWT) method were used. The literature avails a wide range of methods for trend detection in time-series data. There are the MK tests and the regression analyses, among others. Regression analysis is best suited to time series with a normal distribution. The original MK test is known to be sensitive to data dependence, which is typical of hydrometeorological time-series (Abebe et al. 2022). These dependencies in the form of serial correlation or seasonal relationships could lead to a biased analysis. Numerous modifications of the original MK test have been proposed in the literature to tackle the issues of autocorrelation, seasonality, and serial dependence factors in datasets. In this study, the modification proposed by Hamed & Rao (1998), which accounts for seasonality and autocorrelation factors in the dataset, has been applied. The analysis was performed using the Python library pyMannKendall (Hussain & Mahmud 2019).

A second approach used is the DWT, which is a relatively new tool for trend detection studies. It essentially decomposes the time series into a limited number of elementary series depending on both time and frequency (Pandey et al. 2017). A key feature of the wavelet analysis is its ability to reveal various features of a time series, such as trends, periodicity, discontinuities, and change points. The Python library PyWavelets (Lee et al. 2019) was used to perform the wavelet analysis. It consisted of three steps: (1) identifying the mother wavelet to be used for decomposition, (2) establishing the level of decomposition, and (3) determining the signal extension mode. More details about the different steps can be found in Abebe et al. (2022). The library has 14 predefined mother-wavelet families, each with several subcategories. The diversity of the wavelets enables the identification of a wavelet that best fits the features of the signal (time series). In this study, the Daubechies (db) wavelet was chosen over other wavelet families because it is argued to be the most common smooth mother-wavelet for hydrometeorological study (Pandey et al. 2017).

The future change of the different variables relative to the baseline period was also assessed for the two scenarios.

On the monthly scale from 1981 to 2010, the statistics showed a good correlation between the satellite data products and the observed datasets. As shown in Table 2, in most stations the satellite products tended to underestimate rainfall and overestimate temperature. Overall, the datasets showed a good correlation, and the biases can be neglected considering −20% to +20% as an acceptable performance range for satellite products as suggested by Moriasi et al. (2007). Therefore, the CHIRPS and WFDEI datasets were used as reference datasets for the analysis.

The results of the DWT and the MK trend analyses of the monthly and annual rainfall, temperature, PET, and discharge confirmed the existing trend in the dataset. The results, as shown in Table 3, showed a positive trend for all variables. The MK test was conclusive at the 95% confidence interval for both the monthly and the annual scales. For some variables, such as temperature and discharge, the test was statistically significant even at higher confidence intervals (99%). These findings corroborate the general assertion that the Sahel has been going through a wetting process since the mid-1990s, which often is accompanied by an increase in vegetation in some areas, in part due to an increase in precipitation (Nicholson 2005; Olsson et al. 2005; Lebel & Ali 2009; Sanogo et al. 2015). In fact, the Sahel experienced a wet period during the 1950s and 1960s. Starting from the 1970s, rainfall began to decrease, and the decrease intensified in the 1980s. This period is known as La Grande Sécheresse (the Great Drought) in West Africa. Since the mid-1990s, rainfall is reported to be on the increase again. Sanogo et al. (2015), for instance, investigated rainfall trends in West Africa for the period 1980–2010 using daily and monthly rainfall data from 167 and 254 stations, respectively. They reported that the majority of stations in the Sahel demonstrated a statistically significant increasing trend in annual totals. The largest recovery was observed during the flood period (August–October). They also highlighted an increase in extreme rainfall events. The observed trends in the present study are also in line with reports from Aires et al. (2014), who assessed climate change's impact on four representative basins in Africa, namely the Niger, Upper Blue Nile, Oubangui, and Limpopo basins. They reported a distinct temperature and evapotranspiration rise and a significant increase in streamflow across the Niger basin. However, it is worth noting that the observed increase may be the result of a shift in monthly rainfall patterns. For instance, localized instances of intense rainfall will result in an increase in total rainfall but will also lead to inundations, as has been observed in many places in recent years. The devastation caused by heavy rainfall in Niamey in 1998 and the flooding of Dakar, Saint Louis, and Kaolack in Senegal in 1999 and 2000 are some instances of these extremes in the Sahel region (Ozer et al. 2003).

Overall, all scenarios agree on a significant increase in the different hydro-climatic variables, with SSP370 always predicting a greater increase. In most cases, PET and rainfall increase every month, and where decreases do occur, they are limited to a few months when the values of the variables are relatively low. Although climate change predictions in the Sahel are not always conclusive, particularly for rainfall, many studies have concluded that an increase is a likely plausible outcome (Oguntunde & Abiodun 2013; Thompson et al. 2016; IPCC 2021). The IPCC Sixth Assessment Report (AR6) reported continued climate warming, an increase in rainfall extremes, and an increase in evaporation rates in the study area (IPCC 2021). Oguntunde & Abiodun (2013) reported that in all months and seasons, a typically warmer climate is anticipated over the NRB, which reflects the findings of the present study. Thompson et al. (2016) reported that rainfall over the Inner Delta is projected to increase every month except for an already dry January. The findings are also consistent with the predictions made by the Climate Information Portal (SMHI 2024), developed by SMHI on behalf of the World Meteorological Organization, the World Climate Research Programme, and the Green Climate Fund. The portal predicted an increase of about 8.6% and 0.80 °C in rainfall and temperature, respectively, for the Segou region. This overall increase in rainfall will possibly lead to an increase in streamflow.

Given the good performance of the model, it was deemed reliable for assessing the impacts of climate change on future river flows. The model was, therefore, forced with the baseline and the future data for the two scenarios (SSP126 and SSP370). Table 6 provides the mean, the Q5 (discharge equaled or exceeded 5% of the time), and Q95 (discharge equaled or exceeded 95% of the time) discharges at the inlet (Ké-Macina and Douna) and the outlet of the IND (Diré). All flows revealed an increase under both scenarios. Under SSP126, at the entrance of the IND, the mean discharge is expected to increase by 13.8% and 19.3% at Ké-Macina and Douna stations, respectively. For SSP370, the change is slightly higher with 16.4% and 33.7% at Ké-Macina and Douna stations, respectively. At the outlet of the IND (Diré station), the mean discharge is anticipated to increase by 7.3% and 9.8% under SSP126 and SSP370, respectively. Relative to the baseline period, an increase of about 5.2% and 6.2% in the Q95 discharge is anticipated under SSP126 and SSP370, respectively. These findings are comparable to the general statistics reported by Roudier et al. (2014). They reviewed climate change impacts on runoff in West Africa based on 19 published studies and observed that a significant positive change of about 6.1% is expected in the Niger River flows. An interesting finding here was that the increase at the outlet is about half the increase at the inlet. This seems to suggest that multiple processes within the wetland are the cause of this reduction. In essence, in the future, processes such as evaporation would be further exacerbated in the IND as compared with other parts of the basin.

A possible explanation for the overall increase in the discharge may be the increase in rainfall as discussed in the previous section. In this regard, Roudier et al. (2014) observed that variations in runoff are strongly correlated with variations in rainfall, and to a lesser extent, with variations in PET. This confirms evidence of changes in hydrological responses consistent with observed changes in rainfall. Another possible explanation may be the increasing land use change in the basin. Continuous expansion of agricultural lands and desertification contribute to an increased runoff, which in turn leads to increased discharges.

In summary, the present results provide important insights into the still open question of whether the Sahel will get wetter or drier in the future. In contrast to some earlier studies (Liersch et al. 2013; Thompson et al. 2016, 2017), no evidence of a general drying of the region was found. The findings rather provide strong evidence that supports the hypothesis of a positive trend in the basin in accordance with reports of some previous studies (Aich et al. 2014; Angelina et al. 2015).

The approach considered in the present study, like many studies on climate change impact, is limited by some assumptions and uncertainties. As reported by Gosling et al. (2011), each step in this exercise introduces uncertainty. First, the different GCMs, which are the basis of the assessment, are known to contain some uncertainties as the varied process representations and parameterization approaches cause the GCMs to yield different forecasts. Second, the downscaling of the models to a finer spatial and temporal resolution, as well as the bias adjustment, introduces new uncertainties. Third, hydrological models per se are susceptible to a variety of uncertainties. These include, but are not limited to, an inaccurate understanding of hydrological system behavior and missing or incorrect hydrometeorological data. Furthermore, it should be noted that the potential future changes in the hydraulic infrastructures in the basin as well as future land use changes were not considered in this analysis. In fact, future reactions to climate change in water management and land use are difficult to predict (Thompson et al. 2016). These factors will almost certainly have some impact on their own, especially in places where the population is growing fast. For instance, according to the FAO's Global Information System on Water and Agriculture (AQUASTAT), water withdrawals in Mali grew by 58% between 1990 and 2020, and given that the region's food consumption is expected to increase significantly, this value will also likely increase much more in the future.

The hydrological model Niger-HYPE has been calibrated and validated on the Upper Niger and the IND and used to assess the impacts of future climate on streamflow over a future 30-year time slice. The daily rainfall, minimum, and maximum temperature were extracted from ten climate models derived from a combination of the fifth and sixth phases of the CMIP under two scenarios, namely SSP126 and SSP370. The hydrological model was found to be quite effective at predicting streamflow. In fact, in terms of performance criteria (KGE and NSE), the overall performance for the calibrated model may be considered good. The strength of the model is demonstrated by the average values of KGE 0.78 and KGE 0.75 during the 25-year calibration and validation period of all six gauges. In summary, both scenarios projected an increasing trend in the mean annual and monthly rainfall, temperature, and evapotranspiration. The anticipated average increases were 6.5%, 1.25 °C, and 5.3% for rainfall, temperature, and evapotranspiration, respectively. The same trend was observed in the simulated streamflow, which was anticipated to increase on average by 8.5% by the end of 2050. The magnitude of the projected increase varied according to the scenario, with the pessimistic scenario (SSP370) projecting a higher increase.

These findings have many implications for water resource management in the basin. The change in rainfall and discharge will have important impacts on water resources in the highly vulnerable economies of the region. The increase in evapotranspiration, especially in the wetland, will have negative impacts. It will most likely lead to higher crop water demand and the need for additional irrigation withdrawals, especially during the dry season, when a decrease in streamflow is already expected. This will have dire impacts on dry-season agricultural activities. Further impact will almost certainly include degradation of water quality due to increased transport of sediments caused by torrential precipitation. Surface water will also likely undergo a decline due to the increasing temperatures. All these will likely accentuate migration and foster potential conflicts. This emphasizes the need for mitigation and adaptation strategies. Better water management techniques, such as effective irrigation and rainwater harvesting, are essential. Additionally, the development of water infrastructures, such as dams and canals, can serve as effective water storage and management practices, particularly with the increasing number of extreme events. Up to the present, only the Sélingué reservoir, with its total storage volume of 2.2 km³, is the significant infrastructure in the area. There are also the Sotuba and the Markala dams, which are diversion dams. However, notwithstanding the multiple delays in execution, important structures such as the Fomi (6.16 km³) and the Djenné dam (0.357 km³) are already planned. These might help to regulate the flow and possibly mitigate any negative impact of climate change on the hydrology of the IND. Coordinated adaptation measures, which require collaboration (at the national, regional, and international levels) and the development of institutional capability, can also play a key role in improving water management practices, especially for transboundary basins such as the NRB.

The impacts of the change in climate as well as different land use scenarios on future water stocks and fluxes (i.e., water accounting) would constitute a valuable extension of the present study.

The authors wish to thank the Mali Meteorological and Hydrometric Services for providing the hydrometeorological data for the basin. We also acknowledge the Regional Centre AGRHYMET CCR-AOS for providing the Niger-HYPE source code. The authors would also like to thank all anonymous reviewers for their constructive comments.

This research has been designed by B.B., A.Y.B., J.v.d.K., M.M., and L.O.S. The data were processed by B.B. with the help and suggestions of B.M., M.H., H.D.A., and G.I.Y. The manuscript was written by B.B. with inputs from all co-authors. All authors have read and agreed to the published version of the manuscript.

This work forms part of a PhD research funded by the World Bank and the French Development Agency through the Centre d'Excellence Africain pour l'Eau et l'Assainissement (C2EA) programme.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.