Assessment of climate change for predicting water–energy–food–ecosystem nexus vulnerability in a semi-arid basin

The Water–Energy–Food–Ecosystem (WEFE) nexus is an integrated approach that recognizes the interconnections and interdependencies between water, energy, food, and ecosystem sectors (Hanes et al. 2018; Garcia et al. 2019; Karabulut et al. 2019; Momblanch et al. 2019; Özcan 2023). As climate change is expected to have significant impacts on these sectors, understanding the potential effects on the WEFE nexus is crucial for the effective management of the nexus components. In this context, analyzing precipitation and temperature at the watershed scale is critical since precipitation and temperature patterns influence the availability and quality of water resources, which are relied upon for various human activities, including agriculture, industry, and energy production. Climate change alters precipitation patterns and temperature regimes at both global and regional scales (Liu 2016), and these changes can lead to more frequent and intense droughts, floods, and other extreme weather events that can have significant impacts on water resources, ecosystems, and human societies (Apeh & Nwulu 2024). At the watershed scale, changes in precipitation patterns can affect the amount and timing of water availability, which can impact the water supply for human consumption, energy generation (Alakbar & Burgan 2024), irrigation, and other uses. Temperature changes can also affect the water balance of a watershed by altering the rate of evapotranspiration, which can impact the amount of water available for use (Wubneh et al. 2023).

Analyzing trends in and revealing the climate-change impacts on watershed-scale precipitation and temperature is necessary to understand the effects of climate change on the WEFE nexus. Changes in precipitation and temperature can have significant effects on the availability and distribution of water resources, which can in turn impact food production (FAO 2015), energy generation (Cevik 2022), and ecosystem services (Bates et al. 2008). By analyzing trends in precipitation and temperature, we can gain insights into the potential impacts of climate change and develop management strategies to mitigate these impacts. However, there is a lack of a conceptual framework and adequate practices to connect the primary drivers of climate change affecting water resources with the Water–Energy–Food nexus and associated ecosystem processes, which in turn results in knowledge gaps (Liu 2016). Analyzing the impacts of climate change and performing trend analysis on a subbasin scale within a semi-arid watershed is particularly important because climatic characteristics may vary significantly from one subbasin to another. This variability can have significant implications for water availability and usage, making it crucial to account for these differences in developing effective management strategies. Ishida et al. (2017) indicated that when assessing hydrological issues related to climate change, it may be more effective to analyze precipitation across multiple watersheds rather than focusing on just one. This is because the impacts of climate change on precipitation can vary among different watersheds within the same region.

Global climate models (GCMs) simulate climate at a global scale, typically with spatial resolution ranging from tens to hundreds of kilometres. However, this resolution is often insufficient for making predictions at regional or local scales, where climate impacts are felt most acutely (Xu et al. 2019). Dynamical downscaling is a technique used to obtain higher-resolution climate projections from GCMs (Giorgi & Bates 1989). This technique aims to address this issue by using regional climate models (RCMs) to simulate conditions at higher spatial resolutions, typically ranging from a few kilometres to tens of kilometres (Giorgi & Gutowski 2015). Thus, the dynamical downscaling approach can be used whenever small-scale information on future climate or climate change is needed. Consequently, dynamical downscaling finds extensive use in numerous regions and across diverse climate domains (e.g., Hay & Clark 2003; Dulière et al. 2011; Soares et al. 2012; Erler et al. 2015; Ishida et al. 2017; Hu et al. 2018; Komurcu et al. 2018). Other downscaling approaches include statistical downscaling, which involves using statistical techniques to relate large-scale climate variables from GCMs to smaller-scale variables at the regional or local level (Kim et al. 1984). Statistical downscaling methods are often faster and less computationally intensive than dynamical downscaling methods, but they may not capture the full range of regional climate processes and are less suitable for simulating changes in extreme weather events (Jang & Kavvas 2015).

In climate-change impact studies, it is crucial to consider the uncertainties in climate projections. The inertia of the Earth system, along with increased greenhouse gas (GHG) emissions, will lead to further warming in the 21st century, but the degree of this warming remains highly uncertain. This uncertainty is primarily due to three factors: natural variability, model uncertainty, and GHG emission scenario uncertainty (Latif 2011). One approach to account for these uncertainties is to include a collection of simulations, known as an ensemble (e.g. Van Khiem et al. 2014; Kaur et al. 2020; Dong et al. 2021; Gao et al. 2022).

In this study, we assess climate-change impacts on annual precipitation and temperature across seven subbasins in the semi-arid Sakarya Basin, which is vulnerable to the risks of drought and holds significant agricultural, economic, and ecological importance. This assessment is crucial for predicting vulnerabilities in the WEFE nexus. While the impacts of climate change on basin-scale nexus systems have been previously studied (e.g., Liu 2016; Dlamini et al. 2023; Jander et al. 2023; Baccour et al. 2024), our study introduces a distinct approach. We employ a methodology that integrates dynamically downscaled climate data from the Weather Research and Forecasting (WRF) model, specifically calibrated and validated for the Sakarya Basin, with an 18 km spatial and hourly temporal resolution across three future periods (2020–2030, 2055–2065, and 2090–2100). By doing so, we aim to capture both large-scale and small-scale climate dynamics that are critical for understanding localized climate impacts on water, energy, food, and ecosystem sectors. Furthermore, we have incorporated four different climate projections (CCSM4 and MIROC5 under RCP 4.5 and RCP 8.5), enabling us to analyze a wider range of future climate scenarios and uncertainties. After bias correction, we analyze trends in basin-average precipitation and temperature using least squares regression and the Mann–Kendall test, in addition to the statistics, i.e., min, max, and mean, of the basin-average precipitation and temperature based on the annual average precipitation and temperature time-series. The primary aim of this study is to examine the potential impacts of these changes on the nexus sectors' vulnerabilities. These sectors are represented by municipal water demand coverage, energy generation, agricultural productivity, and environmental flow. By incorporating subbasin-specific indicators, we account for variability across the subbasins of the larger Sakarya Basin, allowing for a more localized and precise vulnerability assessment. Our integrated multi-indicator evaluation method combines dynamically downscaled precipitation and temperature data with socio-economic and environmental indicators, i.e., agricultural areas (%), population, provincial gross domestic product (million TL), current state river flow (hm³/y), annual average hydroelectric generation (GWh), and annual average thermal electricity generation (GWh), all together providing a holistic understanding of how changes in climate patterns will affect the interconnected WEFE sectors at both basin and subbasin scales. Additionally, unlike most studies (e.g., Wu et al. 2022; Albatayneh 2023; Dlamini et al. 2023; Jander et al. 2023; Gao et al. 2024), we include an ecosystem component, which is essential for sustaining the other nexus sectors.

Finally, our work informs future WEFE nexus studies by providing inputs to integrated water–energy systems models. This contribution is critical for advancing nexus-based resource management and decision-making, particularly in regions like the Sakarya Basin, where water, agricultural, energy, and ecological demands must be balanced under changing climate conditions.

In conclusion, the novelty of this paper lies in the integration of high-resolution climate projections, a multi-indicator evaluation method, and a subbasin-specific approach to assessing the impacts of climate change on the vulnerability of the WEFE nexus in a semi-arid watershed.

The Sakarya Basin has an average annual precipitation of 479 mm and is thus one of four basins in Türkiye with an average rainfall of less than 500 mm. On the other hand, considering the total annual precipitation, it is one of the basins receiving the highest precipitation at 32 billion m³. Additionally, the Sakarya Basin is one of the highest flow-value basins with an annual flow of 12 billion m³, despite being one of the lowest in terms of rainfall per capita (4,437 m³/person) due to its high population (SYGM 2022).

Within the Sakarya Basin, there are seven subbasins: Upper Sakarya, Porsuk, Ankara, Kirmir, Middle Sakarya, Göksu, and Lower Sakarya, listed from upstream to downstream. The selection of the Sakarya Basin for this analysis is primarily due to the unique climatic and socio-economic characteristics exhibited by its subbasins. Each subbasin presents distinct features that highlight different aspects of the WEFE nexus. This diversity allows for the application of a uniform methodology across regions that demonstrate varying dynamics and challenges. Table 1 provides details on the drainage area, average precipitation, and average temperature values for each subbasin. These average precipitation and temperature values are derived from observed data collected from September 2009 to September 2018.

A study conducted by the Ministry of Forestry and Water Affairs General Directorate of Water Management in 2016 investigated the potential impacts of climate change on water resources in Türkiye (The Ministry of Forestry & Water Affairs 2016). This study discusses the results of three different climate models (HadGEM2-ES, MPI-ESM-MR, and CNRM-CM5.1) and their projections for the Sakarya Basin in Türkiye. The models predict an increasing trend in temperature values during the projection period compared with the reference period. The models also predict changes in precipitation, with the possibility of both increases and decreases, although a decrease in precipitation is expected to be dominant. The study shows that rising temperatures and changing precipitation patterns have led to a decline in streamflow and groundwater recharge. The findings have important implications for water resource management in the country and highlight the need for adaptation strategies to cope with the impacts of climate change.

In the semi-arid Sakarya Basin, known for its high population density, agricultural significance, and intensive industrial activities, investigating the effects of climate change and implementing appropriate measures is crucial. Moreover, the basin's climate varies significantly from north to south, making it crucial to evaluate the impact of climate change on precipitation and temperature at the subbasin level for effective adaptation.

The Sakarya Basin is divided into seven subbasins, each with unique characteristics and vulnerabilities to climate-change impacts on the WEFE nexus. To predict the WEFE nexus vulnerability, we employed an integrated multi-indicator evaluation approach. This method integrates dynamically downscaled climate data with a comprehensive set of socio-economic and environmental indicators. Specifically, we used historical data for the WEFE nexus and future projections for climate variables.

Climate change was analyzed in terms of the projected changes and trends in the basin's average annual total precipitation depths and basin-average temperatures over the seven study subbasins in the Sakarya Basin in the 21st century. The future period was divided into three segments as near century (2020–2030), mid-century (2055–2065), and far century (2090–2100), for a total of 33 years. The results were investigated based on the four dynamically downscaled bias-corrected future climate projections accordingly. In addition to the individual assessment of all climate projection realizations, the analysis was also performed based on the ensemble averages of two emission scenarios (RCP 4.5 and RCP 8.5) and the ensemble average of all GCMs (CCMS4 and MIROC5). Ensemble averages are useful for the assessment of the general trends in the precipitation and temperature, and to take into account the uncertainties resulting from the selected GCMs and scenarios (Ishida et al. 2020).

The least squares regression method and the Mann–Kendall trend test were employed to quantify the trends in the annual basin-average precipitation and temperature. The slopes and the standard errors of the trend lines were calculated through the least squares regression method. The Mann–Kendall trend test is non-parametric and it is used to determine the significance of consistently increasing or decreasing trends (Mann 1945).

Table 2 presents a selection of indicators for assessing the vulnerability of the WEFE nexus in the study subbasins. These indicators encompass both climate-related variables and socio-economic factors. To assess WEFE nexus vulnerability, we used historical socio-economic data under the assumption of a ‘business-as-usual’ scenario, assuming current socio-economic conditions would remain stable. This approach provides a baseline for evaluating the effects of climate variability and change while simplifying the analysis by focusing on climate impacts. In assessing the vulnerabilities of WEFE nexus sectors, represented, respectively, by agricultural productivity, municipal water demand coverage, energy generation, and environmental flow, we conducted expert judgment evaluations. These assessments, specific to each subbasin, were based on statistically analyzed climate data obtained through dynamic downscaling and socio-economic indicators reflecting sectoral importance across subbasins. Instead of direct quantitative calculations, insights were derived through a synthesis of climate data analyses and socio-economic indicators to infer sectoral vulnerabilities.

This study uses four different future climate projection realizations from two general circulation models (GCMs: CCSM4 and MIROC5) based on two emission scenarios (RCP 4.5 and RCP 8.5). CCSM4 is the fourth version of the Community Climate System Model (CCSM4) (Gent et al. 2011). The model was developed by the National Center for Atmospheric Research (NCAR), United States. It has a spatial resolution of 0.9° × 1.25°. MIROC5 is the latest version of the Model for Interdisciplinary Research on Climate which was cooperatively produced by the Japanese research community (Watanabe et al. 2010). The model has a spatial resolution of 1.4° × 1.4°. The two emission scenarios selected were RCP 4.5 and RCP 8.5. RCP 4.5 is a scenario that stabilizes radiative forcing at 4.5 W m⁻² in the year 2100 without ever exceeding that value (Thomson et al. 2011). On the other hand, the GHG emissions and concentrations in the RCP 8.5 scenario increase considerably over time, leading to a radiative forcing of 8.5 W/m² at the end of the century. RCP 8.5 corresponds to the pathway with the highest GHG emissions compared with the total set of RCPs (Riahi et al. 2011).

It is necessary to downscale the GCM results to finer spatial resolutions for a reliable assessment of the regional impact of climate change on precipitation and temperature. Dynamical downscaling is one of the techniques to do this by means of an RCM. In this study, the WRF model is used to dynamically downscale future climate projection realizations. The WRF model is a mesoscale numerical weather prediction system designed for both atmospheric research and operational forecasting applications, developed collaboratively by multiple agencies (Skamarock et al. 2008). It is a fully compressible and non-hydrostatic model.

In this study, the future projection period ranges from 2020 to 2100. The future climate projection realizations were downscaled to 18 km resolution grids over the Sakarya Basin for a period of 33 years (2020–2030; 2055–2065; 2090–2100). The realizations were downscaled as ten-year segments due to the high computational cost of the regional model and time constraints. The historical atmospheric data used in this study is from September 2009 to September 2018 based on the availability of observed meteorological data in the study basin. Observed daily total precipitation and temperature data from 20 different meteorological stations within the Sakarya Basin were obtained from the Turkish State Meteorological Service (MGM). These stations are detailed in Table 3 and their geographic locations are illustrated in Figure 1.

Before downscaling future climate projections, the WRF model needs to be calibrated. ERA-Interim reanalysis data was used for the calibration of the model. ERA-Interim is one of the best reanalysis datasets of its generation and has been extensively downscaled with WRF. It is a common practice to use ERA-Interim-driven WRF simulations to evaluate the ability of the WRF model to simulate observed fields (Gorguner et al. 2019). For this purpose, reanalysis data for the years between 2009 and 2018 were obtained. ERA-Interim data set with a 6 h temporal resolution is available from the Research Data Archive (RDA) at NCAR. The spatial resolution of the dataset is approximately 80 km with 60 levels in the vertical from the surface up to 0.1 hPa.

To find the best combination of physics options, several different combinations of the parameterizations of the WRF model were tested. The model configuration of the WRF model used in this study is shown in Table 4.

The GCM (CCSM4 and MIROC5) precipitation and temperature projections from September 2009 to July 2018 were downscaled, based on the availability of observed data in the monitoring stations, and the observed and downscaled basin-average monthly precipitation and temperature values were calculated. Then, Equation (1) and Equation (2) were used to correct biases in temperature and precipitation, respectively.

In this section, we explore the projected changes and trends in basin-average precipitation and temperature in the 21st century. Additionally, we assess the vulnerability of the WEFE nexus in seven study subbasins, taking into account both climate and socio-economic factors. This comprehensive analysis is structured as follows: annual total basin-average precipitation (3.1), basin-average temperature (3.2), and WEFE nexus vulnerability (3.3).

Minimum and maximum values of annual total precipitation depths in the historical period as well as in the 21st century and each segment of the future period over the seven study subbasins with the mean value in parentheses are given in the Supplementary Material, Tables A1–A4, respectively. Statistical information about the historical period was provided for comparison purposes. The given mean values in these tables indicate the average of the annual total precipitation depth for the relevant time-period. In this way, the changes in the precipitation depths are analyzed by comparing the historical means with the means of each future-period segment. Additionally, the Supplementary Material, Figure B1 in Appendix B presents box-and-whisker plots of total monthly precipitation for the ensemble average across each subbasin.

In the Upper Sakarya subbasin, all climate projection realizations except for CCSM4_RCP8.5 and the ensemble of RCP 8.5 scenarios show a decreasing trend in the basin-average precipitation in the 21st century. Among all of the projections, only the trend in the ensemble of RCP 4.5 scenarios is statistically significant. In the near century, while all CCSM4 projections and the ensemble of RCP 8.5 scenarios have an increasing trend, the rest of the projections have a decreasing trend. However, none of the trends is statistically significant. Similar to the trends in the near century, there is no statistically significant trend in the mid-century. Three of the projections, i.e., CCMS4_RCP4.5, MIROC5_RCP8.5, and the ensemble of RCP8.5 scenarios have a decreasing trend while the rest have an increasing trend. Finally, all RCP 8.5 projections and the ensemble of all projections indicate a decreasing trend in the far century. Again, the trends are away from the monotonic trend according to the p-value of the Mann–Kendall trend test. When the statistics of the historical precipitation are compared with the 21st century, near, mid-, and far century, the mean of the MIROC5 projections and the mean of the ensemble of RCP 4.5 scenarios are lower compared with the historical period with the only exception being the ensemble of RCP 4.5 in the near century. The lowest mean is projected in the MIROC5_RCP 4.5 in all periods.

In the Porsuk subbasin, the only projection with an increasing trend in the 21st century is the CCSM4_RCP8.5 but the trend is not monotonic. All of the ensembles, all MIROC5 projections, and the CCSM4_RCP4.5 scenario have a decreasing trend. Furthermore, CCSM4_RCP4.5 and the ensemble of RCP4.5 denote a statistically significant decreasing trend. In all of the three segments of the future period, although the trends are either decreasing or increasing depending on the GCM and the RCP involved, there is no significant trend in the time series. From the Supplementary Material, Tables A1 to A4, it can be seen that the only projection with a lower mean as compared with the historical period is MIROC5_RCP4.5, though in the far century the ensemble of RCP 4.5 scenarios also projects a lower mean. Among all projections, the lowest mean in the Porsuk subbasin is projected by MIROC5_RCP4.5 in all periods.

The Mann–Kendall trend test detected a significant downward trend in the CCSM4_RCP4.5 and the ensemble average of RCP 4.5 scenarios in the Ankara subbasin in the 21st century. All of the projections except for CCSM4_RCP4.5 show a decreasing trend in the near century. However, none of these trends are statistically significant. Similarly, there is no significant upward or downward trend in the climate projections either in the mid- or far century. Although the trends are either increasing or decreasing in the mid-century, all of the projections indicate a decreasing trend in the far century. In the Supplementary Material, Tables A1–A4 show that the means of the CCSM4_RCP4.5, MIROC5_RCP4.5, MIROC5_RCP8.5, the ensemble of RCP 4.5s, the ensemble of RCP 8.5s and the ensemble of all projections are lower than the mean of the historical period both in the 21st century and the far century. In the near and mid-century, the results are similar to the 21st century and the far century with the only exception being that the ensemble of RCP 4.5 scenarios projects a higher mean in comparison with the historical period. The basin-average annual total precipitation resulting from the MIROC5_RCP4.5 scenario is the lowest among all other projections in all periods.

In the Kirmir subbasin, all the climate projections except for CCSM4_RCP8.5 have a decreasing trend in the 21st century. Moreover, three of these downward trends, i.e., CCSM4_RCP4.5, the ensemble of RCP 4.5s, and the ensemble of all projections, are statistically significant. The results in the near century are very similar to the Ankara subbasin. The only difference is that the downward trend in the MIROC5_RCP4.5 scenario is close to the monotonic trend. This is the only projection that has a statistically significant trend among all projections in all study subbasins in the near century. The number of climate projections with an upward trend in the mid-century is higher compared with the ones with a downward trend. However, none of them has statistically significant p-values. Finally, in the far century, only two of the projections have an increasing trend while the rest have a decreasing trend. Again, there is no monotonic trend. When the mean precipitation values of the historical and future periods are compared, it is seen that all the projections except for CCSM4_RCP4.5 project a lower mean than the historical period in the 21st century and the far century. In the near century, the mean values are lower in the CCMS4_RCP8.5, MIROC5_RCP4.5, the ensemble of RCP 4.5 scenarios, and the ensemble of all projections. In the mid-century, the results are similar to those of the near century and the only difference is that the MIROC5_RCP8.5 has also a lower mean as compared with the historical period in the mid-century.

In the Middle Sakarya subbasin, all the projections except for CCMS4_RCP8.5 have a decreasing trend in the 21st century. Furthermore, the ensemble average of RCP 4.5 scenarios shows a statistically significant downward trend. In the near and mid-century there is no statistically significant upward or downward trend. There are four downward trends in the near century while there are only two in the mid-century. In the far century, on the other hand, CCSM4_RCP.8.5 projection denotes a downward trend with a statistically significant p-value. There is no statistically significant evidence that a trend is present in the other climate projections in the far century. In the 21st century and mid-century, CCSM4_RCP8.5 and MIROC5_RCP4.5 projections have lower means as compared with the historical period. In the near century, the ensemble of RCP 8.5 scenarios is also projected to have a lower mean in addition to the CCSM4_RCP8.5 and MIROC5_RCP4.5 projections. Finally, in the far century, all of the climate projections except for CCSM4_RCP8.5 have lower means than the historical period. The lowest mean among all other projections is projected by MIROC5_RCP4.5 in the 21st century, mid- and far century. In the near century, CCSM4_RCP8.5 has the lowest mean as compared with the other projections.

In the Göksu subbasin, none of the climate projections has a significant trend in the 21st century or any segments of the future period. The most remarkable result is that all projections except MIROC5_RCP4.5 are in an increasing trend in the near and far centuries. Göksu is the only subbasin where almost all of the climate projections indicate an increasing trend in the near and far century but none of them are statistically significant. However, in the mid-century, basin-average annual total precipitation depths tend to decrease. In the 21st century, the only projection that has a lower mean than the historical period is MIROC5_RCP4.5. All of the climate projections in the near century project higher means in comparison with the historical mean. In the mid-century, CCSM4_RCP8.5 and MIROC5_RCP4.5 have lower means as compared with the historical period. In the far century, on the other hand, the means of all projections involving the RCP 4.5 scenario and the ensemble of all scenarios are lower than in the historical period.

Similar to the Göksu subbasin, the Lower Sakarya subbasin does not have any statistically significant p-value in any future period. Only two out of seven projections are on an upward trend in the 21st century. In the near and far century, however, the number of projections with an increasing trend is higher compared with the ones with a decreasing trend. Unlike the near and far century, most of the projections have a downward trend in the mid-century. In the 21st century and mid-century, MIROC5_RCP4.5 is the only projection that has a lower mean compared with the historical mean. In the near century, two of the projections, i.e., CCSM4_RCP8.5 and MIROC5_RCP4.5, project lower means than the historical period. Finally, in the far century, all MIROC5 projections indicate lower means than the mean of the historical precipitation. Moreover, MIROC5_RCP4.5 always projects the lowest mean as compared with the other projections in the Lower Sakarya subbasin.

The results show that all the climate projections with a statistically meaningful p-value point to a downward trend in the basin-average annual total precipitation depths over all study subbasins in the 21st century. There is no projection with an upward trend and with a significant p-value. Thus, the precipitation depths tend to decrease in all subbasins in the 21st century. Except for the Göksu and Lower Sakarya subbasins, there are at least one or more projections with statistically significant decreasing trends in all other subbasins. In addition, Göksu and Lower Sakarya have different trends compared with the other subbasins in all segments of the future period. The CCSM4_RCP8.5 projection, two MIROC5 projections and the ensemble averages of the RCP 8.5 scenarios do not have a significant decreasing or increasing trend based on the results of the Mann–Kendall trend test in any of the subbasins. The p-values of the CCSM4_RCP4.5 projections in the Porsuk, Ankara, and Kirmir subbasins indicate a significant decreasing trend. Similarly, the ensemble averages of the RCP 4.5 scenarios point out a significant decreasing trend in the basin-average annual total precipitation depths of the Upper Sakarya, Porsuk, Ankara, Kirmir, and Middle Sakarya subbasins.Finally, the p-values of the ensemble average of climate projection realizations show a significant decreasing trend only in the Kirmir subbasin. The Mann–Kendall trend test results show neither increasing nor decreasing trend in any of the projections in two subbasins: Göksu and Lower Sakarya.

In the near century, only one subbasin, i.e., Kirmir, has a projection with a statistically significant decreasing trend, which is MIROC5_RCP4.5. In all subbasins except for Porsuk, Göksu, and Lower Sakarya, the number of projections with a decreasing trend is higher than the number of projections with an increasing trend. In the mid-century, there are no projections with a statistically meaningful p-value. Except for Göksu and Lower Sakarya, the number of projections with an increasing trend is higher than the number of projections with a decreasing trend in all subbasins. Finally, in the far century, there is only one projection, i.e., CCSM4-RCP8.5, with a statistically significant decreasing trend, and this trend is projected in the Middle Sakarya subbasin.

Minimum and maximum values of basin-average temperature in the historical period, as well as in the 21st century and each segment of the future period, over the seven study subbasins with the mean value in parentheses are given in the Supplementary Material, Tables A5–A8, respectively. The statistical information about the historical period was provided for comparison purposes. In these tables, the mean value shows the average of the basin-scale temperatures for the related period. In this way, the changes in the basin-scale temperatures are analyzed by comparing the historical means with the means of each future-period segment. Additionally, in the Supplementary Material, Figure B2 in Appendix B presents box-and-whisker plots of average monthly temperatures for the ensemble average across each subbasin.

In the Upper Sakarya subbasin, all climate projection realizations show an increasing trend in the 21st century. Moreover, except for MIROC5_RCP4.5, all of these trends are statistically significant. When each 11-year future-period segment is analyzed individually, however, none of the trends have statistically meaningful p-values. It is also seen that there are climate projections with a decreasing trend. For instance, MIROC5 and the ensemble average of RCP 8.5 scenarios project a decreasing trend in the near century. In the mid-century, MIROC5_RCP4.5 is the only climate projection with a decreasing trend. Finally in the far century, MIROC5_RCP4.5, the ensemble of RCP 4.5 scenarios, and the ensemble of all GCMs have a decreasing trend. When the mean of the historical temperatures is compared with the mean of the future climate projections, it is seen that all future projections have higher means. The increase in the projected annual average temperatures ranges between 0.25 and 4.77 °C. The largest increase is projected by MIROC5_RCP8.5 in the far century while the lowest belongs to the CCSM4_RCP4.5 projection in the near century.

In the Porsuk subbasin, all climate projections point to rising temperatures for the entire 21st century. In addition, all these trends are statistically significant. In the near century, the only statistically significant increasing trend belongs to CCSM4_RCP4.5. Moreover, MIROC5_RCP8.5 and the ensemble average of RCP 8.5 scenarios show a decreasing trend. In the mid-century, there is no statistically meaningful trend. One of the climate projections, MIROC5_RCP8.5, points to a decreasing trend. Similar to the mid-century, none of the projections has a statistically important p-value in the far century. Three of these projections, namely CCSM4_RCP4.5, MIROC5_RCP4.5, and the ensemble of RCP 4.5 scenarios, have a decreasing trend. The average of the basin-scale temperatures for the historical period is always smaller than the future projections. The increase in the temperatures ranges between 0.24 and 4.86 °C. The lowest and the largest increase in the temperature is projected by CCSM4_RCP4.5 in the near century and MIROC5_RCP8.5 in the far century, respectively.

In the Ankara subbasin, temperatures are projected to increase in all climate projections for the entire 21st century. Except for MIROC5_RCP4.5, all climate projections have a statistically significant p-value. In the other future-period segments, however, none of the trends are statistically significant. In the near century, all MIROC5 projections show a decreasing trend while all other projections have an increasing trend. In the mid-century, the only projection with a decreasing trend is MIROC5_RCP8.5. Finally in the far century, MIROC5_RCP4.5 and the ensemble of RCP 4.5 scenarios have a decreasing trend. The mean values of all projections are always above the mean of the historical period. The range of the increase is between 0.20 and 4.78 °C, which are projected by CCSM4_RCP8.5 in the near century and by MIROC5_RCP8.5 in the far century, respectively.

In the Kirmir subbasin, the temperatures show an increasing trend throughout the 21st century. All of the climate projections except MIROC5_RCP4.5 have a trend line with a statistically significant p-value for the entire century. In the near, mid-, and far century, none of the projections has a statistically meaningful trend. In the near and mid-century, one of the projections, i.e., MIROC5_RCP8.5, has a decreasing trend. In the far century, CCSM4_RCP4.5, MIROC5_RCP4.5, and the ensemble of RCP 4.5 scenarios show a decreasing trend. The change in the average annual temperatures compared with the historical period ranges between −0.26 and 4.33 °C. The only decrease in the mean values is projected by MIROC5_RCP8.5 in the near century. The largest increase, on the other hand, also belongs to MIROC5_RCP8.5 but it is projected in the far century.

In the Middle Sakarya subbasin, all of the climate projections point to increasing temperatures in the 21st century. Furthermore, all of them except MIROC5_RCP4.5 have a statistically significant upward trend for the entire century. In the near century, MIROC5_RCP8.5 and the ensemble average of RCP 8.5 scenarios have downward trends. However, none of the trends are statistically important in this period. In the mid-century, there is only one projection with a significant p-value, which is CCSM4_RCP8.5. All of the projections except MIROC5_RCP8.5 have an upward trend. Finally, similar to the mid-century, there is no statistically significant trend in the far century. Three of the projections, i.e., CCSM4_RCP4.5, MIROC5_RCP4.5, and the ensemble of RCP 4.5 scenarios, have a downward trend. The mean values of the climate projections are always higher than the mean of the historical temperatures. The change in the mean values ranges between 0.22 and 4.89 °C. The lowest change belongs to CCSM4_RCP4.5 in the near century while the highest is in MIROC5_RCP8.5 in the far century.

In the Göksu subbasin, all climate projections show an upward trend in temperature with statistically significant p-values for the entire 21st century. Other 11-year future-period segments, however, do not have a statistically meaningful upward or downward trend. In the near century, MIROC5_RCP8.5 and the ensemble average of RCP 8.5 scenarios show a decreasing trend. In the mid-century, the only projection having a downward trend is MIROC5_RCP8.5. In the far century, the number of projections with a downward trend is higher as compared with the ones with an upward trend. Although there are projections with a downward trend, the mean values of these projections are always higher than the mean of the historical temperatures. The change ranges between 0.09 and 4.86 °C. The smallest and the largest increases belong to CCSM4_RCP4.5 in the near century and MIROC5_RCP8.5 in the far century, respectively.

In the Lower Sakarya subbasin, the basin-average temperature has an increasing trend for the entire 21st century. All of the trends except MIROC5_RCP4.5 are statistically significant. In the near, mid-, and far centuries, there is no statistically significant trend. In the near century, almost all of the projections except CCSM4_RCP4.5 and the ensemble average of RCP 4.5 scenarios have a decreasing trend. In the mid-century, however, MIROC5_RCP8.5 is the only projection with a downward trend. In the far century, CCSM4_RCP8.5, MIROC5_RCP8.5 and the ensemble of RCP 8.5 scenarios have an increasing trend, even though there are several projections with a downward trend, and the mean values of the projections in any segment of the future period are higher as compared with the historical period. The range of variation in basin-average temperatures is between 0.04 and 4.72 °C. As in almost all subbasins, the lowest change increase in the temperature is projected to be in CCSM4_RCP4.5 in the near century while the highest is in MIROC5_RCP8.5 in the far century.

The results show that for the most part the temperatures are increasing in all subbasins with significant p-values for the entire 21st century. In the near, mid-, and far century, although the trends are either downward or upward depending on the GCM and the RCP involved, the average of the basin-scale temperatures for all periods are always higher than the mean of the temperatures in the historical period. The only exception to that is Kirmir subbasin where the mean of the MIROC5_RCP8.5 projection in the near century is lower compared with the historical mean. Moreover, it is understood that the estimated increases in temperatures in RCP 8.5 scenarios are higher than in RCP 4.5 scenarios. This is most probably due to the fact that RCP 8.5 is a scenario of comparatively high GHG emissions while RCP 4.5 is an intermediate stabilization emission scenario. In addition, the lowest and highest increases in the temperatures in all subbasins are always projected by CCSM4_RCP4.5 and MIROC5_RCP8.5, respectively.

This section presents predictions for WEFE nexus sector vulnerabilities based on future climate changes and the selected indicators in the seven study subbasins. Table 6 presents the socio-economic indicator values and 21st-century climate trends for the study subbasins. Table 7 provides a summary of WEFE nexus sector vulnerability predictions and compares the seven study subbasins based on their vulnerabilities in terms of water, energy, food, and ecosystem sectors, which are represented by municipal water demand coverage, energy generation, agricultural productivity, and environmental flows, respectively. In the table, H, M, and L represent High, Moderate, and Low vulnerability, respectively. The following paragraphs detail the reasoning behind nexus sector vulnerability predictions for each subbasin.

Table 7

Foresight on WEFE nexus sector vulnerabilities based on future climatic changes in seven study subbasins

View Large

The Upper Sakarya subbasin has a high percentage of agricultural areas (69.4%) and a relatively low population, but its economy heavily relies on agriculture. Decreasing precipitation trends in the 21st century could put pressure on the already limited water resources, potentially leading to water scarcity and reduced crop yields. This scarcity could cascade across sectors: as agricultural productivity declines, economic pressures on farming communities could increase, and competition for water among sectors may intensify. Rising temperatures are expected to further drive water demand for both irrigation and cooling, compounding water scarcity and placing additional stress on farmers who depend on reliable water sources. Thus, agricultural productivity holds a high vulnerability rating in this subbasin. Municipal water demand coverage was assigned a moderate vulnerability level due to the absence of densely populated cities, suggesting current water supply infrastructure may suffice, though this could change with climate-driven pressures on the water supply. There is no hydroelectric power plant, and there is only one thermal power plant in the Upper Sakarya subbasin and therefore, energy generation has low vulnerability due to minimal dependence on large-scale water resources for energy production. However, the ecosystem sector exhibits moderate vulnerability due to agricultural activities. As environmental flows are likely to be impacted, especially during peak irrigation months, there is a risk of reduced water availability for natural habitats, potentially leading to ecological stress and loss of biodiversity.

The Porsuk subbasin has a moderate percentage of agricultural areas (50.4%) and a high population density, with an economy comparable to that of Upper Sakarya, also largely dependent on agriculture. Projected decreases in precipitation could result in water scarcity and reduced crop yields, directly affecting food security and impacting the local economy. High population density in Porsuk intensifies water demand, creating greater competition for limited resources, which may further strain water availability and amplify scarcity issues. Rising temperatures are expected to elevate water demand for household and industrial cooling, compounding pressures on the water sector and further stressing municipal supply systems. Additionally, reduced river flows could disrupt thermal power generation, posing potential risks to the energy sector and adding economic strain due to energy shortfalls. Agricultural productivity was assigned a moderate vulnerability level since the percentage of agricultural areas is moderate. However, municipal water demand coverage is expected to face high vulnerability due to the high population and, consequently, elevated water demand, indicating a substantial risk of water scarcity, especially during peak usage periods. While there is no hydroelectric generation in Porsuk, the presence of thermal power plants, some with dry-cooling systems, results in a moderate vulnerability rating for the energy sector. These plants are still susceptible to the impacts of reduced water availability on thermal power generation, as many rely on cooling water to maintain efficiency, meaning any decrease in river flow may have economic and operational consequences. Finally, the ecosystem sector has been assigned a high vulnerability rating due to the combined presence of agricultural and industrial activities. These activities may exert significant environmental pressures, potentially leading to ecological degradation and risks to biodiversity. Reduced environmental flows could compromise natural habitats, particularly during periods of increased water use for agriculture and industry.

The Ankara subbasin has the highest population density among the seven subbasins, with both agriculture and industry being key to its economy. As part of the province of Ankara, Türkiye's second most populous city and a major industrial hub within the Sakarya Basin, this subbasin plays a crucial role in regional resource demand. The nearby Kirmir subbasin serves as a vital drinking-water source, with water transfers from Kirmir's dams supplementing Ankara's water supply. However, persistent droughts have already limited Ankara's water resources, necessitating additional transfers from the Kızılırmak Basin. Decreasing precipitation trends could exacerbate water shortages, creating substantial pressure on agricultural productivity, municipal water supplies, and environmental flows. Rising temperatures would further intensify water demand for household and industrial cooling, compounding these pressures. The Ankara subbasin's vulnerability is underscored by its high reliance on agriculture, particularly in Polatlı, Türkiye's ‘granary’, justifying a high vulnerability rating for agricultural productivity. Municipal water demand coverage is also rated highly vulnerable due to the large population's drinking-water needs. While hydroelectric power is absent, the thermal power plants in the region use wet-cooling systems, leading to a moderate vulnerability rating in electricity generation. Intense agricultural and industrial activity also raises environmental flow vulnerability to a high level, as these sectors heavily rely on water resources. In contrast, the Kirmir subbasin shows lower vulnerability across WEFE sectors due to fewer high-impact stressors, though decreased precipitation and increased temperatures are expected to strain its water resources over time.

The Middle Sakarya subbasin stands out with its hydroelectric production. There are tens of both dam and river-type hydroelectric power plants located on the Sakarya River main branch. The decreasing trend in precipitation depths can cause a reduction in water availability for energy production, leading to a reduction in electricity generation. In the Middle Sakarya subbasin, agricultural productivity and municipal water demand coverage were assigned moderate vulnerability levels since the share of agricultural areas (31.6%) is moderate and the subbasin has a low population density in comparison with the other subbasins. Electricity generation is highly dependent on water availability in the subbasin, and hence it has a high vulnerability rating. Finally, the environmental flow was assigned a high vulnerability rating since the presence of numerous hydroelectric production facilities in the subbasin can have a significant impact, especially under the changing climate.

In the Göksu and Lower Sakarya subbasins, while there is no statistically significant trend in precipitation, rising temperatures are expected to place increasing pressure on water availability, especially during the growing season, potentially impacting both food production and energy generation. Higher temperatures may lead to elevated evapotranspiration rates, which could increase water demand for irrigation, thereby affecting agricultural productivity. In the energy sector, temperature-driven increases in demand for water in thermal power generation could exacerbate resource competition, especially during dry periods. These factors collectively contribute to a low-to-moderate vulnerability level for the WEFE nexus sectors in these subbasins.

In this study, we analyzed future trends in precipitation and temperature to estimate the vulnerabilities of the WEFE nexus sectors. This paper is novel in its integration of high-resolution climate projections, a multi-indicator evaluation method, and a subbasin-specific approach to assess the impacts of climate change on the vulnerability of the WEFE nexus in a semi-arid watershed. However, it is important to acknowledge that vulnerability levels may vary due to many other factors that were not considered in this study. For example, agricultural productivity could be influenced by soil type, irrigation methods, and crop type, which were not examined. The vulnerability level of the energy sector can be affected by various factors such as the choice of renewable energy, the preference for dry-cooling systems in thermal power plants, and the prevention of evaporation losses in hydroelectric power plants. Additionally, the resilience of the water sector to climate change can be improved through sustainable practices such as preventing loss and leakage in water supply systems and reusing treated wastewater. The implementation of these practices can also contribute to improving the environmental flow. However, it is important to note that a holistic approach that considers the interdependencies between the water, energy, food, and ecosystem sectors and the impacts of climate change is necessary for sustainable development and the long-term resilience of the regions.

This study has found a general trend of decreasing precipitation in most subbasins of the Sakarya Basin during the 21st century, with Göksu and Lower Sakarya showing different patterns. While some increases are projected in Porsuk, Ankara, and Kirmir, most subbasins are expected to experience statistically significant declines. Meanwhile, temperatures are projected to rise across all subbasins, with the largest increases expected under the RCP 8.5 scenario, particularly in the MIROC5_RCP8.5 model.

These results highlight that climate-change impacts can vary greatly even within the same basin, depending on each subbasin's location and topography. This means that it is important to study each subbasin separately, especially in areas with diverse climates and terrains. Additionally, the differences observed between various climate models suggest that using multiple models is crucial to getting a more reliable picture of future climate impacts.

Understanding these changes in precipitation and temperature is key to managing the WEFE nexus. A decrease in precipitation can lead to less water availability, affecting agriculture, energy production, and the health of ecosystems. Rising temperatures might further complicate these issues by increasing water and energy demand, lowering crop yields, and making water scarcity worse. This interconnectedness shows why it is important to include climate projections in WEFE nexus studies to spot potential risks and plan effective responses.

To tackle these challenges, it could be helpful to explore new water management practices, such as improving water-use efficiency, using crops that can withstand climate change, and investing in renewable energy sources that use less water. Also, working together with local stakeholders can lead to more effective and customized adaptation strategies.

The findings from this study will be used to calibrate water–energy system models for each subbasin in the Sakarya Basin. These models will help identify opportunities, challenges, and potential conflicts within the WEFE nexus, leading to the creation of tailored adaptation strategies. By focusing on the specific needs of each subbasin, this approach aims to ensure that the WEFE nexus remains sustainable and resilient as the climate continues to change.

While this study provides valuable insights into the climate-change impacts on the WEFE nexus in the Sakarya Basin, further advancements could enhance the robustness of such assessments. Our methodology currently utilizes two GCMs and two emission scenarios due to computational power and time constraints; however, increasing the number of GCMs and scenarios could further strengthen the analysis. Additionally, while we used CMIP5 data, given its availability for the study period, recent research is increasingly relying on CMIP6 projections. Future studies may consider CMIP6 data to provide even more up-to-date climate projections.

Furthermore, watershed characteristics in this study were quantified through specific indicators. For a more comprehensive vulnerability assessment, a deterministic model could be applied to simulate the watershed and evaluate water availability. We also based our analysis on historical socio-economic data under a ‘business-as-usual’ assumption, presuming stable socio-economic conditions. Future studies could benefit from examining how socio-economic factors might evolve under different scenarios, allowing for a more dynamic understanding of potential impacts on the WEFE nexus.

Overall, the findings of this study are expected to serve as a valuable guide for decision-makers involved in long-term planning within the Sakarya Basin and similar regions. By providing a detailed understanding of projected climate impacts on nexus sectors, these results offer a foundation for developing proactive, evidence-based strategies that address future water scarcity, agricultural productivity, energy generation, and ecosystem health. The subbasin-specific insights can inform adaptive management practices helping policymakers and stakeholders implement targeted interventions that enhance resilience across the WEFE nexus in the face of ongoing climate change.

We would like to acknowledge the Scientific and Technological Research Council of Türkiye (TÜBİTAK) for providing funding for the projects ‘Evaluation of Water, Energy and Food Nexus in Sakarya Watershed’ (project number 116Y166) and ‘A New Approach for Sustainable Water Management: Integration of the Circular Economy Approach to the Framework of Water–Energy–Food–Ecosystem Nexus’ (project number 121R014). We are also grateful to The Turkish Fulbright Commission for making this study possible with a visiting student researcher scholarship. Additionally, we would like to express our gratitude to Dr Ali Ercan for his invaluable guidance and support during this research at the UC Davis Hydrologic Research Lab.

This work was supported by the Scientific and Technological Research Council of Türkiye (TÜBİTAK) [grant numbers 116Y166 and 121R014].

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

The authors declare there is no conflict.