The water–energy–food (WEF) nexus as a tool to develop climate change adaptation strategies: a case study of the Buffalo River catchment, South Africa

Extreme weather events brought on by climate change, such as droughts and floods, have emerged as the most significant concerns confronting southern Africa's fast-growing and rising economies. Temperatures are projected to increase further, and rainfall patterns are anticipated to fluctuate significantly, thus increasing risk and uncertainty, especially in regions with low adaptive capacity (Mpandeli et al. 2018).

Climate change is increasing water stress and exacerbating hydrologic variability in South Africa (Nhemachena et al. 2020). Unlike the rest of the South African regions, the KwaZulu-Natal province is likely to be at risk from more extreme flooding events due to an anticipated increase in the intensity and frequency of rainfall (Graham et al. 2011; Zwane 2019). Moreover, through the hydrologic variability induced by climate change, water availability may be limited (Tabari 2020), and the degree of limitation is dependent on increased water consumption perpetuated by population growth (UNESCO 2015) and human interventions through land and water management (Ashraf et al. 2019). Therefore, it is essential to consider the compounding effects of climatic changes and anthropogenic drivers of water availability to improve water resource management (Ashraf et al. 2019).

Hydraulic infrastructures, like reservoirs and canals, may be a viable solution to strengthen water security. However, they require efficient operation and sustainable allocation strategies to accommodate the demands of various users (Wicaksono & Kang 2019). Sustainable water allocation strategies recognise safe drinking water for basic domestic needs, achieving food and energy security, supporting sustenance agriculture, and meeting the minimum ecosystem needs (Agarwal et al. 2018). Therefore, a transdisciplinary and transformative resource management approach, like the water–energy–food (WEF) nexus (Mabhaudhi et al. 2019), which seeks for an understanding of the linkages, dependencies, synergies, and trade-offs associated with WEF sectors in resource management, is ideal for the optimisation of water allocation (Hui et al. 2021).

The WEF nexus approach addresses the multifaceted and dynamic interrelationships between water, energy, and food systems (Nhamo et al. 2019). Due to its holistic approach to resource management (Wicaksono & Kang 2019), the WEF nexus approach can inform government policies for sustainable development and resource security under climate change (Nhamo et al. 2020). Norouzi & Kalantari (2020) utilised the WEF nexus approach in developing a governance model for the Iranian policymaking system. Implementing the WEF-based model highlighted that the water and food security issues in Iran result from governance shortages, which formed the base of Norouzi & Kalantari (2020) recommendations for the governance improvement process. Similarly, using the WEF nexus approach, Pardoe et al. (2018) assessed the effectiveness of Tanzania's National Adaptation Plan of Action (NAPA) in integrating climate change and specific adaptation strategies into policies and planning documents. The WEF nexus successfully illustrated that the progress brought upon by NAPA on WEF resources is individualistic, and the barriers to effective integration include the lack of cross-sectoral collaboration in practice (Pardoe et al. 2018).

The interconnections among WEF sectors can be conceptually described; however, the actual feedback connections are complex, often invisible, and affected by external factors (Wicaksono & Kang 2019). To interpret and quantify the WEF nexus, several analytical approaches can be applied in South Africa, such as the (a) WEF Nexus Tool 2.0, (b) MuSAISEM, (c) Climate, Land-Use, Energy and Water Strategies (CLEWS), and (d) ANEMI (Mabhaudhi et al. 2018). The CLEWS and the ANEMI models are analytical frameworks that explicitly address WEF resource management and climate change. In contrast to the ANEMI model, which is a single model that performs an inextricably linked assessment of the physical, ecological, and hydrological processes (Davies & Simonovic 2010; Mabhaudhi et al. 2018), CLEWS is a conceptual framework that incorporates analytical land, energy, and water models under various climatic scenarios and allows users to select their preferred analytical models for each WEF component (Welsch et al. 2014; Mabhaudhi et al. 2018; Ramos et al. 2020).

From a water perspective, the WEF nexus shifts from the current ‘silo’ approach of water resource management (Naidoo et al. 2021) and attempts to involve the energy and agricultural sectors in the analysis of the water issues so as to raise awareness of the interdependencies of energy, food, and water security (Shannak et al. 2018). Regions within South Africa, as a developing country, could benefit from this holistic approach as they experience significant trade-offs among WEF sectors, which are exacerbated by climate change (Senzanje et al. 2019).

In the context of a South African river basin, the Buffalo River catchment, which forms part of the uThukela Water Management Area in KwaZulu-Natal, has under-developed water supply infrastructure and consists of unreliable main water supplies, including the Buffalo River, Ngagane River, and Ntshingwayo Dam (Ngubane & Zwane 2019). As a result, this high rainfall receiving area in the KwaZulu-Natal province experiences water supply shortages, has underutilised agricultural potential, and relies heavily on rainfed agricultural produce (LGCCP 2018; Kunene 2019; Ngubane & Zwane 2019; Shabalala et al. 2020). The above-mentioned issues were exacerbated during the 2015 and 2016 drought period, which affected the catchment's livelihood and the ability to provide water to its numerous activities, including irrigation, power generation, domestic, mining, and bulk industries (uMgeni 2020). According to the uMgeni (2020) report, the current water supply schemes within the Buffalo River catchment are anticipated not to cater for their water demands by 2050 due to population growth.

In adapting to the existing water challenges and to alleviate prospective water concerns, Dlamini & Mostert (2019) stated that water allocation plans should be revised to minimise water distribution inequities in the future. An analysis of climate change impacts on surface water availability has been done by Dlamini et al. (2023), which projected precipitation and surface water availability increases in the Buffalo River catchment. However, it was flagged that unmet demands may increase if no changes in the current water allocation and development plans are made. We, therefore, deepen the existing research by applying the CLEWS framework to investigate the impacts of climate change and proposed policy interventions on the Buffalo River catchment's water system's reliability in supporting its future demands. The study used water supply performance indices to scrutinise the integrity of the water supply system and formulate recommended climate change adaptation strategies based on the results.

The CLEWS conceptual framework focuses on the analysis of interactions between climate, land, energy, and water systems, supported by quantitative studies of interactions and use of resources. Therefore, it is interdisciplinary (Ramos et al. 2020). It includes the use of publicly available tools such as the Long-range Energy Alternatives Planning (LEAP) model, Water Evaluation and Planning (WEAP) model, and Agro-Ecological Zoning (AEZ) model. It connects their inputs and outputs and analyses the results at an integrated WEF layer (Byers 2015).

This study was not designed with the modelling of the land-use system as a core component. However, the findings of a global analysis conducted by the Food and Agricultural Organization (FAO) and the International Institute of Applied Systems Analysis (IIASA) using the Agro-Ecological Zones land production planning model were employed. The FAO's global Agro-Ecological Zones (gAEZ) assesses natural resources for finding suitable agricultural land utilisation options (Fischer et al. 2021). Crop suitability and land productivity evaluation findings from the gAEZ are recorded in different databases, each arranged in 5 arc-minute grid cells (Fischer et al. 2021).

The results are organised into separate files (tables) by crop type, input level, water supply, climate scenario, and period. Each crop database includes sub-grid distribution data on suitable extents, potential production, water deficit, and fallow factors, with all data stored according to suitability classes (Fischer et al. 2021). For the Buffalo River catchment, data from simulations and projections of attainable yields and irrigation water requirements (IWR) were extracted and bias-corrected using the linear scaling method, which maintains the mean values of the observed variable (Ghimire et al. 2018). The bias correction process used the historical crop yields and IWR in Table 4.

Adaptation, in response to climate change, is reducing climate risks and vulnerability, mostly by adjustments to existing systems (Portner et al. 2022). Many adaptation options exist and are used to help manage projected climate change impacts, but their implementation depends upon the capacity and effectiveness of governance and decision-making processes (Portner et al. 2022). As such, this study presents a policy-oriented framework for analysing various climate change adaptation strategies for the 2100 horizon year.

Model-based projections of climate change impacts on water resources can differ largely. If this is the case, water resources managers and decision-makers need more confidence in an individual scenario for the future (Kundzewicz et al. 2018). Two alternative courses of action can be envisaged to overcome this uncertainty – the precautionary principle and the adaptive management. The former is a variation of the min–max concept, i.e., to choose the approach minimising the worst outcome. The latter is backed by the observation that, in light of the broad range of results for different climate impact scenarios, adaptive planning should be based on ensembles and multi-model probabilistic approaches (Kundzewicz et al. 2018). As such, for this study, the precautionary principle was utilised in analysing climate change scenarios, whereby the best- and worst-case climate change scenarios were analysed.

Climate projections are based on Representative Concentration Pathways (RCPs), defined in the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC) (IPCC 2014). By considering a set of global climate scenarios regarding greenhouse gas emissions, RCPs provide information on possible development trajectories for the forcing agents of climate change. For analysing and comparing the best- and worst-case scenarios of climate change, two RCP scenarios were considered:

• (a)

  The RCP4.5 scenario, i.e., best-case scenario, is a stabilisation scenario, whereby ‘emissions peak around 2040 then decrease and total radiative forcing reaches 540 ppm by 2100 before levelling off’ (Wayne 2013; Doulabian et al. 2021).

• (b)

  The RCP8.5 scenario, i.e., the worst-case scenario, is a high emission baseline scenario in which ‘emissions rise steadily over the 21^st century, reaching 940 ppm by 2100, and continue rising for another 100 years’ (Vuuren et al. 2011; Doulabian et al. 2021).

This study did not include the development of external climate models. Rather, precipitation estimates for both RCP4.5 and RCP8.5 scenarios were extracted using the Google Earth Engine from the NASA Earth Exchange Global Daily Downscaled Climate projections dataset (NEX-GDDP) Thrasher et al. (2012). The data mining process is detailed in Dlamini et al. (2023).

Different governmental agencies develop various water supply policies and development plans (Nasrollahi et al. 2021). For the Buffalo River catchment, water management recommendations made by uMgeni Water, a South African state-owned water resources management organisation, as well as those stated in the Buffalo River catchment's local municipalities' development plans, were considered and are largely focused on increasing the catchment's total water supply capacity, as seen in Table 5. It is important to note that uMgeni Water does not currently operate infrastructure in the Buffalo River catchment; however, their recommendations are based on projections of population water demands from 2020 to 2050 (uMgeni 2020).

In analysing the WEAP and LEAP models' output data, descriptive statistics such as the mean, percent increase relative to the historical scenario, coefficients of variation, and box and whisker plots were employed. Box and whisker plots were used to show a dataset's stability and general distribution of its variables. The variations of the minimum and maximum variables are displayed by the heights of the box plots, with the median value shown by the line in the middle of the box plots and outliers indicated by the dots above and/or below the box plots.

In evaluating correlation and significance in differences of datasets modelled under various scenarios, the statistical parametric Welch test and F-test, and the Mann–Whitney non-parametric t-test, were employed (Sayekti et al. 2022) after testing for normality using the Shapiro–Wilks test (Khatun 2021). The parametric and non-parametric tests were done with the significance level (α) set at 0.05 so that the null hypothesis (no significant difference) is rejected if the P-value of the dataset is less than 0.05.

Two performance indices were utilised in assessing the Buffalo River catchment's water supply system's performance and analysing the system differences under each climate and policy scenario. These are demand site coverage (Dcov) and reliability (RE) performance indicators. The demand site coverage is the percent of each demand site's water requirement that is met, from 0% (no water delivered) to 100% (delivery of full requirement) (Sieber 2015; Nivesh et al. 2023). The reliability index (RE) is defined as the probability that the water resources system can provide sufficient water supply to meet demands during the entire simulation period (Hashimoto et al. 1982; Sieber 2015; Al-Juaidi & Al-Shotairy 2020). Essentially, it is the percent of timesteps in which a demand site's demand was fully satisfied (Sieber 2015).

As stated in the previous sections, the current study builds on the Buffalo River catchment's WEAP model run performed in the Dlamini et al. (2023) study. Precipitation and evapotranspiration simulations and projections from the (Dlamini et al. 2023) study were computed using the same methodology and datasets as the current study's BAU and PS scenarios, and the results are summarised in this study's Section 2.3, hence they will not be elaborated in this section of the manuscript. However, the outcomes of surface runoff, water availability, and the performance of the water supply system will be compared in the following section under the BAU and PS scenarios, to gauge or assess the degree of reliability of the water supply system in meeting projected water demands under climate change, and possible interventions that could aid in the optimisation of the PS scenarios in improving reliability and allocation of water supplies.

As per the increases in precipitation variations and decreases in evapotranspiration projected under both climate scenarios in the Buffalo River catchment (Dlamini et al. 2023), the current study findings also projected increased volumes of water exiting the catchment via surface runoff (R), as seen in Table 6, especially in the RCP8.5 far future, with 16 and 14.5% increases projected under the BAU and PS scenarios, respectively. Graham et al. (2011) established comparable results by projecting increased R under climate change using regionally downscaled precipitation projections in the Thukela River basin, of which the Buffalo River catchment is a sub-catchment. However, when comparing the PS and BAU scenarios outcomes, the PS produced lower R, which is consistent with the increased water supply capacity under the PS scenario that allows for more volumes of precipitation to be captured.

Under the RCP4.5 scenario, the PS scenario improved irrigation provisions by 1.2%, and under the RCP8.5 scenario, by 3%. For the domestic sector, the quantity of water allocated to it is expected to increase by 0.5% under the PS RCP4.5 scenario, and 3% under the PS RCP8.5 scenario. Also, for this case, the projected surface water availability declines under the PS can be credited to the anticipated increase in water provision for the domestic sector.

An assessment of significant differences in the BAU and PS scenarios was carried out to investigate if the water provision changes made by PS are substantial, and the results are tabulated in Table 7. Under RCP4.5, both parametric tests indicated no significant difference between the unmet demands of the policy scenarios. However, a conflict of results was produced under RCP8.5. As such, the modelled water supply delivered under both policy scenarios in RCP8.5 for the individual domestic, energy, and irrigation sectors was tested for significance. The results listed in Table 7 indicate no significant difference. Henceforth, the changes imposed by the PS scenario are statistically insignificant. This is due to the water strategies' emphasis on increasing water storage capacity, with minimal focus on water allocation changes among demand sites.

PS strategies, when compared to the BAU results, are anticipated to slightly improve the Dcov for the high-priority demand sites (Newcastle and Dannhauser local municipalities), especially Dannhauser, whereby 5, 7 and 8% increases in Dcov are noted in the near-, mid-, and far-future timeframes, respectively. Dcov improvements by 27% are also expected, as a result of PS, in the Utrecht local municipality, this being attributed to the modelled increases in extractions of the Utrecht WTP from the Dorps Dam in the near- and mid-future strategies. However, the PS scenario proved unfavourable in the Nquthu local municipality as an average difference of −1% is calculated when comparing the PS scenario Dcov projections with those of the BAU scenario.

The PS scenario strategies proved beneficial for the Dannhauser local municipality by increasing its RE value by 73%. This is understood to result from increased extractions of the Biggarsberg WTP from Tayside, which directly supplies the Dannhauser local municipality. The Newcastle local municipality's RE doubled, from an average of 20% in the BAU to 40% in the PS. This is ascribable to the proposed Ncandu Dam, which will increase the supply delivered to Newcastle in the far future.

The Nquthu and Utrecht local municipalities' RE remains at 0% under all scenarios. This unreliability is expected as the projected Dcov values for these low-priority regions are under 20%, leaving approximately 80% of the demands unsatisfied annually. The Nquthu local municipality is the highest producer of agricultural produce and requires approximately 29% of the total IWR. The −64% difference in IWR and water supplied for irrigation is largely due to this very low water allocation to the municipality.

In the optimisation of PS, as summarised in Table 8, the focus was mainly on adapting to climatic changes expected under RCP8.5 and increasing supply in low-priority regions, which, as previously established, are the Nquthu and Utrecht local municipalities. For strategies involving increasing water abstractions from reservoirs, an assumed 50% of streamflow remained in the system for consumption losses (CL), which includes 30% for environmental flow release (Hughes & Mallory 2008), and 20% for uncertainty losses (Hughes & Mantel 2010).

For the short- to medium-term strategies in the PS scenario, which cover the near- and mid-future periods of this study, the following changes are made:

• (a)

  Water abstractions from the Utrecht WTP to the Utrecht local municipality were increased by 2 Ml/day, thus enabling the Utrecht WTP to supply its full capacity of 4 Ml/day.

• (b)

  The Dannhauser local municipality is currently allocated 40 Mm³ more than its maximum water requirements per annum projected in 2099. Therefore, as a strategy to increase water allocated to the Utrecht local municipality, the 33 Ml/day water supply from the Ngagane WTP to Dannhauser was decommissioned, and an additional 20 Ml/day was redirected from the Ngagane WTP to the Utrecht local municipality.

• (c)

  To accommodate the 35 Ml/day water inflow losses to Dannhauser local municipality resulting from decommissioning Dannhauser WTP and cutting off the supply from Ngagane WTP, the Durnacol WTP's operational capacity was expanded from 3.5 to 5.5 Ml/day, as well as the Biggarsberg WTP's operational capacity, from 16 to 30 Ml/day.

Strategies for adapting to climate change ought to be premised on a comprehensive understanding of the dynamic interactions and processes that occur within water resource systems, considering the system's socioeconomic and environmental aspects and its hydrological characteristics. As part of this study's efforts to assess climate change impacts and existing policy interventions on the Buffalo River catchment's water system's performance, with the intent of designing improved adaptation strategies, the WEF nexus' CLEWS modelling framework assisted in the quantitative exploration of interactions between water, energy, and food systems, as well as climate change. The WEAP, LEAP, and gAEZ analytical models and assessments provided insight into what could transpire once the water resources development plans are fully implemented under climate change conditions throughout the 21st century.

The findings suggest that, despite the highly anticipated increases in surface water availability in the Buffalo River catchment, water distribution inequalities persist due to the catchment's water allocation plans that are primarily focused on supplying water to the domestic sector of high-priority regions. As a result, there is a significant negative trade-off with water provision to low-priority regions as evidenced by unreliable water supplies to meet their domestic and irrigation requirements. Given the abundance of projected potential for irrigated crop production in these low-priority regions, especially in the Nquthu local municipality, this trade-off is also anticipated to stifle agricultural growth, thus potentially jeopardising the catchment's socioeconomic growth trajectory.

The optimised policy strategies established in this study for climate change adaptation focused on shifting water allocations and expanding existing water infrastructure to accommodate the water demands of low-priority regions under anticipated worsened climatic changes. The developed climate change strategies not only increased surface water availability, but also improved equality in water distribution among sectors, noted by the increased demand for site coverage and reliability of the water system in providing water demands to all local municipalities. As such, it is recommended that policymakers adapt the specifications of optimised policy strategies that correspond to the goals of the proposed policy strategies without design specifications and consider the re-allocation plans proposed by the optimised policy strategies developed in this research. The vast majority of optimised policy strategies necessitate the rehabilitation of transmission pipelines and the construction of reservoirs. Therefore, it is also proposed that detailed feasibility and technical studies be conducted to investigate the practicality of the optimised strategies.

The WEAP model's accuracy depends on the amount of available information and its degree of detailedness. As such, since the analysis of historical precipitation changes was made based on data and information gathered from gridded climatic data, using physically recorded rainfall data, preferably at a quinary scale, is highly recommended. This research is based on the Representative Concentration Pathway scenarios derived from the fifth Coupled Model Intercomparison Project (CMIP5). Future climate change research is encouraged to utilise climate change simulations from the latest project, CMIP6, and to include other components of the water balance, such as groundwater and changes in soil moisture, as well as the inclusion of all dams within the catchment, for more accurate estimates of hydrological changes. Nevertheless, the outcomes from this study can still be used for comparison purposes, as the calibration and validation statistics performed using the WEAP streamflow outputs indicate that the model sufficiently simulated the Buffalo River catchment's hydrology.

The first author is thankful to the Water Research Commission (Project WRC K5/2967//4), the National Research Fund (NRF) and the Nurturing Emerging Scholars Programme (NESP) for financial support. This work forms part of the Sustainable and Healthy Food Systems (SHEFS) Programme, supported through the Welcome Trust's Our Planet, Our Health Programme [Grant number: 205200/Z/16/Z]. This work was also carried out as part of the Nexus Gains Initiative, which is grateful for the support of CGIAR Trust Fund contributors: www.cgiar.org/funders.

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

The authors declare there is no conflict.