1 Introduction

Water, energy and food are three essential resources that human beings depend upon for survival and development. These three resources are interconnected in complex ways (Liang et al. 2020), necessitating a holistic approach to their evaluation. The Water-Energy-Food nexus concept was formally introduced in the Bonn 2011 Nexus Conference (Hoff 2011) as an integrated system that encapsulates the interdependencies between water, energy and food (Conway et al. 2015; Scanlon et al. 2017).

The global temperature is likely to increase to 4.8℃ by 2100 compared to 1995–2014 in the high-emission scenario (IPCC 2021). The increased temperature may lead to more frequent extreme and compound events such as heatwaves and long-term droughts, which could significantly affect constrain food production and energy generation. Climate change may pose great uncertainties and risks to water security, energy security and food security in the future. Therefore, understanding climate change impacts on water, energy and food is crucial for achieving the SDG6 (clean water and sanitation), SDG7 (affordable and clean energy) and SDG2 (zero hunger) (Liu et al. 2018; UN 2018).

Significant efforts have been made to explore and evaluate the Water-Energy-Food nexus via various approaches (de Amorim et al. 2018; D’Odorico et al. 2018; Endo et al. 2020). Mannan et al. (2018) identified the capabilities and hindrances of applying the Life Cycle assessment on Water-Energy-Food nexus. Zhang et al. (2018) discussed the pros and cost of eight quantitative methods. Albrecht et al. (2018) emphasised the importance of integrating quantitative and qualitative methods with social science in studies that incorporate multiple methods. Zhang et al. (2019a) categorised eleven existing nexus methods and tools into three types according to research purposes. Namany et al. (2019) introduced three dynamic decision-making tools and proposed application examples during three decision-making process stages.

However, none of those reviews has analysed the methods and tools used for investigating climate change impacts to water, energy, and food. Several questions remain unanswered: What are existing state-of-the-art analytical methods and tools for studying Water-Energy-Food under climate change? Which ones are more widely used? What are their features and limitations? What is the focus of related research on different spatial scales and topics? How should models be selected according to research topic and spatial scale? How does the related research consider climate change scenarios? What are future prospects?

To address these questions, we have reviewed and analysed research articles published over the past seven years that investigated climate change impacts on Water-Food, Water-Energy and Water-Energy-Food nexus. Promising methods frequently used in each type of study, topics and models for different spatial scales, and climate change scenarios setting methods are identified and discussed. The research challenges and limitations are identified, suggesting potential directions for future research in the domain of Water-Energy-Food interactions under climate change.

2 Methods

We searched peer-reviewed journal articles on the subject of climate change in the Web of Science database that were published after 2017 and related to Water-Food, Water-Energy and Water-Energy-Food. The article selection procedure followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA, Page et al. 2021), and the flow chart of the selection process is shown in Fig. 1.

Fig. 1
figure 1

Flow chart of articles selection process following PRISMA

Full size image

2.1 Review Process

In Web of Science, we used ‘climate change’, ‘water’, ‘food’ and’irrigation’ as the keywords to search the abstract, title and keywords of publications between 2017 and 2022, 1676 articles were found. By replacing the keywords with ‘climate change’, ‘soil moisture’ and ‘crop yield’, 1155 articles were selected; when searching keywords ‘climate change impacts on water and energy’, 5728 articles were found; and 2415 articles were listed through keywords ‘climate change’, ‘water’, ‘energy’ and ‘food’. Altogether, 10,974 articles were selected after the initial search.

2.2 Inclusion and Exclusion Criteria

Articles that met all the following criteria were selected: (1) they contain Water-Food, Water-Energy or Water-Energy-Food, but do not consider water, energy or food separately; (2) the consider climate change impacts; and (3) they use quantitative analytical tools or models for assessment. Besides, eleven articles published in 2023 were added during the revision of the paper.

Based on these criteria, we obtained 45 articles related to Water-Food and climate change, identified 64 articles studying climate change impacts on Water-Energy, and selected 36 articles about climate change and Water-Energy-Food nexus. Thus, a total of 145 articles were identified as suitable.

2.3 Information Extraction

After full-text reading of the 145 relevant articles, we extracted the following information: (1) the purpose of the study (2) the scale of study area; (3) methods used in the study area, based on statistical methods, physics-based modelling, supervised learning or operation optimisation; (4) whether the article used models, combined multiple models or used a single model; (5) whether the article claimed a new method; (6) whether there was a simulation under future scenarios and how the scenarios were set up; and (7) characteristics, major challenges and limitations in the application of methods and models.

3 Diversity of Nexus Methods

Numerous and diverse methods have been used or proposed for evaluating climate change impacts on Water-Food, Water-Energy and Water-Energy-Food nexus. while some studies have combined multiple methods. In this review, we divided the approaches into four categories: Statistical methods, Physics-based modelling, Supervised learning and Operation optimisation.

Methods based on statistics such as formula calculations, regression and statistical tests were grouped into Statistical methods; methods using models based on the representation of physical mechanisms were grouped into Physics-based modelling; methods using machine learning to do simulation were classified into Supervised learning; and methods using optimisation algorithm to determine optimal solution under constraints were classified into Operation optimisation.

3.1 Water-Food Nexus Analysis Methods

In the Water-Food nexus research, 53% of studies (24 of 45) used statistical methods, 64% (29 of 45) used physics-based modelling methods and 9% (4 of 45) used operation optimisation methods. We tabulate and categorise Water-Food analytical methods and tools from the selected 45 articles based on method categorisation and discipline in Table 1.

Table 1 Catalogue of methods used in the climate change impacts on Water-Food studies sample set, categorised by Physics-based modelling (a), Statistical methods (b) and Operation optimisation (c)
Full size table

Studies investigating the influence of climate change on Water-Food nexus mainly focused on how climate change may impact water availability for irrigation, soil moisture and crop yield. Most Water-Food studies utilised multiple methods or coupled models from different disciplines. The input data of the models were commonly multidisciplinary from the areas of meteorology (precipitation, temperature, wind speed, humidity, solar radiation, etc.), environment (CO2 concentration, etc.), geospatial (vegetation, landuse, etc.), economics (GDP, etc.) and society (population, etc.).

Generally, in model coupling studies, a hydrological model was used to simulate runoff or soil moisture, and an agricultural model or a statistical calculation method was used to calculate irrigation water requirements and parameters related to crop yields. For example, Akoko et al. (2020) used Soil & Water Assessment Tool (SWAT) to estimate the current and future water resources availability and Cropwat to calculate irrigation water requirements. Meanwhile, many interdisciplinary models were developed to study the Water-Food nexus. Blanc et al. (2017) integrated water resources model and a crop yield reduction module into the Integrated Global System Modelling framework (IGSM) to assess the influence of climate and socioeconomic changes on irrigation water availability as well as subsequent impacts on crop yields by 2050. Malek et al. (2018) integrated a process-based irrigation module into VICCropSyst to assess the influence of climate change on irrigation losses.

Some model-based studies only utilised a single discipline model or a series of equations. This kind of research mostly focused on soil moisture, irrigation and crop parameters rather than simulations of water availability. He et al. (2021) projected the amount of agricultural water for food production during peak population period (2029–2033) based on a series of equations including FAO’s Penman–Monteith equation. Jha et al. (2020) utilised DSSAT to project the changes in rice yield, water demand and phenological growth due to climate change.

A small number of studies did not use or depend on models but used statistical methods to analyse climate change impacts on Water-Food nexus. Zhang et al. (2020) distinguished the different effects of climate change, planted area crop mix on irrigation water demand based on the LMDI method. Kirby et al. (2017) analysed the historical trends of water use, crop production, food availability and population growth, and project them forward to 2050. Madadgar et al. (2017) developed a multivariate probabilistic model to estimate the probability distribution of crop yields under projected climate conditions.

Operation optimisation methods were less used in the climate change impacts on Water-Food studies, in which more studies using nonlinear optimisation framework (Mitchell et al. 2017; Gohar and Cashman 2018). These studies optimised water or food related strategies under climate change conditions.

3.2 Water-Energy Nexus Analysis Methods

In the Water-Energy nexus research, 33% of studies (21 of 64) used statistical methods, 92% (59 of 64) used physics-based modelling methods, 25% (16 of 64) used operation optimisation methods, and 9% (6 of 64) used supervised learning methods. Most studies utilised models and hydrological models accounted for a large part. The catalogue of Water-Energy analytical methods and tools from the selected 64 articles are tabulated and categorised based on method categorization and discipline in Table 2.

Table 2 Catalogue of methods used in the climate change impacts on Water-Energy studies sample set, categorised by Physics-based modelling (a), Statistical methods (b) and Supervised learning (c) and Operation optimisation (d)
Full size table

Studies analysing climate change impacts on Water-Energy mainly focused on hydropower. Hydropower is vulnerable to the impacts of climate change due to its direct dependence on the timing and magnitude of streamflow. Most studies projected future hydropower generation to evaluate how climate change will affect energy production, or optimised the reservoir operation schemes to minimise the impacts of climate change on streamflow. Generally, hydrological models or supervised learning were adopted to simulate and project future streamflow/inflow to hydropower reservoirs, then energy models or equations were employed to calculate potential hydropower generation, optimisation algorithms or water management models were employed to solve the optimal reservoir operation. For example, Rahmati et al. (2021) simulated future runoff with Artificial Neural Network (ANN) and optimised hydropower generation by Grasshopper Optimisation Algorithm (GOA). Guo et al. (2021) used Genetic Algorithm (GA) to solve the robust optimisation model with the inflows simulated by SWAT under climate and land use change scenarios in the future. Anghileri et al. (2018) contributed a modelling framework combining hydrological modelling, hydropower modelling and multi-objective optimisation technique to assess climate change and energy policies impacts on the operations of hydropower reservoir systems in the Alps.

Meanwhile, a small number of studies used interdisciplinary models to study Water-Energy nexus. Miara et al. (2017) simulated river discharge and temperature as well as electricity generation under climate change using the coupled Water Balance Model and Thermoelectric Power and Thermal Pollution Model (WBM TP2M). Graham et al. (2020) utilised Global Change Assessment Model (GCAM) to investigate the relative contributions of climate and human systems on water scarcity regionally and globally.

3.3 Water-Energy-Food Nexus Analysis Methods

There are relatively fewer selected studies about evaluating climate change impacts on Water-Energy-Food nexus. In the selected research, 36% of studies (13 of 36) used statistical methods, 78% (28 of 36) used physics-based modelling methods, 36% (13 of 36) used operation optimisation methods, and 6% (2 of 36) used supervised learning. Most studies evaluated climate change impacts on Water-Energy-Food through physics-based modelling, among them most utilised interdisciplinary models with the proportion of 44%. Besides, many studies in Water-Energy-Food utilised operation optimisation method. The catalogue of Water-Energy-Food analytical methods and tools from the selected 36 articles are tabulated and categorised in Table 3.

Table 3 Catalogue of methods used in the climate change impacts on Water-Energy-Food studies sample set, categorised by Physics-based modelling (a), Statistical methods (b) and Supervised learning (c) and Operation optimisation (d)
Full size table

Studies investigating climate change impacts on Water-Energy-Food nexus can be divided into three categories: (1) simulations of future Water-Energy-Food nexus change under future climate change scenarios, (2) optimal management options for mitigating future climate change impacts, (3) historical attribution or trend analysis of climate change impacts.

For future simulation research, studies generally utilised interdisciplinary models or coupled different models from multiple disciplines. Sridhar et al. (2021) presented an integrated modelling framework combining Variable Infiltration Capacity (VIC) and System Dynamics (SD) model for analysing the impacts of future climate on irrigation, hydropower, and other supply and demand through a feedback loop. Yang et al. (2018) adopted a two-way coupled agent-based model (ABM-SWAT) to evaluate the water availability for irrigation, hydropower generation, and riverine ecosystem health under joint effect of climate change and water infrastructure development.

Operation optimisation research mainly focused on addressing complex contradictions of Water-Energy-Food nexus to find an optimal strategy and aid sustainable development. Optimisation programming was used in this kind of research, sometimes combining physics-based modelling, supervised learning or statistical methods. Yuan et al. (2018) integrated Life Cycle Assessment (LCA) and linear programming to assess the feasibility of bioenergy and consider future circumstances under climate change impacts. Giuliani et al. (2022) combined HBV hydrological model, ANN and evolutionary multi-objective direct policy search method to demonstrate how local dynamics across Water-Energy-Food systems are impacted by climate change mitigation policies.

Research focused on historical trend utilised statistical methods to analyse datasets. Barik et al. (2017) investigated the Water-Energy-Food nexus in India under drought and monsoon rainfall in the last few decades based on GLDAS and GRACE data.

4 Studies for Different Spatial Scales

According to spatial scale of study area, these selected studies were categorised into large scale studies and small to middle scale studies. The research objectives and typical models of each scale and topic of research are summarised in Table 4. The selected studies related to Water-Food and Water-Energy occupy a higher proportion, while research on water-energy-food is relatively less. This is because the research on this topic involves more interdisciplinary science, so related research is not easy to conduct. Besides there were fewer studies of water-energy-food at global and national scales than at regional scales, because the larger the study area, the more complex the water-energy-food nexus is, the access to the required data also becomes more difficult.

Table 4 Summary of research objectives and typical models of selected studies categorised by research scales and themes. The number of each type of study in different scales, total number of each type of study and the percentage is summarised in the column of ‘Themes’
Full size table

Water-Energy-Food studies at the global scale focused on the competition for water between energy and food. The interlinkages between water, energy and food sectors were explored more in small to middle scale studies. Meanwhile, there were many more optimisations and policy scenarios in research at the regional scale. The Water-Energy research at large scale mainly focused on the impacts of river flow on potential hydropower production, while the optimisation of hydropower system operation is also concerned at small-middle scale research. The focuses of Water-Food research at large and small-middle scale were similar, mainly investigating the water management and crop production strategies. Local water, energy and food management strategies may not be applicable to other regions, so the trade-off between water, energy and food under climate change and the strategies for sustainable development at a larger scale still require continuous efforts from the academic community.

The findings on the global scale may inform future research at different scales. For Water-Energy research, roughly 65% of the world’s current hydropower capacity will face vulnerabilities due to alterations in the 1-in-100-year river flow pattern (Paltán et al. 2021), the most prominent encompassing Europe, northern Africa, the Middle East, and North America (Turner et al. 2017; Paltán et al. 2021). Pursuing a 1.5 °C warming target would mitigate these risks when contrasted with a 2.0 °C scenario (Paltán et al. 2021). For Water-Food research, a projected food deficit might occur prior to 2050 in the scenario of the worst-case climate change, significant water shortages stemming from irrigation in major food-producing nations will hinder these countries from satisfying their domestic food needs (Grafton et al. 2017). An expansion of irrigated land by 100 Mha would be necessary to double food production to meet the projected global food demands by 2050, and an additional 10% to 20% of trade flow will be required, directing water-abundant regions toward water-scarce regions, to maintain environmental flow requirements (Pastor et al. 2019). Expanding irrigation will lead to increased maize production in Europe, but the same cannot be said for rice production in East Asia (Okada et al. 2018). In a scenario of 3 °C warming, a "soft-path" approach with small water storage and deficit irrigation can extend irrigated land by 70 Mha and feed additional 300 million people worldwide, a "hard-path" with substantial annual water storage has the potential to expand irrigation up to 350 Mha and feed 1.4 billion more people (Rosa et al. 2020). The regions that heavily rely on snowmelt as an agricultural water resource are high-mountain regions like the Tibetan Plateau, Central Asia, western Russia, the western United States, and the southern Andes (Qin et al. 2020b). For Water-Energy-Food research, water scarcity reductions driven by human is likely to result in 44% of land area in the world by the end of twenty-first century in certain socioeconomic scenarios (Graham et al. 2020). An additional 1.7 billion people could potentially experience severe water shortages for electricity, industrial use, and household needs if priority becomes for food production and maintaining environmental flow (de Vos et al. 2021).

4.1 Models for Medium to Small Spatial Scale

Watershed hydrological models can be categorised into three types: (1) conceptual models, (2) physics-based models, and (3) data driven models. Conceptual hydrological models are based on physical basis but are in highly simplified forms, they also have the characteristics of statistical regression models (e.g., HBV and Xinanjiang model). The biggest limitation is that they treat the watershed as a whole, ignoring the heterogeneity of spatially distributed watershed characteristic parameters (Devia et al. 2015). Physics-based models adopt spatially varied parameters to reflect the physical mechanism of hydrological process influenced by multiple factors (e.g., SWAT and HEC-HMS). The data-driven models establish statistical relationships between input and output. They work well at the data range, but the simulation performance degrades when applied to epitaxial projection because of the lack of physical basis. Over the past decade, a cutting-edge machine learning methodology, named deep learning, has evolved from the traditional neural network and has outperformed traditional models with considerable improvement in performance (Yuan et al. 2020). However, deep learning cannot completely replace the physics-based models, and the combination of physics-based models and deep learning may open a promising door (Yuan et al. 2020).

Water management models aim to optimise water allocation to fulfil the demands from multiple sectors. Many selected studies established optimisation frameworks for planning and management of water resources. There were some studies employing existing water management models directly, among which WEAP and HEC-ResSim were most used. Crop models are used to simulate crop growth, DSSAT and CropSyst are typical and most common used crop models in selected studies. TIMES and LEAP were relatively frequently employed in investigating climate change impacts on water and energy. Interdisciplinary models like ABM, SD model and WEF Nexus Tool 2.0 were utilised in the Water-Energy-Food nexus studies.

4.2 Models for Large Spatial Scale

Global hydrological models consider more land surface processes like recycling of evapotranspiration. The approach integrates knowledge from multiple disciplines encompassing atmospheric sciences, geography, ecology, oceanography, soil science, global change science, etc. All global hydrological models run in a grid format (Sood and Smakhtin 2015). Typical global-scale hydrological models used in selected studies include WBM, VIC, WaterGAP and LPJmL. Different models have different emphases and characteristics. For example, WaterGAP model is more detailed in water demand simulation including water use for domestic, industry, thermal power production, livestock and irrigation (Döll et al. 2003). LPJmL model puts more emphasis on vegetation and crop simulations (Bondeau et al. 2007). High degree of uncertainty and rough resolution are main limitations of global hydrological models (Sood and Smakhtin 2015).

In selected global-scale studies investigating climate change impacts on Water-Food, GCWM and GFWS were utilised. GCWM mainly focuses on blue and green consumptive water use as well as virtual water of 23 specific crops (Siebert and Döll 2010). GFWS puts more attention on global food and irrigated water availability risks through simulation of food generation and demand, water supply and agricultural water requirement (Grafton et al. 2015).

Integrated Assessment Models (IAMs) are important tools to evaluate human feedback and impacts on climate change and mitigation of greenhouse gases (Schwanitz 2013). The IGSM framework consists primarily of two interacting components (Sokolov et al. 2018): the Economic Projection and Policy Analysis model and the Earth System model. GCAM links water, energy, landuse, earth systems and economics to analyse consequences of policy strategies and interdependencies. IMAGE simulates interactions between biosphere, society and the climate system to assess environmental and sustainable development issues. Delimitation of the system, explanatory power of models, as well as linkage of model evaluation and usefulness are the main challenges for IAMs (Schwanitz 2013).

5 Future Climate Scenarios Setting Methods

Most selected research assessed and projected future water, energy and food systems based on future climate change models. The emission scenarios, climate models, downscaling methods and global warming scenarios in selected articles are summarised and introduced below.

5.1 Emissions Scenarios

For climate change impacts assessment, the Intergovernmental Panel on Climate Change (IPCC) has published Assessment Reports (AR) on climate change based on greenhouse gas emissions scenarios. Future climate projections in the IPCC Fourth Assessment Report (AR4, IPCC 2007) were based on Special Report on Emissions Scenarios (SRES, IPCC 2000) and simulations of the third phase of the Coupled Model Intercomparison Project (CMIP3, Meehl et al. 2005). SRES was superseded by Representative Concentration Pathways (RCPs) in the IPCC fifth assessment report (AR5, IPCC 2014) based on the CMIP5 (Taylor et al. 2012).

The IPCC Sixth Assessment Report (AR6, IPCC 2021) assessed the future climate outcomes based on the combination of socio-economic (SSP1-SSP5) and future radiative forcing scenarios (1.9 to 8.5 W/m2), which called Shared Socioeconomic Pathways (SSPs). The latest SSPs can quantitatively describe the relationship between socioeconomic development and global climate change to reflect the climate change challenges that society will face in the future (Eyring et al. 2016). Basically, some older studies (generally in 2017 and 2018) used SRES of CIMIP3 models. Most selected studies utilised RCPs of CMIP5 models. Some post-2020 studies were starting to use SSPs from CMIP6.

5.2 Climate Models and Downscaling Methods

Global climate model (GCM) is capable and useful for projecting future climate (Overland et al. 2011). Many research institutions have developed GCMs based on their own experiment assumptions and mathematical representations of physical climate system.

Studies at global scale in this review inputted GCMs directly into global models to project climate change impacts (Turner et al. 2017; Pastor et al. 2019). However, GCMs are generally insufficient to provide useful climate predictions on the local to regional scales because of relatively coarse resolution and significant uncertainties in the modelling process (Allen and Ingram 2002; Dibike and Coulibaly 2005). When the climate change impacts studies are carried out at local and regional scales, downscaling methods have been developed to overcome the mismatch of spatial resolution between GCMs and models (Hwang and Graham 2013).

Downscaling techniques are categorised by two approaches (Hwang and Graham 2013):

  1. 1.

    Statistical downscaling using the empirical relationship between GCMs simulated features at the grid scale and surface observations at the sub-grid scale. For example, Bias-Correction Spatial Disaggregation (BCSD, e.g., Zhao et al. 2022) and the Statistical Downscaling Model (SDSM, e.g., Goodarzi et al. 2020) were employed to downscale GCMs in the selected studies.

  2. 2.

    Dynamic downscaling using regional climate models (RCMs) based on physical relations between the climate parameters at large and smaller scale.

Most selected articles using dynamic downscaling method generally applied results from the Coordinated Regional Climate Downscaling Experiment (CORDEX). CORDEX was to create an enhanced modelling framework for generating climate projections at regional scales, enabling impact assessments and adaptation studies globally within the IPCC AR5 (Giorgi et al. 2022).

5.3 Global Warming Scenarios

The Paris Agreement (UNFCCC 2015) aims to keep global mean surface air temperature increasing below 2℃ relatives to pre-industrial levels and targets to limit it to 1.5℃. Some articles simulated future Water-Food or Water-Energy under global warming 1.5℃, 2℃, 3℃ or 4℃ (Donk et al. 2018; Sylla et al. 2018; Meng et al. 2020; Qin et al. 2020b; Rosa et al. 2020; Zhao et al. 2021b). These studies utilised two approaches (James et al. 2017) to assess the regional implications of different degrees of warming: (1) time sampling; (2) pattern scaling.

In time sampling approach, the global warming scenarios are derived by extracting a period of time (usually 30 years) when the driving climate model projects an increase of specified degrees (e.g., 1.5℃ and 2℃) of warming compared to the pre-industrial level (Sylla et al. 2018).

Pattern scaling assumes the relationship between global mean temperature and local change is linear (Huntingford and Cox 2000; Mitchell 2003; James et al. 2017). These patterns can scale changes in global mean annual temperature to local and seasonal changes for climate variables by linear regressions (Qin et al. 2020b).

6 Directions of Future Research and Prospects

Future challenges in climate change and nexus research are identified from five aspects: (1) scale and resolution of study area; (2) internal physical mechanism; (3) extreme climate events; (4) potential competition between sectors; (5) data and model uncertainty.

6.1 Scale and Resolution of Study Area

Most selected studies related to climate change impacts on Water-Food generally focused on watershed, regional and national scale, the analyses not only focused on temporal differences but also spatial difference according to different geographical resolution. In contrast, studies investigating climate change impacts on Water-Energy mainly analysed hydropower, therefore, results were generally shown within hydropower plants, dams and reservoirs without spatial difference. Evaluation studies of climate change impacts on Water-Energy-Food nexus mainly focused on basin and regional scale, the analysis put water, energy, and food into a whole system, but the spatial resolution was often ignored.

It is of great significance for local sustainability management and decision-making to study climate change impacts on Water-Energy-Food on basin and regional scale, but the results may be limited by the boundaries of the study area. For example, the simulated streamflow at the outlet of the study basin is not necessarily the amount of water available in the basin because the water demands in the downstream regions should be considered. The absence of water, energy and food scheduling with other regions may have effects on inaccurate supply and demand simulations, further resulting in inaccurate management strategies. The water transport routes of water resources are sometimes cross-watershed. For example, reservoirs or weirs provide for agriculture, industry, or domestic use through their own pipeline systems. With the impact of climate change, economic globalization and other strong human activities, local Water-Energy-Food nexus is bound to be influenced by global hydrological cycle and non-local human activities. It requires scholars to understand local nexus relationships from a large-scale perspective.

Meanwhile, considering climate change has obvious spatial differences, and the response speed of different regions to climate change is also different, the climate change impacts study on Water-Energy-Food nexus with geographical resolution can show spatial difference of water, energy, food change due to climate change and provide a better reference for sustainability management.

6.2 Internal Physical Mechanisms in Modelling

Most of the studies about evaluating climate change impacts on Water-Food and Water-Energy considered hydrological processes based on physical mechanisms. Research on Water-Energy-Food and climate change impacts consisted of interdisciplinary and transdisciplinary analysis, while the complexity of the system leads to the simplification of many physical mechanisms. Many mathematical or data-driven models were used for investigating climate change impacts on Water-Energy-Food, but the lack of internal physical mechanisms cannot well explain the interactive process between water, energy and food to climate change.

Future research needs to understand the interlinkages and internal physical mechanisms of the nexus sectors and climate change. Meanwhile, science and policy should be integrated to reveal the dynamics of natural processes along with social processes.

6.3 Novel Artificial Intelligence Models

Many selected studies employed Artificial Intelligence (AI) in simulation and operation optimisation. Feedforward and feedback neural networks were used for simulation. Most selected studies used programming and meta-heuristic algorithms, and a small number used reinforcement learning for operation optimisation. In recent years, with the rapid development of AI, many novel AI models have been proposed repeatedly. These AI models will provide a feasible direction for these complex interdisciplinary sciences. For example, deep reinforcement learning (DRL) was developed by combining traditional reinforcement learning with deep learning, and it is capable of handling high-dimensional states and actions (Mnih et al. 2015). DRL has been applied for optimal hydropower reservoir operation (Xu et al. 2020), irrigation optimisation (Alibabaei et al. 2022) and water-energy-food nexus security assessment (Raya-Tapia et al. 2023). The application of DRL on complex water-energy-food system under climate change is still to be investigated.

6.4 Extreme Climate Events

Most projections of future nexus were generally based on temporal continuous climate change scenarios, only few reviewed studies have considered extreme weather events. Climate change will increase the intensity, frequency and spatial extent of extreme climate events (Hasegawa et al. 2021) and compound hazards (Zscheischler et al. 2018). More frequent and extreme events will cause disruptions in the management of water, energy, and food (Núñez-López et al. 2022). For example, relative to moderate-level climate change, an additional 20–36% population may face hunger under a 1-in-100 yr extreme climate event under RCP8.5 (Hasegawa et al. 2021). Compound hazards will cause devastating impacts at a scale far beyond any single disaster in isolation (Zscheischler et al. 2018). For example, increasing compound drought–heatwave risks may affect 90% of the global population and gross domestic product in the future (Yin et al. 2023). Considering the water, energy and food relationship under extreme climate events in any future studies has an important role in developing strategies to ensure water, energy and food security.

6.5 Potential Competition Between Sectors

Previous studies related to climate change impacts on Water-Food and Water-Energy did not consider competition between subsystems because there was no/limited competition between two sectors in early days. When considering three sectors together, competition arises. Competition for water between food and energy sectors is an important part of the Water-Energy-Food nexus (Qin 2021). The competitive relationship is not conducive to the sustainable development. For example, the average total production water footprint in 31 provinces of the Chinese Mainland in the Industry Competition Unsustainability scenario reached 4.08 m3/kg in 2016 (Hua et al. 2022). Considering the economic and social situation, energy production is more profitable than food, so water flows easily into the energy sector. Especially in the context of climate change, water availability is greatly affected. How to ensure food and energy security within limited water resources context should be considered in any future studies.

6.6 Data and Model Uncertainty

In the evaluation of climate change impacts on Water-Energy-Food, numerous data from multiple disciplines including meteorology, agriculture, environment, hydrology, economics, society are needed. The use of different data sets, mismatch of data resolution, the varying quality and availability of data (Perrone et al. 2011), and assumptions and simplifications introduced to deal with data scarcity could lead to very different results. High uncertainty may be caused to exert negative impacts on the nexus analysis and even misrepresent the interactions among nexus sectors (Zhang et al. 2018). What is more, models and analysis tools also introduce uncertainty. Downscaling of future meteorological data, numerous parameters in modelling, limited understanding of nexus processes, the intrinsic indeterminism of complex dynamic systems, and myriad future scenarios will bring uncertainty into final results, making it difficult to identify an optimal policy choice (Gallopín et al. 2001; Antón et al. 2013; Yung et al. 2019).

Endeavours should be made in future studies to identify, analyse and reduce uncertainty in data use and modelling for nexus research to increase the reliability of projection results and build capacity for decision-making in the context of uncertainty.

7 Conclusions

This paper provides a systematic review on the analytical approaches in the evaluation of climate change impacts on Water-Food, Water-Energy and Water-Energy-Food. The key findings are summarised as below:

  1. 1.

    Analytical methodologies used in selected research can be classified into four categories: Statistical methods, Physics-based modelling, Supervised learning and Operation optimisation. Catalogues of methods used in the evaluation of climate change impacts on Water-Food, Water-Energy and Water-Energy-Food are listed respectively based on the classification (see Tables 1, 2 and 3). Such catalogues are helpful to clearly show popular and promising methods in selected studies.

  2. 2.

    The focus of research on different topics at different scales are discussed. Large scale and medium-small scale models are introduced in terms of their characteristics, limitations, providing references for selection of models and issues to consider when using the models. Some models are applicable for different scales but there is no single model suitable for all scales. The classification and discussion of topics and models is helpful to provide guidance on appropriate model selection by considering research scales, objectives and themes (Water-Food, Water-Energy and Water-Energy-Food).

  3. 3.

    Future climate scenarios setting including emission scenarios, climate models, downscaling methods and global warming scenarios are summarised. Climate scenarios are important for simulating interactions between water, energy and food under various future climate change conditions, as well as exploring the effectiveness of mitigation measures or policies. The study has provided references for the setting of climate scenarios and processing of future meteorological data in future research.

  4. 4.

    Despite significant efforts were made in investigating climate change impacts on Water-Energy-Food, limitations of current research still exist, and the challenges for future study are discussed. Current studies do not adequately address the uncertainties generated by data and models. Research about extreme climate events and potential competition in nexus systems is not sufficient. Efforts can be made in the internal physical mechanisms analysis, application of novel artificial intelligence models and spatial differences analysis of nexus issues.