Land-water-energy nexus and long-term climatic trends in smallholder irrigation schemes: evidence from Vhembe District, South Africa

1 Introduction

Irrigation farmers depend on natural resources, particularly land, water, and energy, to sustain their livelihoods. Land, water, and energy are essential, interrelated resources that sustain rural livelihoods (Cho et al., 2023). For instance, land in rural areas has many roles and provides diverse benefits to human sustenance (Jónsdóttir and Gísladóttir, 2023). Land in South Africa is closely tied to identity, belonging, and cultural practice; for many South Africans, access to land is less about livelihood than about reconnecting to historical roots, with indigenous communities often identifying themselves by the land they occupied (Vorster, 2019). In addition, it provides for the establishment of renewable energy sources, including solar and wind energy (Burke et al., 2019; Jones et al., 2015). Meanwhile, water and energy are essential to food production; energy is crucial for water management (extraction and distribution), while energy production also requires water (Rasul and Sharma, 2016). Energy plays a vital role in the production, transportation, manufacturing, preservation, and distribution of food, as well as in the extraction, pumping, treatment, distribution, and collection of water (Cho et al., 2023; Malliaroudaki et al., 2022). These complex connections are referred to as the “land-water-energy nexus” (Cho et al., 2023).

Irrigation smallholders offer a particularly revealing lens on the LWE nexus because, unlike rainfed producers, they draw on all three resources simultaneously and directly: land for cultivation, water abstracted, conveyed, and applied through scheme infrastructure, and energy to pump and deliver that water (Rasul and Sharma, 2016; Cho et al., 2023). This tight coupling means that a shortfall in any one resource immediately constrains the others, making irrigation schemes a tractable case for examining nexus trade-offs that are far more diffuse and harder to observe in rainfed systems, where energy use is comparatively marginal. Findings from irrigation smallholders therefore speak directly to global debates on the agriculture-water-energy nexus (Hamidov and Helming, 2020; Ingrao et al., 2023), because these schemes concentrate the nexus’s interdependencies within a single, bounded system in which all three resources can be measured and compared. Rainfed smallholders undoubtedly face significant and worsening climate exposure, and their vulnerability warrants separate, complementary investigation. However, because their production is not mediated by engineered water and energy infrastructure, they do not, on their own, test the LWE interactions that this study is designed to interrogate.

This study adopts the land-water-energy (LWE) formulation rather than the more established water-energy-food (WEF) nexus that dominates the comparative literature, for two related reasons. First, WEF treats food as a resource stock produced from water and energy inputs, whereas this study’s unit of analysis is the scheme-level input base itself; the same category of measure (availability and accessibility) is required for all three nexus dimensions, and food is an output rather than an input of that kind. Second, in the Vhembe context, land is not merely a background driver of water and energy pressures, as it is typically treated in WEF applications, but an active institutional constraint in its own right, shaped by tenure insecurity, homeland-era fragmentation, and ongoing land reform. Positioning land as a first-class nexus resource, rather than folding it into other resources or treating it as exogenous context, allows the LWENI framework to capture tenure and access-related constraints directly alongside water and energy accessibility. This follows a precedent in the literature for land-inclusive nexus reformulations where land governance is analytically central, including land-water-food (Klümper and Theesfeld, 2017) and water-energy-land-food (WELF) (Ringler et al., 2013) framings.

Sustainable agricultural growth in South Africa is hampered by increasing land degradation and water and energy insecurity. Recent remote-sensing analysis found that 34% of Southern Africa exhibited one or more forms of land degradation, with 5% of the land experiencing intensified degradation from two or three overlapping degradation types, underscoring the scale of the challenge facing agricultural sustainability in the region (Li et al., 2023; Slayi et al., 2024). These resources are under increased pressure from population growth, environmental changes, and seasonal variability (FAO, 2025), resulting in the loss of prime agricultural land. South Africa’s population is projected to reach approximately 76 million by 2050 (UN, 2024; Archer et al., 2019), up from 65 million (Stats SA, 2026), intensifying demand for land, water, and energy resources. As the country’s population grows, it will need more natural resources to increase food production (Boretti and Rosa, 2019). Limpopo’s population grew by 1.1 million, from 5.4 million in 2016 to 6.5 million in 2022, with Vhembe District accounting for the largest share, at 1.6 million total households (Stats SA, 2022). This growth necessitates expanded agricultural production and productivity, particularly in irrigation-dominant areas (Boretti and Rosa, 2019; Hlatshwayo et al., 2021).

South Africa has a total land area of 122 million ha, with 84.8 million ha classified as agricultural land, and only 13.8 million ha (11%) having high agricultural potential (le Roux et al., 2025). Nearly 2 million ha of high-potential agricultural land lies in the former homeland areas, including Vhembe District in Limpopo Province, which comprises the former Venda Bantustan homeland (BFAP, 2022). Yet 440,000 hectares (13%) remain barren and underutilized according to 2025 land cover data (le Roux et al., 2025). This makes Vhembe a particularly relevant case study: its fertile soils and existing irrigation infrastructure sit alongside the structural legacies of homeland-era land allocation, including fragmented smallholder plots and constrained access to water, which continue to limit the district’s agricultural potential (Nhamo et al., 2018).

Smallholder irrigation farmers in South Africa face long-term underutilization despite access to substantial arable land. Key barriers include land tenure insecurity, which discourages investment and expansion (Fanadzo and Ncube, 2018), as well as poor infrastructure, inadequate management and technical skills, limited market access, and financial constraints. These factors contribute to low productivity and scheme failures despite public investment, limiting irrigation potential for food security and rural livelihoods on fragmented communal lands. Land reform in South Africa rests on three pillars set out after 1994: restitution (returning land dispossessed under apartheid-era laws like the 1913 Land Act), redistribution (transferring land to previously disadvantaged people, largely on a “willing buyer, willing seller” basis until recently), and tenure reform (strengthening rights for people living under insecure arrangements, including in former homelands). The major recent development is the Expropriation Act, 2024, which repeals the apartheid-era 1975 Act and provides a framework for expropriation of property for a public purpose or in the public interest, including circumstances in which nil compensation may be just and equitable. Policymakers believed that increasing and revitalizing smallholder irrigation schemes could raise food output, and the government anticipated that new and revitalized schemes would help cut unemployment and poverty by creating. Targeted land reform policies, improved tenure security, infrastructure investment, and support services are essential to unlock this potential.

Land tenure insecurity remains a major institutional constraint on smallholder irrigation development in South Africa. Since 1994, post-apartheid land reform has established many smallholder irrigation schemes (SIS) on communal land, yet their performance has been consistently poor, a failure linked to inappropriate land tenure arrangements, institutional fragmentation, and the fact that land in these schemes is still legally classified as state land, allowing political interference in its administration (Fanadzo and Ncube, 2018; FAO, 1997; Huntington and Lisher, 2025). Assessments of government support programs also show that communal areas have been effectively excluded from key funding instruments (e.g., the Comprehensive Agricultural Support Program) because of their lack of secure tenure, so the benefits of land reform have not translated into investment-ready irrigation infrastructure for smallholders (Fanadzo and Ncube, 2018). A recent field-based study of ten land reform projects in the Eastern Cape found that none of the beneficiaries initially held valid leases, and a decade later, only four of the ten projects had secure rights. This illustrates how persistent tenure insecurity continues to undermine agricultural investment years after land redistribution (Hall and Kepe, 2026).

Climate change further threatens land resources through increased desertification, land degradation, and soil erosion. Land supports renewable energy capture and storage, as well as water development (Burke et al., 2019; Cho et al., 2023; Jones et al., 2015). However, climate change and land-use changes (e.g., deforestation, urbanization) alter water yield, availability, and quality (Zhu et al., 2023). Warmer temperatures and shifting rainfall patterns increase evaporation and runoff, reduce groundwater recharge, and heighten flood and drought risks, while deforestation accelerates erosion and disrupts water flow. Urbanization converts irrigable land to housing, further reducing arable land and altering the water set aside for irrigation.

South Africa is a water-scarce country with average precipitation below 500 mm, compared to the global average of 850 mm (DWS, 2023). Precipitation is unevenly distributed, with Limpopo receiving 560–1,000 mm annually, predominantly summer rainfall (Maluleke et al., 2023; Zuma-Netshiukhwi et al., 2025). Vhembe district receives 400–1,600 mm of rainfall annually, with warm-to-hot temperatures year-round (CSIR, 2024).

Governance, climate, and infrastructure factors influence water security for smallholder irrigation farmers. The National Water Act of 1998 seeks to redress past inequalities through comprehensive frameworks for allocation and user participation (Schreiner and van Koppen, 2020; DWS, 2023). However, many smallholder farmers lack water-user licenses (Mugejo et al., 2022), hindering financing and access. Inadequate infrastructure, poor maintenance, limited farmer involvement, and insecure land tenure exacerbate water insecurity (Chikozho et al., 2020; Mugejo et al., 2025).

Climate change intensifies these challenges through erratic rainfall, shorter seasons, prolonged droughts, intense storms, and rising temperatures (Nyika, 2021; Salonen and Reiser, 2023; Kori et al., 2024). These challenges increase hydrological variability, water stress, and competition across sectors (Ingrao et al., 2023; Rosegrant et al., 2024). During periods of water scarcity, the National Water Act prioritizes human needs and ecosystems over agricultural and economic uses (Nhemachena et al., 2018). Climate shocks reduce water availability for irrigation, energy generation, and food processing (Cho et al., 2023; Jia et al., 2019), while higher temperatures and water application rates elevate energy demands for pumping (Flörke et al., 2018; Thabane et al., 2025). Hydroelectric reliability declines during droughts (Nhamo, Mabhaudhi et al., 2020), compounding vulnerabilities for smallholder farmers facing yield losses, soil erosion, and financial strain (Cheteni and Umejesi, 2022; DuBois et al., 2024; Harvey et al., 2018).

Climate change alters the demand for land, water, and energy (Kushawaha et al., 2021) thereby negatively impacting smallholder irrigation systems and the broader LWE nexus (Herrera-Franco et al., 2023). These pressures threaten food security, poverty reduction, and sustainable development (Olayide and Alabi, 2013; Rahut et al., 2022; Touch et al., 2024). Without intervention, vulnerabilities will worsen, undermining progress toward Sustainable Development Goals (SDG) 2 (Zero Hunger) and 6 (Clean Water and Sanitation) (D’Odorico et al., 2020; FAO, 2021; SDGs, 2019).

In Limpopo Province, particularly in the Vhembe District, agriculture is a key economic driver that relies on land, water, and energy (Maponya, 2021). A related study by Malatsi et al. (2025) evaluated the performance of smallholder irrigation typologies in the Vhembe District, comparing independent and publicly managed schemes using cross-sectional survey data from 95 independent and 165 public irrigators. The results found that smallholders mainly grow grains, vegetables, and horticultural crops, with no significant difference between independent and public irrigators (p < 0.05). Public irrigators make up 64% of the total (p = 0.001) and earn significantly more than independent irrigators (χ² = 25.83, p < 0.001). While 59% of independent irrigators access both formal and informal markets, 30% of public irrigators are limited to informal markets only (p < 0.001). Across all smallholders, the main challenge is a lack of food value addition and commercial packaging. That study was descriptive and comparative in scope, focused on establishing performance differences between scheme typologies rather than on the structural interactions among the resources underpinning that performance. The present study is distinct in three respects. First, in scope, it narrows the focus to the 165 publicly managed smallholder irrigators from that broader sample and moves beyond typology comparison to examine the land-water-energy (LWE) nexus itself, the degree to which land, water, and energy constraints interact with, reinforce, or offset one another at the scheme level, and layers climate change exposure onto this analysis, neither of which the earlier study addressed. Second, in data and methods, while drawing on the same public-scheme respondents, this study constructs separate availability and accessibility indices for land, water, and energy from these data, then models these individually and in combination to characterize LWE scenarios. This index-based, interaction-focused approach differs from the descriptive and bivariate statistical comparisons of typology groups used in Malatsi et al. (2025), and uses the data from that paper. Data used in this study were collected as part of the same survey reported in Malatsi et al. (2025); however, the two papers analyze distinct research questions and variables from that dataset. Third, this study offers a nexus-level diagnostic of how land, water, and energy constraints jointly limit irrigation performance under climate stress within publicly managed schemes; specifically, it provides evidence to inform cross-sectoral resource planning and policy, rather than a within-sector comparison between independent and public scheme types.

This study conducts an integrated assessment of LWE interactions and climate change impacts on smallholder irrigation farmers’ performance. Using selected indices for resource availability and accessibility, it examines individual and combined LWE scenarios to reveal interrelationships. The analysis aims to strengthen cross-sectoral synergies and improve resource productivity at the scheme level, supporting sustainable management despite governance challenges. Understanding these key dynamics is vital for policymakers, researchers, and communities seeking balanced solutions amid population growth, resource constraints, and climate pressures. The assessment highlights the need for a fundamental transformation in LWE management within agri-food systems to achieve sustainability, strengthen rural livelihoods, and bolster food security for smallholder irrigation farmers.

This study is unique in three ways that, to our knowledge, have not been integrated for smallholder irrigation in South Africa. Initially, it measures the LWE nexus not merely as an aggregate of indicators but as an interaction, employing a sustainability-polygon framework that encapsulates the relative equilibrium and distortion among land, water, and energy facets within each scheme, thus rendering trade-offs, rather than just levels, observable and comparable across schemes. Second, it integrates this nexus evaluation with a 44-year (1980–2024) examination of climatic patterns from four automated meteorological stations, enabling interpretation of resource disparities in relation to recorded long-term trends in temperature increases and rainfall fluctuations specific to Vhembe District, rather than relying on broad regional assumptions. Third, it evaluates four public irrigation schemes (Kumbe, Mphaila, Folovhodwe, and Rambuda) in relation to one another and to established Southern African and South African WEF nexus indices (Mabhaudhi et al., 2019; Nhamo, Mabhaudhi et al., 2020), placing scheme-level efficacy within a wider comparative framework. Collectively, these attributes move the study beyond cataloging recognized vulnerabilities to illustrating the extent and severity of the imbalance at the scheme level, where intervention choices are determined.

2 Materials and methods

2.1 Research approach

This research employed a quantitative approach that combined primary and secondary data collection and analysis. Data sets on land, water, and energy availability and accessibility were obtained from the Department of Water and Sanitation (DWS), the Department of Agriculture (DoA), and participants in public irrigation schemes. The study used purposive sampling to select irrigation schemes and participating farmers. The four smallholder irrigation schemes (Folovhodwe, Khumbe, Mphaila, and Rambuda) were purposively selected because they are located within the Vhembe District, represent diverse agroecological and socio-economic conditions, and are actively engaged in irrigated crop production. Within each selected scheme, farmers were purposively chosen based on predefined inclusion criteria: they had to be active smallholder irrigation farmers, have at least one full production season of experience in the scheme, be directly involved in farm management and decision-making, and be willing to participate in the study. This approach ensured that respondents possessed adequate knowledge and practical experience relevant to the study objectives, thereby enhancing the richness and relevance of the information collected. Purposive sampling, while methodologically appropriate for exploratory and contextually rich research, introduces several well-documented biases that must be acknowledged when interpreting the findings and assessing how far conclusions can be generalized beyond the four selected schemes. Five categories of bias are identifiable in the sampling design, including selection bias at the scheme level, volunteer and willingness bias, inclusion criterion bias, research-induced or accessibility bias, and implications for generalizability.

All four schemes share a common typology: smallholder canal- or furrow-based irrigation schemes established under former homeland-era government programs, situated within the same broader Vhembe District smallholder farming context, and operating under similar land tenure arrangements linked to Traditional Authorities (Malatsi et al., 2025). All four schemes fundamentally depend on canal-based bulk water conveyance to balancing or storage infrastructure, even where supplementary individual pumping exists. As a result, the core water-delivery architecture is broadly comparable, even if marginal energy inputs differ. Aggregation to describe general LWE performance across this population is a reasonable unit of analysis, provided this limitation is explicitly acknowledged.

Energy use across the four schemes follows two patterns: Kumbe, Folovhodwe, and Rambuda were designed as gravity-fed systems, but as their canals and balancing dams leak or are nonfunctional, farmers turn to private petrol/diesel pumps as an informal alternative. Mphaila relies on a single, collectively funded, grid-powered pump that is more fragile, since its failure or an unpaid bill would cut off water for the entire scheme. Given Vhembe District’s strong solar irradiance, among the best in South Africa as it is in the tropics, solar-powered pumping is a strong fit for both patterns; it could directly replace fuel pumps in the three gravity-fed schemes and displace Mphaila’s Eskom dependency, with existing (if upgraded) balancing dams potentially serving as storage buffers to avoid costly battery systems. Current energy costs and reliability (grid tariffs, load reduction, and fuel price volatility) remain a challenge. Eskom tariffs rose 76% for customers in 2026, a cumulative rise that becomes a fixed-cost burden on Mphaila’s pooled bill regardless of harvest outcomes, while also strengthening the payback case for solar conversion. This compounds with rising fuel costs for the other schemes’ pump-dependent farmers, all of which matter disproportionately on the thin margins of smallholder plots, where energy access is often a precondition for irrigation.

The required sample size for scheme farmers was calculated using the Equation 1 below:

S = N 1 + N ( e ) 2 (1)

Where:

• S is the sample size.
• N is the population size.
• E is the margin of error (degree of precision).

A 10% margin of error (e = 10%) was used, which is acceptable in social science research when resources are constrained. The population size (N) used in each calculation was the number of individual farmers registered at each scheme, not the combined total across schemes. The combined total of 273 farmers was the sum of the four individual scheme populations (Kumbe = 80, Mphaila = 63, Folovhonde = 45, Rambuda = 85), and the sample size was calculated separately and proportionally for each scheme. Applying the formula to each scheme yields sample sizes of 44 (Kumbe), 38 (Mphaila), 31 (Folovhonde), and 45 (Rambuda), for a total of 158 farmers. The number of farmers interviewed (165) was slightly higher than the calculated figures for each scheme, reflecting a modest oversampling buffer to increase precision, given the specified population sizes and confidence level. Each scheme slightly exceeded its calculated target, with 46 (Kumbe), 40 (Mphaila), 32 (Folovhondwe), and 47 (Rambuda) respondents interviewed. The 165 participants were interviewed using structured questionnaires; their responses provided quantitative data on the availability and accessibility of land, water, and energy.

The LWE nexus normalized indices for smallholder irrigation schemes in Vhembe District were calculated using available data. The scale ranged from 0 to 1. The study measured and modelled progress toward sustainable development using indicators focused on resource accessibility and availability (Nhamo, Ndlela et al., 2020). Irrigated agriculture is a classic example of the LWE nexus, as it uses water, land, and energy, among other inputs, in food production (Hamidov and Helming, 2020). A few authors provide detailed applications of the indices for implementing the LWE nexus approach. Applying the LWE nexus approach to optimize sustainable food systems requires an understanding of resource interactions. The integrated assessment of indices from the LWE nexus perspective across the different irrigation schemes was conducted using sustainability polygons (FAO, 2021).

The geographical distribution of weather stations in the Vhembe District, Limpopo Province (Figure 1). Irrigation schemes were linked to weather stations based on local municipal associations rather than pure geographic distance, reflecting the administrative and agricultural extension structures through which climatic data are typically accessed and applied at the scheme level in the study area. This approach was chosen because rainfall and temperature vary over such distances, particularly in an area with topographic variation such as the Soutpansberg. The municipality’s agricultural extension office reports rainfall and temperature from that station; it is the official reference station for that municipality, and scheme management data were sourced from the municipal office alongside the station’s records. The proximity of each station to each scheme was as follows: Khumbe → Thohoyandou (Thulamela LM, 14 km); Rambuda → Tshiombo (Thulamela LM, 5 km); Mphaila → Levubu (Makhado LM, 25 km); Folovhodwe → Messina Proefplaas (Musina LM, 65 km).

Data on climatic resources were obtained from the Agricultural Research Council: Natural Resources and Engineering (ARC-NRE) (Moeletsi et al., 2022) and the South African Weather Service (SAWS) (www.weathersa.co.za). Monthly climatic data (daily rainfall, minimum and maximum temperatures) were sourced from four weather stations in Vhembe District: Thohoyandou, Mesina Proefplaas, Levubu, and Tshiombo. The climatic dataset covered 44 years (1980–2024).

Monthly climatic data were calculated and compared for summer (Oct to March) and winter (Apr to Sep) seasonal precipitation and maximum and minimum temperatures across weather stations for 1980 to 2024.

The Climatic Moisture Index (CMI) was calculated using the Potential Evapotranspiration (PET) and precipitation. The calculation was based on Willmott and Feddema’s methodology (Willmott and Feddema, 1992) as follows Equations 2, 3:

CMI = P PET − 1 , when P < PET (2)

CMI = 1 − PET P , when P ≥ PET (3)

The CMI indices range from −1 to +1, where wet climatic conditions indicate a positive CMI, and dry conditions indicate a negative CMI. The PET was calculated using monthly climatic data from four weather stations. The trend in seasonal precipitation, minimum and maximum temperatures, and the CMI from 1980 to 2024 was examined using statistical regression (Table 1).

Table 1. Key characteristics of the study irrigation schemes.

2.2 Description of the study area

The study took place in the Vhembe District of Limpopo Province, South Africa (Figure 1), one of the province’s five district municipalities. Vhembe District borders three countries: Zimbabwe in the north, Botswana in the northwest, and Mozambique in the southeast, via Kruger National Park. Vhembe District, like many others in the province, is predominantly rural (Rusere et al., 2019) and has high-potential agricultural areas, where agriculture is a key livelihood activity (Maponya, 2021). Approximately 90% of rural communities found in the Vhembe District depend mainly on agriculture to sustain their livelihoods and generate income (Materechera and Scholes, 2022; Vhembe District Municipality, 2021). Vhembe District covers 2,140,708 ha, of which 247,757 ha is arable, with a high concentration of irrigation schemes (Setshego et al., 2020). The region is suitable for cultivating tropical and subtropical fruits, particularly in the Levubu Valley. Irrigation schemes along river valleys can support the production of horticultural crops, grains, and vegetables (Malatsi et al., 2025).

2.3 Methodological framework

The Land-Water-Energy (LWE) nexus methodological framework has been used for the study (Cho et al., 2023) (Figure 2). The rationale for the study was to establish quantitative relationships among the intricate, broad, and multiple interactions among land, water, and energy resources at the level of smallholder irrigation schemes. The framework guided the selection of relevant, available indicators. These indicators are then normalized, weighted, and aggregated, thereby yielding a unitless index that represents the model being appraised. This index complements the underlying data and presents it coherently, thereby enabling the identification of patterns and trends.

2.4 LWE components representation

The sustainability indicators used to quantify interactions within the LWE nexus focus on both resource availability and accessibility (Table 2). For the land sector, it measures the total extent of arable land resources (availability) and the extent of actually cultivated arable land (accessibility), both in hectares. In the water sector, availability is assessed by water volume per capita (m³/capita). Accessibility is measured by the ratio of hectares per volume of water (ha/m³), indicating how effectively water supports land use (Nhamo, Mabhaudhi et al., 2020; Wolde et al., 2021). For the energy sector, availability refers to the percentage of irrigators with energy infrastructure, and accessibility refers to the percentage with an actual energy supply. Overall, the table provides a structured framework to evaluate resource sustainability by distinguishing between what exists (availability) and what is practically usable (accessibility) across the three interconnected LWE sectors (Wolde et al., 2022).

Table 2. Sustainability indicators used to quantify LWE interactions.

The performance of the LWE nexus indicators is ranked from 0 (unsustainable) to 1 (highly sustainable) (Table 3) (Nhamo, Mabhaudhi et al., 2020). The composite LWENI is calculated as a weighted average of the six normalized indicators (land availability and accessibility, water availability and accessibility, and energy availability and accessibility). In standard practice for constructing such nexus composite indices, each indicator is first normalized, typically using min-max normalization to scale values between 0 and 1, then multiplied by its assigned weight (reflecting relative importance), summed, and finally divided by the sum of the weights (Table 4).

Table 3. LWE nexus indicators performance classification categories.

Source: Nhamo, Ndlela et al. (2020).

Table 4. Demographic characteristics of respondents per scheme.

A sustainability polygon (spider-web diagram or radar chart) jointly compares multiple indicators across alternative options and scenarios. The sustainability polygon tool provides LWE resource management outcomes and integrates sustainability indicators that represent the security of land, water, and energy (Hochman et al., 2017). The tool is valuable for analysing threats, particularly in sustainable irrigation systems (Colloff et al., 2019; FAO, 2021; Taguta et al., 2022). The sustainability polygon clarifies interconnections among dimensions and indices, such as water, energy, and land availability and accessibility (Flammini et al., 2014).

The LWE nexus is not simply the arithmetic sum of the six land, water, and energy indicators. The nexus concept centres on the interdependencies and trade-offs between resources: a gain in one dimension (for example, expanding irrigated land) can come at the direct expense of another (for example, straining energy or water infrastructure), and the sustainability polygon is used precisely because it makes such trade-offs, and the resulting distortion in shape, visible in a way that a single summed score cannot. A scheme with a high composite LWENI but a badly distorted polygon is, in nexus terms, less balanced, and arguably more fragile, than a scheme with a lower composite score but a more even polygon; both the composite index and the shape of the polygon are therefore reported and interpreted jointly throughout this study.

2.5 LWENI normalization, weighting, and aggregation method

The calculation and normalization of LWE indices follow a composite indicator methodology established by OECD/European Union/Joint Research Centre – European Commission (2008). A conceptual framework first guides the selection of relevant indicators, which are then normalized, weighted, and aggregated to yield a unitless index. Raw indicator data are normalized using the min–max method, whereby the normalized value for region i and indicator j is calculated as Equations 4–6:

N ij = ( X ij obs − X j mkn ) / ( X j max − X j mkn ) (4)

Where X°ᵇˢ is the observed value, and Xᵐᵃˣ and Xᵐᵏⁿ are the maximum and minimum values of indicator j across the sample, converting all data to a unitless scale from 0 to 100 so that indicators in different units (ha, m³/capita, %) can be meaningfully combined (Simpson et al., 2022). For example, the water availability indicator (m³/capita) for a given region would be rescaled so that the region with the highest per-capita water volume scores 100 and the lowest scores 0, making it directly comparable with the land indicator (ha) and the energy access indicator (%). Each sub-pillar score is then calculated as the weighted arithmetic mean of its normalized indicator values:

S P k = Σ ( w j × N ij ) (5)

Where wj is the assigned weight for indicator j, and the weights within each sub-pillar sum to 1; pillar scores for Land, Water, and Energy are then calculated as the arithmetic average of their underlying availability and accessibility sub-pillar scores (Simpson et al., 2022). Weights for individual indicators are often derived using the Analytic Hierarchy Process (AHP) and a pairwise comparison matrix (PCM), which normalize indicators and establish the composite index (Wolde et al., 2022). An alternative normalization approach is z-score standardization, which measures each indicator’s deviation from the mean under a Gaussian assumption:

z ij = ( X ij − μ j ) / σ j (6)

After z-score normalization, Principal Component Analysis (PCA) can be applied to obtain orthogonal components that are summed to form a final index, thereby reducing the influence of correlated variables (He et al., 2024). Because aggregation choices entail uncertainty, sensitivity analyses such as Monte Carlo simulations are used to assess whether using an arithmetic versus a geometric mean, the latter being less compensatory across dimensions, materially changes regional rankings (Simpson et al., 2022).

The step-by-step methodology for calculating a composite Land-Water-Energy (LWE) nexus index for four irrigation schemes (Folovhondwe, Kumbe, Mphaila, and Rambuda) uses six normalized indicators: water availability, water accessibility, energy availability, energy accessibility, land availability, and land accessibility.

Raw indicator values are first converted to a common 0–1 scale using min-max normalization, enabling indicators measured in different units (e.g., m³ of water vs. hectares of land) to be combined:

Iij = ( Xij − Xmin , j ) / ( Xmax , j − Xmin , j ) (7)

Where Iij is the normalized index of indicator j for scheme i; Xij is the observed raw value; and Xmin,j and Xmax,j are the minimum and maximum observed values of indicator j across all schemes.

The composite index used in this study applies equal weights directly to the min-max normalized indicator values (Equation 7). For scheme i, the composite index is the equal-weighted arithmetic mean of its six normalized indicators: CIi = (1/n) × Σ (j = 1 to n) Wj × Iij, with Wj = 1/n = 1/6 ≈ 0.17 for each of the n = 6 indicators (water availability, water accessibility, energy availability, energy accessibility, land availability, and land accessibility). Applied directly to Iij, this method preserves genuine differences between schemes and yields the differentiated composite LWENI values.

2.6 Climatic trend analysis

The climate trend analysis examined winter (Apr to Sep) and summer (Oct to March) precipitation patterns, as well as trends in winter and summer minimum and maximum temperatures, based on records from 1980 to 2024. The Statistical Package for the Social Sciences (SPSS) 2024 (data analysis system) was used to analyze data.

Trends in winter and summer precipitation, minimum and maximum temperatures, and the CMI over 1980–1924 were assessed using linear regression for each variable year, with the model’s T-test used to assess the overall significance and the year coefficient (e.g., mm/year or °C/year) reported alongside adjusted R² and p for each variable.

2.7 Climatic variables-LWE nexus conceptual framework

Smallholder irrigated agriculture sits at the intersection of climatic, hydrological, and energy systems. Climate variability affects farm productivity not through a single, direct pathway; instead, it propagates through a chain of intermediate physical and socio-technical processes that together constitute a land–water–energy (LWE) nexus. This framework conceptualizes the chain as a five-stage causal sequence: climate drivers, climate stress/aridity, water availability, water and energy access, and agricultural outcomes, with contextual factors modulating the strength of each linkage. The framework makes explicit the mechanisms by which two climatic variables, rising maximum temperature and increasingly erratic rainfall, translate into farm-level productivity losses via coupled water-accessibility and energy-demand pathways, rather than treating them as independent or additive impacts.

The conceptual framework proposes that climate variability affects smallholder irrigated agriculture through a mediated, nexus-structured pathway rather than a direct or additive one. Rising temperatures and erratic rainfall jointly produce aridity (CMI decline), which reduces water availability. This reduction bifurcates into coupled water-accessibility and energy-demand pathways that jointly determine irrigation feasibility and cost, and these converge to reduce agricultural productivity. Contextual socioeconomic and institutional factors moderate the strength of this chain throughout, and several feedback loops, particularly around groundwater depletion and adaptive capacity, mean that the system’s response to climate stress is not static but can compound over time in the absence of intervention.

Rising maximum temperatures and increasingly erratic rainfall together drive a declining CMI, reflecting growing aridity. This declining CMI reduces water availability by diminishing surface storage and groundwater recharge, which in turn affects the nexus along two parallel pathways: reduced physical water accessibility (requiring deeper wells and longer conveyance infrastructure) and increased energy demand for water abstraction (as pumping from greater depths or over greater distances requires more energy input). Both pathways converge to reduce agricultural productivity through higher irrigation costs and a constrained water supply, with the potential for a reinforcing feedback loop in which reduced agricultural output and associated land degradation further diminish infiltration and water availability, compounding aridification over time.

3 Results and discussion

3.1 Demographic characteristics of the irrigation schemes

Gender distribution varied across the schemes. Kumbe and Mphaila had male majorities (52 and 65%, respectively), while Folovhodwe and Rambuda had slight female majorities (56 and 57%, respectively). Overall, the gender distribution leaned toward males at two schemes and toward females at the other two, with no scheme showing an extreme imbalance.

Respondents were predominantly middle-aged to older adults. Across all four sites, the largest group was 46–64 years (38–47%), followed closely by 36–45 years (33–45%). Younger adults (18–35 years) made up a smaller share (13–20%), and those over 65 years were a small minority, ranging from 9% in Kumbe to 0% in Rambuda.

Educational attainment was generally low to moderate. Secondary school was the most common level of education across all sites (40–66%), with the highest proportion in Folovhodwe (66%). Tertiary education was notable in Kumbe and Mphaila (26 and 33%), but much lower in Folovhodwe and Rambuda (13 and 17%). No formal education was most prevalent in Mphaila (23%), compared with 3–9% elsewhere. Primary schooling only was more common in Rambuda (26%) and Folovhodwe (19%) than in Kumbe (13%) or Mphaila (5%).

Most respondents operated small farms. In Kumbe, the vast majority (85%) farmed plots of 1–3 ha, with very few (4%) below 1 ha. In contrast, Mphaila had a much higher share of very small farms (< 1 ha; 57%) and a lower share in the 1–3 ha range (38%). Folovhodwe and Rambuda showed an intermediate pattern, with 62–72% farming 1–3 ha plots and smaller proportions farming either <1 ha or >3 ha. Across all sites, very few respondents farmed more than 5 ha (0–13%).

The sample was highly experienced overall. Most respondents at each site had more than 10 years of farming experience (53–70%), with Kumbe showing the highest proportion (70%). Respondents with less than 1 year of experience were rare (0–6%), indicating a largely established group of experienced farmers rather than newcomers.

The sample is best characterized as middle-aged to older small-scale farmers with generally low- to secondary-level formal education and substantial farming experience (often 10 + years), operating mostly on plots of 1 to 3 hectares. Gender balance and specific educational and farm-size patterns vary somewhat by site, with Mphaila standing out for its higher proportion of very small farms and lower formal education, and Kumbe and Rambuda showing relatively higher educational attainment and more balanced farm sizes.

3.2 LWENI results

The composite index applied equal weights directly to the min-max normalized indicator values (Equation 7):

C I i = Σ n j = 1 W j · I ij = ( 1 / 6 ) × Sum i

Worked example for Folovhondwe Irrigation Scheme:

C I Folovhondwe = ( 1 / 6 ) ( 0.75 + 0.65 + 0.15 + 0.05 + 0.90 + 0.87 ) = ( 1 / 6 ) ( 3.37 ) = 0.56

The LWENI for four smallholder irrigation schemes in the Vhembe District: Folovhondwe, Kumbe, Mphaila, and Rambuda (Table 5). Each scheme is scored on six normalized indicators (0 to 1) across the three nexus dimensions of water, energy, and land, with each dimension assessed for both availability and accessibility. The composite index (CI) is calculated as the arithmetic mean of the six indicator scores (sum ÷ 6), yielding a single unitless value between 0 and 1 that represents each scheme’s overall nexus sustainability performance.

Table 5. LWENI by irrigation scheme (equal-weighted aggregations of normalized indicators).

Bold values in the bottom row indicate the composite Land-Water-Energy Nexus Index (LWENI) for each irrigation scheme, calculated as the equal-weighted arithmetic mean of the six normalized indicators listed above (Sum ÷ 6).

Mphaila records the highest composite index (CI = 0.74), driven largely by perfect land availability (1.00), strong land accessibility (0.90), and notably higher energy scores (availability = 0.70; accessibility = 0.55) than all other schemes, suggesting a more balanced and integrated resource profile. Folovhondwe (CI = 0.56) and Kumbe (CI = 0.59) perform similarly, with strong land dimensions (0.87–0.90 availability) but critically weak energy components (0.05–0.20), indicating near-total energy infrastructure deficiency. Rambuda records the lowest CI (0.53), characterized by the weakest water scores (availability = 0.52; accessibility = 0.47) and energy accessibility (0.15), pointing to compound resource insecurity.

Across all four schemes, land dominates the resource profile, while energy, particularly accessibility, is the weakest LWENI across all four schemes (0.05–0.15), consistent with findings from Ethiopian irrigation schemes within the moderate-sustainability range identified by Simpson et al. (2022).

For water, the scores reported here (0.47–0.88) are higher than those observed in dryland smallholder contexts. A nexus assessment of irrigation schemes in the Limpopo Basin found water availability scores below 0.40 for schemes dependent on rain-fed or ephemeral river sources, compared with 0.65–0.88 for schemes with more regulated or perennial water access. This contrast helps explain Rambuda’s weaker water performance (0.52 availability) than Kumbe (0.88), likely reflecting differences in source reliability. He et al. (2024) demonstrated that water accessibility scores are systematically lower than water availability scores across developing-country irrigation schemes because physical access infrastructure (canals, pipes, pumps) frequently lags behind the legal or hydrological availability of the resource. This pattern is clear here, with all four schemes recording lower accessibility than availability scores for both water and land dimensions.

Land scores in this dataset (0.75–1.00 availability; 0.75–0.90 accessibility) are relatively high. This aligns with findings from South African government-supported smallholder irrigation schemes, where Permission to Occupy (PTO) arrangements and traditional authority land allocations provide de facto tenure security sufficient to sustain cultivation, even without full title deeds, thereby explaining the high land scores despite the absence of formal freehold tenure in many public schemes. Nhamo, Mabhaudhi et al. (2020a) reported similar patterns in irrigation schemes in the Limpopo Province, noting that community-based land governance through traditional authorities yields relatively high effective land security scores in nexus assessments, despite formal tenure insecurity.

The overall CI range of 0.53–0.74 indicates moderately sustainable nexus performance, consistent with the threshold classifications proposed by Simpson et al. (2022), who define CIs below 0.40 as critically unsustainable, 0.40–0.60 as moderately unsustainable, 0.61–0.80 as moderately sustainable, and above 0.80 as highly sustainable. Under this classification, Mphaila (0.74) approaches the upper boundary of moderate sustainability, whereas Rambuda (0.53) falls within the moderately unsustainable range, underscoring the need for targeted investment in energy and water infrastructure.

These moderate CI values (0.53–0.74) have practical implications for farmers’ livelihoods and food security. This moderate sustainability threshold means farmers can produce crops for part of the year but face recurring constraints that limit yield potential.

3.3 LWENI interpretation

The calculated LWENI is presented as a spider graph to provide a general overview of the numerical relationships among the indicators (Figure 3). The graph shows the interactions and interdependencies among the LWE sectors for each irrigation scheme, providing an integrated overview of resource status in the context of use and management. The deformed shape of the centre-pieces (Figure 3) indicates an imbalance in resource management. The center of the spider graph indicates how close it is to the axis, signifying unsustainable resource management.

Position along each axis reflects sustainability: values near the center indicate unsustainable resource management, while values near the outer edge indicate highly sustainable management.

The spider diagram shows notable variations in LWE resource performance across the four irrigation schemes (Figure 4). Mphaila performs strongest overall (CI = 0.74), driven by near-perfect land availability and accessibility and comparatively higher energy scores. Folovhondwe and Kumbe show solid land metrics but critically weak energy components. Rambuda records the weakest scores overall (CI = 0.53), driven by the lowest water availability and accessibility and weak energy accessibility.

These uneven profiles reflect common trade-offs and imbalances in smallholder irrigation systems, where water constraints often limit nexus performance despite varying land strengths (Mabhaudhi et al., 2019; Taguta et al., 2026). Multiple factors contribute to water insecurity among irrigated farmers, including allocation issues, insufficient infrastructure, suboptimal water system performance, limited farmer participation in water management, and land tenure insecurity (Schreiner and van Koppen, 2020; Mugejo et al., 2022).

This local assessment aligns closely with broader WEF nexus findings for African irrigation schemes and small-scale farmers (Simpson et al., 2023), which frequently document moderate-to-low composite indices and irregular radar/spider chart shapes due to resource imbalances. Applications of the iWEF tool in South Africa’s Inkomati-Usuthu Water Management Area and Zimbabwe’s Mzingwane Catchment yielded composite WEF nexus indices of approximately 0.209 and 0.189, respectively, indicating suboptimal sustainability, with weaknesses in water availability, accessibility, and productivity patterns that mirror the diagram’s lower water and energy scores (Taguta et al., 2026). Similar radar visualizations for southern Africa show a strong emphasis on food self-sufficiency but deficits in the water and energy pillars, driven by unreliable irrigation infrastructure, climate variability, and siloed management typical of smallholder schemes (Mabhaudhi et al., 2018; Nhamo et al., 2018).

Comparisons with other African smallholder and irrigation contexts highlight shared vulnerabilities and opportunities for integrated interventions. Studies on Eastern Cape and Limpopo smallholders emphasize land-water tensions, governance challenges, and underperformance in small-scale irrigation, underscoring the need for nexus approaches such as solar-powered systems and improved water productivity to achieve more balanced outcomes (Musetsho et al., 2024; Tantoh and McKay, 2023). Continent-wide and regional analyses recommend expanding efficient irrigation while addressing energy and land trade-offs through coordinated planning, consistent with the diagram’s indication that targeted improvements, for instance, water infrastructure in lower-performing schemes, could enhance overall WEF security (Mabhaudhi et al., 2019).

3.4 Winter seasonal precipitation

Figure 5 shows winter (April–September) seasonal precipitation (mm) recorded at four AWS. It highlights the strong seasonal fluctuations typical of semi-arid subtropical locations, with precipitation ranging from below 100 mm in arid years to over 600 mm in pluvial years. Significant characteristics include a pronounced increase around 2005 (notably at Thohoyandou and Levubu, surpassing 500–600 mm), many moderate-to-high seasons in the early 1980s and mid-1990s, and more recent heightened totals in 2022–2023. Levubu often receives more precipitation because the Soutpansberg Mountains enhance rainfall, whereas Messina is generally drier. This geographical and temporal variability reflects local topography, the prevalence of convective thunderstorms, and the impact of large-scale phenomena like the El Niño-Southern Oscillation (ENSO). Winter (April–September) is the arid season in this subtropical area, characterized by low precipitation (averaging <20–50 mm monthly, with July generally the driest) (Adeola et al., 2019; Shikwambana et al., 2024).

The winter seasonal graph shows a general decline and significant variability among stations during the 2000s, marked by distinct dry intervals (e.g., lows from the mid-1980s to the 1990s and volatility after 2010). Levubu and Thohoyandou show elevated seasonal peaks and variability, indicating wetter microclimates relative to the arid Messina (green) (Nenwiini and Kabanda, 2013; Odiyo et al., 2019). Winter-specific trends are harder to identify directly but align with broader regional patterns of low baseline precipitation, interspersed with sporadic intrusions. No significant rising trend in winter conditions is apparent; rather, the data suggest the continuation of dry winter conditions amongst interannual fluctuations (Shikwambana et al., 2021). Similar research on the Limpopo Province regularly indicates diminishing or stable-to-declining annual and seasonal precipitation, with the winter (dry season) remaining particularly parched and susceptible to further decline due to climate change. Research using climatic data from 1960 to 2018 in the Mopani and Vhembe districts indicated a substantial reduction in annual rainfall (~38% in some assessments), an increased frequency of drought, and elevated potential evapotranspiration, thereby intensifying dry-season water stress (Shikwambana et al., 2021).

Winter precipitation is scarce and shows a limited, not statistically significant upward trend (CSIR, 2024). Research on extreme rainfall in Vhembe indicates more frequent heavy rainfall events, mostly in summer, while successive dry days (associated with winter) show reductions in some indices; overall, drier conditions continue to prevail (Munzhedzi et al., 2024; Shikwambana et al., 2021). This aligns with variable patterns at ARC-NRE stations (e.g., Tshiombo/Levubu variations) and is consistent with farmer perceptions and hydrological assessments, indicating predominantly moderate dryness and mixed station trends (Odiyo et al., 2019). In conclusion, winter precipitation at the Vhembe stations is minimal.

3.5 Summer seasonal precipitation

Seasonal precipitation (October–March) was recorded at four AWS (Figure 6). Thohoyandou has moderate precipitation totals, typically ranging from 500 to 1,200 mm, with significant maxima in the early 1980s and mid-1990s. However, recent decades show a discernible trend toward lower and more irregular values, particularly after 2015. Messina Proefplaas typically records lower precipitation, often below 800 mm, reflecting its more northern/semi-arid location, with some higher years but an overall muted trend. Levubu has the highest and most varied precipitation totals, with pronounced surges exceeding 2000 mm between 2000/01 and 2012–2014, reflecting its wetter, subtropical microclimate in the Levubu valley; recent years indicate a degree of recovery after the drier mid-2010s period. Tshiombo shows significant fluctuations, marked by pronounced peaks in the late 1990s to early 2000s, including a notable high between 2000 and 2003, followed by a decrease and subsequent oscillations at lower levels. The data reveal significant seasonal fluctuations characteristic of the area, with no consistent linear trend, and indicate clusters of wetter and drier intervals.

Vhembe District summer (October–March) precipitation reflects this unpredictability, often showing non-significant long-term decreases or shifts toward more intense but irregular episodes. Research using similar station data by Nenwiini and Kabanda (2013) indicated substantial negative rainfall trends in several Vhembe locations from the mid-1990s onwards, associated with deforestation and overarching drying patterns, with Sen’s slopes reflecting decreases of several mm/year, especially in humid regions. Again, Shikwambana et al. (2021) reported a 38% reduction in annual precipitation in Vhembe and Mopani from 1960 to 2018, with significant interannual variability and more frequent drought. However, exceptional events may obscure the statistical significance of the overall trend.

Recent analyses by Munzhedzi et al. (2024) indicated substantial increases in total precipitation at certain northern locations, coupled with a reduction in consecutive dry days. This suggests a transition to fewer, more intense summer rainfall events rather than a uniform drying trend, consistent with pronounced spikes in the graph, particularly at Levubu and Tshiombo. Forecasts and further catchment analyses (Levubu River) reveal significant uncertainty in future totals; however, they suggest a trend of delayed start, extended early-season droughts, and general subtropical aridity, hence increasing agricultural hazards in this summer-rainfall area.

3.6 Linear regression for annual precipitation

The linear regression for annual precipitation was conducted on data records at four AWS (Table 6). Year-to-year annual precipitation showed a statistically significant increasing trend (3.89 mm/year, p < 0.04). However, the model explained only a small fraction of the year-to-year variability (adjusted R² = 0.01).

Table 6. Linear regression for winter and summer precipitation in the Vhembe District.

The observed increase in annual precipitation, while statistically significant, is weak and explains almost none of the year-to-year variation in rainfall (adjusted R² = 0.01). In practical terms, this means the long-run trend offers little guidance for what any given season will bring. Farmers experience not a steady rise in water availability, but a highly erratic sequence of wetter and drier years driven by factors the model does not capture, such as interannual climate variability, large-scale atmospheric circulation, and multi-year oscillations like ENSO. This unpredictability, rather than a decline in total rainfall, undermines water security for smallholder irrigators in the Vhembe District. The trend is statistically real but weak, and vulnerability is what drives irrigation vulnerability.

Water restrictions affect smallholder irrigation farmers by reducing crop yields due to limited access to irrigation and water-saving technologies, by forcing competition with other sectors for scarce water, and by creating potential conflicts within communities. South Africa’s constitutional water right guarantees sufficient water for domestic use only, not for agriculture, so during droughts, communal and smallholder farmers needing water for irrigation often must buy it from private suppliers (Makaya et al., 2020).

Climate change is expected to intensify this variability rather than simply reduce total rainfall, producing more frequent and more extreme dry spells alongside episodes of intense rainfall. During these dry-year episodes, reduced river flow and increased evaporation from a drier atmosphere can still constrain reservoir water available for irrigation (Arfasa et al., 2024) even where the multi-decade trend in annual totals is nominally positive.

3.7 Winter minimum and maximum temperatures

The minimum and maximum winter (Apr to Aug) temperatures were recorded from 1980 to 2024, at four weather stations (Figure 7). The top cluster of lines (highest winter temperatures) typically ranges from 24–31 °C, with Messina Proefplaas and Thohoyandou often exhibiting elevated peaks, especially after 2010, when maxima sometimes surpass 30 °C. The lowest cluster of winter minimum temperatures ranges from 9–14 °C, with Thohoyandou and Levubu generally showing higher temperatures than Tshiombo.

The data show seasonal variability and decadal fluctuations rather than a significant linear trend; however, several maximum-temperature series show discernible upward trends, particularly in Messina and Thohoyandou, alongside relatively stable or slightly increasing minima in some instances. This indicates milder winters with fewer cold extremes in some areas of the region.

Messina Proefplaas and Thohoyandou have higher maximum temperatures than the more inland or elevated locations of Tshiombo and Levubu, reflecting microclimatic variations within the Levubu River catchment region (hotter, drier lowveld versus somewhat milder subtropical zones). Minimum temperatures show less divergence while following similar fluctuation patterns. These data align with extensive regional research indicating statistically significant warming in Limpopo province, with annual minimum temperatures increasing by around 2.3 °C and maximum temperatures by roughly 1.6 °C from 1960 to 2018, with Vhembe recognized as a climate change hotspot (Shikwambana et al., 2021). Winter warming, historically around 0.18 °C per decade, reduces frost risk but exacerbates evaporation and drought stress on crops, and alters agricultural seasons for smallholder farmers dependent on rainfed agriculture (Shikwambana et al., 2021).

3.8 Summer minimum and maximum temperatures

The time-series comparison of maximum and minimum temperatures across four weather stations from the 1980/81 to 2023/24 seasons (Figure 8). Maximum temperatures (upper panel) show clear differentiation between sites. Messina Proefplaas consistently records the highest maxima, often ranging between 32–36 °C with notable peaks around 2015–2018 and 2020. This can be explained by considering the locations of the various automatic weather stations: the one at Messina is near the Limpopo River at an altitude of 535 m above sea level, whereas the others are at 700 m above sea level and in well-vegetated, fruit-growing areas of the bushveld. The other three sites (Thohoyandou, Levubu, and Tshiombo) cluster more closely around 27–31 °C, displaying similar season-to-season fluctuations and a slight upward trend over the last decade. Extreme summer maximum-temperature events, such as hot, dry conditions, threaten agricultural and water resources by rapidly depleting soil moisture through increased evaporation.

Minimum temperatures (lower panel) follow a similar pattern but at lower values: Messina again leads (typically 19–22 °C), while the remaining sites hover between 16–20 °C. All series exhibit substantial inter-seasonal variability, with visible dips in the early 2000s and recoveries after 2010 (Moeletsi et al., 2022; Shabalala et al., 2019).

Overall, the data suggest a broadly warming trend across the region, particularly in maximum temperatures during recent years, though Messina Proefplaas stands out as the warmest location throughout the record. This aligns with regional climate patterns in northern South Africa, where rising temperatures can affect subtropical fruit and crop production (Shikwambana et al., 2021). Messina’s elevated readings likely reflect its more inland/arid microclimate compared to the other sites, which may benefit from slightly higher elevation or proximity to moisture sources. These differences highlight the importance of location-specific climate monitoring for agricultural planning and adaptation strategies amid climate variability (Moeletsi et al., 2022).

3.9 Linear regression of maximum temperatures

The linear regression of maximum temperature in the Vhembe District, Limpopo Province, from 1980 to 2024 (Table 7). The results show that the model explained 10% of the variance in the maximum annual temperature, with an adjusted R² of 0.10. The model was statistically significant (p < 0.001), with maximum temperature showing a significant upward trend of 0.04 °C per year.

Table 7. Linear regression for maximum temperature in the Vhembe District.

The results are consistent with temporal changes in maximum temperatures at selected locations in Southern Africa, where an upward trend was observed, reaching a maximum of 0.01–0.09 °C over 38 years (Mupangwa et al., 2023). Similarly, in South Australia, climatic data from various weather stations showed that summer and winter maximum temperatures increased by 1.1 °C and 0.7 °C, respectively, from 1970 to 2021 (Ferreira et al., 2023). High temperatures and low precipitation are likely to reduce water availability for rainfed agriculture (Maúre et al., 2018; Mupangwa et al., 2023), and irrigation demand is likely to increase (Arfasa et al., 2024). Rising temperatures are projected to reduce agricultural water supply in the semi-arid and sub-humid highland areas of southern Africa (Nhemachena et al., 2020). The decrease will, in turn, affect crop productivity in irrigated systems, and yields are projected to decline (Kusangaya et al., 2014).

3.10 Climate moisture index (CMI)

A linear regression of CMI on year confirmed a statistically significant downward trend (B = −0.007, SE = 0.003, β = −0.337, t = −2.478, p = 0.017), with year explaining approximately 10% of the variance in CMI (R² = 0.10). This finding aligns with the annual-level trend analysis reported above, which likewise identified a significant decline in CMI over 1980–2024. This aligns with the annual-level trend analysis above, confirming a consistent, modest aridification signal across both temporal aggregations (Table 8).

Table 8. Linear regression for CMI in the Vhembe District.

The district’s CMI shows a statistically significant negative trend of 0.01 units/year, indicating a gradual, consistent drying. The annual decline is small, but over decades it erodes the district’s effective water balance, the surplus of precipitation over potential evapotranspiration (PET) that sustains soil moisture, streamflow, and groundwater recharge. A diminishing CMI signals that moisture deficits, when atmospheric water demand (PET) exceeds supply (precipitation), are becoming more frequent, persistent, or severe. The water balance is affected by reduced effective recharge to groundwater aquifers, streamflow, and reservoir inflows, as well as shorter soil moisture windows for natural vegetation and rainfed agriculture between rainfall episodes. Because CMI integrates precipitation and PET, the decline may reflect falling precipitation, rising PET (driven by warming and increased evaporative demand), or both, a pattern increasingly observed in regions experiencing climate warming without significant precipitation change. As the CMI declines, agriculture relies more on supplementary irrigation and the gap between agricultural water needs and rainfall-supplied moisture widens. This compounds pressure: water availability is decreasing while irrigation demand is rising to compensate for the natural moisture shortfall (Table 9).

Table 9. Correlation analysis between LWENI components and climate variables.

A continued decline in CMI would likely increase reliance on groundwater and surface storage to meet irrigation demand; because groundwater recharge is tied to the same precipitation-PET balance, this could plausibly raise the risk of unsustainable groundwater drawdown over time. However, this study did not directly measure groundwater levels, extraction volumes, or irrigation demand, so treat this relationship as a hypothesis for future monitoring rather than a demonstrated outcome.

Based on the temperature trend reported here and in regional literature, higher winter and summer maximum temperatures (23–36 °C range) are expected to increase soil moisture depletion, heat stress during critical growth stages, and pressure on river and reservoir water supplies. These climatic pressures may compound the resource constraints already reflected in low LWENI scores. This study did not directly measure soil moisture, irrigation performance, or crop yields; the risks to yield, scheduling, and resource conflict described here come from regional literature, not this study’s data.

3.11 Correlation analysis between LWENI components and climate variables

The statistically significant relationships (p < 0.05) are between maximum temperature trend and the two water indicators, water availability (R² = 0.98, p = 0.022) and water accessibility (R² = 0.99, p = 0.013).

With only four schemes, this analysis is exploratory, and these coefficients should not be interpreted as evidence of a meaningful or causal relationship. Descriptively, the two water indicators covary strongly with the local maximum-temperature trend in this small sample (water availability R² = 0.98, water accessibility R² = 0.99), but statistical inference from n = 4 is severely limited: a single additional or differently sited observation could substantially change or reverse these coefficients. Treat this pattern as a hypothesis to test with a larger sample, not as evidence that local warming is associated with higher water availability or accessibility.

The overall LWENI composite shows no significant correlation with any climate variable (all p > 0.4). Resource endowment (land/water/energy) and current climate trends are also statistically independent across these four schemes, at least at this sample size. This justifies developing a separate climate-exposure resilience indicator.

Energy and land indicators show weak, non-significant correlations with all climate variables, consistent with energy and land being driven by scheme-specific infrastructure and land tenure history rather than by the recent climate trends measured here.

Climate change is understood here as a stress multiplier acting through infrastructure and energy demand, rather than as a force that has already visibly reshaped the LWE balance. Its future threat is real and well documented (via temperature trends), but its differential impact across schemes depends more on whether a scheme has functioning storage and independent (non-grid) energy access than on which station shows the steepest warming trend.

Therefore, the LWENI can be adapted to incorporate climate vulnerability and resilience through four nested interventions of increasing complexity: appending a standalone Climate Exposure Index as a fourth pillar; replacing static water indicators with climate-adjusted water productivity measures; adding a Climate Resilience Sub-Index that scores energy independence, storage capacity, repair capacity, and adoption of climate-smart practices; and ultimately constructing a dynamic, scenario-updatable LWENI that projects sustainability trajectories under defined warming scenarios. Together, these adaptations would transform the LWENI from a diagnostic snapshot of current resource endowment into a forward-looking planning instrument that can guide investment sequencing, infrastructure prioritization, and climate adaptation strategy across Vhembe District and analogous semi-arid smallholder irrigation contexts in southern Africa.

4 Conclusion and recommendations

The LWENI assessment and long-term climatic trends were combined. This analysis looked at four purposively selected irrigation schemes in Vhembe District. The schemes range from moderately unsustainable to moderately sustainable. Composite scores for these schemes range from 0.53 to 0.74. Three schemes- Folovhodwe, Kumbe, and Rambuda have scores between 0.53 and 0.59. These fall within the moderately unsustainable range. Only Mphaila scores 0.74, which approaches moderate sustainability.

The schemes and participating farmers were purposively selected. These findings should be seen as a comparative assessment of the four selected schemes. They are not statistically representative of all smallholder irrigation schemes in Vhembe District. Individual indicators still show imbalanced resource management at each scheme. There is also degraded infrastructure. Energy supply is unreliable. Increasing climatic variability affects each scheme.

Climate data from 1980–2024 further contextualizes these challenges, showing a statistically significant but weak upward trend in precipitation (3.89 mm/year) amid high variability. Summer and winter show extreme peaks that cause floods and lows that signal droughts, leading to erosion, sedimentation, and water stress.

Linear regression shows a statistically significant but weak upward trend in annual precipitation. The increase is 3.89 mm per year (F = 4.253, p < 0.04, adjusted R² = 0.01). This explains only about 1% of the year-to-year variance. Maximum temperatures show a clear warming trend of 0.04 °C per year (adjusted R² = 0.10, p < 0.001). Minimum temperatures decline slightly. The overall trend is toward warming. This warming is expected to pose risks to crop growth. Winter-grown staples like maize and vegetables are especially at risk, as they depend heavily on irrigation. This study did not directly measure crop growth or yield.

Two limitations of this assessment should be noted explicitly. First, the composite LWENI reported here (Table 5; Section 3.2) is the equal-weighted arithmetic mean of the six min-max normalized indicators for each scheme [CIi = (1/n) × Σ Wj × Iij], applied directly to the normalized indicator values. This is the only aggregation method used to generate the differentiated scores of 0.53–0.74 reported throughout this study.

Aggregating four schemes into a single LWE assessment simplifies real differences in their energy setups. Folovhodwe, Kumbe, and Rambuda use gravity-fed canals. They supplement these with private fuel pumps that are informally and unevenly distributed. Mphaila relies on one collectively funded, grid-powered pump. This is a very different risk profile. A pooled scheme-level index cannot fully capture this difference. Aggregation is still a reasonable unit of analysis. The schemes share typology, tenure arrangements, and a canal-based water delivery system. However, each scheme’s unique energy vulnerabilities should guide interpretation and use of the results.

While land tenure provides some stability, deficiencies in water access and energy reliability, coupled with rising temperatures, declining CMI, and erratic precipitation, create a precarious environment for smallholder agriculture. Together, these factors threaten productivity, food security, and livelihoods in this semi-arid region. To enhance sustainability, targeted interventions should be scheme-specific rather than uniform. Prioritize water infrastructure rehabilitation, such as balancing dams and distribution canals, at Rambuda, where strong land availability (0.95) is undermined by the weakest water and energy access among the four schemes, driving its lowest composite score (0.53). This should be paired with farmer training in efficient water management and greater involvement in decision-making. Energy solutions should likewise be matched to each scheme’s configuration: reliable solar-powered pumping systems, leveraging Vhembe District’s strong solar irradiance, could directly replace the informal, unevenly distributed fuel pumps used at Folovhodwe, Kumbe, and Rambuda, while at Mphaila, solar conversion would reduce dependence on Eskom and its rising tariffs, a cumulative increase that currently falls as a fixed-cost burden on the scheme’s pooled electricity bill regardless of harvest outcomes. Solar pumping, in particular, offers real resilience benefits because it is decoupled from the grid and unaffected by load-shedding schedules.

However, for smallholder farmers, the upfront cost of a solar system and the maintenance it requires over its lifetime can be a significant barrier to adoption. Kumbe’s comparatively strong scores for water availability and accessibility suggest that its canal and balancing-dam management practices could serve as a reference model for rehabilitation efforts at the other three schemes. Climate-smart practices, including drought-resistant crops, improved irrigation technologies, and soil moisture conservation, should be promoted alongside strengthened early-warning systems for floods and droughts. Policy support for secure land tenure, integrated LWE nexus planning that accounts for scheme-level energy heterogeneity rather than treating energy access as a uniform constraint, and investment in resilient infrastructure will be critical to elevating LWENI scores, particularly at Rambuda and Folovhodwe, and to building long-term adaptive capacity in Vhembe District irrigation schemes.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Masoud Irannezhad, Australian University - Kuwait, Kuwait

Reviewed by: Inoussa Zagre, UNFCCC, Germany

Putu Doddy Heka Ardana, Universitas Ngurah Rai, Indonesia