Nutrition-sensitive climate risk across food production systems

Our analysis focuses on five essential micronutrients—calcium, folate, iron, vitamin A, and vitamin B12—and dietary energy. Folate, iron, vitamin A and vitamin B12 were selected because they are among six ‘sentinel micronutrients’ that cause severe morbidity and for which deficiency is prevalent [3]. In addition, calcium was chosen because deficiency is widespread, particularly in low and middle-income countries [26].

The availability and dietary sources of the five micronutrients and dietary energy were derived from the Global Nutrient Database, which estimates the availability of 156 nutrients in 195 countries and territories [22], most recently available for 2017. To link nutrient supply data to climate hazards, we divided all food items into six food groups: (1) aquatic products; (2) fruits & vegetables; (3) legumes & nuts; (4) cereals & tubers; (5) livestock products (including dairy and eggs); and (6) other crops (mostly oil crops, sugar crops, and spices). This grouping aligns with commonly used dietary food groupings [27], although we separated aquatic products from other foods given their distinct production environment and nutritional benefits [28]. Additional details on nutrient supply data and limitations can be found in the supplemental materials.

Due to the lack of process-based or statistical models for many food production systems, our climate hazard assessment focuses on the probability of future conditions deviating significantly from their historical range, compared to a baseline period [21]. This approach addresses the challenge of establishing science-based thresholds for all food products by recognizing the potential limitation of local food production systems outside their accustomed climate range [21, 29], serving as an indicator of the adaptation demands posed by climate extremes.

We assessed climate hazards by quantifying the likelihood of adverse climate conditions in 2041–2060 relative to historical baseline period (1961–1990) [21]. We used the historical and SSP3-7.0 (medium-high emissions) simulations from 30 models in the Coupled Model Intercomparison Project 6 (CMIP6) database [30], recognizing that the standard ‘business-as-usual’ scenario (SSP5-8.5) is currently considered statistically improbable [31]. For each of the six food groups, we developed spatial and seasonal masks (table 1) to calculate national-average anomaly time series relative to the baseline mean:

$$\begin{align*}{x_{t,f}} = \frac{{\mathop \sum \nolimits_i^{} {w_{i,m,f}}{x_{t,i,m}}}}{{\mathop \sum \nolimits_i^{} {w_{i,m,f}}}}\end{align*}$$

$$\begin{align*}\Delta {x_f} = {x_{2041 - 2060,f}} - {x_{1961 - 1990,f}}\end{align*}$$

where x_t,i,m represents the climate variable for each year, grid cell, and month, and w_i,m,f the area and seasonal weights for each food group.

Climate hazard is the probability of a growing season in which a climate variable—temperature¹⁰, precipitation, or the combination thereof—exceeds a threshold $\alpha$:

Finally, we derived a single climate hazard value for each nutrient and country by weighting the climate hazards for the six food groups by the relative contribution of each food group to the total supply of each nutrient [22]:

$$\begin{align*}{H_{\text{n}}} = \frac{{{{\mathop \sum \nolimits}}_{\text{f}}^{}{H_{\text{f}}} \times {S_{{\text{f,n}}}}}}{{{{\mathop \sum \nolimits}}_{\text{f}}^{}{S_{{\text{f,n}}}}}}\end{align*}$$

where S is the supply of nutrient n from food group f.

Although trade is an important driver of nutrient supply in many countries [36, 37], we focus in our risk calculations on the share of a country’s nutrient supply that is exposed to climate hazards domestically. We calculated this using the SSR of each nutrient. The SSR indicates the extent to which a country’s total use of a nutrient is supplied by domestic production:

$$\begin{align*}{\text{SSR = }}\frac{{{\text{production}}}}{{{\text{total use}}}}.\end{align*}$$

The term ‘total use’ encompasses all forms of utilization over and above direct food consumption, i.e. animal feed, waste, and all other forms of utilization¹¹. To calculate an exposure score for each nutrient, we weighted the SSR for the six food groups by the relative contribution of each food group to the total supply of each nutrient:

$$\begin{align*}{E_{\text{n}}} = { }\frac{{\sum\limits_{\text{f}}^{}{\text{SS}}{{\text{R}}_{\text{f}}} \times {S_{{\text{f,n}}}}}}{{\sum\limits_{\text{f}}^{}{S_{{\text{f,n}}}}}}.\end{align*}$$

Exposure to hazard is then the product of exposure and hazard:

$$\begin{align*}E{H_{\text{n}}} = {E_{\text{n}}} \times {H_{\text{n}}}.\end{align*}$$

For inclusion in the climate risk score calculations, exposure to hazard scores were first standardized and normalized.

We calculated the vulnerability component of climate risk using key indicators to represent sensitivity and adaptive capacity to disruption of nutrient supply. Where possible we chose values for 2017, to align with nutrient supply data.

Nutrient vulnerability NutV_n is the average between the prevalence of inadequate nutrient intake (def_n; micronutrients [23]:—2011; dietary energy [38]:—2017) and dietary diversity div_n, as measured by the species evenness score across the six food groups [22, 39]—2017. Economic vulnerability EconV is the average between the latest available percentage of the population with income below the national poverty line pov [24], latest available per capita Gross Domestic Product at Purchasing Power Parity gdp [24], and the 2017 cost of a healthy diet cost [24]. The combined vulnerability score V_n is the average between NutV_n and EconV_n.

$$\begin{align*}Nut{V_{\text{n}}} & = \left( {def^{^{\prime}}_{\text{n}} + div^{^{\prime}}_{\text{n}}} \right)/2; \nonumber\\ EconV &= \left( {pov^{\prime} + GDP^{\prime} + cost^{\prime}} \right)/3\end{align*}$$

$$\begin{align*}{V_{\text{n}}} = \left( {NutV^{^{\prime}}_{\text{n}} + EconV^{\prime}} \right)/2\end{align*}$$

where apostrophes indicate standardized and normalized values. For nutrient and economic vulnerability, an average score was still calculated if either of the indicators was missing; for the combined vulnerability score, no score was calculated if either of the input values was missing.

Domestic nutrition-sensitive climate risk is the arithmetic mean between the exposure to hazard score and the vulnerability score:

$$\begin{align*}{R_{\text{n}}} = \left( {EH^{^{\prime}}_{\text{n}} + V^{^{\prime}}_{\text{n}}} \right)/2.\end{align*}$$

This method assigns equal weight (50%) to both vulnerability and exposure to hazard, creating a risk score that reflects the climate change risk to the domestically produced nutrient supply. This assessment does not account for risk through global markets, which is addressed in other studies [e.g. 40–42]. As a final step, we normalized and standardized the resulting risk scores, resulting in a nutrition-sensitive climate risk score ranging from 0 to 1. If data on either exposure to hazard or vulnerability was unavailable, the risk was not calculated.

To identify patterns in the specific nutrition-sensitive climate risk that countries face, we applied a K-means clustering analysis to the variables used in the climate risk calculations. We excluded countries that have more than half of these variables missing. Determination of the optimal number of clusters was facilitated through the use of statistical metrics such as the gap statistic and the Davies–Bouldin Index.

A list of key climate resilience and nutrition interventions for each of the climate risk profiles was developed through a broad search of peer-reviewed literature using a combination of keywords, including ‘nutrition policy’ ‘food security policy’ ‘climate change,’ and ‘policy recommendations.’ The search was limited to articles published in the English language from the year 2005 to present. We included articles that (a) are published in peer-reviewed journals, (b) identify action in either a case study, cross-sectoral, or multi-sectoral approach, and (c) provide concrete recommendations for policy action at one or multiple scales. The identified recommendations were then condensed to focus on those that tie directly to elements included in our analysis. Following several rounds of feedback and revision by our multidisciplinary team of co-authors, the list was narrowed and grouped into thematic categories. The relevance of each intervention for each climate risk profile was assessed based on the input variables to the cluster analysis.

Temperature extremes serve as a primary indicator of climate impacts because long historical records exist and projections are consistent between climate models. Temperature extremes are projected to become increasingly prevalent by the middle of the century (figure 1 (a), SI figure 1). They emerge most rapidly in tropical regions with low interannual climate variability, and in countries with large continental landmasses. In contrast, their emergence is slower in mid-latitudes and regions dominated by monsoon climates. The pace and magnitude of temperature increase differs substantially across the production areas of different food groups (SI figure 1). In countries with vast land areas (e.g. Brazil, India, the United States) different foods are produced in distinct regions, resulting in distinct climate trajectories. Notably, aquatic environments (SI figure 1 (f)) are governed by different dynamics than terrestrial ones. In some countries (e.g. Australia) marine temperature extremes emerge more rapidly than terrestrial ones, while in others (e.g. Peru, with a highly variable coastal upwelling system) they lag behind.

Zoom In Zoom Out Reset image size

For terrestrial crops, extreme heat becomes particularly detrimental when coupled with dry conditions [43–45]. The spatial distribution of the probability of compounded hot and dry conditions (figure 1(b), SI figure 2) differs substantially from that of hot conditions alone (figure 1(a)) and is primarily driven by projected declines in national-average precipitation. Consistent with earlier work on fewer crops [21], we find that this interaction is particularly pronounced in the Mediterranean region, northern and southern Africa, Central and South America, and Australia. Globally, terrestrial crop systems are projected to face compound hot and dry conditions approximately once every three years by the end of the century (figures 1(c)–(f), SI figure 2), the impacts of which will vary depending on the crop type and system (e.g. irrigated vs. rainfed). In the remainder of this paper, we base calculations of climate risk on probabilities of compound hot and dry conditions for terrestrial crops (SI figure 2), and hot conditions alone for aquatic products and livestock products (SI figure 1).

Climate risk assessments that exclusively focus on staple crops such as maize, wheat, and rice place a premium on dietary energy compared to micronutrients (figure 2). Whereas Cereals & Tubers and foods in the ‘Other Crops’ food group (largely oil crops and sugar) provide most of global energy supply, livestock products and fruits & vegetables provide the bulk of global calcium and vitamin A. Folate and iron derive from a larger diversity of foods, which also include legumes & nuts and smaller quantities of aquatic foods. Vitamin B-12 derives exclusively from animal-source foods, either aquatic or terrestrial.

Zoom In Zoom Out Reset image size

Even without the impacts of climate change, food systems in large parts of the world fail to deliver access to diets that meet minimum nutrient requirements. Of the 167 countries with data on micronutrient deficiencies [23], 20% or more of the population is deficient in at least one or the micronutrients evaluated here in 141 countries and in all five micronutrients in 22 countries (SI figure 11). South and Southeast Asia and sub-Saharan Africa have higher-than-average nutrient deficiencies, with substantial shortfalls in calcium, iron, and Vitamin A supply (figure 2). These places have comparatively lower supplies of nutrients from non-staple foods, especially livestock products and fruits & vegetables.

Large regional differences also exist in the relative contributions of different food groups to nutrient supply. For example, livestock products provide a larger share of the nutrient supply of North America, Europe, and Oceania—in conjunction with lower deficiencies in calcium and vitamin B-12—whereas Fruits & Vegetables play a larger role in East Asian diets—in conjunction with lower deficiencies in folate. It is worth noting that regional averages mask important sub-regional variation (SI figures 4–9). Aquatic Products, for example, contribute relatively little to regionally averaged nutrient supply for four of the five micronutrients considered here, but make high contributions in specific countries [46], for instance nearly 15% of iron in the Maldives, and more than a fifth of calcium and vitamin A in Cambodia.

Combining climate hazard estimates for different food groups (figure 1) with insights into their contributions to nutrient supply (figure 2) we calculate integrated climate hazards confronting domestic production of each nutrient (figure 3). Despite the relatively simple hazard indicators utilized in this study, we find substantial variation in integrated hazard scores across different nutrients. Hazard scores are generally highest for vitamin B-12 and calcium, which have large contributions from the animal-source foods that in our model are connected only to temperature extremes. Several regions face high hazards across all micronutrients, notably the Mediterranean region and Central America. Other regions, such as sub-Saharan Africa and South America, face especially high climate hazards in one or two specific nutrients. Globally, 75% of calcium, 30% of folate, 39% of iron, 68% of vitamin A, 79% of vitamin B12, and 54% of energy production is projected to face climate extremes at least every other year by the middle of the century.

Zoom In Zoom Out Reset image size

To assess nutrient-specific climate vulnerability we examine a combination of dietary and economic indicators (figures 4(a) and (b); see Methods). We assume populations to be vulnerable to a (temporary) loss in nutrient supply if they are already consuming the nutrient below recommended levels (figure 2; SI figure 10), if they are highly dependent on a single food source for their nutrient supply (that is, low nutrient supply diversity; SI figure 12), or if they face high economic barriers to accessing healthy diets (SI figure 13). We find nutrient supply diversity to be lowest in large parts of Africa, and Central and South Asia. Economic and nutrient indicators of vulnerability generally are highly correlated, except for calcium which in high-income (Western) countries is almost exclusively derived from a single food source (dairy), exhibiting low supply diversity (SI figure 12). In aggregate, countries in sub-Saharan Africa are most in need of investments to reduce vulnerability to climate shocks (figure 4(a)): out of the 23 countries that make up the top ten most vulnerable countries for each of the micronutrients, eighteen are located in this region.

Zoom In Zoom Out Reset image size

The growing role of international trade in nutrient supply [36, 37] complicates assessment of climate risk: countries can both be impacted through domestic climate hazards or through changes in international markets [29, 40, 47]. To estimate domestic nutrient climate risk, we define exposure as the share of the total nutrient supply derived from domestic production (figure 4(c), SI figure 14; see methods). By this definition, the countries with the highest average exposure are Argentina, Uganda, and Brazil, whereas Saint Kitts and Nevis, Saint Lucia, and the United Arab Emirates are most import-dependent. Combining then climate hazard, vulnerability, and exposure into a holistic domestic nutrient climate risk score (figure 5, SI figure 15), we estimate that 82 countries—containing more than half of the current global population (57%)—face high climate risk (exceeding the 75th percentile) to the domestic supply of at least one micronutrient (figure 5(a)). 48 of these countries face high climate risk for two or more micronutrients (figure 5 (b)), containing 1.7 billion people, or 22% of today’s world’s population. Ten countries—Afghanistan, Bolivia, Guatemala, Guinea–Bissau, Haiti, Honduras, Madagascar, Mozambique, Nicaragua, and Zambia—face high climate risk to domestic supply in all five micronutrients. It is worth noting however that these results present a relative ranking of climate risk: low values do not indicate the absence of risk.

Zoom In Zoom Out Reset image size

In addition to domestic climate risk, importing countries can also be exposed to climate hazards through international markets [36, 37, 40, 41]. Given the dynamic and evolving nature of international trade networks [48], a true estimate of the climate risk associated with imports would require a model that can account for substitution based on supply, demand, prices, and policies, which is beyond the scope of this paper. However, based on our analysis we are still able to identify countries that are both highly exposed to global markets and highly vulnerable to loss of nutrition (SI table 1), creating risk from trade-related climate impacts. This primarily concerns arid countries in North Africa and the Middle East, small island states, and smaller countries in Africa. These countries will be highly affected by climate shocks to production in major exporting countries [29]. By the middle of the century, micronutrient production in the top 10 producing countries faces adverse climate conditions on average every other year (ranging from 42% for folate to 59% for vitamin B12), suggesting that investments are urgently needed to both strengthen domestic production systems in importing countries and increase the resilience of global markets (see Discussion).

The components of nutrition-sensitive climate risk presented here can help inform context-specific strategies for supporting climate-resilient healthy diets. Based on a clustering analysis, we identify eight archetypal climate risk profiles (figure 6; table 2) for which different policy and investment objectives could be prioritized to achieve national goals on nutrition under climate change. The distribution of input variables for each of the identified clusters is shown in SI figure 18, with a complete list of countries in SI table 1.

Zoom In Zoom Out Reset image size

We identify three clusters with high vulnerability (figure 4(a)): one with high levels of exposure to climate hazards in all nutrients (‘High domestic climate risk, broad’), one with high levels of climate hazards in select food groups (‘High domestic risk, targeted’), and one where present-day nutritional needs are particularly high (‘Acute nutritional needs’). Amongst large producers, we identify one cluster with low nutrient supply diversity in all nutrients (‘Low-diversity large producers’), one cluster with low nutrient supply diversity in calcium specifically (‘High-capacity livestock-dependent producers’), and one cluster that is especially prone to combined heat and drought extremes (‘Drought-prone producers’). Finally, we identify two clusters of import-dependent countries with high costs of healthy diets, one facing high climate hazards in all food groups (‘Import-dependent island states’) and one facing high hazards and nutrient deficiencies in one or two nutrients (‘High-diversity traders with targeted deficiencies’).

Based on our climate risk assessment, we identified key climate resilience and nutrition interventions from the literature and assessed where they may be most relevant for each of the climate risk profiles (table 3). Climate interventions focused on food production and domestic supply chains—ranging from diversifying production and adopting climate-adaptive practices to strengthening infrastructure—may be most salient for countries with high levels of domestic production facing high climate hazards (‘High domestic climate risk, broad’, ‘High domestic climate risk, targeted’, ‘Low-diversity large producers’, ‘High-capacity livestock-dependent producers’) [5, 49–54]. Additionally, nutritionally vulnerable countries may consider investing specifically in low-trophic, nutritious aquaculture. Aquaculture of particular species offers the potential to contribute to nutrition, climate adaptation, and sustainability goals, but requires intentional investment to support those outcomes [55–57].

Interventions outside food supply chains can also enhance food and nutrition security outcomes under climate change. For example, both social safety nets and nutrition education programs can be expanded to focus on nutritious and climate-resilient foods—especially for those facing high nutrient deficiencies and high climate hazards (‘High domestic climate risk, broad’, ‘High domestic climate risk, targeted’, ‘Acute nutritional needs’, ‘Import-dependent island states’) [50, 51, 54, 58, 59]. Countries that have not done so already, could adopt and update dietary guidelines and food environment policies to support diversity, climate-resilience, and sustainability outcomes, as currently underway in the Nordic countries [50, 60, 61]. Meanwhile expanding access to financial services, tools, insurance, and loans—while important in all countries with high numbers of small-scale food actors—may be particularly important in places with high rates of domestic production and high climate hazards (‘Drought-prone producers’, ‘High domestic climate risk, broad’) [50, 51]. Lastly, investing in R&D targeting climate-resilient food innovation should be a priority in countries with strong financial resources (especially ‘High-capacity livestock-dependent producers’) [5].

Given that trade can both expose countries to and shield them from production and market shocks [36, 37, 40], climate-resilience of nutrient supply is a transboundary issue. Bilateral and multilateral trade agreements will need to be renegotiated in a climate just framework that recognizes the universal Right to Food [49, 62]. Countries can put in place policies to diversify their imports—both in terms of food groups and trade partners (‘Import-dependent island states’, ‘High-diversity traders with targeted deficiencies’)—or to (re)build stockpiles of diverse food sources to buffer against shocks in food prices and availability (‘High domestic climate risk, broad’, ‘High domestic climate risk, targeted’, ‘Acute nutritional needs’) [66]. Finally, all countries can develop and strengthen forward-looking national commitments on climate change and nutrition, for example through their National Pathways, Nationally Determined Contributions, and National Adaptation Plans [51, 54, 65].

Our framework provides an initial, national-level assessment of nutrition-climate interactions, but results need to be interpreted with a few data and modeling gaps in mind. Our extremes-based approach did not incorporate information about the sensitivity of different foods and food production systems to different climate conditions, and has proven to be sensitive to the choice and superposition of climate variables (compare e.g. figure 3 to SI figure 3, and figure 5 to SI figure 17). In particular, using only temperature-based indicators for aquatic and livestock products may have led to an overestimation of their relative climate hazards. However, there is currently insufficient data to estimate climate sensitivity for all foods relevant for nutrient supply, especially to compound exposure. As of now, process-based crop models only exist for major crops like maize, soybeans, wheat, barley, and millet [67], while the impacts on livestock, inland fisheries, and aquaculture remain poorly quantified [68].

In addition, key processes, like the impacts of elevated CO₂ levels on yields and nutrient composition [69], secondary impacts through feeds [70], impacts on labor [71], or post-production impacts [5], were not included. Elevated CO₂ levels can both stimulate photosynthesis and water use efficiency [72] and reduce the concentration of important nutrients such as protein, iron, and zinc [69] leading to competing impacts on nutrient supply that depend on the crop or variety and environmental factors [68], precluding inclusion of these effects within our model. While it may not be feasible to develop process-based models for every crop, livestock, and aquatic food system, our analysis of food group contributions to nutrient supply (figure 2) could help identify ‘keystone’ production systems for priority model development. Alternatively, integrative indicator approaches such as fuzzy logic modeling [73, 74] could be used to cross-evaluate food group-specific variables, especially when grounded in regional understanding of climate change and sensitivity.

For several countries—notably in Central and East Africa and many small island states—one or more of the indicators used in our nutrition-sensitive climate risk framework were missing, precluding a holistic climate risk assessment. Failing to address these data gaps can perpetuate inequities, as resilience investments are likely to go to places and systems assessed in the research and policy literature [75]. A further update on analysis and recommendations could also address how variability between food groups compares to variability within food groups. For example, the Aquatic Products group contains a large diversity of species that vary widely in both nutrient composition [28] and climate sensitivity [74]. Similarly, future work applying this framework at national or regional levels could include subnational indicators of nutrient availability, climate hazards, and climate vulnerability [e.g. 46,71,76] to further identify climate risk hotspots. Finally, further research and evaluation are needed to test options and effectiveness for the recommendations presented in table 3, including how they could generate co-benefits with other SDGs, including targets for biodiversity and emission reductions [77]. In particular, micronutrient deficiencies coexist with other food-related health outcomes such as diabetes and cardio-vascular disease [1]. Our results on food group climate hazards (SI figures 1–2) could inform further analysis on climate impact pathways for diet-related non-communicable diseases [78, 79].