Global relationships between crop diversity and nutritional stability

Agricultural systems face mounting challenges from market volatility, land degradation, pests, and climate change, which complicate sustaining a growing, healthy population. Traditional food security policies emphasize yield and calories, but nutrition-sensitive agriculture recognizes that calories do not equate to food security, elevating the importance of nutritional diversity. Crop diversification is often promoted to improve dietary diversity and nutritional status, yet evidence linking crop diversity to nutritional outcomes is mixed, prompting calls for systemic, multilevel approaches that capture how agrobiodiversity influences nutrient provision at larger scales. Existing metrics such as potential nutrient adequacy integrate multiple nutrients into a single measure but do not assess the resilience of nutrient provision through disturbances over space and time. The authors define nutritional stability as the capacity of a food system to provide nutrients despite disturbance and propose that quantifying this stability can aid resilience planning aligned with SDG2, which links crop diversity, resilient farming, and nutritious diets. Leveraging a network-based framework, the study links crops to their constituent nutrients in bipartite networks and quantifies how nutrient availability changes as crops are removed, yielding a unitless robustness metric of nutritional stability. The study applies this approach to 55 years of FAO data (1961–2016) for 225 crops across 184 countries to assess: (1) the relationship between crop diversity and nutritional stability; (2) temporal changes in both; and (3) differences between nutrients from domestic production versus production plus imports. The central finding is a positive, saturating relationship between crop diversity and nutritional stability, alongside regional and temporal patterns indicating that imports play a key role in maintaining stability.

Design and data sources: The study constructs crop–nutrient bipartite networks for each country, year (1961–2016), and supply source, yielding 22,400 initial networks (reduced to 19,044 across 184 countries with ≥20 years of data). Two supply sources are evaluated: production alone (P) and production plus imports (PI). Annual national crop production data are from FAOSTAT. Nutrient composition per 100 g edible portion for 23 nutrients across 225 food categories is from the Global Expanded Nutrient Supply (GENUS) database. Seventeen nutrients are analyzed: calcium, carbohydrates, copper, folate, iron, magnesium, niacin, phosphorus, potassium, protein, riboflavin, thiamin, vitamin A, vitamin B6, vitamin C, zinc, and fiber (dietary fat and fatty acids omitted; some globally important nutrients like vitamin D, B12, and iodine are not in GENUS).

Network construction and link weighting: For each country–year–source, a bipartite graph connects crop nodes to nutrient nodes if the crop contains the nutrient. To minimize negligible contributions, links are weighted using the proportion daily value (PDV) per serving: PDV = 100N(Prod/(Pop*DV)), where N is grams of nutrient per 100 g of food (GENUS), Prod is national production (FAOSTAT), Pop is population (UN), and DV is the adult daily value (FDA). Links with PDV below 0.1 are removed, a threshold chosen after exploring stability across cutoffs 0–0.7 (variance plateau up to ~0.1; mean stability declines monotonically).

Nutritional stability metric (R): For each network, the robustness curve is generated by randomly removing crops sequentially and recording the number of remaining connected nutrients after each removal. This procedure is repeated 1000 times to obtain an average robustness curve. The curve is normalized (max nutrients and max steps scaled to 1), and nutritional stability R is the area under this curve (unitless). Additional removal procedures are also evaluated: (1) most-to-least connected crops first and (2) least-to-most connected. For each network, the study also calculates crop diversity (number of unique crops), nutrient diversity (number of unique nutrients present), and average crop degree (number of nutrient links divided by number of crops).

Statistical analyses: Relationships between crop diversity and nutritional stability are modeled with non-linear mixed-effects models, comparing saturating (ax/(β+x)), logarithmic, and exponential forms; the saturating function is selected via AIC. Region-specific models assess differences in parameter estimates (asymptote α and half-saturation β). Temporal trends in crop diversity, nutritional stability, and average crop degree are analyzed with linear mixed-effects models with an interaction between supply source (P vs PI) and year as fixed effects, random intercepts for source nested within country, and AR(1) correlation to account for temporal autocorrelation. Differences by macroeconomic grouping (developing status, small island developing states) are tested using linear mixed-effects models with interaction between grouping and supply source and country as random effect. Analyses are conducted in R (nlme, lme4); the stability algorithm is implemented in Python 3.7.4.

• Nutritional stability vs crop diversity: A positive, non-linear (saturating) relationship exists across countries, with significant regional differences in the function parameters (P < 0.05). Gains in stability slow after networks include roughly 7–16 unique crops. Across networks, 83% possessed all 17 nutrients considered.
• Crop and nutrient diversity: Crop diversity is positively (non-linearly) related to nutrient diversity, and nutrient diversity is linearly related to nutritional stability.
• Temporal trends in crop diversity: Crop diversity increased since 1961 in all regions except Oceania. Imports strongly contributed to increases: imported crop diversity rose by 43% in Asia and 35% in Europe over 55 years; in Europe, most diversity gains were import-driven rather than from domestic production.
• Spatial variation in stability: Substantial cross-country variation exists. The United States, Brazil, and much of Europe exhibit high and stable R over time. Many countries in the Middle East, Southeast Asia, and Africa show low stability and high interannual variance. Small island developing states and developing countries have significantly lower nutritional stability for both P and PI (P < 0.05 to P < 0.001).
• Temporal trends in nutritional stability: Despite increased crop diversity, stability remained stagnant or decreased in all regions except Asia. Regional specifics: Asia’s stability increased by ~8% from PI but declined by ~1% from production alone; Africa’s production-based stability decreased by ~4% (imports-based stability unchanged); Americas and Oceania experienced production-based declines of ~7% and ~4%, respectively; Europe remained relatively high and stable.
• Role of crop degree: Although 72% of countries increased crop diversity (PI), 87% of these saw decreases in average crop degree, associated with declines in stability. Crop degree declined across all regions and supply sources, indicating added crops often contribute fewer or redundant nutrient links, generating diminishing returns for stability.
• Imports and exposure: Imports generally increase crop diversity and nutritional stability, implying many countries’ stability is market exposed and potentially vulnerable to trade disruptions.

The saturating relationship between crop diversity and nutritional stability explains why increased diversity does not always yield proportional stability gains: once key nutrient-providing crops are present, additional crops add fewer new nutrient links. Declines in average crop degree indicate shifts toward less nutrient-dense crops or additions that duplicate existing nutrients, reducing marginal benefits for stability. Countries with low diversity are particularly vulnerable: shocks such as droughts, pests, or trade disruptions can cause rapid loss of nutrient availability. The pronounced role of imports underscores both benefits and vulnerabilities—trade can bolster stability but increases exposure to market volatility, trade disputes, and price shocks. Regional disparities, especially within Africa, likely reflect differences in trading capacity, domestic availability affected by conflict or political instability, and climate-related disasters. The study advances resilience assessment for nutrition-sensitive agriculture by focusing on nutrient availability robustness rather than calorie-based or intake-based metrics, and highlights that managing both production and trade portfolios is crucial for maintaining stable nutrient supplies.

This study introduces a generalizable, unitless robustness metric to quantify national nutritional stability from crop–nutrient networks and applies it globally over 55 years. It demonstrates a positive but saturating relationship between crop diversity and stability, documents widespread increases in crop diversity often driven by imports, and reveals that nutritional stability has largely stagnated or declined outside Asia due to diminishing returns and declining crop degree. Policy and planning implications include prioritizing crop additions that supply underrepresented (“vulnerable”) nutrients, evaluating reliance on imports, and designing trade and production strategies to bolster resilience. Future research should integrate nutrient availability with intake data to connect stability to nutritional outcomes, consider non-random crop loss scenarios and abundance-weighted links, include animal-source foods where feasible, and model trade dynamics and climate-driven multi-crop failures to identify targeted adaptation and diversification strategies.

• Scope limited to nutrient availability from crops; does not measure dietary intake, utilization, or health outcomes, and thus does not equate to food security.
• Animal-source foods excluded due to data resolution and variability, potentially underestimating or mischaracterizing nutrient availability patterns across countries.
• Network links treated as present/absent after applying a PDV threshold; while PDV weighting reduces negligible contributions, binary treatment may oversimplify quantity effects and could underestimate or overestimate stability.
• Equal probability of crop removal in the primary robustness simulations; real-world loss likelihoods differ by factors such as climate vulnerability, market value, pest pressure, and production levels. Alternative, non-random or abundance-weighted removal scenarios were not fully explored.
• GENUS nutrient scope omits some nutrients (e.g., vitamins D and B12, iodine); nutrient content variability due to environment or management is not captured.
• FAO data limitations and missingness, particularly in low-income countries and earlier years, may affect accuracy.
• The metric captures robustness of nutrient availability but not identity-specific management priorities without additional analyses.