Abstract

Micronutrient deficiencies from poor dietary diversity remain a global health challenge. This issue is exacerbated by climate change-driven extreme weather events that impact crop quantity and quality. While process-based crop models effectively simulate plant nutrient (N, P, K) dynamics for productivity projections, they lack the ability to assess crop nutritional content. This Perspective proposes a framework for integrating nutrient dynamics into crop models for informing nutrition security strategies amid climate change. We delineate key biological pathways influencing nutrient uptake, translocation and density in response to elevated CO2, temperature and low precipitation. We highlight the scarcity of comprehensive datasets, underscoring the need for urgent, collaborative research to amass foundational data and models to ensure nutritional integrity in an uncertain climate.

You have full access to this article via your institution.

Main

Macronutrients (for example, carbohydrates, proteins and fats) and micronutrients (for example, iron, vitamin A, vitamin D, iron, potassium, magnesium, calcium and zinc) play pivotal roles in human health, underpinning vital physiological functions ranging from enzymatic activities and cognitive development to immune responses and bone health. Poor diets, defined in terms of food quantity, quality and safety, serve as an important risk factor in the global burden of disease1, particularly in vulnerable populations such as children and women of reproductive age2. Vitamin A, iron and zinc are examples of micronutrients with a moderate to high prevalence of global deficiency and severe consequences stemming from poor dietary intake3. While the highest concentrations of these vitamins and minerals are often found in nutrient-rich foods, such as animal-sourced foods and dark-green leafy vegetables, these dietary sources are often unaffordable, unavailable or perhaps not culturally appropriate in many low- and middle-income countries4,5.

Increasing dietary intake of essential nutrients through evidence-based nutrition interventions includes biofortifying or fortifying staple foods, targeted supplementation programmes to reach at-risk groups and nutrition counselling that promotes dietary diversity6. However, the scalability and effectiveness of these interventions face several challenges, including potential overnutrition (especially for iron)[7](https://www.nature.com/articles/s41558-025-02470-3#ref-CR7 “Tan, X., Tan, P. Y., Gong, Y. Y. & Moore, J. B. Overnutrition is a risk factor for iron, but not for zinc or vitamin A deficiency in children and young people: a systematic review and meta-analysis. BMJ Glob. Health https://doi.org/10.1136/bmjgh-2024-015135

(2024).“), economic feasibility, socio-cultural barriers and compliance[8](https://www.nature.com/articles/s41558-025-02470-3#ref-CR8 “Reerink, I. et al. Experiences and lessons learned for delivery of micronutrient powders interventions. Matern. Child Nutr. https://doi.org/10.1111/mcn.12495

(2017).“). For example, the potential physiological side effects of iron supplements pose considerable challenges9, while the consistency required for effective zinc supplementation can be difficult to maintain10. Evidence on the efficacy of fortification strategies on nutritional status of vulnerable populations, which are mediated by a wide variety of context-specific logistical, technical and financial factors, is mixed11,[12](#ref-CR12 “Hombali, A. S., Solon, J. A., Venkatesh, B. T., Nair, N. S. & Peña-Rosas, J. P. Fortification of staple foods with vitamin A for vitamin A deficiency. Cochrane Database of Systematic Reviews https://doi.org/10.1002/14651858.CD010068.pub2

(2019).“),13.

It is widely anticipated that climate change will pose further challenges for food security and dietary diversity from production to consumption due to changes in crop productivity, volatility in yields and prices, food safety, market dynamics and shifts in the nutritional quality of the global food supply14. There is evidence to suggest that elevated atmospheric carbon dioxide concentrations ([CO2]) will dilute the density of key nutrients, such as protein, iron and zinc15, across a range of crops16, with stronger dilution anticipated in C3 crops—which constitute the majority of cultivated species that are directly consumed (for example, rice, wheat)—compared with C4 crops (for example, maize, sorghum)17. However, leguminous C3 crops may exhibit less nutrient dilution compared with other C3 crops due to their inherent nitrogen-fixing capabilities18. The impacts of climate change on food security and nutrition are especially concerning for vulnerable communities who do not have the capacity to absorb climate-related shocks and face economic, physical or social barriers to access healthy and nutritious foods.

To better project climate change impacts on macro- and micronutrient content, there is a need to develop tools that can adequately represent the biophysical dynamics responsible for crop nutrient concentration. Process-based crop models that simulate crop growth and development by mathematically representing key biological and environmental processes can assist in projecting risks across future climate change scenarios and test the efficacy of a range of interventions designed to address those risks across environments, plant genetics, management scenarios and cropping systems19,20. Current crop models focus predominantly on plant macronutrients such as nitrogen (N), and to a lesser extent phosphorus (P) and potassium (K) when projecting yield, leaving a critical gap in our understanding of how climate change could affect the content of key nutrients relevant for human health (that is, human macronutrients and micronutrients mentioned above). Existing modelling studies that evaluate climate change impacts on crop protein or micronutrient content usually use the ratio of carbon to nitrogen in the grain (C/N ratio) as a proxy for nutritional quality (related to carbohydrate formation and protein synthesis), growth potential and disease resistance21. Field-scale process-based crop models currently lack modules that represent nutritional content dynamics within and across plant tissue at various stages of crop development. This is attributable not only to limited theoretical development but also to a scarcity of empirical data necessary for calibrating and evaluating these new models. To bridge this gap, we provide a simple framework to simulate nutrient content within field-scale crop models based on nutrient–yield relationships from the literature to establish a theoretical foundation for incorporating process-based nutrient accumulation routines. For the purposes of this simple framework, we defined ‘nutrient’ as crop macronutrients (that is, carbohydrates, protein and fat) or micronutrients (that is, calcium, magnesium, zinc, potassium, iron). However, we recognize there are other compounds such as phytonutrients (alkaloids, carotenoids, flavonoids, phenolics and terpenoids), anti-nutrients (that is, phytates and lectins) and heavy metals (that is, arsenic, cadmium, lead and copper) that require closer investigation with regards to human health. This is a necessary first step to enhance our predictive capabilities under future climate scenarios. In addition, we explore innovative solutions to overcome data limitations, such as leveraging machine learning to predict nutrient dynamics in complex systems, and advocate for comprehensive, interdisciplinary research to generate foundational datasets to advance the field. Such efforts will be crucial in maintaining nutritional integrity at a global scale with climatic uncertainty.

Understanding climate change impacts on nutrient dilution

Plants respond to global change drivers through various metabolic and physiological shifts that often affect their nutrient uptake and redistribution capabilities, thereby altering crop nutritional composition and stoichiometry22,23. A deeper understanding of these mechanisms and factors that modulate crop nutrient processes under various climatic conditions are important to consider in models designed to project climate change impact on nutrient availability in crops. There is also a critical gap in crop modelling research on the climatic impacts on more nutrient-dense foods, such as legumes, nuts and green leafy vegetables, as opposed to major cereals, such as maize, rice and wheat14.

One factor of interest relates to the indirect effect of rising [CO2] on stomatal closure. Since transpirational mass flow—the process by which water carries dissolved nutrients throughout the plant—is affected by stomatal behaviour, decreased transpiration from [CO2]-induced stomatal closure may limit nutrient absorption24. This is consistent with previous reports that showed that nutrients associated with mass flow, such as iron and zinc, tend to exhibit larger concentration decreases in response to elevated [CO2] (ref. 25). Alternatively, elevated [CO2] can reduce the N demand of RuBisCO—the key enzyme for carbon fixation—as increased photosynthetic efficiency from reduced photorespiration lowers N requirements (mainly in C3 plants)26. However, this benefit from elevated [CO2] may be temporary as C3 plants acclimate with metabolic adjustments restoring N demand27. In addition, elevated temperatures can accelerate RuBisCO turnover, increasing N needs28, while vapour-pressure-deficit-driven stomatal changes can limit [CO2] assimilation rate29. These factors may offset initial elevated [CO2] gains, underscoring the complex interaction between climate variables and nutrient dilution. A meta-analysis of 57 studies, consisting of 1,015 observations, reported declines in the concentrations of protein, nitrate, magnesium, iron and zinc in vegetables by 9.5%, 18%, 9.2%, 16% and 0.4%, respectively, as conditions shift from low-to-ambient CO2 (<450 ppm) to elevated levels (>540 ppm; ref. 30). Another factor may be directly related to the influence of concurrent increases in temperature and [CO2] relative to protein and mineral needs for photosynthesis and growth. Warming and water stress are found to increase N allocations, compensating, in part, [CO2] effects on protein content31,32,33. A recent open-field experiment showed an increase in wheat grain protein content by 11% to 31%, paired with a reduction in grain starch content by 23% to 29%, under warming by 1.3 °C (ref. 34). Additional data and modelling are needed to confirm these responses33,35.

In addition to the effects of higher [CO2], reduced precipitation can impact plant nutrient relations irrespective of temperature changes. For example, a meta-analysis by He and colleagues demonstrated that drought stress decreases the concentration of N and P in plant tissue in the short term36. However, there is evidence to suggest that elevated [CO2] may mediate the impacts of drought on nutrient content in some crops37. Alternatively, flooding or waterlogging as a result of excess precipitation may reduce plant nutrient quality and uptake via soil oxygen depletion limiting root nutrient absorption (hypoxia), increasing nutrient leaching and/or introduction of contaminants in the soil38.

Separate from any physiological response at the plant level, climate-induced changes in elemental soil profiles have been examined, given that soils are critical for plant growth and development and ultimately crop nutritional composition. Soil serves as the primary medium for nutrient exchange, and its physical and chemical properties dictate nutrient availability. Soil temperature remains a key understudied component of plant stoichiometry, but its role in ion uptake, root growth and the characterization and function of the soil microbiome is widely recognized39. It has been shown that nutrients that are transferred by diffusion (for example, P) would experience greater mobility in warmer soil40. However, it should be recognized that toxic elements such as heavy metals (for example, arsenic) can also increase in the rhizosphere with soil warming41. In addition, temperature-driven shifts in soil moisture dynamics and redox potential can alter nutrient solubility and microbial activity, influencing organic matter decomposition and nutrient cycling42. Changes in cation exchange capacity and pH buffering under warming scenarios may further impact the retention and mobility of nutrients and contaminants43.

Overall, these changes in the root–plant–atmosphere interface are profound and have clear consequences for understanding nutrition dynamics. As such, there is a need to address how these metabolic changes could be modelled to better estimate quantitative and qualitative shifts in food sources across a wide range of crops. Assessments of [CO2] influence on soil chemistry and nutrient availability, and a more inclusive description of speciation, solubility and potential toxicity, are required as a basis for such nutritional analyses.

Integrating nutrient dynamics into process-based crop models

Most existing process-based crop models calculate plant macronutrient (N, and to a lesser extent P and K) uptake using crop-soil routines with species-specific parameters. Quantifying these parameters depends on the reliability and availability of high-quality nutrient data, as well as knowledge of the relationships between nutrient concentration and transport to agricultural outputs, such as aboveground biomass or yield. This is a challenge for agricultural modelling because these relationships may co-vary with yield and other driving factors (for example, abiotic stresses) or are not always evident44. If the relationship is known, one approach is to incorporate the relationship into crop models such as those developed, intercompared and applied within the Agricultural Model Intercomparison and Improvement Project community45,46. A widely used example is the Decision Support System for Agrotechnology Transfer (DSSAT) platform47. DSSAT consists of a common interface and code structure enabling execution of over 45 process-based crop models that simulate daily dry-matter accumulation for different stages of a plant’s development cycle. It is a tool that has been used to examine climate change impacts on crop production from local to global scales, and has proved to be successful at capturing seasonal plant growth via crop-specific parameters such as radiation use efficiency, optimal temperature and soil water ranges, and various thresholds at which plant stress response functions are initiated48,49.

Recently, a nutrient-quality module examining two commonly used proxies for strawberry fruit quality, soluble solid content and titratable acidity, was incorporated into the DSSAT CSM-CROPGRO-Strawberry model50,51. This nutrient-quality module was based on the relationship between the nutrient-quality variable and strongest correlated driving environmental factor, determined to be average temperature during fruit cohort growth. Here we first propose a similar, simplistic approach to examine crop nutrient concentrations on the basis of the growth dynamics from process-based crop models. We then outline the implementation of a more complex and mechanistic nutrient transport approach within a process-based model.

The first approach integrates a simple statistical relationship between nutrient concentration and yield into the crop models based on existing literature. The application is similar to dose–response relationships, such as the Weibull function for approximating crop relative yield loss as a response to tropospheric ozone (O3) stress[52](https://www.nature.com/articles/s41558-025-02470-3#ref-CR52 “Tai, A. P. K., Sadiq, M., Pang, J. Y. S., Yung, D. H. Y. & Feng, Z. Impacts of surface ozone pollution on global crop yields: comparing different ozone exposure metrics and incorporating co-effects of CO2. Front. Sustain. Food Syst. https://doi.org/10.3389/fsufs.2021.534616

(2021).“),53. Here yield is used as the covariate because linking mean nutrient concentrations with yield data allows for improved experimental comparability and synthesis44, although other correlated driving factors may be used to determine a response50. As an initial approximation, we use the linear relationships between wheat yield and both iron and zinc concentrations derived from multiple studies collated by Miner and colleagues44. After conducting a linear regression on 13 datasets from 7 studies to determine the iron and zinc nutrient–yield relationships, the slopes were aggregated into a single proxy for use in wheat models54,55,56,57,58,59,60 (Fig. 1). This simplistic approach serves as an important starting point for integrating crop nutrient projections into field-scale crop models, particularly when compared with the status quo that includes no response. Furthermore, while wheat probably does not serve as a primary source of iron and zinc in most diets, the principles discussed here can be applied to other crops as this research frontier advances.

Fig. 1: Relationship between wheat grain yield and nutrient concentration.

The relationship between wheat grain yield and nutrient concentration for iron (top, solid lines) and zinc (bottom, dashed lines) observed across multiple field experiments from the literature. The nutrient–yield response for each reference study is derived from the relationships provided by Miner et al.44, which encompasses data from multiple genotypes (with the exception of Clemensen et al.54) and sites and varying rates of nitrogen fertilizer or different crop rotations. The black line indicates the linear fit across all studies included in the analysis, and the shaded area indicates the 95% confidence interval of the regression. The equation and coefficient of determination (R2) correspond to the linear fit (black line).

The equations in Fig. 1 provide the parameter coefficients (equation slope and intercept) to be incorporated into wheat crop models to output iron and zinc nutrient grain concentrations (y) on the basis of simulated yield (x). The relationship shows that while a generalized wheat yield response may be representative of iron concentrations (R2 = 0.65), there is less agreement regarding the zinc concentrations (R2 = 0.50). This is due to higher variability in the zinc measurements across the literature, which highlights the constraints of using a simplified statistical approach. While the nutrient equations provide an acceptable approximation across multiple scenarios (R2 > 0.50), it is best to calculate the nutrient–yield relationship coefficients for the specific scenario (cultivar, experiment and treatment) being examined, when possible, to provide better representation of the system to be modelled. More complex relationships between yield and nutrient concentrations could be explored using advanced statistical techniques or machine learning approaches in the future. For example, machine learning algorithms could be leveraged to uncover complex, nonlinear relationships and interaction effects in the data (for example, combined CO2–heat–drought responses on nutrient translocation to seed and/or grain), potentially revealing emergent properties that simpler statistical models might overlook.

While this statistical relationship is useful for nutrient-quality approximation based on simulated yield within the model, it is independent from the simulation of plant growth and development. In the long term, integrating nutrient dynamics into process-based crop models will require the appropriate crop-specific coefficients that govern plant nutrient uptake from the soil (including dynamics within the soil system depending on the nutrient), nutrient translocation across the plant, and eventual micronutrient concentration in the edible product. This approach could use a relationship that accounts for daily dynamic nutrient accumulation and translocation throughout crop growth and development. For most crop models, this would consist of integrating four main process routines as depicted in Fig. 2, including:

1. 1.

Simulating nutrient uptake in roots per phenological stage on the basis of the initial pool of nutrient available in the soil

1. 2.

Simulating nutrient accumulation in relation to daily plant dry-matter accumulation and harvestable product

1. 3.

Simulating the soil factors affecting the solubility and movement of available nutrients in the soil (for example, pH, cation exchange capacity, soil organic matter, soil water content and mass flow per day)

1. 4.

Simulating the impact of temperature, water stress or other atmospheric factors (for example, [CO2] and O3) on nutrient uptake per phenological stage

Fig. 2: Inputs, processes and outputs to assess nutrient dynamics in crop models.

The infographic highlights the four key integration process routines needed to simulate dynamic nutrient accumulation and translocation within crop models. Each process will require the calibration of crop-specific factors that regulate nutrient absorption from the soil (while considering the nutrient dynamics within the soil system), the nutrient translocation across the plant and the final concentration of micronutrients in the harvestable product.

The crop model output, when paired with the framework of plant uptake models61, would allow for exploration of the translocation of nutrients into the grain or harvestable organs. For example, Brunetti and colleagues61 coupled a dynamic plant uptake model with the HYDRUS model to effectively simulate carbamazepine translocation and transformation in three vegetables. This would be an ideal approach (or combining the modules into a single model instead of coupling the models) if data were available for each response factor, but it is not currently feasible given data availability and collection limitations. In addition, the nutrient translocation relationship between biomass and harvestable product is yet to be determined for many crops, but the bioaccumulation mechanisms within plant uptake models may be able to provide guidance once this relationship is known. Taken together, integrating nutrient dynamics into process-based crop models would allow for models that are apt at simulating the effects of elevated [CO2] and other climate change variables at the field level, which are essential for understanding the larger regional dynamics that underpin food and nutrition security assessments.

Data needs

To successfully design and integrate nutrient uptake modules into process-based crop models beyond staple crops, comprehensive data across a variety of ecological conditions are paramount. This includes, but is not limited to, detailed soil characteristics such as pH levels, organic matter content and cation exchange capacity, metrics that influence the availability of micronutrients in the soil. In addition, accurate information on the concentration of specific micronutrients within different soil types and locations is essential to model their availability accurately. Species- and cultivar-specific genotypic information is also critical, as different crop varieties exhibit varying efficiencies in nutrient uptake and assimilation, impacting the concentration of micronutrients in the edible parts of the plant. This is especially true in the context of the Vision for Adaptive Crops and Soils in Africa project[62](https://www.nature.com/articles/s41558-025-02470-3#ref-CR62 “Office of Global Food Security The Vision for Adapted Crops and Soils (US Department of State, 2025); https://2021-2025.state.gov/the-vision-for-adapted-crops-and-soils/

“), which seeks to identify and promote a diverse array of neglected and underutilized crop species for their nutrient density and climate resilience. Environmental data encompassing a wide range of factors, including temperature, precipitation patterns and [CO2] and O3 levels, are necessary to understand their effects on plant growth and nutrient uptake. In addition, standardization of parameters across various cropping systems (for example, types and quantity of fertilizer used and other management practices such as irrigation and crop rotation), study location (that is, greenhouse versus field) and combined environmental variables (for example, both heat and water stress) should be considered given that each introduces substantial heterogeneity in crop nutrient responses63,64,65,66. Understanding the dynamics of nutrient translocation within the plant, from roots to shoots and reproductive organs, requires detailed physiological and phenological data on the species and cultivars of interest. Factoring in how post-harvest processes (for example, storage and transport, industrial processing, preparation and cooking) impact the nutrient content of consumed foods is crucial to understand the impact of these biophysical processes on human health.

Given the challenges associated with collecting such extensive and varied data, prioritizing a nutrient sensitivity analysis will help identify the most critical input parameters affecting macro- and micronutrient content variability, allowing for a streamlined and focused data collection strategy. Such an approach can adequately capture the essential elements influencing nutrient dynamics. It is important to note, however, that the specific sensitivities of crop nutrient concentrations to environmental variables are probably unique to each crop and may be different across varieties of the same crop67. For illustrative purposes, we undertook a literature review of 17 studies on 2 staple crops, wheat and rice, and their nutrient sensitivity of 6 relevant nutrients to environmental variables given data availability68,69,70,71,72,73,74,[75](#ref-CR75 “Sattar, A. et al. Individual and combined effect of terminal drought and heat stress on allometric growth, grain yield and quality of bread wheat. Pak. J. Bot. https://doi.org/10.30848/pjb2020-2(5)

(2020).“),76,77,78,79,80,81,82,[83](#ref-CR83 “Yue, L. et al. The mechanism of manganese ferrite nanomaterials promoting drought resistance in rice. Nanomaterials https://doi.org/10.3390/nano13091484

(2023).“),84 (Fig. 3). For wheat, there is consensus that [CO2], drought and heat stress will have consequences on most nutrients, with the exception of K. Importantly, stress combinations induce a unique set of complex responses at metabolic and molecular levels, which differ from those observed under individual stress scenarios85. For example, Galani and colleagues86 found that climate factors could increase the ability of wheat to meet adult daily dietary requirements by 6% to 12% for protein, zinc and iron, but decrease those of magnesium, manganese and P by 3% to 6%, and of K by 62%. For rice, there were more-variable nutrient responses across each stress scenario and limited evidence on most micronutrients. In addition to these parameters, O3, soil health and management practices (that is, irrigation, tillage and fertilization) are other key factors to be considered in the development of these models87,88.

Fig. 3: Climate–nutrient interactions in rice and wheat.

The overall effect observed across rice- and wheat-field and simulation studies published in the literature68,69,70,71,72,73,74,[75](#ref-CR75 “Sattar, A. et al. Individual and combined effect of terminal drought and heat stress on allometric growth, grain yield and quality of bread wheat. Pak. J. Bot. https://doi.org/10.30848/pjb2020-2(5)

(2020).“),76,[77](#ref-CR77 “Wang, Y., Frei, M., Song, Q. & Yang, L. The impact of atmospheric CO2 concentration enrichment on rice quality