Global food systems face significant threats from climate change, which could have far-reaching consequences for human well-being and social stability. While it is widely recognized that climate change will affect food distribution, weather and bioproduction processes in agriculture, the extent to which farmers can effectively adapt to these changes remains unclear and controversial. New study seeks to fill this gap by systematically empirically estimating the impact of adaptation by global producers, which provides key insights for ensuring food security in a changing climate.

An innovative empirical approach to the study of adaptation

Traditional process models used to estimate the impacts of climate change on agriculture often assume that producers optimize yields according to predetermined decision rules, such as “no adaptation” or “optimal variety change.” These models are based on data from scientifically controlled experimental fields, which may limit their ability to represent real-world decisions made by producers in different socioeconomic contexts, influenced by financial constraints, market failures, and human error.

To overcome these challenges, the study authors developed a unified empirical approach that measures the impact of climate change on staple crop production, taking into account costs, benefits and producer adaptation rates as observed in practiceThey used one of the most comprehensive datasets on subnational crop yields, covering 12,658 regions in 54 countries for six staple crops, accounting for two-thirds of global caloric crop productionThis approach, called “reduced-form,” empirically assesses the overall impact of adaptive adjustments (e.g., changing varieties, adjusting crop timing, changing fertilization) without explicitly modeling each individual mechanism. The assumption is that producers facing similar climatic, income, and infrastructure conditions make similar management decisions.

The costs of adaptation cannot be directly observed, but can be inferred by assuming that profit-maximizing producers only undertake adaptations where the benefits exceed the costs. This means that the study takes into account not only the benefits but also the costs associated with adaptation measures.

Key findings: Persistent losses despite adaptation

The study estimates that global production will decrease by 5.5 × 10^14 kcal per year for every 1 °C increase in global mean surface temperature (GMST)This corresponds to approximately 120 kcal per person per day, or 4.4 % of recommended daily intake for every 1°C increase in GMST.

It is important that adaptation and income growth will mitigate 23 % of global losses by 2050 and 34 % by the end of the century in the high emissions scenario (6 % and 12 % in the moderate emissions scenario). This underlines the key role of adaptation. However, despite these adaptation benefits, significant residual losses persist for all staple crops except riceOverall, the central estimates of losses by the end of the century under the high emissions scenario (RCP 8.5) are substantial: maize -27.8 %, rice -6.0 %, soybean -35.6 %, cassava -29.8 %, sorghum -21.7 % and wheat -28.2 %. These results differ from some process models that often predict global productivity gains.

Geographic and income differences in impacts

The impacts of climate change on crop yields vary around the world. Temperature changes generally dominate in determining the sign of local impactsHot climates are already more adapted to heat, so further warming has less of an impact, while colder areas may benefit from warming.

Surprisingly, the study found that global impacts are most affected by losses in modern "breadbaskets of the world" with favorable climates, where adaptation is currently limitedThese regions, while rich, are optimized for high average yields, not for resilience to climate change. Loss of production in these key regions has an amplified impact on global calorie production.

At the same time, they are losses in low-income regions are also significant, especially for populations dependent on crops such as cassava. While average losses are estimated at 28 % in the lowest income decile, losses can climb to 41 % in the highest income deciles.

Impact of economic development and implications

Rising incomes allow producers to respond to environmental conditions in different ways, although outcomes vary by crop. For example, maize and sorghum yields become more sensitive to temperature with increasing income, while rice and cassava become less sensitive. Higher incomes and access to irrigation are associated with greater resilience to extreme rainfall across all crops.

The authors also calculated the social cost of carbon (SCC) component resulting from changes in global yields, which ranges from from 0.99 to 49.48 USD per ton of CO2 depending on assumptions. Their estimates are lower (2- to 5-fold) than some previous ones because they take into account autonomous adaptation, which reduces perceived harm.

Overall, the study results underline the urgency and importance of developing new agricultural technologies and ensuring access to adaptive technologies for producers worldwideto maintain global food security. Despite the extensive adaptation already demonstrated by the global population, particularly in warmer and poorer regions, expected losses in staple food production remain significant. Spring

The study is published in the journal Nature

Glossary of key terms

• Adaptation: Changes in agricultural practices, technologies or crops that producers make in response to climate change, with the aim of mitigating adverse impacts or exploiting potential benefits.
• AgMIP (Agricultural Model Intercomparison and Improvement Project): A global project aimed at comparing and improving agricultural models to improve projections of the impacts of climate change on global and regional food security.
• CMIP5 (Coupled Model Intercomparison Project Phase 5): An international initiative that collects and standardizes outputs from global climate models (GCMs) to assess climate change.
• Cross-validation: A statistical technique used to assess how well a statistical model generalizes to an independent set of data. In this study, it is used to select relevant weather and adaptation variables.
• Damage function: An empirically derived relationship that quantifies the continuous distribution of caloric production loss as a function of the change in global mean surface temperature (GMST).
• Degree days: A measure of heat accumulation that takes into account the time and intensity of crop exposure to certain temperatures above or below a threshold. It is used to predict crop growth and development.
• DSCIM (Data-driven Spatial Climate Impact Model): A model developed to project the impacts of climate change that integrates empirical economic results with geographically detailed climate and socioeconomic data.
• Fixed effects: A statistical method used in panel data to control for unobservable, time-invariant variables that may affect the outcome, such as soil quality in an administrative unit.
• GCM (Global Climate Model): A mathematical model of Earth's climate, used to simulate future climate change in response to greenhouse gas emissions.
• GMST (Global Mean Surface Temperature): The average temperature of the Earth, which is used as a key indicator of global warming.
• CO2 fertilization: A process in which increased concentrations of carbon dioxide in the atmosphere can lead to increased plant growth and crop yields due to improved photosynthesis.
• Longitudinal data: Data collected repeatedly from the same entities (in this case, regions) over time, allowing changes and trends to be tracked.
• Monte Carlo simulations: Computer simulations that use random sampling to obtain numerical results, especially when solving problems with multiple variables and uncertainty.
• OOS performance (Out-of-sample performance): The ability of a model to accurately predict values for data that was not used in its training, which is important for assessing the robustness and generalizability of the model.
• Partial SCC (Partial Social Cost of Carbon): The part of the total social cost of carbon that refers to economic damages caused by changes in global agricultural crop yields due to CO2 emissions.
• Process-based models: Agricultural models that simulate the biological and physical processes of crop growth (e.g. photosynthesis, transpiration) in response to environmental conditions.
• Reduced-form econometric approach: A method that directly estimates relationships between variables without explicitly modeling all underlying mechanisms. In this study, it is used to capture the net consequences of producers' adaptive behavior.
• Representative Concentration Pathways (RCPs): Four scenarios of greenhouse gas concentrations used for climate modeling and projections. RCP 4.5 is a moderate emissions scenario and RCP 8.5 is a high emissions scenario.
• Revealed preference: An economic theory that states that the preferences of consumers or producers can be inferred from their observed behavior and choices in the marketplace.
• Shared Socioeconomic Pathways (SSPs): Scenarios of future socioeconomic changes (e.g. population growth, economic development, technological change) that are used in combination with RCPs to project climate change and its impacts.
• Subnational administrative units: Geographic areas, such as counties, states, or provinces, that are smaller than countries and for which agricultural data is available.
• Vapour pressure deficit: The difference between the amount of moisture in the air and how much moisture the air could hold when saturated. A key factor in plant transpiration and water stress.
• Yields: The amount of crop (e.g. calories per hectare) produced on a given area of land.