Abstract
Agricultural labor is highly vulnerable to heat stress. However, existing studies exploring the impacts of future weather conditions on agriculture have primarily focused on crop yields, often overlooking the effects on labor. Here, we incorporate labor response to heat stress into an established model-coupling framework to examine the effects of future weather conditions on agricultural market dynamics, considering both biophysical crop yield and labor productivity responses. Under a scenario with high heat stress and high sensitivity of labor to heat stress, global agricultural labor productivity could decrease by 18% by the end of the century, with especially pronounced effects in Africa, South Asia, and Southeast Asia. Our results highlight the critical role of labor in agricultural production under future weather conditions and its high sensitivity to the choices of General Circulation Models and labor-heat response functions.
Similar content being viewed by others
Introduction
Growing evidence of the adverse impact of heat stress on human health1,2, labor productivity3,4 and labor participation5 underscores the need to include labor response to heat stress in climate impact studies, especially in agriculture6. Agricultural labor is particularly vulnerable to heat stress due to the high work intensity and outdoor working environment7. In 2021, primary agriculture accounted for 25% of global employment8, with this share exceeding 40% in regions that have been heat-stressed, such as Africa9, India10, and Southeast Asia11. Due to high labor involvement and high heat exposure, agriculture in low and medium UN-defined human-development-index countries accounts for almost half of the global total estimated work hours lost due to heat stress in 202012.
Despite its importance, labor responses have been largely overlooked in studies on the impacts of future climate scenarios on agriculture (Table S1). Most existing studies focus on quantifying crop yield responses13 to future climate scenarios and investigate the associated economic and environmental consequences14,15,16, while the role of labor has been neglected. There is also a stream of literature that quantifies physiological labor responses to heat stress exposure17,18,19,20,21. For instance, it is suggested that resting time at the workplace is necessary when the Wet Bulb Globe Temperature (WBGT) rises to 25.5 °C for heavy work18 to avoid the risk of core body temperature exceeding 38 °C. WBGT is a widely used measure of human heat stress exposure, and labor productivity is expected to decrease as WBGT increases beyond a certain threshold3.
Building on the physiological literature, studies have explored the economic consequences of heat-induced labor productivity loss, often focusing on broader economic sectors. Empirical analysis has shown heat-induced reductions in manufacturing output due to labor productivity loss in India22 and decreased daily work time for outdoor industries in the United States5. Studies calculating WBGT with historical weather data have suggested a 10% loss in global labor productivity during peak months over the past few decades19 and a potential loss of 295 billion work hours in the global economy in 202012. Under Representative Concentration Pathway 8.5 (RCP 8.5, a high-emission scenario), heat-induced labor productivity loss globally is projected to reach 37% in the hottest months in 210019, with South Asia, Southeast Asia, and Africa being most affected. Studies have also examined economy-wide consequences of future heat-induced labor productivity loss, such as GDP losses23,24,25,26,27, fluctuations in consumption and trade23, and supply chain disruptions28.
Lima et al.29 provided, to our knowledge, the first study that integrated the two diverse sets of literature and addressed both crop and labor responses in assessing climate impacts on global agriculture. Introducing labor-heat responses into a Computable Generable Equilibrium (CGE) model, they highlighted that labor responses contribute to more than 40% of the combined global welfare loss under a 3 °C global average temperature change. Similarly, Orlov et al.30 used a CGE model to assess climate impacts via crop and labor responses under RCP 7.0, highlighting that heat-induced labor productivity loss could offset the economic gains from increased crop yields. However, their assessments were based on a static economy, neglecting future socioeconomic changes and labor market evolution.
Capturing socioeconomic changes and agricultural labor market evolution is essential to quantify the long-term heat-induced labor productivity loss and the consequential economic impact. Global agricultural employment peaked in the early 2000s at ~1.1 billion people and then decreased by 21% to 0.87 billion in 2021, despite rising population, labor force, and agricultural output during the same period8. Human exposure to heat stress depends not only on weather conditions but also on the size and spatial distribution of the population31, with the adverse impact of heat stress on humans being more pronounced in regions with larger populations12. Labor productivity growth and the transition from agriculture to the manufacturing and service sectors have fueled the long-term evolution of the agricultural labor market32,33, and are therefore expected to affect labor exposure to heat in agriculture. Heterogenous agricultural labor market evolution across regions, along with scenarios of more intensive, frequent, and regionally heterogeneous heat waves later in the century34,35,36,37,38, will influence the distribution and trajectory of future agricultural production and employment39, ultimately affecting agricultural labor heat stress exposure.
Our study seeks to expand the literature on climate impacts on agriculture by incorporating labor-heat responses into the established model coupling framework (Fig. 1; “Methods”). Following Nelson et al.14 and Zhao et al.15, studying impacts of future climate scenarios on agriculture typically involves three steps. The first step is to select a future scenario (e.g., RCPs) under which General Circulation Models (GCMs) provide projections of climatic variables, such as temperature, precipitation, and humidity. In this study, we utilize projections from two GCMs, GFDL-ESM2M (GFDL)40 and HadGEM2-ES (HadGEM2)41, under RCP 6.0. The second step involves translating these climate variables into biophysical responses in crop yield using global gridded crop models (GGCMs); we use the GIS-based EPIC model42 (GEPIC) as the default model in our analysis, with the Lund-Potsdam-Jena managed Land43 (LPJmL) included in sensitivity tests. Importantly, we expand this step by incorporating labor-heat response functions (LHRs) to convert temperature, relative humidity, and air pressure to biophysical labor productivity responses (Fig. S1). To account for uncertainty in acclimatization capacity and access to protective adaptations, we employ two LHRs (Fig. S2a): a High response function, based on Dunne et al.19, and a Low response function, based on Roson and Sartori20. The High LHR represents labor groups more vulnerable to heat stress due to limited acclimatization ability or lack of protective adaptations, while the Low LHR reflects the response of labor with greater resilience to heat stress. The last step applies the biophysical responses to global economic and multisector dynamic models to explore agroeconomic outcomes. We employ a recent version of the Global Change Analysis Model (GCAM) (“Methods”) that incorporates evolving agricultural labor markets, as detailed in Sheng et al.44. By comparing systematically designed experiments (Table S2), we evaluate the impact of future climate scenarios on agriculture and its sensitivity, highlighting the crucial role of labor responses in this assessment.
Output from the previous stage enters as input for the following stage on the modeling chain. In biophysical responses, GGCM applications represent existing modeling efforts that address crop response to future climate scenarios (i.e., Nelson et al.14, and the LHR is the addition we highlight in our study, and key variables are highlighted in red. Climate inputs to the labor-heat response model include near-surface air temperature, near-surface relative humidity, and surface air pressure. Detailed data processing is provided in Supplementary Methods and Fig. S1.
The reference GCAM scenario, which accounts for heterogeneous productivity growth and the transition of regional agricultural labor markets, outlines an agroeconomic pathway to the end of the century assuming no changes in biophysical crop yield and labor productivity to future climate scenarios, i.e., the Reference experiment. The Crop-only experiment includes the effects of future climate scenarios on crop yields but excludes labor productivity, while the Crop and Labor experiment considers both biophysical crop yield and labor productivity responses. For clarity, we focus on two key scenarios in the main Results: the Higher heat-induced labor productivity loss scenario, generated based on climatic projections from HadGEM2 and the High labor-heat response function; and the Lower heat-induced labor productivity loss scenario, combining climatic projections from GFDL with the Low labor-heat response function. Crop responses are based on the GEPIC model for both the Higher and the Lower scenarios. Results from additional scenarios are provided in Supplementary Information (SI). We aggregate GCAM regions into IPCC R10 regions to communicate the results effectively, when applicable, and regional mapping can be found in Table S3.
Results
Agricultural labor productivity loss and employment impact
Global mean labor productivity in crop production is projected to decrease by 5% (Lower) to 18% (Higher) by the end of the century due to heat stress under RCP 6.0 (Fig. S2b). The impact is highly heterogeneous across regions. In the Higher scenario (HadGEM2-High LHR), projected labor productivity losses are highest in Southeast Asia (27%), Africa (23%), and South Asia (20%), while Europe (1%) and Reforming economies (2%) experience relatively small impact (Fig. 2a). In the Lower scenario (GFDL-Low LHR), the losses are relatively moderate, with crop labor in South Asia (8%), Africa (5%), and Southeast Asia (6%) remaining the most heat stressed. Across crops, global heat-induced labor productivity loss is higher for crops that require warm and humid environments, such as oil palm (7% to 32%), sugar crops (7% to 24%), and rice (7% to 23%), and lower for crops that require temperate climate and moderate humidity, like wheat (Fig. S3).
a shows heat-induced labor productivity changes in crop production by 2100 at the water basin level. b shows changes in agricultural labor productivity loss and employment over time. c shows the decomposition (share) of employment increase by region and agricultural sectors by 2100. Others denote the rest of GCAM crops. ROW denotes the rest of the world. Data source: labor productivity changes in a and b are calculated based on climatic projections under RCP 6.0 from ISIMIP2b; other outcomes in b and c are GCAM results.
The projected heat-induced labor productivity loss increases nonlinearly over time (Fig. 2b) due to dynamic changes in climate variables and the nonlinear transformation in labor-heat response functions (Fig. S2a). In the Higher scenario, global heat-induced labor productivity loss relative to the Reference grows from 3% by 2040 to 7% by 2060, 15% by 2080, and reaches 18% by 2100. The Lower and Higher scenarios show comparable productivity losses by 2040, but a 13-percentage point difference by 2100 (Fig. 2b), due to larger differences in the projected WBGT across HadGEM2 and GFDL later in the century (Fig. S4). Heat-induced agricultural labor productivity loss leads to lower agricultural output per worker, resulting in relatively higher agricultural labor input and slowing down the labor transition from agriculture to the rest of the economy. Considering the impact on labor (Crop and Labor vs. Crop-only), global agricultural employment in 2100 is projected to increase by 4% (20 million people) (Lower) to 18% (92 million people) (Higher). This increase in employment drives a corresponding rise in global agricultural wages, ranging from 1% (Lower) to 10% (Higher) by 2100 (Fig. S5).
Heat stress disproportionally affects densely populated agricultural regions, with Africa (48%), South Asia (31%), and Southeast Asia (13%) jointly accounting for 92% (84 million people) of global heat-induced agricultural employment increase in 2100 in the Higher scenario (Crop and Labor vs. Crop-only) (Fig. 2c). Higher demand for agricultural labor in Crop and Labor compared to Crop-only contributes to increases in agricultural wages by 34% in Africa, 20% in South Asia, and 34% in Southeast Asia, in the Higher scenario (Fig. S5). Regions in their late stage of agricultural transformation, such as North America and Pacific OECD countries, with agricultural employment share lower than 1% and moderate heat stress impact, contribute less than 0.02% of the total employment increase. South Asia (45%), Africa (38%), and Southeast Asia (11%) remain the top 3 contributing regions to global agricultural employment expansion (20 million people) in the Lower scenario, and agricultural wages increase by 6% in these regions (Fig. S5). Across agricultural commodities, vegetables, fruits, and rice jointly account for more than half of the agricultural labor employment expansion in both Higher and Lower scenarios (Fig. 2c). These crops involve a substantial amount of agricultural labor increase due to relatively higher labor intensity and increasing demand, and labor producing these crops is exposed to a high level of heat stress as these crops are usually planted in warm and humid regions.
Long-term agroeconomic implications of heat stress-induced labor productivity loss
Despite agricultural employment expansion, heat-induced labor productivity loss still results in lower crop production and consumption and higher crop prices at the global level (Figs. 3a and S6). In the Higher scenario, the global production of key crops (i.e., corn, rice, wheat, and soybean) and other crops in 2100 is projected to decrease by 2% (93 million tonnes, Mt) and 6% (704 Mt), respectively, due to the heat stress-induced labor productivity loss (i.e., Crop and Labor vs. Crop-only). In the Lower scenario, global production reduction in key crops (15 Mt) and other crops (108 Mt) is smaller. Reduced global production contributes to higher crop prices. Due to the lower labor productivity, crop prices are 1% (Lower) to 4% (Higher) higher for key crops and 1% (Lower) to 7% (Higher) higher for other crops by 2100. In the Higher scenario, reduced production also affects inputs in agricultural production. Globally, with a 2% increase in crop yield and a 2% reduction in production, key cropland decreases by 4% (27 million hectares, Mha). Conversely, a 16% decrease in labor productivity with the same 2% reduction in production leads to a 17% increase (14 million people) in employment for key crops. For other crops, a 2% decrease in yield and a 6% reduction in production reduce the land area by 4% (50 Mha). A 22% decline in labor productivity paired with a 6% production reduction increases labor demand by 21% (77 million people) for other crops. Under the Lower scenario, the market response to heat-induced labor productivity loss is smaller compared to the Higher scenario, with a 4% rise in labor and minimal changes (±1%) in other market outcomes at the global level, for both key crops and other crops.
a shows responses in key crop production, producer price, consumption, labor inputs, and land inputs. The green lines denote the World means. Boxplot is across 32 GCAM regions for aggregated key crops (corn, rice, wheat, and soybean). Whiskers represent the values of the 5th to the 95th percentile; boxes range from the 25th to the 75th percentile. The thick black lines denote the medians. Colored dots denote the aggregated regional means. b shows changes in trade and food prices. Net import to consumption share and net import level presented in this figure are for Key crops only. Food price indexes are weighted by food consumption, including both crop and livestock consumption. Net import to consumption (%) and Net import (Mt) are measured in level difference, while changes in the food price index are measured as a percentage change. Results in a and b are from comparing the Crop and Labor scenario with the Crop-only scenario. Data source: GCAM results.
Heat stress disproportionally affects tropical regions, leading to heterogeneous regional agricultural market responses to heat-induced labor productivity loss. In the Higher scenario, key crop production in Crop and Labor compared to Crop-only is projected to be 11% (63 Mt) lower in Africa, 8% (29 Mt) lower in Southeast Asia, and 3% (20 Mt) lower in China. For other crops, Latin America is projected to see the highest production reduction, with a decrease of 16% (366 Mt). This considerable reduction is primarily driven by a 24% (335 Mt) decrease in sugar crop production in Brazil due to high heat-induced labor productivity loss for labor that produces sugarcane. These regional production declines are associated with notable increases in regional crop prices (Figs. 3a and S6). Compared to Crop-only, key crop prices in Crop and Labor are projected to rise by 10% in Southeast Asia and prices for other crops are expected to increase by 18% in Latin America. Changes in regional production and biophysical productivity jointly affect the regional labor and land demand. In Western Africa, at the end of the century, a projected 33% decline in labor productivity and a 16% reduction in crop production are expected to increase labor demand for key crops by 26% (2 million people). Additionally, key cropland is expected to decrease by 15% (1 Mha) due to the reduced production and a 1% decline in crop yield caused by heat-induced labor productivity loss. Other crops show similar and larger market changes due to higher heat-induced labor productivity loss than key crops, both globally and regionally. Overall, the Higher scenario, characterized by higher heat-induced labor productivity loss, results in larger impacts on the regional agricultural market compared to the Lower scenario.
In the context of increasingly integrated global supply chains, heat-induced labor productivity loss in hot and humid regions further affects markets in the rest of the world through trade. In the Higher scenario, reduced key crop production in Africa, Southeast Asia, and China lead to key crop production expansion in 2100 by 3% in North America (25 Mt), Europe (14 Mt), and Reforming economies (5 Mt) to meet the global demand (Fig. 3a). Compared to Crop-only, key crops’ net import-to-consumption ratio, indicating the crop trade dependency, is 2–3 percentage points higher in Crop and Labor for Southeast Asia, China, and Africa (Fig. 3b). This results in higher net imports of key crops in Southeast Asia (8 Mt) and China (15 Mt). In Africa, despite a higher net import-to-consumption ratio, the reduction in key crop consumption levels for biomass and feed leads to a reduced net import (−2 Mt). Increased key crops net import in Crop and Labor is met by increased net export from North America (15 Mt), Europe (8 Mt), and reforming economies (5 Mt). Heat-induced labor productivity loss reduces the net export of other crops from Latin America (610 Mt) and Southeast Asia (44 Mt), resulting in decreased net imports of other crops in Africa (68 Mt) and Europe (55 Mt). Although international trade mitigates some of the impacts of heat-induced labor productivity loss on regional crop supply, the food price index in Crop and Labor is higher compared to Crop-only across regions, particularly in Southeast Asia (8%) and Africa (6%) (Fig. 3b).
Combined impact of future climate scenarios on agriculture from labor and crop responses
We evaluate the impact of future climate scenarios on agriculture by comparing agroeconomic outcomes from Crop and Labor (∆Combined) and Crop-only (∆Crop) relative to the Reference. The additional impact of including heat-stress impacts on agricultural labor (∆Labor) is measured as the difference between ∆Combined (combined impact) and ∆Crop (crop impact) (Fig. 4). In the Higher scenario, the ∆Labor for key crops is greater than or equal to the ∆Crop in magnitude across production, consumption, and price (Fig. 4a) and has the same direction as the ∆Crop. Therefore, the ∆Combined on key crop market outcomes are more than doubling compared to considering the ∆Crop alone. Overall, at the global level, considering both crop and labor responses to future climate scenarios introduces a 4% reduction in key crop production and a 5% increase in price. Due to minor crop productivity changes in the Higher scenario, the ∆Combined on key crop employment and land area are primarily driven by changes in labor productivity. Omitting labor responses in the assessment can underestimate the price impact by 4 percentage points and the employment impact by 19 percentage points while overestimating the production impact by 2 percentage points and land use impact by 4 percentage points for key crops. In the Lower scenario, the ∆Labor is smaller in magnitude compared to the ∆Crop for production, consumption, land use, and crop prices of the key crop. For key crop prices, ∆Labor is positive and ∆Crop is negative in the Lower scenario. ∆Labor on other crops is larger in magnitude than that on key crops, due to the relatively higher labor intensity.
a shows global climate impact (%) on the crop market. The black line denotes the combined (Crop&Labor) climate impact on the crop markets relative to the Reference scenario. Green bars denote the climate impact through crop yield response (Crop-only), and the orange bars denote the difference between Crop and Labor and Crop-only. b shows regional climate impact (%) on the key crops market in the Higher scenario. The results for Other Crops in the Lower scenario are in Fig. S7. Data source: GCAM results.
Further decompositions reveal heterogeneous ∆Crop and ∆Labor across regions (Figs. 4b, and S7, S8) and crops (Fig. S9). For example, in the Higher scenario, ∆Crop and ∆Labor are both negative in Africa, leading to a 14% reduction in key crop production in Africa by the end of the century, with 70% of the production reduction coming from ∆Labor. In Southeast Asia, ∆Crop and ∆Labor are both negative and comparable in magnitude. The ∆Combined leads to a reduction of key crop production in Southeast Asia by 12% in 2100, 61% of which comes from ∆Labor. In contrast to Southeast Asia, ∆Crop and ∆Labor in South Asia largely offset each other and result in a small key crop production reduction (1%). For production responses of key crops, we observe offsetting ∆Crop and ∆Labor in Reforming economies and the Middle East, with a negative ∆Crop dominating in Reforming economies (−6%) and a positive ∆Crop dominating in the Middle East (+9%). Changes in production further induce changes in employment and land use. In the Lower scenario, the ∆Combined on the key crops market is mainly driven by the crop impact (Fig. S8).
Among key crops, the ∆Labor is most pronounced in rice production (Fig. S9). In the Higher scenario, the ∆Combined on global rice production (−5%) and price (−8%) are mainly due to the ∆Labor. In the Lower scenario, the ∆Labor on rice is also relatively large, exceeding one-third of the ∆Crop. Across regions, the ∆Labor shows a pronounced share of the ∆Combined on production and consumption in Africa and for consumption and price in Southeast Asia (Fig. S8).
Discussion
Existing assessments of climate impact on agriculture have primarily focused on crop responses, often overlooking the potentially considerable and adverse impacts on labor. To address this gap, we incorporate labor-heat responses into the established model coupling framework14 used for studying crop yield responses to future climate scenarios. By leveraging climatic projections and biophysical yield impact data, together with recent advancements in representing regional evolving agricultural labor markets in a global multisector dynamic model, GCAM, our study enhances the understanding of changes in the agricultural sector under future climate scenarios. Here, we highlight a few key insights.
First, omitting labor responses underestimates the negative impact of future climate scenarios on agriculture
The main scenarios in our study, which rely on biophysical yield responses projected by GEPIC, showed global negative impacts on crop production under RCP 6.0 (Fig. S10). In contrast, LPJmL scenarios suggest global positive impacts through crop yield responses, though both outcomes vary across regions (Fig. S11). However, the impact through labor productivity responses and the corresponding market responses is generally negative, as higher WBGT increases the risk of extreme heat exposure and the proportion of the agricultural labor force subject to heat stress. Previous studies have primarily focused on crop yield responses under high RCP scenarios. For instance, in Nelson et al.14, the model comparison demonstrated a 20% increase in mean crop prices in response to a 17% decrease in mean crop yield under RCP 8.5. Accounting for labor responses is likely to exacerbate these market impacts.
Our results further highlight the heterogeneity in heat-induced labor productivity across crop sectors, driven by differences in crop-specific growing environments. By incorporating regional crop planting and harvesting calendar information, along with historical harvested area data, we identified varying degrees of heat-induced labor productivity losses among workers producing different crops. For instance, farmers cultivating sugar crops, oil palm, and rice are particularly affected (Figs. S3 and S9). More broadly, regional production patterns and production mixes will be affected since regional and sectoral production costs are influenced by changes in factor productivity when labor-heat responses are considered.
Second, impacts of future climate scenarios on agriculture through crop and labor responses are highly sensitive to the choice of modeling inputs
In our study, across all combinations of GCMs, GGCMs, and LHRs, projected changes (∆Combined) in global crop production, relative to the Reference scenario, range from −4% to 0% for key crops, and -5% to 2% for other crops, which contribute to changes in projected crop price by −4% to 5% for key crops and −6% to 7% for other crops in 2100 (Fig. S12, S13). The regional implications are more pronounced. Previous studies have also highlighted the uncertainty in biophysical crop yield responses across GCMs and crop models45. While our estimates of exogenous agricultural labor productivity loss generally align with the literature19,29,35, showing that labor in Africa, Southeast Asia, and South Asia is more vulnerable to future higher temperatures and humidity, the magnitude is sensitive to both GCM and labor-heat response function, as demonstrated in our results.
Consequently, the combined impact of future climate scenarios on agriculture and its decomposition into labor and crop responses are highly dependent on the model input choices (Fig. S10–13). A High LHR generally leads to a more negative impact on agriculture compared to a Low LHR. The direction of crop yield responses is different across crop models, with LPJmL tending to show a crop yield increase, resulting in more favorable crop production outcomes than GEPIC.
It is also crucial to note that crop and labor responses may interact. For instance, when labor productivity decreases, crop producers might shift to more capital-intensive technologies, resulting in different crop yields. One advantage of using GCAM for this analysis is that it endogenously captures the dynamic interactions between crop and labor responses. Additionally, changes in crop yield and labor productivity influence agricultural production in distinct ways, given that land and labor supply behave differently, with pronounced regional variations. Thus, the combined impact on agriculture is not a simple summation of crop and labor responses, as illustrated in Figs. S10, S11 and S14.
Third, future socioeconomic changes and agricultural labor market evolution are important determinants of future agricultural employment and their heat stress exposure
These factors determine the long-term distribution of agricultural employment across sectors and regions, and overlooking these dynamics may lead to biased estimates of human heat exposure scales and the associated economic consequences. For example, projected population and economic growth in Africa suggest substantial increases in agricultural production and employment, which will require considerable agricultural employment44. Missing the socioeconomic changes and agricultural transformation could misrepresent the distribution of agricultural production and lead to misleading assessments of climate impacts on agricultural markets. Moreover, heterogeneous trends in socioeconomic changes coupled with disproportionate regional warming over time contribute to a nonlinear impact of future climate scenarios on agriculture through labor responses.
Our work lays important foundations for future studies, but certain limitations remain. This study focuses on the agriculture market response to future climate scenarios through labor and crop pathways while highlighting the sensitivity of agroeconomic outcomes to RCP, GCM, and LHR choices. However, our study explores only part of the sensitivity space, whereas the impacts of future climate scenarios on agriculture may be sensitive to modeling choices across multiple factors. These include socioeconomic and climate scenarios6,23,44, as well as biophysical labor productivity scenarios, which depend on the selection of heat stress metrics46,47 and labor-heat response functions. Future studies should broaden this analysis by incorporating alternative heat stress metrics and additional RCP and SSP scenarios.
While we introduce exogenous labor productivity growth to represent technological innovation and mechanization, we did not capture how potentially labor-heat responses evolve over time. Existing labor-heat response functions, often calibrated based on context-specific field data, may have limited global applicability48,49. Since the adoption of capital-intensive protective strategies (e.g., air-conditioned tractors or robotic harvesting) varies across regions and periods50, omitting these dynamics may overestimate the impact of human heat exposure and underestimate the corresponding capital demand. Future studies should incorporate regional labor-heat response functions that account for technological advancement and adaptation.
Additionally, our study did not account for responses in the livestock sector and the rest of the economy to future climate scenarios. Heat stress reduces animal welfare and productivity51,52,53,54, negatively impacting livestock production55,56, which may lead to ripple effects on water consumption, feed demand, and land use change57,58. Furthermore, heat-induced agricultural employment expansion slows the labor transition from agriculture to the rest of the economy. Future studies are needed to further examine such transition considering the heat stress impact on labor force participation and macroeconomy5,27,59,60. Addressing these dynamics is crucial for understanding the interconnected impacts of heat stress on agriculture and the broader economy.
Methods
Biophysical responses to future climate scenarios
In our study, we employ daily gridded (0.5-degree) bias-adjusted atmospheric projections under RCP.6.0, from the 2b round of Inter-Sectoral Impact Model Intercomparison Project simulation (ISIMIP2b)61, to represent weather conditions under future climate scenarios. We choose projections based on CMIP5 output from two GCMs, GFDL40 and HadGEM241. These two GCMs are widely used15,23 and span well the uncertainties around the ensemble global mean temperature across key GCMs62. HadGEM2 projects a higher global mean temperature than GFDL across all RCPs, and the differences increase over time under RCP 6.0 (Figs. S4). For GGCM, we rely on the projections from GEPIC42, accounting for CO2 fertilization. GEPIC projections indicate that crop yield is negatively affected overall but varies considerably across regions and crops, consistent with discussions in the literature13,45,63. Recognizing the uncertainty in GGCMs, we provide results with crop yield response to future climate scenarios from LPJmL40 in SI (Figs. S11–S14).
To obtain heat-induced labor productivity loss, we first evaluate heat stress experienced by labor using WBGT. WBGT is widely used in occupational health studies to quantify heat stress on humans and provide working safety guidelines studies64,65. We estimate daily gridded WBGT using daily gridded surface air temperature, air pressure, and relative humidity following an approximation from Dunne et al.19 that assumes adaptation with shading and optimization of activity scheduling and clothing to avoid peak temperatures and direct sunlight.
Next, we evaluate labor productivity response to WBGT. Detailed data processing can be found in Section S1.1 and Fig. S1. To account for uncertainties in labor productivity’s response to heat stress, we employ two labor-heat response functions (LHR) to represent a High response function, based on Dunne et al.19, and a Low response function, based on Roson and Sartori20 (Figs. S2a). An LHR with lower labor productivity loss with a given WBGT can represent a greater adaptation potential with mechanization and/or better acclimatization.
In this study, heat-induced labor productivity impact is confined to labor in crop sectors. We project 4 exogenous heat-induced labor productivity loss scenarios based on 2 GCMs and 2 LHRs. HadGEM2 projections yield higher global mean WBGT (Fig. S4) and, therefore, higher global heat-induced labor productivity loss compared to GFDL. A high LHR (High), by definition, leads to higher labor productivity loss with the same level of WBGT compared to a low LHR (Low). Across these 4 scenarios, HadGEM2-High and GFDL-Low span the heat-induced labor productivity loss in this study (Fig. S2b).
GCAM
GCAM is an integrated multisector model developed and maintained at the Pacific Northwest National Laboratory’s Joint Global Change Research Institute (see http://jgcri.github.io/gcam-doc/ for model documentation). GCAM has detailed representations of the economy, energy, agriculture, land use, and water in 32 geopolitical regions. Land allocation, water use, and agricultural production are modeled at 384 land subregions (intersection of water basins and regions). In each land subregion, representative profit-maximizing agricultural producers in crop, livestock, and forestry sectors make production decisions by determining production input uses given constant-return-to-scale technologies and a vector of input and output prices. The model is calibrated to historical data in the base year of 2015 and runs in 5-year steps to 2100, mainly driven by future socioeconomic and technological changes.
GCAM has been widely used in developing future agroeconomic pathways66,67,68,69 and studying climate impacts on agriculture14,15,70. In this study, we adopt the version of GCAM developed by Sheng et al.44, in which agricultural labor markets were incorporated into GCAM v7.0. Specifically, we construct an exponential agricultural labor supply function, specifying the relationship between the agricultural labor share of the labor force and the agricultural wage rate71,72. We leverage the urbanization projections to exogenously represent the labor transition between the agriculture sector and the rest of the economy. The demand for agricultural labor is jointly determined by the production level and labor intensity (labor input needed per unit of output) of adopted production technology. Based on various data sources, including ILO8, USDA73, FAO74, and GTAP75, we calculate labor intensity and capital intensity (capital input needed per unit of output) for each agricultural commodity in the base year. We further introduce exogenous labor productivity growth that is calibrated to GDP in the Shared Socioeconomic Pathways 2 (SSP2) Middle-of-the-Road scenario, and this growth features the decreasing labor intensity due to human capital accumulation, technology innovation, and mechanization in the agriculture sector over time. The global average agricultural labor productivity is set to more than quadruple from 2015 to 2100, with the highest productivity growth in Africa (Figs. S15).
To allow endogenous response in labor productivity, capital productivity, and labor-capital substitution, we introduce a parameter to differentiate the labor and capital intensities across production technologies within each commodity. Different labor and capital intensities across technologies incentivize crop producers to adopt more profitable production technology, given the labor and capital prices at the market equilibrium. For example, when labor wages increase, labor-intensive technologies are less profitable compared to capital-intensive technologies, and profit-maximizing crop producers are incentivized to adopt more capital-intensive technologies. This way, at the commodity level, labor productivity and capital productivity respond to input prices endogenously, and labor and capital substitution are achieved through technology transition. A more detailed description of GCAM and agricultural labor market representation can be found in Section S1.3 and the Supplementary Section 4 in Sheng et al.44. Across all experiments and scenarios of this study, the population76 and income77 projections from the SSP2 scenario were used. Biophysical crop yield growth assumptions in the Reference are based on FAO projections78,79.
Scenario design
Comparing Crop-only with Reference reveals the climate impact on agriculture through crop yield responses (∆Crop), which has been widely emphasized in the literature. Comparing Crop&Labor with Crop-only reveals the additional climate impact on agriculture through labor response (∆Labor). Note that a comparison between a Labor-only experiment and the Reference experiment also provides an estimate of climate impact through labor response (∆Labor* in Table S2). Given that we want to build on top of the literature addressing climate impact through crop response, we focus on ∆Labor estimate in the Results section and present ∆Labor* in the SI. To explore the sensitivity of climate impact on agriculture, we associate Crop-only and Crop&Labor experiments with different scenarios with alternative model choices of climate projections, crop responses, and labor responses (Fig. 1 and Table S2). Across choices of 2 GCMs, 2 GGCM, and 2 LHRs, we generate 4 estimates of ∆Crop, 8 estimates of ∆Labor, and 8 estimates of ∆Combined. In the Results, we focus on 2 ∆Crop, 2 ∆Labor, and 2 ∆Combined from the Higher and Lower scenarios. Additional results are provided in SI. For clarity, in the Results, we group GCAM crops to key crops and other crops (see Tables S4–S6 for GCAM crop mapping), and aggregate regions to the IPCC R10 regions (see Table S3 for regional mapping).
Data availability
GCAM simulation results and processed data80 used for analysis in this study are available at Zenodo https://zenodo.org/records/15168636.
Code availability
GCAM is an open-source model available via GitHub at https://github.com/JGCRI/gcam-core, and the specific version of the model used in this study is archived at http://github.com/realxinzhao/paper-nfood2024-AgLaborEvolution-GCAM and available via Zenodo at https://doi.org/10.5281/zenodo.13852265. Additional input for alternative experiments is available via Zenodo at https://zenodo.org/records/14783094. The R code for generating the main figures is archived at https://github.com/dsheng1026/paper-CEE2025-HeatStress-DisplayItems.
References
Mora, C. et al. Global risk of deadly heat. Nat. Clim. Change 7, 501–506 (2017).
Lüthi, S. et al. Rapid increase in the risk of heat-related mortality. Nat. Commun. 14, 4894 (2023).
Kjellstrom, T., Kovats, R. S., Lloyd, S. J., Holt, T. & Tol, R. S. J. The direct impact of climate change on regional labor productivity. Arch. Environ. Occup. Health 64, 217–227 (2009).
Flouris, A. D. et al. Workers’ health and productivity under occupational heat strain: a systematic review and meta-analysis. Lancet Planet. Health 2, e521–e531 (2018).
Graff Zivin, J. & Neidell, M. Temperature and the allocation of time: implications for climate change. J. Labor Econ. 32, 1–26 (2014).
Hertel, T. W. & De Lima, C. Z. Viewpoint: climate impacts on agriculture: searching for keys under the streetlight. Food Policy 95, 101954 (2020).
Khayat, M. E. et al. Impacts of climate change and heat stress on farmworkers’ health: a scoping review. Front. Public Health 10, 782811 (2022).
ILO. ILOSTAT. Last accessed on 2023-05-03. (2023)
Ceccherini, G., Russo, S., Ameztoy, I., Marchese, A. F. & Carmona-Moreno, C. Heat waves in Africa 1981–2015, observations and reanalysis. Nat. Hazards Earth Syst. Sci. 17, 115–125 (2017).
World Bank. Climate Investment Opportunities in India’s Cooling Sector. https://doi.org/10.1596/38340 (World Bank, 2022).
Li, X.-X. Heat wave trends in Southeast Asia during 1979–2018: the impact of humidity. Sci. Total Environ. 721, 137664 (2020).
Romanello, M. et al. The 2021 report of the Lancet Countdown on health and climate change: code red for a healthy future. Lancet 398, 1619–1662 (2021).
Zhao, C. et al. Temperature increase reduces global yields of major crops in four independent estimates. Proc. Natl. Acad. Sci. USA 114, 9326–9331 (2017).
Nelson, G. C. et al. Climate change effects on agriculture: Economic responses to biophysical shocks. Proc. Natl. Acad. Sci. USA 111, 3274–3279 (2014).
Zhao, X. et al. Global agricultural responses to interannual climate and biophysical variability. Environ. Res. Lett. 16, 104037 (2021).
Molina Bacca, E. J. et al. Uncertainty in land-use adaptation persists despite crop model projections showing lower impacts under high warming. Commun. Earth Environ. 4, 1–13 (2023).
Pilcher, J. J., Nadler, E. & Busch, C. Effects of hot and cold temperature exposure on performance: a meta-analytic review. Ergonomics 45, 682–698 (2002).
Kjellstrom, T., Holmer, I. & Lemke, B. Workplace heat stress, health and productivity – an increasing challenge for low and middle-income countries during climate change. Glob. Health Action 2, 2047 (2009).
Dunne, J. P., Stouffer, R. J. & John, J. G. Reductions in labour capacity from heat stress under climate warming. Nat. Clim. Change 3, 563–566 (2013).
Roson, R. & Sartori, M. Estimation of climate change damage functions for 140 regions in the GTAP 9 data base. J. Glob. Econ. Anal. 1, 78–115 (2016).
Nelson, G. C. et al. Global reductions in manual agricultural work capacity due to climate change. Glob. Change Biol. 30, e17142 (2024).
Somanathan, E., Somanathan, R., Sudarshan, A. & Tewari, M. The impact of temperature on productivity and labor supply: evidence from indian manufacturing. J. Polit. Econ. 129, 1797–1827 (2021).
Orlov, A., Sillmann, J., Aunan, K., Kjellstrom, T. & Aaheim, A. Economic costs of heat-induced reductions in worker productivity due to global warming. Glob. Environ. Change 63, 102087 (2020).
Orlov, A. et al. Global economic responses to heat stress impacts on worker productivity in crop production. Econ. Disasters Clim. Change 5, 367–390 (2021).
García-León, D. et al. Current and projected regional economic impacts of heatwaves in Europe. Nat. Commun. 12, 5807 (2021).
Matsumoto, K., Tachiiri, K. & Su, X. Heat stress, labor productivity, and economic impacts: analysis of climate change impacts using two-way coupled modeling. Environ. Res. Commun. 3, 125001 (2021).
Dasgupta, S. et al. Heat stress and the labour force. Nat. Rev. Earth Environ. 5, 859–872 (2024).
Sun, Y. et al. Global supply chains amplify economic costs of future extreme heat risk. Nature https://doi.org/10.1038/s41586-024-07147-z (2024).
de Lima, C. Z. et al. Heat stress on agricultural workers exacerbates crop impacts of climate change. Environ. Res. Lett. 16, 044020 (2021).
Orlov, A. et al. Human heat stress could offset potential economic benefits of CO2 fertilization in crop production under a high-emissions scenario. One Earth 7, 1250–1265 (2024).
Jones, B. et al. Future population exposure to US heat extremes. Nat. Clim. Change 5, 652–655 (2015).
Peter Timmer, C. Chapter 8 The agricultural transformation. in Handbook of Development Economics vol. 1 275–331 (Elsevier, 1988).
Charlton, D., Rutledge, Z. & Taylor, J. E. Chapter 77 - Evolving agricultural labor markets. In Handbook of Agricultural Economics (eds. Barrett, C. B. & Just, D. R.) vol. 5 4075–4133 (Elsevier, 2021).
ILO. Working on a warmer planet: the impact of heat stress on labour productivity and decent work (2019).
Vargas Zeppetello, L. R., Raftery, A. E. & Battisti, D. S. Probabilistic projections of increased heat stress driven by climate change. Commun. Earth Environ. 3, 183 (2022).
Russo, S., Marchese, A. F., Sillmann, J. & Immé, G. When will unusual heat waves become normal in a warming Africa?. Environ. Res. Lett. 11, 054016 (2016).
Im, E.-S., Pal, J. S. & Eltahir, E. A. B. Deadly heat waves projected in the densely populated agricultural regions of South Asia. Sci. Adv. 3, e1603322 (2017).
Dong, Z. et al. Heatwaves in Southeast Asia and Their Changes in a Warmer World. Earths Future 9, e2021EF001992 (2021).
Christiaensen, L., Rutledge, Z. & Taylor, J. E. Viewpoint: the future of work in agri-food. Food Policy 99, 101963 (2021).
Dunne, J. P. et al. GFDL’s ESM2 global coupled climate–carbon earth system models. part I: physical formulation and baseline simulation characteristics. J. Clim. 25, 6646–6665 (2012).
Jones, C. D. et al. The HadGEM2-ES implementation of CMIP5 centennial simulations. Geosci. Model Dev. 4, 543–570 (2011).
Liu, J., Williams, J. R., Zehnder, A. J. B. & Yang, H. GEPIC – modelling wheat yield and crop water productivity with high resolution on a global scale. Agric. Syst. 94, 478–493 (2007).
Bondeau, A. et al. Modelling the role of agriculture for the 20th century global terrestrial carbon balance. Glob. Change Biol. 13, 679–706 (2007).
Sheng, D. et al. Labour market evolution is a key determinant of global agroeconomic and environmental futures. Nat. Food https://doi.org/10.1038/s43016-024-01088-6 (2025).
Rosenzweig, C. et al. Assessing agricultural risks of climate change in the 21st century in a global gridded crop model intercomparison. Proc. Natl. Acad. Sci. USA 111, 3268–3273 (2014).
Buzan, J. R. & Huber, M. Moist heat stress on a hotter earth. Ann. Rev. Earth Planetary Sci. 48, 623–655 (2020).
Kong, Q. & Huber, M. Explicit calculations of wet-bulb globe temperature compared with approximations and why it matters for labor productivity. Earth's Future 10, e2021EF002334 (2022).
Wyndham, C. H. Adaptation to heat and cold. Environ. Res. 2, 442–469 (1969).
Foster, J. et al. An advanced empirical model for quantifying the impact of heat and climate change on human physical work capacity. Int. J. Biometeorol. 65, 1215–1229 (2021).
Vanschoenwinkel, J., Mendelsohn, R. & Van Passel, S. Do Western and Eastern Europe have the same agricultural climate response? Taking adaptive capacity into account. Glob. Environ. Change 41, 74–87 (2016).
Collier, R. J. & Gebremedhin, K. G. Thermal biology of domestic animals. Annu. Rev. Anim. Biosci. 3, 513–532 (2015).
Belhadj Slimen, I., Najar, T., Ghram, A. & Abdrrabba, M. Heat stress effects on livestock: molecular, cellular and metabolic aspects, a review. J. Anim. Physiol. Anim. Nutr. 100, 401–412 (2016).
Polsky, L. & Keyserlingk, M. A. G. von. Invited review: effects of heat stress on dairy cattle welfare. J. Dairy Sci. 100, 8645–8657 (2017).
Liu, W., Zhou, J., Ma, Y., Chen, S. & Luo, Y. Unequal impact of climate warming on meat yields of global cattle farming. Commun. Earth Environ. 5, 1–10 (2024).
Rahimi, J., Mutua, J. Y., Notenbaert, A. M. O., Marshall, K. & Butterbach-Bahl, K. Heat stress will detrimentally impact future livestock production in East Africa. Nat. Food 2, 88–96 (2021).
Thornton, P., Nelson, G., Mayberry, D. & Herrero, M. Impacts of heat stress on global cattle production during the 21st century: a modelling study. Lancet Planet. Health 6, e192–e201 (2022).
Bernabucci, U. et al. Metabolic and hormonal acclimation to heat stress in domesticated ruminants. Animal 4, 1167–1183 (2010).
Godde, C. M., Mason-D’Croz, D., Mayberry, D. E., Thornton, P. K. & Herrero, M. Impacts of climate change on the livestock food supply chain; a review of the evidence. Glob. Food Secur. 28, 100488 (2021).
Dasgupta, S. et al. Effects of climate change on combined labour productivity and supply: an empirical, multi-model study. Lancet Planet. Health 5, e455–e465 (2021).
Dell, M., Jones, B. F. & Olken, B. A. Temperature shocks and economic growth: evidence from the last half century. Am. Econ. J. Macroecon. 4, 66–95 (2012).
Lange, S. & Büchner, M. ISIMIP2b bias-adjusted atmospheric climate input data. https://doi.org/10.48364/ISIMIP.208515 (2017).
Warszawski, L. et al. The inter-sectoral impact model intercomparison project (ISI–MIP): project framework. Proc. Natl. Acad. Sci. USA 111, 3228–3232 (2014).
Hasegawa, T. et al. Risk of increased food insecurity under stringent global climate change mitigation policy. Nat. Clim. Change 8, 699–703 (2018).
Epstein, Y. & Moran, D. S. Thermal comfort and the heat stress indices. Ind. Health 44, 388–398 (2006).
Lemke, B. & Kjellstrom, T. Calculating workplace WBGT from meteorological data: a tool for climate change assessment. Ind. Health 50, 267–278 (2012).
Zhao, X. et al. The impact of agricultural trade approaches on global economic modeling. Glob. Environ. Change 73, 102413 (2022).
Graham, N. T. et al. Agricultural market integration preserves future global water resources. One Earth 6, 1235–1245 (2023).
Zhang, Y., Waldhoff, S., Wise, M., Edmonds, J. & Patel, P. Agriculture, bioenergy, and water implications of constrained cereal trade and climate change impacts. PLoS ONE 18, e0291577 (2023).
Zhao, X., Mignone, B. K., Wise, M. A. & McJeon, H. C. Trade-offs in land-based carbon removal measures under 1.5 °C and 2 °C futures. Nat. Commun. 15, 2297 (2024).
Snyder, A. et al. The domestic and international implications of future climate for U.S. agriculture in GCAM. PLoS ONE 15, e0237918 (2020).
Hill, A. E., Ornelas, I. & Taylor, J. E. Agricultural labor supply. Annu. Rev. Resour. Econ. 13, 39–64 (2021).
Ray, S., Haqiqi, I., Hill, A. E., Taylor, J. E. & Hertel, T. W. Labor markets: a critical link between global-local shocks and their impact on agriculture. Environ. Res. Lett. 18, 035007 (2023).
USDA. TFP indices and components for countries, regions, countries grouped by income level, and the world. Int. Agric. Product. 1961–2020 (2022).
FAO. FAOSTAT database on agriculture. Last accessed on 2022-09-14 (2019).
Aguiar, A., Chepeliev, M., Corong, E. L., McDougall, R. & Mensbrugghe, D. van der. The GTAP database: version 10. J. Glob. Econ. Anal. 4, 1–27 (2019).
Kc, S. & Lutz, W. The human core of the shared socioeconomic pathways: population scenarios by age, sex, and level of education for all countries to 2100. Glob. Environ. Change 42, 181–192 (2017).
O’Neill, B. C. et al. The roads ahead: Narratives for shared socioeconomic pathways describing world futures in the 21st century. Glob. Environ. Change 42, 169–180 (2017).
World Agriculture towards 2030/2050: The 2012 Revision. https://doi.org/10.22004/ag.econ.288998 (2012).
Alexandratos, N. et al. World agriculture: towards 2030/2050. Interim report. Prospects for food, nutrition, agriculture and major commodity groups (2006).
Sheng, D. Supplementary data: model results for ‘Omitting labor responses underestimates the effects of future heat stress on agriculture’ https://zenodo.org/records/15168636 (2025).
Acknowledgements
The authors are grateful for the support from the U.S. Department of Energy, Office of Science, as part of research in the Multi-Sector Dynamics, Earth and Environmental System Modeling Program. The Pacific Northwest National Laboratory is operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830. The views and opinions expressed in this paper are those of the authors alone. We extend our gratitude to Siwa Msangi, Ying Zhang, Siddarth Durga, Noah Prime, and other PNNL-JGCRI colleagues for their valuable comments and suggestions.
Author information
Authors and Affiliations
Contributions
D.S., X.Z., and J.E. conceptualized the research. D.S., X.Z., and P.P. contributed to the model development for this study. D.S. carried out the simulations and visualization of results. D.S. and X.Z. drafted the manuscripts. D.S., X.Z., J.E., P.P., S.M., B.O., C.T., and M.W. contributed to discussing the results, writing, and/or reviewing the paper.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Communications Earth & Environment thanks Anton Orlov and Martina Bozzola for their contribution to the peer review of this work. Primary Handling Editors: Martina Grecequet. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Sheng, D., Zhao, X., Edmonds, J.A. et al. Omitting labor responses underestimates the effects of future heat stress on agriculture. Commun Earth Environ 6, 400 (2025). https://doi.org/10.1038/s43247-025-02318-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s43247-025-02318-w
