Maize migration mitigates the negative impact of climate change on China’s maize yield

Crop migration as an adaptation to modulate climate change’s impact on crop yields presents both benefits and risks. We explored how maize migration in China modulates yield responses to climate change and quantified the potential economic benefits of maize migration as an adaptation strategy. We employed a panel data model to identify and measure the factors driving the relocation of maize area, linear regression to quantify the effects of maize migration on climate exposure and irrigated area, and an econometric model to estimate the effects of maize migration on yield. The results show that rise in temperature has a significant negative effect on maize area and that precipitation has a significant positive effect. The migration of maize area is driven by socio-economic factors including agricultural gross domestic product, power of farming machines, and fertilizer input. Moreover, expanded irrigation reduces the adverse effects of high temperatures on maize yield, thereby influencing adaptive crop migrations. The beneficial effects of maize migration are primarily achieved by reducing the adverse effects of extreme heat and strengthening the positive effects of irrigation. However, the extent of this adaptation is jointly affected by agricultural policies, irrigation infrastructure, and economic factors. Current market-oriented agricultural policies may be effective in guiding spatial shifts in maize distribution to align with climate-driven changes, potentially decreasing the vulnerability of China’s maize yield to the impact of climate change. China’s food security policies need to consider climate-driven spatial shifts in crop cultivation and enhance food subsidy policies to highlight the benefits of investment in climate change adaptation, such as adjusting cropping acreage and irrigation to farmers in North China.


Introduction
In China, maize is one of the most important staple crops, accounting for 39.1% of the total grain production and significantly contributing to the country's food security (Han et al 2022, Peng et al 2023).Nevertheless, the expected impact of climate change is likely to exacerbate food security challenges, harming maize productivity (Zhang et al 2017, Hou et al 2021, Pickson et al 2022).Therefore, it is imperative to understand how adaptation measures moderate the impacts on maize production for designing and implementing targeted policies to reduce climate change risks and facilitate adaptation.
Extreme temperatures due to climate change are predicted to reduce average yields for several major crops (Rezaei et al 2023).However, these impacts vary across space, with some cold areas acquiring benefits from increase in moderate temperatures and some hot areas suffering harms from increase in extreme temperatures (Rising and Devineni 2020).Changes in productivity caused by climate change drive farmers to substitute crops and move to new areas (Cui 2020).Thus, farmers can adapt to climate change by spatially adjusting their cropping patterns (Costinot et al 2016).Sloat et al (2020) showed that crop migration mediates rising temperatures in crops' growing seasons.This implies that empirical studies that ignored crop migration adaptation might have overestimated the impact of adverse climate change on crop yields.However, the beneficial effects of dynamic crop migration as an adaptation to modulate the impact of climate change on crop yields remain unquantified.
This study explores how maize migration in China modulates maize yields' response to climate change and quantifies the potential production benefits of maize migration.China's maize acreage showed a year-on-year increase of 0.37 million ha from 1949 to 2022.Since 2000, maize cultivation in northern China has expanded rapidly, with the acreage under cultivation expanding by 42% (NBSC 2022).The geographical centroid of maize production has distinctly shifted towards the northeast at a rate of 15.65 km yr −1 between 2000 and 2015, with the most significant distance being 259.10 km (Fan et al 2018).
This study makes the following contributions: first, we consider the adaptation of crop spatial distribution adjustments when assessing the impact of climate change on crop yields.Previous studies evaluated the impact of climate change on crop yields, assuming that crop spatial distribution remained unchanged, but ignored farmers' adaptation to climate change by altering crop spatial distributions (Schlenker and Robert 2009, Chen et al 2016, Zhang et al 2017).Although some studies have already examined the impact of climate change on the spatial distribution of crops (Zhao et al 2016, Ewert et al 2015, Cui 2020), there has been little research exploring how changes in crop spatial distribution impact crop yield response to climate change (Leng and Huang 2017).
Second, in addition to climatic factors, socioeconomic factors such as agricultural policies, urbanization, irrigation infrastructure, and production techniques play a significant role in crop migration (Hu et al 2019b).These factors directly affect the profitability of maize cultivation, thereby indirectly affecting farmers' decisions on adjusting maize area (Cui 2020).However, few studies have distinguished between climate factors and socioeconomic factors regarding their respective impacts on the spatial distribution of maize in China.
Our analysis comprises the following steps: (1) identifying the extent to which each of the driving factors affects crop migration; (2) quantifying the effects of maize migration on extreme heat exposure and the irrigation ratio; (3) assessing the role of maize migration in modulating the impact of climate change on maize production.

Study area
Maize is planted throughout China, with regional differences in varieties and growing season.Five major maize production regions-Northeast, Northwest, Southwest, South, and North-were included in the study area (supplementary information 1 figure S1).Maize is categorized by season (Chen et al 2016).Spring maize, typically planted in April and harvested in late September, is cultivated in the mountainous regions of Northeast, Northwest, and Southwest China.Summer maize is grown in June, has a slightly shorter growing season than spring maize, and is primarily produced in the North China Plain.Autumn maize is primarily planted in the mountainous regions of Southwest China (Liu 1993).

Data
The high-resolution maize distribution data for the years 2000-2019 were provided by the National Ecological Science Data Centre (https://cstr.cn/31253.11.sciencedb.08490;Luo et al 2020).The data were drawn from GLASS LAI remote sensing images and were provided on a 1 km × 1 km resolution grid.The 500 m irrigation cropland distribution data for the years 2000-2019 were obtained from Zhang et al (2022).We upscaled these data to a 10 km × 10 km resolution grid for consistency with the weather data.The grid data on the spatial distribution of maize and irrigation are displayed in the supplementary information (figure S2).The accuracy of the gridded data in this study was further validated based on county-level and the provincial statistical data.The county-level maize planting area data were obtained from the Chinese Academy of Agricultural Sciences, which has been widely used in relevant research on agricultural production in China (Chen et al 2016, Wang et al 2024).The provincial statistical data were obtained from the National Bureau of Statistics (www.stats.gov.cn).The results showed that the gridded areas data are highly aligned with these two data sources (supplementary information figure S3).
Weather data were obtained from the China Meteorological Data Service Centre (http://data.cma.cn/en).The center has been publishing the daily weather data of over 800 weather stations across China since 1951.We employed the kriging spatial prediction technique to interpolate daily weather data from each meteorological station onto a 10 × 10 km grid, aligning it with the spatial distribution of maize (Cressie and Wikle 2015).
The historical data of each county's agricultural input data and each provinces' maize yield between 2000 and 2019 were obtained from the Chinese Academy of Agricultural Sciences and a series of statistical books by the Provincial Bureaus of Statistics.

Maize area model
We used a panel data model to measure the factors driving the relocation of maize area (table 1).With the log-transformed factors, the regression equation is as follows: where ln is the natural log, i is the index of the 10 × 10 km grid, and t is the annual index.area it refers to the maize area for grid i in year t.C it represents climate variables, including temperature (TEM) and total precipitation (TSP) during the maize growing season.Referring to Cui (2020), temperature and precipitation are the most critical climatic variables affecting the suitable crop planting area.E denotes the socio-economic variables, including the proportion of the county's agricultural gross domestic product (Agdp), per ha Labor (Labr), per ha machinery (Mach), irrigation area (Irrg), cropland area (Crland), and Urbanisation level (Urban) (see table S1 for more details about these variables).These socio-economic variables were also widely adopted in similar studies (Li et al 2015, Hu et al 2019a, Fan et al 2020, Qian et al 2022).δ i represents grid fixed effects to control for time-invariant grid-specific characteristics such as geographic location and soil quality.θ t represents year fixed effect to control for technological progress, policy change, and maize price fluctuation (Wang et al 2022).ε it is the error term.α m is the estimated coefficient for climate factors, indicating that when temperature and precipitation change by 1%, the planting area of maize will change by α%.β n is the estimated coefficient for socio-economic variables, indicating that when socio-economic variables change by 1%, the planting area of maize will change by β%.
As maize policies altered farmers' marginal returns to maize cultivation, these policies may have influenced the extent to which farmers respond to climate change (Roberts and Schlenker 2013).China started a nationwide maize stockpiling program in 2007.A key feature of this policy is that the government collects maize from farmers at minimum support prices (Huang and Yang 2017).In 2016, China government implemented the direct payment maize subsidy policy tied to planting acres.Under the new policy, maize price is determined by the market conditions (Wu and Zhang 2016).
We proposed a causal framework to examine whether maize policies intensify or weaken the impacts of temperature on maize area.Given the relatively minor impact of precipitation change on maize area, we focused on how maize policies affect temperature-driven changes in maize area.We constructed a model incorporating the interaction between project implementation and temperature.Conceptually, we used the interaction model to estimate the impact of maize policies on the relationship between temperature and maize area.The interaction model is expressed as follows: where  1).γ 1 represents the impact of policy implementation on maize acreage.γ 2 represents the impact of maize policies on the sensitivity of maize acreage to temperature.

Climate impact isolation
To separate the impact of climatic and socioeconomic factors on the migration of maize area, we followed Li et al (2015) and predicted maize area using the real-world socio-economic factors, holding the climatic factors constant at the 2000 value: where HA C i,t is the exponential conversion of the predicted maize area from equation (1).A i,2000 and C i,2000 represent the actual maize area and the climate variables, respectively, in 2000 for grid i. C i,t denotes the climate variables in t for grid i. α is the coefficient estimated in equation (1).

Maize migration impact on climate exposure and irrigated area experienced by maize
Linear regressions of growing season temperature and precipitation over time were analyzed at the grid cell level.We used per gridded maize area as the regression weight.This approach ensures that grid cells with a larger proportion of land dedicated to cultivating maize is given greater weights, effectively capturing the entire temperature and precipitation distribution experienced by maize (Sloat et al 2020): where C it represents climate variables, including temperature (TEM) and total precipitation (TSP) during the maize growing season; t represents the year; ε is the error term; weights represent regression weights, where dynamic maize area from 2000 to 2019 is used.β 0 indicates the linear trend of temperature and precipitation experienced by maize.
We also considered a counterfactual scenario as regression weights-static maize grid areas in 2000: where C it , t and ε are the same as in equation ( 4).Weights represent regression weights, where static maize area in 2000 is used.β c indicates the linear time trend of temperature and precipitation experienced by maize if it remains at the distribution observed in 2000.
We also used quantile regression to estimate equations ( 5) and ( 6).This analysis was conducted at the 95% quantile of temperature, corresponding to 28 • C. Temperatures above 28 • C are defined as extreme high temperatures, as this temperature represents a threshold for maize growth (Wang et al 2024).β indicates the time trends in the warm bound (95th percentile) of growing season temperature for maize areas.

Maize yield model
We employed an econometric model to estimate the effects of maize migration on yield.Following Leng and Huang (2017), we used two gridded maize area weights to aggregate gridded climate data to the province-level: one using transient weights incorporating change in the maize's spatial distribution from 2000 to 2019 and the other based on a static maize grid map in 2000.The static maize grid map assumed that the spatial distribution of maize remained constant over the study period.The grid irrigation data was similarly aggregated.
The econometric specification of the maize yield function employed in this study follows that of Wu et al (2021).Thus, we examined the impact of climate change on maize yield production across 31 Chinese provinces using the specification below: where ln is the natural log and y it is the yield for province i in t.The tem it and pre it represent temperature and precipitation, respectively, during the maize growing season.The irr it is the irrigated maize area.X it denotes the control variables including fertilizer and machinery inputs.The trend it indicates the time trend to control for technological progress.δ i represents province fixed effects to control for timeinvariant characteristics such as geographic location and soil quality.
We employed two aggregated province-level climate datasets to fit the econometric models: the transient maize maps from 2000 to 2019 and the static maize map in 2000.Considering group-wise heteroscedasticity, cross-sectional correlation, and autocorrelation within the panels, we used a feasible generalized least square to estimate the model (Wu et al 2021).Statistical tests can be found in the supplementary information.

Climatic factors promote the migration of maize to the northeast
Regression results from the maize area model indicate that growing season temperature has the strongest negative effect on maize area, with each 1% increase in growing season temperature leading to a 2.533% decrease in maize area (table 1).Growing season precipitation has a slight positive effect on maize area.Expanding irrigated areas weakens the sensitivity of maize growth to precipitation (Zhou and Turvey 2014, Kang et al 2017).
Machinery and irrigation area are positively correlated with maize area.Agricultural machinery, as a substitute for farm labor, helps to improve crop productivity (Ma et al 2022).Irrigation compensates for rainfall deficiencies in Northern China.Urbanization has a negative effect on maize area (every 1% increase in urbanization decreases maize area by approximately 0.09%).
Figure 1 shows that the geographical centroid of maize area was found to have a distinct migration towards the northeast at the longest distance of 256 km between 2000 and 2019.The effects of climatic factors on the maize centroid's migration indicates a distinct migration towards the northeast.
Table 2 shows that how the temperature impacts on maize area depend on the treatment assigned by agricultural policies.Market-based policies affect farmers' responses to climate change to the greatest extent, increasing the negative effect of temperature on maize area by about 0.72%; while stockpiling program increases the negative effect of temperature on maize area by only 0.21%.This shows that current market-oriented agricultural policies may be effective in guiding spatial shifts in maize distribution to align with climate-driven changes.We added five types of sensitivity checks to ensure the robustness of the estimated climate effects on maize area, including controlling for maize revenue, production costs and lagged terms of precipitation, winsorizing data, controlling for province-by-year fixed effects, and considering spatial correlation in the error term (see SI 7).As shown in table S7, the estimated climate effects are consistent across specifications, indicating that the results are robust.
We used socio-economic data from farmer surveys, including fertilizers, seeds, pesticides and agricultural machinery power cost.The results indicate that the estimated coefficients are still robust (see supplementary table S8).

Climate-driven maize migration reduces exposure to extreme temperatures and strengthens the irrigation ratio
The negative effect of global warming on maize yields is moderated by spatial shifts in the maize production area.Figures 2(a) and (b) show the trends in crop-specific growing season temperatures from 2000 to 2019 (solid red lines).The growing season temperature experienced by maize is significantly lower than the counterfactual days.Extreme temperature (28 • C), a key factor damaging crop yields, is also significantly lower than in the counterfactual scenario.Figure 3 illustrates that average temperature increases significantly faster during the maize growing season in southern China than in the north.Therefore, the temperature exposure of maize decreases as maize production shifts from the south to the north.
Figure 2(c) illustrates the changes in water resources during the maize growing season.The precipitation trends experienced by maize acreage exhibits a decreasing trend (figure 2(c)) and is less than the counterfactual.Increased drought hazards associated with climate change and increased exposure of maize to droughts owing to production area expansion in the northern region primarily account for these changes.
Figure 2(d) shows that the ratio of irrigated maize expanded by 12% compared to the counterfactual scenario.This indicates that irrigation expansion is an important driver of rain-fed crop migration.In 1996, the Central Rural Work Conference of the Chinese government launched irrigation construction projects in 300 key counties, which contributed to a 7% increase in irrigation coverage (Wang et al 2024).

Climate-driven maize migration reduces production losses from climate change
Table 3 presents the sensitivity of maize yield to each climate variable using two econometric models based on maize area map weights.We found that if changes in the spatial distribution pattern of maize are not considered, the negative effect of growing season temperature would be overestimated, whereas the positive effect of the irrigation ratio would be underestimated.When considering crop migration, a 1% increase in temperature and expansion in irrigation result in a 0.398% decrease and a 0.11% increase in maize yield, respectively.In the absence of crop migration, a 1% rise in temperature causes a more severe 0.533% reduction in yield, while an increase in irrigation area by 1% elevates the yield by 0.7%.The no difference in the estimated elasticity of precipitation between the two models may be attributed to the compensation of irrigation for the decline in precipitation.
Figure 4 depicts the impact of climate change on maize production in China.At a national level, crop migration had a mitigation effect of 13.49 million tons and transformed an expected loss of 42.13 million tons with no adaptation to a loss of 28.64 million tons from 2000 to 2019.The avoided loss was equivalent to 15% of the China's total maize production in 2000 (NBSC 2022).

Discussion
Our results indicate that the maize production system becomes less vulnerable to climate change as the northward movement of maize area mitigates growing season temperature and strengthens the positive impact of irrigation.Furthermore, the geographical centroid of maize area was found to have a distinct migration towards the northeast at the longest distance of 256 km between 2000 and 2019.North China achieves a higher grain growth mainly due to a rapid increase in maize production.The development of animal husbandry, the impact of international soybean trade, and the drive of economic benefits have promoted the expansion of maize planting scale.
Second, our findings reveal that agricultural machinery has a significant positive impact on maize  2017).In 2016, the government revised the maize stockpiling policy into 'market-oriented purchase' and 'subsidy' , aiming to give play to the role of maize market price formation mechanism (Wu and Zhang 2016).These two policies create different incentives for farmers to adjust the cultivated area and spatial distribution of maize.Specifically, China's stockpiling program prevents domestic prices from soaring, potentially discouraging climate-driven expansion in maize acreage (Cui and Zhong 2024).The marketoriented policy allows Chinese farmers to independently make decisions about what to grow and how to manage agricultural production based on their knowledge of climatic conditions and market demand.Large-scale farmers in North China are more active in expanding their planting acreage in virtue of abundant arable land and favorable climate (Wang et al 2018).
Our study has several limitations.Firstly, our empirical models capture only current adaptation practices in response to within-year temperature shocks.Secondly, province-level data on yield cannot capture the heterogeneity in maize yield within a province.Our results reflect the role of maize migration in mitigating the adverse effects of climate change at the national level.However, we cannot identify the role of maize migration in mitigating the negative effects of climate change on maize production in China's major maize-producing provinces, such as Heilongjiang and Henan.Future studies need to incorporate finer-scale crop yield datasets and consider the role of maize migration within specific provinces.

Conclusion
Our findings indicate that rise in temperature has a significant negative effect on maize area and that precipitation has a significant positive effect.Socioeconomic factors dominating the migration of maize area include agricultural gross domestic product, power of farming machines, and fertilizer input.From 2000 to 2019, the migration of maize production area mitigated 13.49 million tons in climate change damage to maize production, equivalent to 15% of the total maize production in 2000.Current market-oriented agricultural policies may be effective in guiding spatial shifts in maize distribution to align with climate-driven changes, potentially decreasing the vulnerability of China's maize yield to climate change.

Figure 1 .
Figure 1.Predicted shift in maize area centroids in China, which have a ln(y)-ln(x) relationship.Note: Blue dot: actual migration of maize area; red dot: actual migration of maize area by climate change.

Figure 2 .
Figure 2. Trends in growing season temperature, warming bound (>28 • C), precipitation, and irrigation ratio over time.Red lines indicate observed temperature trends that are influenced by changes in crop area and climate, whereas green dashed lines represent a counterfactual scenario in which maize areas remain static at the 2000 distribution level.

Figure 3 .
Figure 3. Spatial differences in temperature trends from 1980 to 2019.Note: Trends are calculated using linear regression.

Table 1 .
Estimated maize area function.

Table 2 .
The impact of agricultural policies on temperature-maize area relationship.

Table 3 .
The sensitivity of maize yields to the climate variables.Dependent variable = ln (maize yield).Migration model uses a transient weight incorporating change in the maize's spatial distribution from 2000 to 2019; the counterfactual model uses a static maize grid map in 2000.* p < 0.05; * * p < 0.01; * * * p < 0.001. Notes: Wu Q and Zhang W 2016 Of maize and markets: China's new maize policy agricultural policy review CARD Agric.Policy Rev. 3 7-9 (available at: www.card.iastate.edu/ag_policy_review/article/?a=59) Zaveri E and Lobell D 2019 The role of irrigation in changing wheat yields and heat sensitivity in India Nat.Commun.10 4144 Zhang C, Dong J and Ge Q 2022 Mapping 20 years of irrigated croplands in China using MODIS and statistics and existing irrigation products Sci.Data 9 407 Zhang P, Zhang J and Chen M 2017 Economic impacts of climate change on agriculture: the importance of additional climatic variables other than temperature and precipitation J. Environ.Econ.Manage.83 8-31 Zhao J, Yang X, Liu Z, Lv S, Wang J and Dai S 2016 Variations in the potential climatic suitability distribution patterns and grain yields for spring maize in Northeast China under climate change Clim.Change 137 29-42 Zhou L and Turvey C G 2014 Climate change, adaptation and China's grain production China Econ.Rev. 28 72-89