Measuring and Enhancing Food Security Resilience in China Under Climate Change

Abstract

As global warming intensifies, extreme weather phenomena such as heatwaves, flash droughts, torrential floods, cold waves, and blizzards are becoming increasingly frequent. Against this backdrop, traditional static food security assessment methods fail to capture the dynamic transmission patterns of agricultural productivity risks and their regional heterogeneity. Therefore, it is imperative to reconstruct a resilience analysis paradigm for food production systems, dynamically investigate the mechanisms through which climate change affects China’s agricultural productivity and discern the interactive effects between technological evolution and climate constraints. This will provide theoretical foundations for building a climate-resilient food security system. Accordingly, this study establishes a multidimensional resilience measurement index system for China’s grain productivity by integrating agricultural factor elasticity analysis with disaster impact response modeling. Through production function decomposition and hybrid forecasting models, we reveal the evolutionary patterns of China’s grain productivity under climate risk shocks and trace the transmission pathways of risk fluctuations. Key findings indicate the following: (1) Extreme climate events exhibit significant negative correlations with grain production, with drought and flood impacts demonstrating pronounced regional heterogeneity. (2) A dynamic game relationship exists between agricultural technological progress and climate risk constraints, where the marginal contribution of resource efficiency improvements to productivity growth shows diminishing returns. (3) Climate-sensitive factors vary substantially across agricultural zones: Northeast China faces dominant cold damage, North China experiences drought stress, while South China contends with humid-heat disasters as primary regional risks. Consequently, strengthening foundational agricultural infrastructure and optimizing regionally differentiated risk mitigation strategies constitute critical pathways for enhancing food security resilience. (4) Future research should leverage higher-resolution, county-level data and incorporate a wider range of socio-economic variables to enhance granular understanding and predictive accuracy.

1. Introduction

Food security serves as a crucial foundation for national economic and social stability. Ensuring a stable food supply is intrinsically linked to the well-being of the people and the long-term stability of society, thereby granting it the highest priority in national governance [1,2]. In recent years, governments worldwide have increasingly focused on issues arising from climate change, as it directly impacts agricultural production. Global warming exerts a profound influence on China’s food security [3]. It leads to more frequent extreme weather events, such as droughts, floods, and heatwaves, which directly threaten crop yields and quality [4]. For instance, China’s per capita grain availability experienced periods of stagnant growth in the 1980s, highlighting the necessity of deeply understanding the adverse effects of climate change to better address its challenges [4]. Furthermore, climate change exacerbates the uneven distribution of agricultural resources, posing significant challenges to water availability, soil fertility, and the stability of agricultural ecosystems [5]. Given the multiple uncertainties introduced by global warming, in-depth research on the impact mechanisms of climate change on China’s food security and the prediction of future trends are of paramount importance [6].

Current research primarily focuses on the dynamic impact mechanisms of climate change on food production capacity and their regional variations. On the one hand, substantial empirical evidence reveals the significant negative impacts of extreme climate events on agricultural production. Studies by Chou Jieming et al. [1], Cui Ningbo et al. [7], and Hasegawa et al. [8] have all confirmed that events like droughts, floods, and abnormal temperatures significantly threaten the growth cycles and final yields of major food crops. Parry et al. [9] further pointed out that rising temperatures and precipitation instability are common drivers of agricultural yield reduction. On the other hand, research emphasizes the spatial heterogeneity of these impacts. Comparative analyses by Su Fang et al. [10] indicate that the sensitivity of rice-growing regions in southern China and wheat-producing areas in the north to climate change is significantly higher than that of the corn belt in northeastern China. Estimations of climate potential productivity based on cropping structures by Lu Yanyu et al. [11] highlight differences in regional climate carrying capacity. Huang Jin et al.’s [12] investigation into drought responses in China’s main grain-producing areas using the ensemble empirical mode decomposition method also provides a methodological basis for understanding regional differentiation. Therefore, identifying key sensitive factors and dominant risk types in different agro-ecological zones—such as low-temperature damage in the Northeast, drought stress in the North China Plain, and heat-humidity stress in South China—is a prerequisite for assessing the resilience of food security.

Furthermore, exploring strategies to enhance the climate adaptability of food systems has become another vital dimension of research. Notable advancements have been made in various areas: in forecasting models, Zou Cui et al. [13] successfully developed a provincial yield model that integrates policy variables, while Lu Wencong et al. [14] created a global equilibrium model for simulating market risks, and Chen Quanrun et al. [15] improved prediction accuracy by integrating meteorological events with expert insights. In the realm of resource coordination, Gao Jing et al. [16] demonstrated that optimizing energy-food-water systems can effectively strengthen regional resilience. In the area of supply chain loss reduction, Xie Xiaoliang et al. [17] designed a multi-level collaborative model to minimize losses, while Liu Xiuli et al. [18] emphasized the need to integrate dietary structural changes with circulation loss data to calibrate demand forecasts. Mereu et al. [19] employed crop models to simulate the responses of key crops under changing climatic conditions to provide a basis for adaptive variety selection and agronomic management optimization. However, existing research still lacks systematic assessments of resilience, predictions of dynamic risk transmission, and the proposal of precise improvement pathways. In particular, limitations in data availability—such as a shortage of long-term, high-quality extreme climate event data—coupled with biases in indicator construction, which manifest as an inadequate integration of multidimensional climate factors and their nonlinear effects, restrict a deep analysis of risk transmission mechanisms. Furthermore, limitations of traditional analytical methods have made it challenging to capture the complex interactions among climate, technology, and resources effectively. Abbass et al. [20], Wheeler et al. [21], and Liu Litao et al. [22] have all emphasized the urgency of investing in adaptation and mitigation measures and building “smart food systems.” Furthermore, Park Young-hee [23] has analyzed the pressure transmission mechanisms from a systemic perspective, while Stein Holden et al. [24] revealed the complexities of risk indirectly transmitted through markets. Therefore, there is an urgent need to integrate resilience analysis, disaster response modeling, and advanced forecasting technologies—such as hybrid models—to construct a multidimensional resilience measurement system that can dynamically reveal the laws of production capacity evolution and risk transmission pathways. This system would provide a solid theoretical foundation for formulating differentiated regional prevention and control strategies and optimizing technological evolution directions.

In summary, the impact of climate change on food production security is complex and far-reaching, involving multiple stages including production, processing, storage, transportation, and consumption [25]. To address this challenge, cooperative efforts and initiatives are necessary to enhance climate change monitoring, early warning systems, and adaptive measures, as well as to promote the sustainable transformation of food systems. Based on a systematic analysis of the intrinsic relationships between climate risks and food security, the core bottleneck facing current research lies in the insufficient integration of long-term reliable climate and agricultural data, which leads to regional risk assessment biases [26]. Traditional single indicators struggle to capture the cumulative effects of multiple climate stresses, and conventional models fail to elucidate the dynamic interactions among climate, technology, and markets. This paper innovatively proposes a “Production Function Analytical-Hybrid Prediction” coupling framework, which integrates a three-dimensional index system encompassing crop disaster sensitivity, technological adaptability, and resource bearing capacity to establish a climate resilience assessment standard applicable across regions. It further introduces a combination of Gaussian Process Regression and time-series forecasting models to accurately interpret the transmission pathways of extreme climate events from production to market. Ultimately, this framework culminates in region-specific customized solutions—including frost-resistant variety breeding in Northeast China, intelligent irrigation systems in North China, and heat disaster prevention technology packages in South China—resulting in a closed-loop decision-making process that transitions from precise risk diagnosis to effective prevention and control strategies. This framework provides an intelligent decision-making system that combines theoretical foundations with practical value, aimed at overcoming the climate adaptation bottleneck in major production areas.

2. Overview of Food Security Resilience Theory

In the context of the relationship between climate change and food security, the Cobb-Douglas (CD) production function model can be employed to quantify the impact of climatic factors on food output [27], facilitating an understanding of how climate change affects food security. As extreme weather events become more frequent due to climate change, both the stability and yield of food production are adversely impacted. Below are the steps and formulas for an improved CD production function model used to study the effects of climatic factors on food production:

Cobb-Douglas Model

$$Y=A·K^{\alpha }·L^{\beta }·C^{\gamma }$$

(1)

where: Y represents food output, A is a technology level coefficient that accounts for factors such as agricultural technology and planting techniques. K denotes capital input, including infrastructure such as machinery and irrigation systems. L signifies labor input, and C represents climate variable factors, such as temperature and disasters. The parameters $\alpha$, $\beta$ and $\gamma$ represent the output elasticities of capital, labor, and climatic factors, respectively, indicating the degree to which each factor affects food output. The primary steps for modeling are as follows:

Step 1: Define A, $\alpha$, $\beta$ and $\gamma$ as the parameters to be estimated. Utilizing historical climate data and crop yield statistics, regression analysis and other statistical methods can be employed to estimate these parameters, allowing for a more accurate assessment of the effects of climate change on food output.

Step 2: Calculate the marginal product of climate variables to evaluate the impact of climate change (such as increased temperature or irregular precipitation) on food output. The marginal product model for climate factors is given by:

$${\displaystyle \frac{\partial Y}{\partial C}}=\gamma ·A·K^{\alpha }·L^{\beta }·C^{\gamma -1}$$

(2)

where:

Equation (2) quantifies the marginal impact on food output due to changes in climate variables, aiding in the analysis of the potential risks to food security under different climatic scenarios.

Step 3: Assess the output elasticities ($\alpha$, $\beta$ and $\gamma$) to evaluate the relative importance of climatic factors in food production. The main criteria for judgment are as follows:

(1) If $\gamma$ is large, it indicates that food output is highly sensitive to climatic conditions, suggesting that climate change could lead to significant fluctuations in output.

(2) If $\alpha +\beta +\gamma =1$, it signifies constant returns to scale, meaning proportional changes in all inputs will result in proportional changes in output.

(3) If $\alpha +\beta +\gamma >1$, it indicates increasing returns to scale.

(4) If $\alpha +\beta +\gamma <1$, it represents decreasing returns to scale.

Step 4: Based on different climate scenarios (such as projected increases in temperature or decreases in precipitation), utilize the model to predict food output. Future climate change scenarios (such as data from IPCC climate models) can be input into the model to estimate the potential impacts of extreme weather or long-term climate change on food security.

Step 5: Analyze the relative impacts of climate change, capital, and labor on food output based on actual climate data and model results. This analysis provides data support for addressing the food security risks posed by climate change. Governments and policymakers can reference this model to take appropriate measures (such as improvements in agricultural technology and the formulation of climate adaptability policies) to mitigate the threats of climate change to food security.

3. Construction of Food Security Resilience Measurement Indicator System

3.1. Data Collection and Preprocessing

Data Sources:

(1) Select the most recent timely data from the years 2002 to 2022, utilizing sources such as the China Statistical Yearbook, China Environmental Statistical Yearbook, China Grain Yearbook, China Labor Statistical Yearbook, China Agricultural Products Price Survey Yearbook, China Rural Statistical Yearbook, China Agricultural Statistical Yearbook, and relevant data published by the Ministry of Agriculture and Rural Affairs of China [28]. References will also include reports such as the Statistical Bulletin of National Economic and Social Development, Civil Affairs Statistical Yearbook, China Grain Development Report, and Compilation of National Agricultural Cost and Benefit Data and Statistics on Agriculture in New China over 60 Years.

(2) Due to limitations in data availability, there are missing values for certain years. To ensure data integrity and the accuracy of analysis, linear interpolation has been employed to reasonably fill in the missing data for specific years. The data processing was conducted using Python software 3.9, applying linear interpolation to minimize the bias caused by missing values in subsequent analyses.

3.2. Selection of Indicator Variables

Based on a review of the literature and the specific context of China, nine primary indicators have been selected as the foundational framework for evaluating food security. The details are presented in

Table 1. Grain security evaluation index system.

Primary Indicator	Indicator Measurement	Positive/Negative
Typhoon Damage Level $D_1$	Area affected by typhoons (thousand hectares)	−
Frost Damage Level $D_2$	Area affected by frost (thousand hectares)	−
Drought damage Level $D_3$	Area affected by drought (thousand hectares)	−
Flood damage Level $D_4$	Area affected by floods (thousand hectares)	−
Degree of Agricultural Mechanization $D_5$	Total agricultural machinery power (10,000 kW)	+
Area of Grain Sown $D_6$	Total sown area for grain (thousand hectares)	+
Irrigation of Arable Land $D_7$	Irrigated arable land area (thousand hectares)	+
Fertilizer Application Intensity $D_8$	Fertilizer application amount (10,000 tons)	+
Temperature $D_9$	Temperature (°C)	−

:

3.3. Indicator Variable Contents

The following nine primary indicators have been selected to systematically assess the impact of climate change and social factors on food security [29]: (1) Typhoon Damage Level ${\mathit{D}}_1$ reflects the extent of destruction caused by extreme weather events such as typhoons on agricultural production, helping to assess the negative impact of typhoon disasters on grain yield, thereby indicating the risks that climate change poses to food security. (2) Frost Damage Level ${\mathit{D}}_2$ quantitatively measures the impact of low-temperature frost disasters on crops, especially winter crops. It is a key indicator for analyzing the stability of grain yields in relation to climate change. (3) Drought Damage Level ${\mathit{D}}_3$ indicates the extent to which drought affects agricultural production, quantifying the pressure of water scarcity on grain yield and assisting in the assessment of drought resistance in food production. (4) Flood Damage Level ${\mathit{D}}_4$ assesses the impact of floods and excessive precipitation on farmland and food production, particularly in monsoon regions, providing insights into the potential risks of extreme precipitation for grain supply. (5) Degree of Agricultural Mechanization ${\mathit{D}}_5$ reflects agricultural production efficiency, indicating the food production capacity per unit area of land and serving as a core indicator of agricultural technology and production efficiency. (6) Area of Grain Sown ${\mathit{D}}_6$ measures the level of agricultural mechanization, assessing the efficiency and modernization of agricultural production, which plays a significant role in improving grain yield and production stability. (7) Irrigation of Arable Land ${\mathit{D}}_7$ directly influences the fundamental factors of food supply. The size of the sown area reflects the scale of food production and the national support for food security. (8) Fertilizer Application Intensity ${\mathit{D}}_8$ evaluates the role of irrigation in agricultural production; increased irrigated area contributes to drought resistance and ensures the stability of food production. (9) Temperature ${\mathit{D}}_9$ reflects the input intensity of food production. Judicious use of fertilizers enhances yield, but excessive application may lead to land pollution risks, serving as an indicator for evaluating the sustainability of food production.

These indicators comprehensively measure the impact of climate change and socio-economic factors on food security from multiple angles, including natural disasters, production efficiency, resource utilization, and input factors.

3.4. Standardization of Indicator Variables

To facilitate comparability among the different indicators, various standardization methods are applied to normalize positive and negative indicators, eliminating the effects of dimensions and units. This enables the indicators to vary in the same direction, thereby better reflecting the state of food security. Generally, a higher value in positive indicators indicates a higher level of food security, whereas the opposite applies to negative indicators. Let $D_i$ denote the value of an indicator in year i, and the standardized value be denoted as $r_i$. Define:

$$\Delta ={\displaystyle \frac{maxD-minD}{minD}}$$

(3)

Then:

1. Normalization model for positive indicators:

$$r_i=\left\{\begin{array}{cc}{\displaystyle \frac{D_i-minD}{maxD-minD},}\hfill & \Delta \ge 10\hfill \\ {\displaystyle \frac{D_i}{maxD+minD},}\hfill & \Delta <10\hfill \end{array}\right.$$

(4)

2. Normalization model for negative indicators:

$$r_i=\left\{\begin{array}{cc}{\displaystyle \frac{maxD-D_i}{maxD-minD},}\hfill & \Delta \ge 10\hfill \\ {\displaystyle 1-\frac{D_i}{maxD+minD},}\hfill & \Delta <10\hfill \end{array}\right.$$

(5)

Equations (3) to (5) will be employed for the standardization of each indicator to facilitate comparison and analysis on a common scale.

3.5. CDC Model

Let the parameter reflecting natural disasters be C. The expression for the new model is given by:

$$Y_i=\prod _{i=1}^9{X}_i^{{\beta }_i}\mu$$

(6)

where $\mathit{Y}$ represents unit area food yield; ${\mathit{X}}_1$ through ${\mathit{X}}_9$ denote the areas affected by typhoons, frost, drought, floods; total agricultural machinery power; total sown area for grain; irrigated arable land area; amount of fertilizer applied; and temperature respectively, all as input factors. The parameters fi₁ through fi₉ denote the output elasticities of each factor, while $\mathit{C}$ epresents the effect parameter of changes in natural disasters, which is the selected input factor for natural disasters. fl indicates the output elasticity of natural disasters. When studying the impact of natural disasters on food yield, special emphasis is placed on analyzing the effects brought by the input factor $\mathit{C}$. By linearizing the model, the logarithmic linear form of the model is given as follows:

$$lnY={\beta }_0+\sum _{i=1}^5{\beta }_{i}ln{X}_i+\sum _{i=1}^5{\gamma }_{i}ln{C}_i+\mu$$

(7)

4. Analysis of Evidence on China’s Grain Production

To systematically analyze the interactive impacts of climate risks and production factors on grain output, this section constructs a multidimensional analytical framework based on the production function model. It incorporates panel data from the past two decades across the nation and three major grain-producing regions, examining the historical evolution, trend forecasting, and spatiotemporal disparities.

4.1. Historical Evolution Analysis of Grain Production

Figure 1 illustrates the actual changes in national grain production from 2002 to 2022, alongside the corresponding linear-log fitting model. Analyzing from a historical perspective, the unit area grain yield in China exhibited an overall upward trend during the period from 2002 to 2022, although it was characterized by significant fluctuations and phase changes [30]. Specific results are as follows:

In the period from 2002 to 2004, the transition from market-oriented agricultural policies to protective regulation marked a critical juncture. Following China’s accession to the World Trade Organization, low-priced international grains flooded the domestic market, impacting local planting income and creating instability in farmers’ production expectations. Concurrently, a continual reduction in arable land resources, compounded by extreme weather events such as the catastrophic floods in the Huai River Basin, exposed the vulnerabilities of the traditional agricultural model. Although the level of agricultural mechanization improved, issues such as aging irrigation infrastructure and weak disaster prevention systems limited production efficiency, resulting in a “V-shaped” volatility in unit area yields, initially declining then rising. The fluctuations in production during this period highlighted the fragility of the agricultural production system during the policy void, setting the stage for subsequent protective policies.

From 2004 to 2014, this period was marked as a golden growth phase driven by policy incentives. The institutional reforms, epitomized by the abolition of the agricultural tax, ushered in a new phase of policy-driven grain production. The “policy triangle” comprising fiscal subsidies, price support, and agricultural insurance effectively stimulated production enthusiasm: subsidies for agricultural machinery facilitated a leap in mechanization rates, minimum purchase price policies stabilized planting income expectations, and the promotion of superior seeds and irrigation infrastructure enhanced technological empowerment. These measures prompted a stair-step increase in unit area yields, with an average annual growth rate significantly surpassing that of the preceding period. However, the drawbacks of an overly input-dependent development model gradually became apparent—the excessive use of fertilizers and pesticides led to soil degradation, resulting in conflicts between grain stockpiling and ecological pressures, ultimately culminating in the conundrum of “increased production without increased income” around 2013, signaling diminishing marginal returns for this strategy.

The period from 2014 to 2020 marked a transition towards quality-driven transformation in supply-side reform. Faced with resource and environmental constraints, as well as structural contradictions in “three increases” (production area, yield, and input), policy orientation began to shift toward quality and efficiency. The implementation of the “dual reduction” initiative on fertilizers and pesticides directly restrained extensive production practices, while the promotion of high-standard farmland construction and water-saving irrigation technologies enhanced disaster resilience. Reforms in corn storage policies pressured adjustments to planting structures, with yield breakthroughs in crops such as soybeans partially offsetting the impacts of reduced acreage. During this phase, the growth rate of unit area yield noticeably slowed, yet the amplitude of fluctuations decreased, reflecting a transformation in policy adjustment from a singular focus on yield to a balance between ecological benefits. Supply-side structural reforms in agriculture aimed to eliminate ineffective supply and enhance the provision of green, high-quality products, extending food security towards sustainable development.

From 2020 to 2022, this period experienced dual impacts from the COVID-19 pandemic and frequent extreme weather events, resulting in a notable decline in the stability of grain production. The growth rate of grain output further decelerated and even exhibited slight decreases, indicating that after reaching a certain high level, further increases in grain production face heightened environmental and economic pressures, necessitating new policy guidance and technological support.

The fitted values for unit area grain production across the country were derived from various methods, including gradient descent, ridge regression, lasso regression, and elastic net regression, with the optimal fitting results selected. Transitioning from ridge regression to elastic net, the regularization strategies evolved from a single constraint to a multi-objective coordination approach. The fitting results are expressed as follows:

$$ln\left(y\right)=8.57+\sum _{i=1}^9{\beta }_{i}ln\left(x_i\right)$$

(8)

where the variables ${\mathit{X}}_1\sim {\mathit{X}}_9$ correspond to the area affected by typhoons, frost, drought, floods, total agricultural machinery power, total sown area for grain, irrigated arable land area, amount of fertilizer applied, and temperature, respectively. The parameter values are presented in

Table 2. Parameter estimates (${\beta }_1$–${\beta }_9$) of the linear-log model for national grain production.

${\mathit{\beta }}_1$	${\mathit{\beta }}_2$	${\mathit{\beta }}_3$	${\mathit{\beta }}_4$	${\mathit{\beta }}_5$	${\mathit{\beta }}_6$	${\mathit{\beta }}_7$	${\mathit{\beta }}_8$	${\mathit{\beta }}_9$
−0.016	−0.014	−0.013	−0.012	0.026	0.019	0.073	0.018	−0.013

:

In this context, ${\beta }_i$ reflects the influence of the corresponding variable ${\mathit{X}}_i$ on grain yield: a positive ${\beta }_i$ indicates a positive impact on grain yield, with larger values signifying a greater beneficial effect. Conversely, a negative ${\beta }_i$ denotes a negative impact on grain yield, where a larger absolute value implies a more substantial adverse effect. Furthermore, the ${\mathrm{R}}^2$ (coefficient of determination) is 0.974, indicating a very good fit for the model, while the $\mathrm{RMSE}$ (Root Mean Squared Error) is 60.9, suggesting that the predicted values deviate minimally from the actual values. The fitted model for the nation demonstrates that water conservancy construction is a crucial pillar for food security, with the contribution of irrigated arable land to yield significantly exceeding that of other factors. The fitted model for the nation demonstrates that water conservancy construction (which includes the development of infrastructure such as dams, reservoirs, and irrigation canal networks) is a crucial pillar for food security, with the contribution of irrigated arable land to yield significantly exceeding that of other factors. This underscores the strategic significance of the principle “to stabilize production, one must first manage water.” Regarding disaster impacts, the negative effects of typhoons and droughts are widespread; however, regional variations are masked by the overall dataset, indicating a need for localized disaster prevention and control policies. The potential threats posed by climate warming at the national level are less pronounced than those suggested by regional models, yet long-term vigilance is warranted against the trend of increasing extreme weather events. Additionally, the mere expansion of sown areas contributes limited benefits to average yield improvements, and may even lower the overall average yield, as expansion typically occurs onto less productive ‘marginal land’. Future efforts should prioritize technological innovation and the efficiency of resource utilization over reliance on mere scale expansion.

Figure 2 illustrates the actual changes in grain production in South China from 2002 to 2022, along with the corresponding linear-log fitting results. During this period, the unit area yield of grain crops in South China exhibited an initial decline followed by an increase, ultimately stabilizing within a high platform with fluctuations.

From 2002 to 2003, the unit area grain yield in South China saw a significant decrease, coinciding with a phase of policy adjustment and transformation in China’s agricultural sector. At that time, the marketization reform of the national grain circulation system was intensifying, leading to depressed grain prices and persistently low profitability for farmers engaged in grain production. This severely undermined farmers’ enthusiasm for grain cultivation, prompting many to switch to other cash crops or abandon their fields. Moreover, the implementation of ecological protection policies, such as “grain-for-green”, significantly reduced the area under grain cultivation, exacerbating structural imbalances in grain production. Concurrently, due to relatively underdeveloped agricultural mechanization and infrastructure, South China’s ability to mitigate and adapt to disasters was weak. After facing prolonged droughts, floods, and localized frost disasters, the stability of grain production drastically declined.

From 2004 to 2015, the unit area grain yield in South China entered a period of rapid growth, marked by a pivotal restructuring of the food security strategy at the top level, supported by a series of beneficial agricultural policies. The abolition of the agricultural tax and the establishment of direct subsidy systems redefined the distribution of interests between the government and farmers. The implementation of minimum purchase price policies stabilized market expectations and encouraged the reinvestment of production factors into grain production. Amid an upgrade wave in infrastructure, the networking of water conservation projects significantly enhanced disaster resilience, while the construction of high-standard farmland unlocked productivity in mid- to low-yield fields. Breakthroughs in the breeding of super rice led a technological revolution, and innovations in supportive cultivation models elevated the yield ceiling. During this stage, the rate of mechanization surpassed critical thresholds, while cooperative models effectively bridged the gap between smallholder farms and modern agriculture. Although disasters such as typhoons and floods occurred, the compensation mechanisms for disaster losses and emergency management systems became increasingly sophisticated, significantly improving resilience in capacity recovery.

In the 2016 to 2022 period, the unit area grain yield in South China experienced high volatility. The agricultural supply-side reform shifted the development logic from a focus on increased production to prioritizing quality and efficiency. The implementation of crop rotation and fallow systems triggered profound adjustments in planting structures, where reductions in double-cropped rice areas and expansions in specialty dryland crops created a yield offset effect. The reduction in fertilizer and pesticide usage compelled a transformation in production methods, and widespread adoption of biological control and precision fertilization technologies redefined pathways to increase yield. The aging labor force spurred innovations in socialized service provisions, with management models maintaining baseline yields through standardized operations. In the context of climate change, the frequency of extreme weather events surged, and phenomena such as El Ni?o led to cascading disaster reactions; however, smart warning systems and the promotion of resilient varieties effectively cushioned the fluctuations in yield. The essential nature of yield stabilization during this phase reflects a dynamic balance between hard constraints of resource and environmental factors and the soft support of technological substitutes. The fitted results using the logarithmic linear model are as follows:

$$ln\left(y\right)=8.56+\sum _{i=1}^9{\beta }_{i}ln\left(x_i\right)$$

(9)

The variables ${\mathit{x}}_1\sim {\mathit{x}}_9$ correspond to the following factors: the area affected by typhoons, the area affected by frost, the area affected by drought, the area affected by floods, total agricultural machinery power, total sown area for grain, irrigated arable land area, amount of fertilizer applied, and temperature. The estimated values of the parameters ${\beta }_1\sim {\beta }_9$ are presented in

Table 3. Parameter estimates (${\beta }_1$–${\beta }_9$) for South China region: impacts of floods and typhoons.

${\mathit{\beta }}_1$	${\mathit{\beta }}_2$	${\mathit{\beta }}_3$	${\mathit{\beta }}_4$	${\mathit{\beta }}_5$	${\mathit{\beta }}_6$	${\mathit{\beta }}_7$	${\mathit{\beta }}_8$	${\mathit{\beta }}_9$
−0.031	−0.011	−0.023	−0.046	0.028	0.017	0.012	0.012	−0.025

:

Additionally, the ${\mathrm{R}}^2$ (coefficient of determination) is 0.953, indicating a good fit for the model, $\mathrm{RMSE}$ (Root Mean Squared Error) is 58.9, suggesting that the predicted values deviate minimally from the actual values. The fitted model for South China reveals that agriculture in the region faces significant impacts from floods and typhoons. The destruction caused by floods during the rainy season is particularly pronounced, with inadequate drainage systems leading to widespread crop damage. Typhoon disasters cause persistent reductions in output through the direct destruction of crops and infrastructure. In terms of strategies to increase production, advancing mechanization is of paramount importance, particularly as labor shortages may necessitate improvements in production efficiency. In contrast, the marginal benefits of irrigation and fertilizer usage are relatively low, likely due to the region’s abundant rainfall and limited soil characteristics. Moving forward, it is essential to strengthen disaster warning systems and promote flood-resistant crops while optimizing solutions to address the conflict between mechanization and land fragmentation.

Figure 3 illustrates the actual changes in grain production in North China from 2002 to 2022, alongside the corresponding linear-log fitting results. An analysis using a historical perspective shows that the unit area grain yield in North China exhibited a generally upward trend characterized by fluctuations, with distinct variations across different periods. This change reflects the interactive influences of policy environments, agricultural inputs, technological development, and natural disasters. Between 2002 and 2005, grain production in North China was largely in a declining fluctuation phase. During this time, agricultural enthusiasm was low, infrastructure development was relatively underwhelming, mechanization was insufficient, and agricultural capacity to withstand natural disasters was limited. Additionally, the nation was undergoing market-oriented adjustments in the grain circulation system, leading to depressed grain prices. Policies such as “grain-for-forest” reduced the area planted in grains, severely discouraging farmer engagement in grain production and further exacerbating the downward trend in grain yields. During this period, extreme weather events, particularly severe droughts, significantly weakened the crop growth environment and contributed to further declines in yields.

Starting in 2004, the agricultural production environment in North China experienced substantial improvements, leading to the rapid recovery of grain yields, which entered a phase of fast growth. The government launched a series of robust agricultural support policies, including exemptions from agricultural taxes, grain production subsidies, minimum purchase price policies, and subsidies for agricultural machinery. These measures effectively enhanced farmers’ motivations to cultivate grain. Additionally, significant increases in agricultural inputs and the promotion of technology led to a notable rise in mechanization, with advancements in irrigation infrastructure greatly improving the capacity to withstand droughts and other natural disasters. Agricultural technological progress, centered on the promotion and application of high-quality, high-yield crop varieties, effectively drove steady increases in grain production and further safeguarded food security.

From 2013 to 2019, the growth rate of grain yields began to slow, entering a phase of fluctuating adjustment. Influenced by increasing pressures from resource and environmental constraints and the government’s policies aimed at structural adjustments in grain production, the planting structure in North China began to shift from a focus on maximizing grain output to emphasizing quality and sustainable development. Some regions implemented crop rotation and fallowing practices, resulting in changes to the structure and area planted in grains. At the same time, frequent natural disasters, particularly droughts, floods, and low-temperature frost events, exacerbated short-term yield fluctuations. However, thanks to advancements in irrigation infrastructure and agricultural technology promotion, overall grain production remained at a high level, demonstrating strong resilience to disasters.

Entering the period from 2020 to 2022, grain production stabilized at a high level. In the context of the COVID-19 pandemic and changing international circumstances, the government placed greater emphasis on food security, further enhancing the policy support system and strengthening technological backing. Agricultural production methods shifted towards greener and more efficient practices, with accelerated high-standard farmland construction enhancing the stability and sustainability of grain production. Although extreme weather occasionally affected yield stability, overall, through policy guidance, technological support, and optimized production structures, the grain production capacity in North China remained stable and gradually improved during this high plateau phase. The fitted results using the logarithmic linear model are as follows:

$$ln\left(y\right)=8.49+\sum _{i=1}^9{\beta }_i\ln \left(x_i\right)$$

(10)

In this model, the variables ${\mathit{x}}_1\sim {\mathit{x}}_9$ correspond to the following factors: the area affected by typhoons, the area affected by frost, the area affected by drought, the area affected by floods, total agricultural machinery power, total sown area for grain, irrigated arable land area, amount of fertilizer applied, and temperature. The estimated values of the parameters ${\beta }_1\sim {\beta }_9$ are shown in

Table 4. Parameter estimates (${\beta }_1$–${\beta }_9$) for North China region: drought and temperature effects.

${\mathit{\beta }}_1$	${\mathit{\beta }}_2$	${\mathit{\beta }}_3$	${\mathit{\beta }}_4$	${\mathit{\beta }}_5$	${\mathit{\beta }}_6$	${\mathit{\beta }}_7$	${\mathit{\beta }}_8$	${\mathit{\beta }}_9$
−0.011	−0.042	−0.047	−0.025	0.026	0.021	0.050	0.020	−0.041

:

Additionally, the ${\mathrm{R}}^2$ (coefficient of determination) is 0.965, indicating an excellent fit for the model, while the $\mathrm{RMSE}$ (Root Mean Squared Error) is 53.8, suggesting that the predicted values deviate minimally from the actual values. The fitted model for North China reveals that grain production in the region is primarily threatened by drought and extreme temperatures. Drought is the leading factor contributing to yield declines, as frequent water shortages severely restrict crop growth. Furthermore, winter frost and abnormally high temperatures exacerbate the risk of reduced yields, highlighting the dual pressure of climate change on agriculture in North China. In terms of positive drivers, expanding irrigated land is a core strategy for enhancing yields, with improvements in water conservation infrastructure effectively mitigating the impacts of drought. However, the benefits of increasing agricultural mechanization and fertilizer usage appear limited, indicating that agricultural production efficiency in North China is nearing a bottleneck given current technological conditions. Thus, there is a pressing need to shift towards the development of water-saving technologies and the breeding of resilient crop varieties.

As shown in Figure 4, between 2002 and 2007, the unit area grain yield in Northeast China exhibited a relatively stable growth trend. However, around 2005, a certain degree of fluctuation began to appear, with actual yields slightly declining compared to fitted values, reflecting the inability to maintain steady growth in grain production. During this period, natural disasters such as typhoons and frost significantly impacted grain yields, leading to declines in some years. For instance, in 2004 and 2005, the occurrence of extreme weather events, such as frost, resulted in decreased grain production.

Starting in 2007, grain yields began to show a rising trend, reflecting gradual improvements in agricultural technology and infrastructure in Northeast China, particularly in investments and advancements in irrigation technology and mechanical power. From 2008 to 2013, the unit area grain yield in Northeast China experienced a dramatic increase, with actual yields significantly surpassing fitted values. This sharp growth during this period can be attributed to several key factors. Firstly, policy support played a crucial role: after 2008, the government implemented a series of agricultural support policies, including subsidies for grain production and the promotion of mechanization. The implementation of these policies significantly enhanced agricultural productivity in Northeast China, leading to substantial increases in unit area yields. Secondly, agricultural technological innovations, including the widespread application of modern agricultural techniques—particularly improvements in fertilizer use, mechanization, and irrigation systems—contributed to marked increases in grain production efficiency. Thirdly, favorable climatic conditions during this period, with minimal occurrences of extreme weather events such as frost and flooding, provided a conducive environment for grain production. Therefore, the significant rise in grain yields during this time resulted from a positive interaction among policies, technology, and climatic conditions.

However, after 2014, the growth rate of grain yields in Northeast China noticeably slowed. Although production levels remained high, the pace of increase significantly diminished. This slowdown can be attributed to the constraints imposed by land resources, as agricultural land in the region approached saturation, limiting further enhancements in productivity. As land utilization efficiency gradually approached its limit, the growth of grain yields stabilized. Additionally, during this period, the climate experienced considerable fluctuations, including frequent droughts and flooding, which directly affected agricultural productivity. In some years, yield declines were observed due to drought impacts, contributing to a slowdown in growth.

From 2018 to 2022, the unit area grain yield in Northeast China exhibited a relatively stable growth trend, albeit with some fluctuations. This period was characterized by sustained support for agricultural technology and policies; however, the driving forces behind yield growth were noticeably weakened due to climate change and limitations in land resource utilization efficiency. Despite ongoing investments in irrigation technology, fertilizer application, and mechanization helping to maintain growth, the incremental improvements became smaller, making it more challenging to achieve the substantial increases in yields seen during 2008–2013. Additionally, the exacerbation of climatic fluctuations became increasingly significant, with Northeast China experiencing more extreme weather phenomena such as drought and frost, which led to yield declines in certain years—particularly in 2019 and 2020, when drought conditions resulted in reduced unit area yields.The fitted results using the logarithmic linear model are as follows:

$$ln\left(y\right)=8.12+\sum _{i=1}^9{\beta }_{i}ln\left(x_i\right)$$

(11)

In this model, the variables ${\mathit{x}}_1\sim {\mathit{x}}_9$ correspond to the following factors: the area affected by typhoons, the area affected by frost, the area affected by drought, the area affected by floods, total agricultural machinery power, total sown area for grain, irrigated arable land area, amount of fertilizer applied, and temperature. The estimated values of the parameters ${\beta }_1\sim {\beta }_9$ are shown in

Table 5. Parameter estimates (${\beta }_1$–${\beta }_9$) for Northeast China region: frost and drought constraints.

${\mathit{\beta }}_1$	${\mathit{\beta }}_2$	${\mathit{\beta }}_3$	${\mathit{\beta }}_4$	${\mathit{\beta }}_5$	${\mathit{\beta }}_6$	${\mathit{\beta }}_7$	${\mathit{\beta }}_8$	${\mathit{\beta }}_9$
−0.002	−0.072	−0.021	−0.008	0.011	0.034	0.025	0.012	−0.006

:

Additionally, the coefficient of determination ${\mathrm{R}}^2$ is 0.9897, indicating an excellent fit for the model, $\mathrm{RMSE}$ (Root Mean Squared Error) is 54.0, signifying that the predicted values deviate minimally from the actual values. The fitted model for Northeast China reveals that grain production in the region is primarily threatened by frost and drought. Frost is a critical factor contributing to yield declines, particularly spring frost, which adversely affects early crop growth and significantly slows down the growth rate, thereby impacting overall production. Drought is also an important constraint on yield growth, with frequent drought conditions leading to insufficient water supply, which negatively impacts the normal growth and development of crops, especially in years with unstable precipitation. Relatively, the adverse effects of rising temperatures on crop growth are also becoming more pronounced. During high-temperature seasons, the grain filling period of crops is compressed, resulting in reduced yields. On the positive side, the expansion of irrigated land plays a crucial role, as improved water resource allocation effectively alleviates the impacts of drought and ensures the availability of water necessary for crop growth. However, enhancements in agricultural mechanization and fertilizer usage have had a limited effect on boosting yields, suggesting that agricultural production in Northeast China is nearing a bottleneck given the current technological constraints. Moving forward, there is a pressing need to strengthen the development of water-saving irrigation technologies and drought-resistant, cold-tolerant crop varieties to address the challenges posed by increasingly complex climate changes and pressures on land resources.

4.2. Grain Production Trend Analysis

This study employs a hybrid GPR + ARIMA model to forecast the trends in grain production. The core modeling process follows a logical chain of “data reconstruction—dual model synergy—dynamic integration.” Initially, historical yield data undergoes standardization and stationarity correction, with differencing operations employed to eliminate non-stationarity disturbances. Subsequently, a Gaussian Process Regression (GPR) model is constructed, utilizing radial basis functions to capture the nonlinear trends and uncertainty characteristics of yield changes. Meanwhile, the model’s residual sequence is used to establish an ARIMA model, compensating for linear cyclical fluctuations. Finally, based on the principle of minimizing validation set errors, dynamic weights are allocated to integrate both models, resulting in a hybrid forecasting system that combines nonlinear fitting capabilities with periodic correction functions. This method leverages the probabilistic modeling advantages of GPR to analyze the long-term yield potential driven by technological advancements, while incorporating the linear correction mechanisms of ARIMA to quantify the damping effects of external shocks like climate fluctuations. Consequently, the predicted results exhibit a typical pattern of “rapid initial increase—subsequent stabilization,” aligning with the realities of diminishing marginal returns from agricultural technological advancements and the increasing risks posed by climate change [31].

The forecast results are presented in Figure 5, illustrating a “preceding increase followed by stabilization” evolution in grain production trends. This phenomenon is fundamentally a result of the dynamic balance between the driving force of technological advancement and the resistance posed by climatic conditions within the agricultural production system. In the early phase (2023–2026), the sustained increase in grain production is primarily driven by systemic breakthroughs in agricultural technological innovation. Improvements in genetic resources enhance crops’ photosynthetic efficiency and resilience, significantly extending the release period of yield growth’s technological dividends. The widespread adoption of precision agriculture optimizes the allocation efficiency of water and fertilizer resources, allowing for deep exploration of production potential in resource-constrained areas. Meanwhile, the iterative upgrades in mechanization and smart technology reduce field management losses, creating a synergistic effect that enhances overall production efficiency. These technological advancements, driven by research and development investments, technology diffusion, and the deep integration of production practices, propel unit area yields along an exponential growth trajectory, evidenced by the steep upward curve in historical stages.

However, as time progresses, the multifaceted impacts of climate change gradually emerge as dominant constraints. The frequency of extreme weather events leads to decreased spatiotemporal matching of light, heat, and water resources, exposing crops to heat stress and significantly increasing the risk of drought or waterlogging. The overlap of pest and disease generations and the evolution of resistance due to climate warming weaken the effectiveness of existing pest control technologies. Changes in precipitation patterns and the intensification of extreme precipitation events exacerbate soil erosion and nutrient loss, raising the ecological costs of maintaining stable yields. Such climate risks not only result in direct production fluctuations, but also alter the foundational environment for technology application, accelerating the diminishing marginal returns on agricultural technological innovations. For instance, the adaptive thresholds of drought-resistant varieties are frequently breached under sustained drought pressure, while the regulatory capacity of intelligent irrigation systems approaches saturation following extreme precipitation shocks.

At a deeper level, the interplay between technological advancement and climate constraints exhibits asymmetric characteristics. The yield-enhancing effects of technological innovation are cumulative and progressive, requiring long-term research investments to achieve breakthroughs. In contrast, the suppressive effects of climate deterioration tend to be abrupt and have exponential amplification characteristics, with negative impacts often surpassing the buffering thresholds of technology. This dynamic equilibrium leads to a flattening of the yield curve in the later stages, reflecting a shift in the agricultural production system from “technology-driven growth” to a “risk-constrained steady state.” The key to breaking this trend lies in developing a climate-smart technology system that enhances cultivar resilience, establishes disaster-resilient management systems, innovates eco-intensive production models, and reshapes a new synergistic relationship between technological advancement and climate adaptation.

4.3. Analysis of the Spatiotemporal Evolution of Grain Production

Grain production is a crucial indicator of agricultural productivity and a key factor in ensuring national food security. From 2002 to 2024, influenced by multiple factors such as agricultural technological advancements, climate change, and policy adjustments, the spatiotemporal evolution of grain production across various provinces in China (excluding Hong Kong, Macau, and Taiwan) has demonstrated significant regional disparities [32], as illustrated in Figure 6.

The changes in grain production across different provinces from 2002 to 2024 exhibit significant disparities. Notably, provinces such as Inner Mongolia, Hebei, Ningxia, and Gansu have demonstrated remarkable increases in grain yields, primarily attributable to agricultural technological innovations and supportive policies. In Inner Mongolia, the grain yield per unit area rose from 3240 kg in 2002 to 5620 kg in 2024, reflecting an impressive increase of 73.3%. This growth is primarily driven by the introduction of modern agricultural facilities and advancements in irrigation technology. Similarly, Hebei’s grain yield per hectare increased from 43,780 kg to 6980 kg, a growth of 59.6%, closely linked to sustained investments in agricultural technology, mechanization, and water resource management. Additionally, Ningxia and Gansu reported growth rates of 58.4% and 57.3%, respectively, benefiting from the development of water-saving irrigation and agricultural mechanization. In contrast, traditional grain-producing regions, such as Northeast China, have experienced a gradual slowdown in growth, revealing diminished potential for yield increases under the dual pressure of increasingly constrained land resources and declining marginal returns on technological advancements.

With changes in grain production, the spatial pattern of grain production is also undergoing significant transformation. The growth rates in traditional grain-producing areas, such as Northeast and North China, are slowing down annually, while emerging high-yield areas, including the Northwest region and some southern provinces, are experiencing accelerated grain production growth. For instance, Heilongjiang continues to see an increase in grain yield, but the overall increase from 2002 to 2024 is only 45.2%, indicating limited growth potential and the need for more effective methods to enhance yield rates. In contrast, Hebei and Tianjin have exhibited substantial increases in yields, with Hebei’s total increase reaching 59.6%, demonstrating rapid growth in grain production in North China driven by modern agricultural technologies and policies. The Northwest region, particularly Inner Mongolia, Ningxia, and Gansu, has consistently increased grain yields thanks to superior irrigation technology and improved agricultural infrastructure, gradually replacing parts of Northeast and North China as new high-yield areas. The grain yield increases in the Northwest region are significant, showing greater stability in growth rates during certain years.

Over the past two decades, the distribution of high-yield grain production areas in China has been undergoing a notable shift. While Northeast and North China maintain high grain yield levels, their growth rates have exhibited a consistent decline over the past twenty years, with some years even experiencing stagnation in yield growth. In contrast, the Northwest region, particularly Inner Mongolia, Ningxia, and Gansu, has emerged as a new high-yield area due to ongoing investments in modern agricultural technology, water-saving irrigation, and mechanization. The yield increases in the Northwest region far exceed those of traditional grain-producing areas, suggesting its growing significance in future food production. Especially under the influence of climate change, the arid regions in the Northwest are expected to further enhance their grain production capabilities through technological innovation and effective water resource utilization, thereby altering the national spatial pattern of grain production.

5. Policy Recommendations

First, strengthen the collaborative construction of an agricultural infrastructure system across regions. Based on the research findings, the types of natural disasters and infrastructural deficiencies in the three major production areas exhibit significant spatial differentiation. Therefore, region-specific enhancement projects should be implemented to address infrastructural shortcomings in the context of climate change: Northeastern regions should focus on improving frost protection facilities and soil fertility conservation systems; North China Plains should deepen the integration of water-saving irrigation and dryland farming technologies; and coastal South China should construct a comprehensive typhoon-flood prevention network. Concurrently, advancing the construction of an intelligent monitoring and early-warning platform for agricultural meteorological disasters is essential to achieve dynamic perception of disaster conditions and coordinated emergency responses [33].

Second, deepen differentiated risk prevention and control strategies based on regional characteristics. Tailored governance plans should be developed according to the spatial differentiation of dominant disasters: Northeast China should concentrate on breeding cold-resistant varieties and creating low-temperature warning systems; North China should innovate water-saving planting systems and establish ecological compensation mechanisms; and South China should promote the development of flood-tolerant crops and composite disaster-resistant models. A cross-regional joint prevention and control mechanism should be established to enhance system resilience through shared risk mapping and coordinated resource allocation.

Third, break through resource constraints by fostering technological collaborative innovation. Based on the diminishing returns revealed in the production functions, a triadic innovation system combining “variety-equipment-management” should be constructed: accelerate the improvement of photosynthetic efficiency and the development of resistance gene editing technologies, and develop data-driven precise operation systems. An integrated intelligent monitoring network encompassing aerial, terrestrial, and satellite technologies should be established. Additionally, promoting the digital transformation of agricultural knowledge services is imperative to facilitate the precise implementation of climate-adaptive technologies.

Fourth, improve the institutional framework for enhancing resilience. The research findings suggest that climate risk transmission exhibits cross-regional chain reactions. Establishing an agricultural climate risk special fund linked with a catastrophe insurance mechanism can integrate climate resilience indicators into the ecological compensation assessment system [34]. Relevant technical standards in land spatial planning should be revised to reinforce binding indicators for food security resilience, thus forming a synergistic governance structure characterized by “legal constraints, policy incentives, and market-driven approaches.”

Fifth, construct a national resilience system for food security. Given the diminishing marginal benefits of technological advancement and the continuous intensification of climate impacts, a systematic response is required. By implementing the strategic directives on food security from the 20th National Congress of the Communist Party and through infrastructure upgrades, targeted regional prevention, breakthroughs in technological innovation, and optimized institutional collaboration, the climate adaptability of the agricultural production system can be systematically enhanced. This approach would establish a food security resilience system characterized by a “solid foundation, controllable risks, technology empowerment, and robust institutional frameworks.”

6. Conclusions and Future Prospects

By integrating production function decomposition with hybrid forecasting models, this study innovatively constructs a multidimensional measurement indicator system for food security resilience. It dynamically reveals the evolutionary patterns and risk transmission pathways of China’s food production capacity under climate change. Methodologically, the proposed ‘production function analysis–hybrid prediction’ coupling framework offers a new paradigm for regional resilience assessment. This framework emphasizes the dynamic interplay between technological progress and climate constraints, thereby enriching the conceptual understanding of food security research.

However, this study is subject to certain limitations. First, constraints in data availability, such as the shortage of long-term, high-quality data on extreme climate events, may have led to biases in the indicator construction; although interpolation methods were employed to address gaps, this could introduce uncertainties. Second, while the model captured the impacts of major climate factors, it did not fully account for non-linear effects or all pertinent socio-economic variables.

Future research could expand to include a wider array of climate scenarios and socio-economic indicators, apply machine learning methods to enhance predictive accuracy, and extend this framework to other developing countries for comparative analysis to test its generalizability. Overall, the findings of this study can serve as a reference for food security policies in regions globally facing similar climate risks, though the indicators and model parameters must be carefully adjusted to local conditions.