The Impact of Surface Ozone and Particulate Matter 2.5 on Rice Yield in China: An Econometric Approach

Abstract

Based on real county-level surface ozone and PM2.5 data, an econometric model was constructed to comprehensively consider its impacts on rice yield in China. At the same time, other economic, climate, and regional variables were included in the econometric model to accurately measure the interaction between surface ozone and PM2.5 on rice yield in China. The results showed that an increase in ozone concentration in different periods and an increase in PM2.5 in the growth period would reduce the rice yield. The interaction coefficient between ozone and PM2.5 suggests that PM2.5 partially mitigated the amount of ozone absorbed by rice plants, which in turn had a positive effect on rice yield. Therefore, while controlling PM2.5 and reducing near-surface ozone concentration, it is necessary to comprehensively consider its impact on rice yield.

1. Introduction

Academia and policymakers have highly valued food security in recent years, with crop yield being a central concern. Understanding the factors that influence crop yield is thus a critical research. Several scholars have analyzed the important factors that affect crop yield from different perspectives. Among these factors, environmental contamination plays a pivotal role. Studies have demonstrated that air pollution, particularly PM2.5 and surface ozone, poses substantial risks to both human health and agricultural productivity [1,2,3,4].

Surface ozone is widely recognized as a significant factor in influencing crop yields [5,6,7,8]. Elevated ozone levels can impair plant health by reducing chlorophyll content, damaging leaf membranes, and diminishing photosynthetic efficiency [9,10,11,12]. For instance, Tai et al. [13] hypothesized that ozone pollution is expected to have a more severe impact on global food security compared to climate change. Feng et al. [14] quantified the costs of ozone-induced losses in rice (USD 7.5 billion). Tang et al. [15] reported that the impact of ozone on wheat yield loss in China and India reached 6.4–14.9% and 8.2–22.3%, respectively. Nevertheless, the need for rice output is expected to rise in the coming decades, driven by population growth, especially in key rice-consuming regions such as Asia, Africa, and Latin America [16]. Elevated levels of ground-level ozone may result in significant declines in rice production, potentially impacting the global rice market [17]. Moreover, surface ozone levels in Asia are gradually rising relative to those in North America and Europe [18]. Therefore, understanding the impact of rising ozone levels on rice production is crucial for ensuring global food security [3].

In addition to ozone, PM2.5 is another major air pollutant in China, linked to severe public health issues and economic burdens [19]. In response, the Chinese government introduced the Technical Policy for the Comprehensive Prevention and Control of Fine Particulate Matter Pollution in Ambient Air in 2013. This policy aims to improve the ecological environment and quality of the environment and ensure human health and ecological security. Since 2013, the Chinese government has undertaken significant initiatives aimed at cutting PM2.5 levels by 25% in key urban regions by 2017 [20]. However, the reduction in PM2.5 has inadvertently contributed to increased ozone concentrations, as lower PM2.5 levels enhance solar radiation and reduce the loss of HO2 free radicals, both of which promote ozone formation [21]. Consequently, China now faces a dual challenge of high PM2.5 and ozone pollution, creating a complex regional air pollution scenario [22].

Despite the growing recognition of the individual impacts of PM2.5 and ozone on crop yields, few studies have comprehensively examined their interactions. Photosynthesis plays a crucial role in determining crop productivity, which is largely regulated by the stomata on plant leaves. For instance, PM2.5 can alter the environmental conditions necessary for crop development by absorbing and dispersing sunlight [23], to reduce photosynthesis and have a negative impact on crop growth. However, crops absorb ozone during photosynthesis, which destroys photosynthesis, reduces its efficiency, and reduces crop yield. It is noteworthy that the blocking effect of PM2.5 on stomata also reduces the amount of ozone absorbed during photosynthesis, that is, PM2.5 has a certain protective effect on crop yield. If we only consider the impact of changes in ozone concentration or PM2.5 on rice yield, this may lead to estimation deviation. We, therefore, need to consider the interaction between ozone and PM2.5 on rice yield.

Few studies have been conducted to assess the impact of pollution on agricultural productivity in China, with the majority relying on experimental or simulation-based approaches [17]. Dose–response functions estimated from experimental data are frequently used to predict the physical damage to crops from various pollutants [24]. Experimental studies have demonstrated that ozone exposure negatively impacts crop productivity [25]. While these methods provide valuable insights, they often lack the flexibility to capture real-world agricultural practices and farmer adaptations. In contrast, econometric approaches, which integrate atmospheric, socioeconomic, and biophysical data, offer a more comprehensive understanding of the impact of pollution on crop yields. For example, Yi et al. [26] employed an econometric framework to analyze the effects of ground-level ozone on winter wheat yields in China, providing valuable insights into farmer responses to pollution-related challenges.

Building on the methodology of Yi et al. [26], this study adopts an econometric framework to evaluate the combined impact of PM2.5 and ozone on rice production in China, utilizing field observations and integrating atmospheric, socioeconomic, and biophysical data. Specifically, we explore the detrimental effects of PM2.5 and ozone on rice yields, as well as the potential mitigating effect of their interaction. By addressing this research gap, our study aims to provide a more nuanced understanding of the complex relationship between air pollution and agricultural productivity, offering critical insights for policymakers and stakeholders working to ensure food security in the face of rising pollution levels.

2. Materials and Methods

2.1. Data and Variables

The data in this study primarily include surface ozone data, PM2.5, weather data, and socioeconomic variable data, and each data point is introduced in detail.

2.1.1. Ozone

Previous research has often relied on simulated surface ozone data [27,28]. In contrast, this study collected ground-level ozone measurements from 1412 monitoring stations operated by the China National Environment Monitoring Center from 2013 to 2015. The inverse distance weighted (IDW) technique was applied to interpolate the station-level data into 0.10412° × 0.10412° grids [29]. We used four common methods to measure the surface concentration:

$$\begin{array}{c}M7={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n{C}_{O_3}^i}, (C_{O_3}^i \mathrm{is} \mathrm{measured} \mathrm{from} \mathrm{9:00} \mathrm{to} \mathrm{16:00})\hfill \\ M12={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n{C}_{O_3}^i}, (C_{O_3}^i \mathrm{is} \mathrm{measured} \mathrm{from} \mathrm{8:00} \mathrm{to} \mathrm{20:00})\hfill \\ Sum06={\displaystyle \sum _{i=1}^n(C_{O_3}^i-60)}, (\mathrm{when} C_{O_3}^i\ge 60 ppb)\hfill \\ Aot40={\displaystyle \sum _{i=1}^n(C_{O_3}^i-40)}, (\mathrm{when} C_{O_3}^i>40 ppb, \mathrm{from} \mathrm{9:00} \mathrm{to} \mathrm{16:00})\hfill \end{array}$$

where $C_{O_3}^i$ represents the surface ozone concentration at hour i and n is the number of hours in the time we want to measure. M7 represents the mean daily ground-level ozone concentration measured between 9:00 and 16:00, while M12 denotes the average daily ozone level recorded from 8:00 to 20:00. Sum06 indicates the total ozone accumulation when hourly levels surpass 60 ppb. Similarly, Aot40 measures the cumulative ozone concentration during daylight hours (9:00–16:00) when hourly readings exceed 40 ppb and global radiation is above 50 Wm⁻². These five indicators were comprehensively used to verify the robustness of the conclusions, and M7 was used as the benchmark index.

Figure 1 and Figure 2 illustrate the spatial distribution of ozone concentrations across the sampled counties, as measured by M7 during the ozone-sensitive phase of rice growth in 2013 and 2015, respectively. Figure 3 depicts the variations in ozone concentration distribution, also measured by M7, during the same sensitive period from 2013 to 2015.

In general, areas with elevated ozone levels in 2013 were predominantly located in the eastern part of the country, with the southern and central regions also showing significant concentrations. Regions with high ozone concentrations in 2015 were not in the eastern or central regions. It is also clear that ozone concentrations in the eastern and central regions decreased. This feature is further illustrated in Figure 3.

2.1.2. PM2.5

Figure 4 and Figure 5 present the spatial distribution of PM2.5 levels, as measured during the ozone-sensitive phase of rice growth (defined as the critical phenological stage from tillering to flowering) in 2013 and 2015, respectively. This period, typically occurring between June and August in major rice-growing regions, was chosen due to its heightened vulnerability to ozone exposure, as prior studies have demonstrated that ozone exerts its most pronounced detrimental effects on rice yield during these stages. Figure 6 illustrates the variations in PM2.5 concentrations over the same sensitive period from 2013 to 2015.

According to Figure 4 and Figure 5, the characteristics and changes in PM2.5 in 2013 and 2015 cannot be observed intuitively. However, from Figure 6, it is observed that compared with 2013, the PM2.5 value in most regions decreased in 2015, which may be attributed to the enforcement of China’s Technical Policy for the Comprehensive Prevention and Control of Fine Particulate Matter Pollution in Ambient Air. At the same time, PM2.5 increased in several regions in 2015 compared to 2013. Figure 6 illustrates this feature more clearly.

2.1.3. Weather

In line with the methodologies of Carter et al. [17] and Yi et al. [29], meteorological data, encompassing daily minimum, maximum, and average temperatures, along with precipitation records from 825 weather stations across the country, were sourced from the China Meteorological Data Service Center (CMDC). Due to discrepancies in the spatial distribution of weather stations relative to county boundaries, spatial interpolation techniques were employed to generate county-level weather data based on observed climate information. Initially, the inverse distance weighted (IDW) approach was utilized to interpolate climate data from the 825 stations into a spatial grid with a resolution of 500 m. The average values of each grid within a county were calculated to represent the climatic conditions of that area. Both temperature and precipitation were included in the statistical analysis. Growth degree days (GDDs), defined as the cumulative temperature above the threshold required for plant growth [30], were also incorporated.

2.1.4. Socioeconomic Variables

The socioeconomic variables encompassed metrics such as the area under rice cultivation, production output, market prices, total production, and the price index for agricultural inputs. Data on county-level rice yields and cultivated areas for 2014 and 2015 were sourced from the database maintained by the Institute of Agricultural Information at the Chinese Academy of Agricultural Sciences. This database covers 19 provinces (or municipalities), representing 91.42% of the country’s total rice production in 2015. These provinces (municipalities) are Heilongjiang, Jilin, Liaoning, Henan, Inner Mongolia, Ningxia, Gansu, Shandong, Jiangsu, Anhui, Shanghai, Zhejiang, Jiangxi, Hubei, Hunan, Sichuan, Guangdong, Guangxi, and Yunnan. Of these provinces (municipalities), 822 were used as samples. Given that farmers typically base their planting decisions on the previous year’s output prices, the lagged rice price was employed as a proxy for anticipated market prices [29].

Table 1. Summary statistics.

Variable	Unit	Mean	Std.	Min	Max
Yield	ton/ha	6.6302	2.5564	0.3481	86.6878
M7 (ozone-sensitive period)	ppb	40.2516	7.5385	19.7574	68.3679
M12 (ozone-sensitive period)	ppb	38.8049	7.1419	19.3875	66.8271
Sum06 (ozone-sensitive period)	ppm	4899.13	3982.41	30.0151	23,930.9
Aot40 (ozone-sensitive period)	ppm	2134.98	1290.60	6.0840	7773.01
M7 (March–May)	ppb	39.2897	6.6168	20.6642	63.2793
M12 (March–May)	ppb	37.8921	6.2471	19.7537	60.4312
Sum06 (March–May)	ppm	13,149.4	9692.27	30.6432	53,688.1
Aot40 (March–May)	ppm	5681.89	3172.66	49.2145	19,022.9
PM2.5	μg/m3	56.8455	40.9722	5.0102	443.149
PM2.5 (growth period)	μg/m3	57.2214	40.5122	5.4924	431.759
PM2.5 (M7-sensitive period)	μg/m3	57.2213	40.4816	6.1877	427.590
PM2.5 (M12-sensitive period)	μg/m3	57.2254	40.4872	5.8104	424.778
PM2.5 (Sum06-sensitive period)	μg/m3	57.2271	40.4986	5.8058	426.015
PM2.5 (Aot40-sensitive period)	μg/m3	57.2213	40.4816	6.1877	427.590
PM2.5 (M7-March–May)	μg/m3	57.2087	40.4705	6.1372	426.282
PM2.5 (M12-March–May)	μg/m3	57.2142	40.4807	5.8166	426.540
PM2.5 (Sum06-March–May)	μg/m3	57.2150	40.4801	5.9627	427.405
PM2.5 (Aot40-March–May)	μg/m3	57.2087	40.4705	6.1372	426.282
Rice sown area	10,000 ha	2.4319	4.2877	0.0001	93.537
Ratio of the means of production price index and rice price index lag	—	0.9948	0.0257	0.9270	1.0542
Growth degree days (8–32 °C)	1000 °C	1.9897	0.3943	0.4389	2.8335
Precipitation	m	0.6359	0.2076	0.1336	1.5963
Number of observations	2466

provides an overview of the descriptive statistics for the variables included in the study.

2.2. Econometric Modeling

A double fixed-effects model was employed to assess the influence of ground-level ozone pollution on rice yields, while controlling for PM2.5-related variables. The core assumption of the fixed-effects model is that unobserved individual heterogeneity is correlated with the explanatory variables but does not change over time. By controlling for individual fixed effects, the model can eliminate the potential influence of time-invariant unobserved variables (such as soil quality, climatic conditions, etc.) on the outcomes, thereby providing a more accurate estimation of the causal effects of ozone and PM2.5 on crop yields. Additionally, the fixed-effects model assumes no correlation between the explanatory variables and the error term, i.e., it satisfies the strict exogeneity assumption. To validate the model’s applicability, we conducted a Hausman test, the results of which indicated that the fixed-effects model is superior to the random-effects model (p < 0.01), further supporting the choice of the fixed-effects model. Furthermore, we controlled for potential heteroscedasticity by using robust standard errors to ensure the reliability of the estimation results.

The use of fixed-effects models is well-supported by prior research examining the impacts of environmental factors on agricultural productivity. For example, studies by Yi et al. [29] have employed fixed-effects models to analyze the effects of air pollution and climate variables on crop yields [31]. By adopting a fixed-effects model, our study aligns with established methodologies in the field, ensuring comparability and consistency with the existing literature. The economic model was constructed as follows:

$$\begin{array}{l}\ln Y_{it}={\alpha }_0+{\alpha }_1{O}_{i,t}+{\alpha }_{2}Pm{2.5}_{i,t}+{\alpha }_3{O}_{i,t}\times Pm{2.5}_{i,t}+{\alpha }_{4}are{a}_{i,t}+{\alpha }_{5}areasq{u}_{i,t}+\\ {\alpha }_6{\displaystyle \frac{P_{i,t}^l}{P_{i,t-1}^r}}+{\alpha }_{7}gd{d}_{i,t}+{\alpha }_{8}gddsq{u}_{i,t}+{\alpha }_{9}rai{n}_{i,t}+{\alpha }_{10}rainsq{u}_{i,t}+{\epsilon }_{i,t}\end{array}$$

where $Y_{i,t}$ denotes the rice yield in county i and year t, and ${\alpha }_j$ denotes the parameters to be estimated; $O_{i,t}$ denotes the M7, M12, Sum06, and Aot40, respectively in county i and year t; ${Pm}_{{2.5}_{i,t}}$ denotes the PM2.5 during the growth period in county i and year t; $O_{i,t}\ast {Pm}_{{2.5}_{i,t}}$ is the interaction term of the corresponding period of ozone and PM2.5; $are{a}_{i,t}$ denotes the rice sown area in county i in year t; $P_{i,t}^l$ and $P_{i,t-1}^r$ represent the price index of the agricultural means of production and the rice price index of the previous period, respectively. This ratio reflected the cost of rice planting.

2.3. Yield Loss

Wang and Mauzerall [32] used the (relative yield loss) method to estimate the rice yield loss in China. RYL is defined as follows:

$$RYL=1-Y/Y_{base}$$

where $Y_{base}$ is the estimated mean yield of the reference exponential index. The reference level is 25 ppb for M7 [33] and 35 μg/m³ for PM2.5¹. China’s mean annual and 24 h PM2.5 concentration limits, which were published in 2012 and implemented in 2016, are 35 μg/m³ and 75 μg/m³, respectively. Crop production loss (CPL) was calculated based on the RYL and actual output as follows:

$$CPL=RYL/(1-RYL)\times Output$$

The output represents rice yield. The annual rice yield for each province was obtained from the China Statistical Yearbook. The national average RYL for each crop was calculated as follows:

$$RY{L}_{na}={\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(CPL)}_i}}{{\displaystyle \sum _{i=1}^n({(CPL)}_i+{(Output)}_i)}}}$$

3. Results and Discussion

3.1. Baseline Results

The findings from the regression analysis are summarized in

Table 2. Effect of ozone and PM2.5 on rice yield.

Variable	(1)	(2)	(3)	(4)
M7 (ozone-sensitive period)	−0.00342 *** (0.00101)	−0.00387 *** (0.00110)	−0.00556 *** (0.00150)	−0.00587 *** (0.00155)
PM2.5 (growth period)			−0.00161 *** (0.00055)	−0.00143 *** (0.00053)
M7 × PM2.5			0.00003 *** (0.00001)	0.00003 *** (0.00001)
Rice sown area	−0.17078 *** (0.01458)	−0.16791 *** (0.01458)	−0.16994 *** (0.01456)	−0.16732 *** (0.01457)
Rice sown area squared	0.00167 *** (0.00033)	0.00163 *** (0.00032)	0.00166 *** (0.00033)	0.00162 *** (0.00032)
Price index ratio	−0.42198 * (0.22440)	−0.22838 (0.34816)	−0.39911 * (0.22127)	−0.20511 (0.34659)
Growth degree days (8–32 °C)	−0.36254 (0.23109)	−0.38349* (0.22739)	−0.37241 (0.23126)	−0.39367 * (0.22760)
Growth degree days (8–32 °C) squared	0.06957 (0.05954)	0.06771 (0.05815)	0.07139 (0.05972)	0.06997 (0.05815)
Precipitation	0.27552 ** (0.12572)	0.20132 (0.12526)	0.25619 ** (0.12318)	0.18913 (0.12388)
Precipitation squared	−0.23925 *** (0.09186)	−0.21271 ** (0.09115)	−0.22877 ** (0.09000)	−0.20607 ** (0.08985)
County fixed effect	Yes	Yes	Yes	Yes
Year fixed effect	No	Yes	No	Yes
2014		−0.00724 (0.01742)		−0.00603 (0.01731)
2015		0.03770 *** (0.01355)		0.03718 *** (0.01334)
Constant	3.12240 *** (0.30762)	3.10636 *** (0.36120)	3.21542 *** (0.31364)	3.09391 *** (0.36627)
Number of observations	2466	2466	2466	2466
R2	0.7577	0.7607	0.7587	0.7614

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

. A range of control variables were incorporated into the model. In models (1) and (2), the focus was solely on the effect of ozone on rice production, with PM2.5 excluded as a moderating variable. The first column controls only the individual empty top effect, and the second column controls the time fixed effect. In both models, higher ozone concentrations, as measured by M7, were found to significantly reduce rice yields. Specifically, the results in the second column indicate that a one-unit rise in M7-based ozone exposure leads to a 0.387% decline in rice production, with statistical significance at the 1% level.

In the third and fourth columns, we control for PM2.5; the third column only controls the individual fixed effect and the fourth column controls the time fixed effect. After accounting for PM2.5, individual fixed effects, and practice fixed effects, the findings in the fourth column reveal that a one-unit rise in ozone concentration led to a 5.87% reduction in rice output. The one-unit increase in PM2.5 concentration is associated with a 0.143% reduction in rice yield. In other words, for every 10 µg/m³ increase in PM2.5 concentration, the average rice yield decreases by 1.43%, which is equivalent to a reduction of 0.095 ton/ha per hectare. It was found that if PM2.5 factors are not considered, the estimation results will underestimate the degree of production loss caused by the increase in ozone concentration. At the same time, the results indicate that the coefficient of the interaction term between ozone and PM2.5 is positive and statistically significant at the 1% level, suggesting that PM2.5 partially attenuates the adverse effects of ozone on rice yield. In fact, the existing literature has confirmed the negative impact of ozone on rice yield based on agricultural conditions in countries such as Japan, South Korea, and India [34,35,36]. The research conclusion of this article further provides evidence from the actual situation in China.

Table 3. Impact of ozone exposure metrics on winter wheat productivity.

Variable	(1)	(2)	(3)
M12 (ozone-sensitive period)	−0.00604 *** (0.00163)
PM2.5 (growth period)	−0.00157 *** (0.00058)
M12 × PM2.5	0.00004 *** (0.00001)
Sum06 (ozone-sensitive period)		−0.00758 *** (0.00211)
PM2.5 (growth period)		−0.00037 ** (0.00017)
Sum06 × PM2.5		0.00005 ** (0.00002)
Aot40 (ozone-sensitive period)			−0.02887 *** (0.00768)
PM2.5 (growth period)			−0.00049 ** (0.00021)
Aot40 × PM2.5			0.00016 ** (0.00007)
Rice sown area	−0.16750 *** (0.01459)	−0.16933 *** (0.01460)	−0.16844 *** (0.01459)
Rice sown area squared	0.00162 *** (0.00032)	0.00164 *** (0.00033)	0.00164 *** (0.00033)
Price index ratio	−0.21372 (0.34635)	−0.23969 (0.34972)	−0.25981 (0.34964)
Growth degree days (8–32 °C)	−0.37486 * (0.22632)	−0.36236 (0.22601)	−0.37781 * (0.22713)
Growth degree days (8–32 °C) squared	0.06559 (0.05789)	0.06354 (0.05793)	0.06687 (0.05814)
Precipitation	0.18600 (0.12465)	0.22138 * (0.12494)	0.20745 * (0.12420)
Precipitation squared	−0.20233 ** (0.09017)	−0.21926 ** (0.09094)	−0.21449 ** (0.09039)
County fixed effect	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes
2014	−0.00629 (0.01747)	−0.00505 (0.01710)	−0.00703 (0.01737)
2015	0.03541 *** (0.01341)	0.03127 ** (0.01343)	0.03113 ** (0.01341)
Constant	3.08305 *** (0.36472)	2.88505 *** (0.35404)	2.95207 *** (0.35720)
Number of observations	2466	2466	2466
R2	0.7612	0.7596	0.7604

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

displays the outcomes of various ozone exposure metrics suggested in prior studies. In the first, second, and third columns, M12, Sum06, and Aot40, respectively, are used as measurement indicators of ozone concentration. PM2.5, individual fixed effects, and time fixed effects were controlled for in the model. Each unit increase in ozone concentration measured by M12, Sum06, and Aot40 caused rice yield losses of 0.604%, 0.758%, and 2.887%, respectively.

3.2. Robustness Check

Some scholars have suggested calculating the near-ground ozone concentration level within three months before crop harvest to capture the adverse impact of ozone pollution on crop growth [37]. Therefore, this study also calculates the near-ground ozone concentration level within three months before rice harvest, and then analyzes its impact on rice yield. The outcomes of the regression analysis are summarized in

Table 4. Robustness test results.

Variable	(1)	(2)	(3)	(4)
M7 (March–May)	−0.00371 *** (0.00098)	−0.00525 *** (0.00122)	−0.00648 *** (0.00163)	−0.00796 *** (0.00182)
PM2.5 (growth period)			−0.00195 *** (0.00064)	−0.00184 *** (0.00064)
M7 × PM2.5			0.00004 *** (0.00001)	0.00004 *** (0.00001)
Rice sown area	−0.17146 *** (0.01456)	−0.16768 *** (0.01451)	−0.17066 *** (0.01454)	−0.16709 *** (0.01450)
Rice sown area squared	0.00167 *** (0.00033)	0.00163 *** (0.00032)	0.00166 *** (0.00033)	0.00162 *** (0.00032)
Price index ratio	−0.53227 ** (0.20769)	−0.29605 (0.35075)	−0.49640 ** (0.20329)	−0.24845 (0.34881)
Growth degree days (8–32 °C)	−0.35177 (0.22791)	−0.36919 * (0.22461)	−0.35762 (0.23035)	−0.37678 * (0.22630)
Growth degree days (8–32 °C) squared	0.06718 (0.05909)	0.06290 (0.05763)	0.06853 (0.05960)	0.06490 (0.05798)
Precipitation	0.26485 ** (0.12481)	0.16003 (0.12414)	0.23612 * (0.12183)	0.13718 (0.12258)
Precipitation squared	−0.22454 ** (0.09091)	−0.18216 ** (0.08941)	−0.20685 ** (0.08847)	−0.16803 * (0.08763)
County fixed effect	Yes	Yes	Yes	Yes
Year fixed effect	No	Yes	No	Yes
2014		−0.00877 (0.01728)		−0.00712 (0.01721)
2015		0.04565 *** (0.01367)		0.04616 *** (0.01356)
Constant	3.23023 *** (0.29895)	3.13532 *** (0.36413)	3.32986 *** (0.30955)	3.21320 *** (0.36959)
Number of observations	2466	2466	2466	2466
R2	0.7575	0.7617	0.7587	0.7626

Note: Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

. The influence coefficients of M7, M12, and Sum06 within three months before the rice harvest were negative and passed the significance test, while the coefficients of other explanatory variables showed no significant change in value or significance, indicating that the negative impact of ozone pollution on rice yield can also be identified by using the ozone concentration value within three months before the rice harvest.

The results in

Table 5. Impact of varying ozone exposure metrics on winter wheat productivity.

Variable	(1)	(2)	(3)
M12 (March–May)	−0.00811 *** (0.00188)
PM2.5 (growth period)	−0.00196 *** (0.00068)
M12 × PM2.5	0.00005 *** (0.00002)
Sum06 (March–May)		−0.00392 *** (0.00099)
PM2.5 (growth period)		−0.00049 ** (0.00020)
Sum06 × PM2.5		0.00003 *** (0.00001)
Aot40 (March-May)			−0.01417 *** (0.00337)
PM2.5 (growth period)			−0.00065 *** (0.00024)
Aot40 × PM2.5			0.00009 *** (0.00003)
Rice sown area	−0.16729 *** (0.01452)	−0.16913 *** (0.01457)	−0.16834 *** (0.01454)
Rice sown area squared	0.00162 *** (0.00032)	0.00164 *** (0.00033)	0.00163 *** (0.00032)
Price index ratio	−0.25093 (0.34845)	−0.31543 (0.35031)	−0.33569 (0.35121)
Growth degree days (8–32 °C)	−0.35765 (0.22506)	−0.34767 (0.22675)	−0.35267 (0.22572)
Growth degree days (8–32 °C) squared	0.06030 (0.05776)	0.05888 (0.05824)	0.05995 (0.05795)
Precipitation	0.13915 (0.12311)	0.19788 (0.12455)	0.17943 (0.12367)
Precipitation squared	−0.16984 * (0.08795)	−0.20215 ** (0.08999)	−0.19040 ** (0.08914)
County fixed effect	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes
2014	−0.00695 (0.01724)	−0.00547 (0.01703)	−0.00688 (0.01719)
2015	0.04476 *** (0.01354)	0.03435 *** (0.01330)	0.03662 *** (0.01328)
Constant	3.19196 *** (0.36720)	2.97072 *** (0.35620)	3.02985 *** (0.35875)
Number of observations	2466	2466	2466
R2	0.7623	0.7603	0.7612

Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

show that the influence coefficients of M12, Sum06, and Aot40 within three months before the rice harvest are negative and pass the significance test, while the coefficients of other explanatory variables have no significant change in value or significance, which further shows that the negative impact of ozone pollution on rice yield can also be identified using the ozone concentration value within three months before the rice harvest.

3.3. Yield Loss Measure

Table 6. Ozone-induced relative yield loss (%) and rice yield loss.

	RYL	CPL
2013	8.00%	1759.8076
2014	6.92%	1525.8286
2015	7.07%	1575.5551

presents the calculated rice production losses due to ozone exposure from 2013 to 2015. Based on the M7 indices, the rice production losses during 2013–2015 were 8%, 6.92%, and 7.07%, respectively, corresponding to yield losses of 1759.8076, 1525.8286, and 1575.5551 tons. These estimates, however, do not account for the potential compounding effects of climate change, such as rising temperatures, altered precipitation patterns, and increased frequency of extreme weather events, which could exacerbate ozone-induced crop stress. For instance, higher temperatures can accelerate ozone formation and increase plant sensitivity to ozone damage, while water stress from irregular rainfall may further reduce rice resilience. Additionally, adaptive agricultural practices, such as the use of ozone-resistant crop varieties, optimized irrigation techniques, and adjusted planting schedules, could mitigate some of these losses. Incorporating these factors into future analyses would provide a more comprehensive understanding of the ozone’s impact on rice productivity and enhance the practical utility of the findings for policymakers.

Comparing these results to prior research, Zhao et al. [38] estimated relative yield losses of single-cropping rice in China between 2015 and 2018 from 10.6% to 13.5%, using ozone monitoring data from national stations and crop yield records. Variations in findings across studies may stem from differences in assumptions about ozone metrics, dose–response relationships [39], and the inclusion of climate change and adaptation factors. Despite these differences, this study validates that PM2.5 acts as a moderating factor in the relationship between ozone and rice productivity, highlighting the need for integrated air pollution and climate change mitigation strategies in agricultural policy.

4. Conclusions

Food security is a major concern in China. Protecting rice yield to the greatest extent possible has long been an important issue. Meanwhile, numerous studies have demonstrated the detrimental impacts of ground-level ozone on rice productivity from various perspectives. However, the current study ignores some relatively important influencing factors, which may lead to deviations in the research conclusions and policy suggestions.

This study comprehensively investigated the potential interactions between near-surface ozone and PM2.5 on rice yield. Using real near-surface ozone, PM2.5, weather, and economic variables, we empirically tested this interaction by constructing an econometric regression model. This study revealed that PM2.5 acts as a mediating factor in the relationship between ground-level ozone and rice productivity. Specifically, PM2.5 mitigated the adverse impacts of ozone on crop yields. Moreover, according to the estimation in this study, considering PM2.5, the rice yield during 2013–2015 will decrease by 8%, 6.92%, and 7.07%, respectively, that is, the total rice yield will decrease by 1759.8076 tons, 1525.8286 tons, and 1575.5551 tons, respectively.

Considering the current situation of China’s ozone pollution and its impact on ambient air quality, as well as the potential harm to human health, food security, and ecological security, China should prioritize the prevention and control of ozone pollution by establishing a national multi-pollutant collaborative control strategy with PM2.5 and ozone as the core pollutants. Additionally, deepening scientific research on the chemical mechanisms and meteorological factors driving PM2.5 and ozone pollution is essential to develop targeted mitigation measures. Advanced monitoring networks and interdisciplinary studies should be expanded to better understand the combined effects of these pollutants on rice productivity and other agricultural systems.

In implementing PM2.5 regulation measures, it is crucial to account for its moderating influence on the relationship between ozone levels and rice productivity. This includes promoting ozone-resistant crop varieties and adaptive agricultural practices, such as optimized planting schedules and improved irrigation techniques, to mitigate the impacts of air pollution. Strengthening public awareness and fostering collaboration between policymakers, researchers, and farmers will ensure that pollution control strategies are both effective and aligned with the needs of agricultural production.