The environmental impact of agriculture is influenced by several factors, including the use of fertilizers, pesticides, irrigation, and credit availability, which can drive both direct and indirect carbon emissions (Shao, 2024). This study is anchored in the theory that modern agricultural practices, while boosting productivity, can lead to significant environmental externalities, particularly in the form of greenhouse gas (GHG) emissions. Drawing on resource-based theory, this study examines how agricultural inputs such as fertilizers, pesticides, water, and credit impact CO₂ emissions. The theory posits that efficient resource use can mitigate environmental degradation, while overuse or inefficient practices tend to amplify carbon emissions (Rehman et al., 2019).

Furthermore, climate change impact suggests that temperature fluctuations directly and indirectly influence crop yields and agricultural practices, which, in turn, affect CO₂ emissions (Demirhan, 2020). This study also considers the sustainability and development theory to evaluate the balance between agricultural growth and environmental sustainability, exploring whether economic gains can coexist with reduced carbon emissions.

Figure 2 represents a cyclic process where agricultural practices designed to enhance productivity simultaneously drive carbon emissions. This cyclic process illustrates the complex association between agricultural practices, carbon emissions, and the environmental impact on maize production. The consumption of fossil fuels has increased across those producing fertilizers and pesticides. These industrial activities emit significant levels of carbon dioxide (CO₂) in the environment. The emissions cycle connects to climate change and contribute to erratic climate patterns, including extreme temperatures, floods, droughts, and crop diseases.

Furthermore, agricultural support systems, government and institutional funding, promote modern agricultural practices through financial assistance to farmers. This support encourages farmers to increase the use of fertilizers, pesticides, and water, all of which are essential for high-yield farming. As a result, the excessive use of agricultural inputs can reduce soil fertility, contaminate groundwater, and further increase carbon emissions.

Modern agriculture involves advanced machinery and technologies that enhance maize productivity, as well as agricultural GDP. However, this productivity boost comes at an environmental cost, as the adoption of modern agricultural technology also leads to higher CO₂ emissions. These emissions contribute to climate change, which in turn impacts crop yields. The red arrows represent the flow of carbon emissions through various processes, while the black arrows indicate the cycling of resources and environmental impact, emphasizing the interconnectedness between agricultural productivity, carbon emissions, and environmental sustainability.

The study analyzes panel data from six major corn-producing provinces in China: Hubei, Sichuan, Henan, Heilongjiang, Hebei and Shandong as shown in Figure 3. The main goal is to evaluate the impact of maize productivity along with various influencing factors including pesticides use, fertilizers use, agricultural water usage, agricultural credit, climate Change, and agricultural gross domestic product (AGDP), all impact CO₂ emissions. The data used in this study was derived from the China Agricultural Statistical Yearbook (CASY) for the years 1990 to 2022.¹ Temperature records were obtained from models processed by Cubic Data Academy, which utilized baseline data from the National Oceanic and Atmospheric Administration's National Centers for Environmental Information (NCEI) (https://www.ncei.noaa.gov/data/global-summary-of-the-day/archive/) which offers highly accurate daily meteorological data from observation stations. These data are further processed by Cubic Data Academy to produce national temperature datasets, widely recognized for their reliability in climate studies. To address missing data within the panel datasets, we applied a meticulous mean imputation method, selected for its appropriateness given the nature and pattern of the missing data. The integrity of our results was validated through sensitivity analysis, which explored how different imputation methods and assumptions about missing data might affect our findings, detailed in Table 1. CO₂ emissions were estimated based on historical terminal energy consumption data, as the primary emissions are generally linked to fossil fuel combustion (IPCC, 2007). Following methodologies from Li and Yin (2024) and Jiehua and Zhang (2020), our calculation for total CO₂ emissions included nine energy sources, arranged in the following order: diesel, natural gas, coal, electricity, gasoline, kerosene, coke, crude oil, and fuel oil, as used in the following equation.

C_it signifies the total carbon emissions of province i in year t; E_ijt denotes the usage of the jth type of energy in province i in year t; η_j signifies the carbon emission coefficient of the jth type of energy; i represents the province i; j represents the jth energy source. Since the original data on energy consumption is presented in physical quantities, necessitating conversion into standard statistical units to compute carbon emissions accurately. The conversion factors for these emissions are derived from the “2006 IPCC Guidelines for National Greenhouse Gas Inventories.” These factors take into account the average lower heating values and established carbon emission factors for each energy type. The calculated carbon emission coefficients for the nine energy types—coal, gasoline, diesel, natural gas, kerosene, fuel oil, crude oil, electricity, and coke—are 1.9003 kg-CO₂/kg, 2.8604 kg-CO₂/kg, 3.0202 kg-CO₂/kg, 2.9251 kg-CO₂/kg, 3.0179 kg-CO₂/kg, 3.0959 kg-CO₂/kg, 3.175 kg-CO₂/kg, 2.1622 kg-CO₂/m³, and 0.801 kg-CO₂/kWh, respectively. The econometric approaches used in the empirical analysis are outlined in Figure 4.

Table 2 presents a wide range between the minimum and maximum values of variables, suggesting the presence of outliers—data points that significantly differ from the overall trend when displayed in their logarithmic forms. Low standard deviation values suggest that the data points are closely grouped around their means, indicating that the dataset is stable and shows little variation from the average. Many of the variables display negative skewness, which points to an asymmetry that leans to the left, with a higher concentration of values toward the upper end of the scale. Kurtosis measures how the tails of a distribution compare to those of a normal distribution, where positive kurtosis indicates heavier tails and a sharper peak. Higher kurtosis values reflect more pronounced peaks and heavier tails, while lower values suggest a flatter distribution with lighter tails. The results of the Jarque-Bera (JB) test reveal that the data distribution is asymmetrical and deviates from the classic bell shape of a normal distribution, indicating variations in skewness and kurtosis. Figure 5 displays the logarithmic transformations of the variables examined across the provinces from 1990 to 2022. Table 3 presents the results of the correlation analysis, demonstrating strong positive relationships between CO₂ emissions and the various independent variables examined in the study.

The results in Table 4 confirm the presence of cross-sectional dependence (CSD), indicating that standard shocks affect the variables across the specified provincial group. Table 5 presents the findings of second-generation unit root tests, including the CIPS and CADF test statistics for both level and first difference specifications. According to the results of the CIPS test, variables such as carbon emissions, fertilizer use, agricultural water usage, agricultural credit, and temperature are stationary at the level, whereas corn production, pesticide use, and agricultural GDP reach stationarity at the first difference. The CADF test results show that all variables except for LTM are stationary at the first difference. Table 6 displays Kao cointegration results, which shows a significant long-term relationship among the variables in our model. Co-integration tests are essential for investigating long-term relationships among variables (Gengenbach et al., 2006).

Before conducting the multiple regression analysis, the several diagnostic tests were employed to ensure the validity of results. Multicollinearity was assessed using the Variance Inflation Factor (VIF), with all values found to be below 10 as shown in Table 7, indicating no econometric issues (Wooldridge, 2009). Additionally, the Breusch-Pagan/Cook-Weisberg test was conducted to check for heteroskedasticity, which can lead to an underestimation of the intercept and an overestimation of the slope coefficients (Street, 2003). The results showed no evidence of heteroskedasticity, as the p-value was >0.05, confirming that the variance of the residuals was constant (Table 8). The Wooldridge test was employed to test for autocorrelation, and the results indicated its presence in the panel data, as the p-value (0.0006) was below 0.05, leading to the rejection of the null hypothesis (Table 9). To address this issue, the PMG-ARDL model and dynamic panel models were used, ensuring the autocorrelation was accounted for and reliable results were obtained.

Table 10 presents the PMG-ARDL results, which show that the use of fertilizers, agricultural water, and agricultural credit significantly contribute the increase in CO₂ emissions over the long term. Specifically, a one-unit increase in fertilizer use, agricultural water consumption, and agricultural credit corresponds to increases in CO₂ emissions of 1.49, 1.23, and 0.47 units, respectively, in the long run. It is important to note that the effects of fertilizer and agricultural water usage are more substantial compared to other variables in the model. In contrast, corn production, pesticide use, temperature, and agricultural GDP show a negative relationship with CO₂ emissions in the long term. Additionally, agricultural water use and agricultural credit positively impact CO₂ emissions significantly, while corn productivity also correlates positively with CO₂ emissions. In the short term, increases in corn production, fertilizer use, and agricultural GDP each contribute to significant environmental degradation, resulting in respective increases in CO₂ emissions of 10%, 17%, and 10%, respectively. Furthermore, the Error Correction Term (ECT) is notably negative, which indicates strong support for the model's validity. This suggests that the model adjusts quickly toward equilibrium, with an adjustment speed of 15%. This rapid adjustment underscores the model's effectiveness in reflecting short-term deviations from long-term equilibrium.

The findings from Table 11 display the PMG-ARDL short-term analysis results across provinces (the long-term results remain same as shown in Table 10). In the short term, corn production notably increases CO₂ emissions in Henan and Sichuan provinces, which are major corn-producing regions with high pesticide use and agricultural water consumption, both contributing to increased emissions. In contrast, in Heilongjiang province, corn production has a negative impact on CO₂ emissions. This difference may be attributed to factors such as urbanization in Northeast China, which has led to a shift of rural laborers to non-agricultural sectors. This shift has promoted increased mechanization in agriculture, enhancing agricultural efficiency and potentially reducing carbon emissions associated with inputs like pesticides and fertilizers. The flat terrain of the Heilongjiang Plain, along with its high level of mechanization in grain production, may also help mitigate emissions. Furthermore, fertilizer usage significantly contributes to increased CO₂ emissions in both Heilongjiang and Sichuan provinces. Agricultural water consumption drives CO₂ emissions in Hubei and Sichuan, while agricultural credit is linked to higher emissions in Heilongjiang and Hubei. Changes in temperature due to climate change generally raise carbon emissions across all examined provinces, except Hubei. Agricultural GDP is positively correlated with carbon emissions in all provinces, except Shandong and Sichuan. Additionally, Figures 6, 7 illustrate trends in pesticide use, fertilizer use, agricultural water consumption, agricultural credit, a temperature, and agricultural GDP over the period 1990 to 2022 across the studied provinces. Figure 8 illustrates the annual corn production in six major corn-producing provinces in the study from 1990 to around 2022. The trends suggest that most provinces have experienced overall growth in corn production over the past few decades, with varying degrees of stability or slight declines after reaching peak levels in recent years. Differences in peak times and fluctuations among provinces may be due to factors such as regional climate conditions, government policies, technological advancements, and investment in agricultural infrastructure. Figure 9 illustrates the share of CO₂ emissions by sector in the world and China. The power sector is the largest contributor, accounting for 57% of total emissions, industrial sector at 26% and agricultural sector contribute to 1% share in total carbon emissions from energy use in China in 2022 (Statista, 2024). Globally, the power sector contributes 28% of total emissions, industrial sector at 23% and agricultural sector contribute to 17% share in total carbon emissions (IEA, 2022).

To ensure the robustness of our findings, we conducted a panel quantile regression analysis as shown in Table 12. The results indicate that corn production negatively affects CO₂ emissions. In contrast, fertilizer use, agricultural credit, and agricultural water consumption have statistically significant and positive impacts on carbon emissions, supporting the findings from the PMG-ARDL model (Table 10). Providing financial assistance to farmers appears to encourage the adoption of modern agricultural practices, which can enhance crop yields but also increase carbon emissions. Our results thus suggest a strong link between agricultural financing and carbon emissions. Additionally, the interaction between temperature and agricultural GDP was found to have a negative impact on carbon emissions, consistent with the PMG-ARDL model. To further evaluate data reliability, we applied a Panel Least Squares regression analysis (Table 13). The findings confirm that carbon emissions are significantly influenced by corn production, agricultural credit, and fertilizer use. Temperature has a negative effect on carbon emissions, while agricultural GDP contributes positively. The R-squared value indicates that approximately 48% of the variation in carbon emissions can be explained by the model's explanatory variables. The research underscores the complex relationship between environmental quality and corn production, with a particular focus on carbon emissions. Our findings indicate a long-term negative relationship between CO₂ emissions and corn production. However, the long-term increase in carbon emissions is significantly driven by the use of fertilizers, agricultural water, and agricultural credit. The results from the PMG-ARDL model (Table 10) align with those from the Panel Quantile Regression (Table 12) and Panel Least Squares (Table 13), all of which show strong positive correlations between agricultural loans, fertilizers use, agricultural water use, and carbon emissions.

The results of this study, as shown in Table 10, provide key insights into the long-term and short-term impacts of corn production and various agricultural inputs on CO₂ emissions in China's major corn-producing provinces. In the long run, corn production shows a negative impact on CO₂ emissions, suggesting that increased corn production over time is associated with a reduction in CO₂ emissions. This finding aligns with Zhang et al. (2022), who reported that agricultural intensification, through higher crop yields, can lead to more efficient resource use, thereby lowering agriculture's carbon footprint. However, this negative relationship differs from the findings of Ullah et al. (2021), who observed a positive correlation between corn production and CO₂ emissions, attributing it to the high input demands of modern corn farming, such as fertilizers and irrigation. This discrepancy may arise from differences in the scope and scale of the studies; while Ullah et al. (2021) focused on short-term impacts in specific regions, our study takes long-term perspective, potentially capturing the benefits of technological advancements and improved farming practices that reduce emissions over time. Additionally, regional variation in agricultural practices and policies may contribute to these contrasting findings.

The positive and significant effect of fertilizer use on CO₂ emissions aligns with the findings of Rehman et al. (2019), who highlighted that the extensive use of nitrogen-based fertilizers significantly contributes to CO₂ emissions due to both the production processes of these fertilizers and their application in the field. Similarly, Shao (2024) emphasized that fertilizer use is a major contributor to agricultural CO₂ emissions globally and advocated for more efficient fertilizer use to mitigate these emissions. While several research have investigated the effects of crop productivity and irrigated land area (Zhai et al., 2017; Pickson et al., 2021; Zhang et al., 2022), fewer studies have explored the relationship between carbon emissions, water energy consumption, and water energy management (Anser et al., 2020; Wang et al., 2020). The link between agricultural water usage and carbon emissions, specifically in terms of quantity, remains underexplored. Our research on agricultural water use indicates a significant positive influence on carbon emissions, consistent with Rehman et al. (2020), suggesting that excessive irrigation, combined with fertilizer use, increases atmospheric evaporation, contributing to long-term carbon emissions. Our analysis shows a positive correlation between agricultural credit and carbon emissions. Agricultural water use similarly exhibits a positive and significant impact on CO₂ emissions. This result aligns with Li and Yin (2024), who found that irrigation practices, especially in water-scarce regions, contribute significantly to CO₂ emissions due to the energy required for pumping and distribution. Inefficient water management practices exacerbate these emissions in China's agricultural sector, a finding echoed by Anser et al. (2020), who recommended more sustainable water use in agriculture. Agricultural credit has a significant and positive impact on CO₂ emissions in the long run, supporting Koondhar et al. (2021a), who reported that increased access to agricultural credit often results in higher CO₂ emissions due to the expansion of high-input farming practices. Onyeneke et al. (2024) also found that agricultural credit enables farmers to invest in more intensive practices, which, while increasing productivity, also heightens emissions. However, our results suggest a stronger relationship between agricultural credit and CO₂ emissions than reported in other studies. This may be due to the specific nature of corn production in the studied provinces, where credit may is often used to finance the purchase of fertilizers, pesticides, and irrigation equipment-all of which contributing to higher emissions.

The negative impact of temperature on CO₂ emissions contrasts with the general expectation that higher temperatures lead to increased emissions. Demirhan (2020) found that rising temperatures often increase the use of water and energy for cooling in agriculture, and his finding suggest that, in the long run, high temperatures could be associated with the adaptation of less energy-intensive farming practices or shifts in crop types requiring fewer inputs. This counterintuitive result might also reflect the specific climatic conditions of the regions studied, where rising temperatures may reduce the need for energy-intensive practices like heating or might prompt a shift to less resource-intensive crops or farming methods. This underscores the complexity of the relationship between climate variables and CO₂ emissions and highlights the importance of regional studies to fully understand these dynamics.

The agricultural GDP does not have a significant impact on CO₂ emissions in the long run. This finding is somewhat surprising, as Huang et al. (2019) found a positive relationship between agricultural GDP growth and CO₂ emissions, attributing it to the expansion of high-input agricultural practices. Some scholars have highlighted the decoupling effect between agricultural GDP and carbon emissions, suggesting that increases in agricultural value-added have limited long-term impacts on emissions due to enhanced efficiency and reduced energy dependency (Ghosh et al., 2023; Amin and Rahman, 2024). Additionally, some studies report that agricultural growth can reduce carbon emissions (Mahmood et al., 2019; Samargandi, 2017). Cole et al. (1997) found that technological advancements in agriculture have led to a 32% reduction in CO₂ emissions. Zaharani et al. (2011) investigated that modern agricultural strategies such as efficient water use may contribute to a safe and sustainable environment. Conversely, Ullah et al. (2021) showed that agricultural value-added negatively impacts carbon emissions, while Waheed et al. (2018) identified agriculture as a significant source of CO₂ emissions. Wang and Zhang (2021) suggested that increases in agricultural GDP might also enable investments in more sustainable practices, potentially mitigating emissions. The lack of a significant relationship in this study could indicate China's agricultural policies have promoted sustainable practices (e.g., crop rotation, conservation tillage, and integrated pest management) that help reduce greenhouse gas (GHG) emissions even as productivity increases (Zou et al., 2024; Yang et al., 2014). Furthermore, technological advancements and the adoption of precision agriculture have enabled Chinese farmers to use inputs more efficiently. For instance, modern irrigation systems and optimized fertilizer application reduce CO₂ emissions, mitigating the environmental impact typically associated with agricultural growth (Ma et al., 2024b; Chen et al., 2024; Long et al., 2018). Moreover, government support for renewable energy in agriculture, including solar-powered irrigation and bioenergy from agricultural waste, has also played a key role in reducing carbon emissions (Zhu and Huo, 2022). Ecological methods such as Integrated Rice-Animal Farming (IRF) with fish and ducks have not only reduced the need for fertilizers and pesticides but also helped preserve environmental diversity in China (Yifan et al., 2023). Diversified cropping systems, efficient fertilizer use, intensified crop rotations, and integrated practices increase agricultural productivity while lowering carbon footprints (Cheng et al., 2011). Yang et al. (2022) found that the trend in agricultural carbon emissions shifted from increasing to continuously decreasing after 2015 at the national level, mainly due to improvements in major grain-producing areas. They concluded that advancements in agricultural infrastructure and human capital have positively contributed to agricultural eco-efficiency growth. This finding suggests that economic growth in agriculture does not automatically translate into higher emissions but rather depends on the nature of the growth and the types of practices it supports.

The Dumitrescu and Hurlin (2012) approach indicates causal relationships among variables (Table 14). The findings reveal significant unidirectional and bidirectional causal links between the variables. Unidirectional causation is observed between carbon emissions and pesticide use, carbon emissions and temperature, and carbon emissions and agricultural GDP. Additionally, a bidirectional causal relationship exists between carbon emissions and several other variables, including corn production, fertilizer application, agricultural water usage, and agricultural credit.