Introduction

The economic impact of climate change continues to gain interest in food production research. This is because its externalities are complex and uncertain to forecast. For instance, climate change improves economic welfare (Tol 2008, 2009, 2018). However, its gains are sunk, and an overwhelming share of the global welfare loss from climate change is borne by relatively poor economies that rely on climatic factors for survival (Tol. 2008, 2009, 2018). For communities that depend on rain-fed agriculture, episodes of extreme drought followed by high heat waves disrupt irrigation systems and agricultural water productivity, thus creating a mismatch between water demand and supply for agriculture production (Bhardwaj et al. 2018; Aryal et al. 2019). Inadequate rainfall disrupts seeds germination and the maturity period of crops, and their harvest (Basnayake et al. 2006; Haque et al. 2016), whereas excess downpours cause flooding, reduce soil fertility, and destabilize hydrologic balances and crop yields (Agovino et al. 2019; Pickson and He 2021; Mahdu 2019; Simpson 2016). Existing evidence on rice production shows that the rice crop is profoundly susceptible to extreme cold and hot temperatures (Dabi and Khanna 2018). The flowering stage of rice crops at high temperatures above 35 °C can cause unfruitfulness (Haque et al. 2016). High water evaporation reduces the water required for rice growth through transpiration (Mahdu 2019). Although maize can sustain high temperatures (Wu et al. 2021), its development is susceptible to temperatures greater than 30 °C (Kang et al. 2017; Lobell et al. 2013; Zhou and Turvey 2014).

Indeed, climate change is continuously threatening the food security of developing countries (Zainal et al. 2014; Pickson and Boateng 2021; Chandio et al. 2022a), particularly in African and Asian regions where food shortage is prominent (Zewdie 2014; Wang et al. 2018). China, for example, is the largest producer of rice (Food and Agriculture Organisation and Organisation for Economic Co-operation and Development 2018; Liu et al. 2020a, b), the second-largest producer of maize, and the leading importer of maize globally (Food and Agricultural Organisation 2017). Approximately 80% of rice, wheat, and maize production in China is dedicated to human consumption (Kearney 2010; Foley et al. 2011; Li et al. 2015; Wu et al. 2021). Notwithstanding, climate extremes pose a significant threat to the country’s future maize, wheat, and rice production (Wang et al. 2011; Huang et al. 2010). Studies revealed that the frequency of climate warming could severely affect the production of rice, wheat, and maize in China over the next two to eight decades (Xiong et al. 2007; Solomon et al. 2007; Lin et al. 2017; Lv et al. 2018; Chen and Pang 2020). Xiong et al. (2014) and Liu et al. (2020a, b) showed that the rising temperature in China had adversely affected the growth of several food crops like wheat, maize, and soybeans. In contrast, Pickson et al. (2020) argue that rising carbon emissions and temperature are among the climate change factors that hinder cereal crops in China. This situation is not only peculiar to the Chinese economy, but farmers in countries such as India, Nepal, Pakistan, and Bangladesh face similar adverse impacts of climate change (Baig et al. 2022; Chandio et al. 2022b; Gul et al. 2021; Islam and Wadud 2020).

In this study, we explored the role of climate change in rice and maize production in China. China is a particular case in the food security debate in several ways. First, the rapid growth of China’s population and the increasing food demand has called for the need to intensify the food supply to feed the growing population (Du et al. 2004; Zhai et al. 2014). It is predicted that the country’s food security is likely to shrink from 94.5% in 2015 to approximately 91% in 2025 (Huang et al. 2017) if policies to sustain food production are not implemented. Besides, evidence reveals that China may need 776 million tonnes of grain to meet its increasing food demand by 2030 (Li et al. 2014). With climate change imposing a significant impact on food production, a better understanding of how climate change affects cereal crops like rice and maize has become a priority for the Chinese government. As the debate continues, and following Howard et al. (2016), we provided a comprehensive analysis of both short- and long-run effects of climate fluctuation on rice and maize production. Second, we paid particular attention to the nonlinearity of rice and maize production. For this reason, we used the autoregressive distributed lag (ARDL) model and the quantile regression (QR) in our empirical analysis. These analytical techniques are more effective and robust tools for analyzing how rice and maize production respond to the changing climatic conditions in China over time than the crop simulation models commonly used in prior studies (Wang et al. 2014; Zhang et al. 2016; Yang et al. 2017; Lin et al. 2017; Lv et al. 2018; Tian et al. 2020; Fei et al. 2020; Tian et al. 2020; Chen and Pang 2020; Jiang et al. 2021; Zhang et al. 2021). Crop simulation models are location-specific experiments that require extensive datasets regarding farm management strategies, crop growth and development, and soil samples. It is, therefore,challenging to attain conclusive outcomes at a regional or national level (Folberth et al. 2012; Wu et al. 2021). The temperature effect simulations are difficult since the effects of varying temperature intervals can also be affected by the precipitation and irrigation conditions (Schauberger et al. 2017). By using the ARDL model, we account for linear associations between the explained and explanatory variables. The QR model considered distributional asymmetry in the association between the explained and explanatory variables. The QR methodology assists in comprehending (and thus forecasting) the connection between cereal crops and climatic variables based on the propensity of rainfall, temperature, and carbon emissions in the low, medium, and upper quantiles. Finally, our policy recommendations provide alternative mechanisms for sustaining maize and rice production in the presence of climate change.

Econometric methodology

Model specification

Following Pickson et al. (2020), Chandio et al. (2020a), Pickson et al. (2021), and Wu et al. (2021), two separate models for rice and maize production are presented whereby rice and maize production are intuitively related to the levels of temperature, carbon emissions, and rainfall coupled with other control variables such as fertilizer consumption, agricultural irrigated lands, and the cropped lands for respective crop production. The regression models are specified as follows:

$${\mathrm{lnRICE}}_{\mathrm{t}}={\theta }_{\mathrm{o}}+{\theta }_{1}{\mathrm{lnFERZ}}_{\mathrm{t}}+{\theta }_{2}{\mathrm{lnAGIL}}_{\mathrm{t}}+{\theta }_{3}{\mathrm{lnCARI}}_{\mathrm{t}}+{\theta }_{4}{\mathrm{lnCO}}_{2\mathrm{t}}+{\theta }_{5}{\mathrm{lnTEM}}_{\mathrm{t}}+{\theta }_{6}{\mathrm{lnRAIN}}_{\mathrm{t}}+{\varepsilon }_{\mathrm{t}}$$
(1)
$${\mathrm{lnMAIZ}}_{\mathrm{t}}={\theta }_{o}+{\theta }_{1}{\mathrm{lnFERZ}}_{\mathrm{t}}+{\theta }_{2}{\mathrm{lnAGIL}}_{\mathrm{t}}+{\theta }_{3}{\mathrm{lnCAMA}}_{\mathrm{t}}+{\theta }_{4}{\mathrm{lnCO}}_{2\mathrm{t}}+{\theta }_{5}{\mathrm{lnTEM}}_{\mathrm{t}}+{\theta }_{6}{\mathrm{lnRAIN}}_{\mathrm{t}}+{\varepsilon }_{\mathrm{t}}$$
(2)

In Eqs. (1) and (2), FERZ represents fertilizer consumption, AGIL connotes agricultural irrigated lands, \({\mathrm{CO}}_{2}\) represents carbon emissions, TEM indicates average temperature, RAIN connotes average precipitation, \({\upvarepsilon }_{t}\) is the white noise error term, and ln denotes the natural logarithm. In Eq. (1), RICE and CARI signify rice production (the explained variable) and the cropped lands under rice production, respectively. In Eq. (2), MAIZ stands for maize production (the dependent variable), and CAMA represents the cultivated areas under maize production.

Data characteristics

This study used a quarterly time series dataset from 1978Q1 to 2015Q4. The availability of reliable data influenced data coverage. The data for rice production, maize production, cultivated areas under rice production, and cropped lands under maize were extracted from the various issues of the China Statistical Yearbook published by the National Statistical Bureau of China. Also, agricultural irrigated lands, fertilizer consumption, average temperature, carbon emissions, and average rainfall were sourced from the World Bank database. In this study, average temperature, carbon emissions, and average rainfall were used as indicators of climate change. Table 1 summarizes the explained and explanatory variables with their expected signs.

Table 1 Summary of explained and explanatory variables with their expected signs
Full size table

Estimation techniques

Trend analysis of climatic variability

Following previous studies, we used the Mann–Kendall test (Mann 1945; Kendall 1948) and Sen’s slope estimator (Sen 1968) to estimate the actual trend and the degree of the observed trend in climatic variables (see, for example, Pickson et al. 2020; Pickson and Boateng 2021). The Mann–Kendall (M–K) test (a nonparametric method) offers two benefits. First, this nonparametric technique is not dependent on the normality of the dataset. Second, it is unresponsive to unexpected interruptions due to the heterogeneous time-series data. The null hypothesis of the M–K test of no trend in the time series data is tested against the alternative proposition that a pattern exists. Thus, a positive Z value of the M–K test shows a rising trend in the meteorological dataset, while a negative Z value shows a declining trend. The absolute Z value is then compared to the critical value of \({Z}_{1-\alpha /2}=1.96\) to verify the trend. When the absolute Z value is greater than the critical value, the null hypothesis that there is no trend is refuted, implying a pattern in the meteorological dataset (Gocic and Trajkovic 2013; Koudahe et al. 2017; Akinbile et al. 2020; Pickson et al. 2020; Pickson and Boateng 2021). Besides, Sen’s slope method, which is a nonparametric approach, was used to estimate the slope of an existing trend and predict the future direction of the trend. Thus, Sen’s slope estimator is used to compute the magnitude of the movement in a time series (Sen 1968). Sen’s gradient (\({Q}^{^{\prime}}\)) represents the rate of change across the whole timeframe. In the meteorological dataset, a positive sign of \({Q}^{^{\prime}}\) value implies a rising pattern, whereas a minus sign of \({Q}^{^{\prime}}\) value shows a diminishing trend (Gocic and Trajkovic 2013; Koudahe et al. 2017; Akinbile et al. 2020; Pickson et al. 2020; Pickson and Boateng 2021).

ARDL cointegration test

This study used the ARDL model, an alternative cointegration method that Pesaran et al. (2001) advanced, to examine the long-run association among the underlying variables in the rice and maize production models. The ARDL technique has many advantages. First, the ARDL model is the more robust method for determining the cointegration between small samples. Also, the ARDL methodology is applicable regardless of whether the regressors are I(1) and/or I(0) and if they are all integrated of the same order. Despite this advantage, we employed the recently developed Ng-Perron unit root test by Ng and Perron (2001) to determine the integration levels of the variables of interest. Ng and Perron (2001) proposed four test statistics \({(\mathrm{MZ}}_{\mathrm{\alpha }}, {\mathrm{MZ}}_{\mathrm{t}},\mathrm{ MSB},\mathrm{ and MPT})\) to overcome the constraints of the augmented Dickey-Fuller (ADF) and the Phillips-Perron (PP) tests (Yildirim et al. 2015). The MZa and MZt tests are remakes of the Phillips-Perron Za and Zt tests by Phillips and Perron (1988) and Phillips (1987), the MSB test is an improved version of the R test by Bhargava (1986), and the MPT test is a rendition of the ADF-generalized least squares (GLS) test by Elliot et al. (1996). Lastly, the ARDL technique allows various variables to have varying optimum numbers of lags.

The ARDL approach is a two-step process that can identify the long-term relationship among the underlying variables (Pesaran and Pesaran 1997). The first step is the Fisher test of the null hypothesis that there is no long-term co-movement among the underlying variables. The second step involves concurrently estimating the coefficients of long-run and short-run relationships with the error term of the ARDL model. The error correction model version of the ARDL framework helps determine how the system will adjust to reach equilibrium.

The following conditional ARDL models are estimated by applying the bounds test procedure for examining the long-term co-movement between the cultivation of cereal crops (rice and maize) and their respective explanatory variables.

$$\begin{array}{l}{\Delta\mathrm{lnRICE}}_{\mathrm t}=\delta_{\mathrm o}+\Omega_1{\mathrm{lnRICE}}_{\mathrm t-1}+\Omega_2{\mathrm{lnFERZ}}_{\mathrm t-1}+\Omega_3{\mathrm{lnAGIL}}_{\mathrm t-1}+\Omega_4{\mathrm{lnCARI}}_{\mathrm t-1}+\Omega_5{\mathrm{lnCO}}_{2\mathrm t-1}+\\\Omega_6{\mathrm{lnTEM}}_{\mathrm t-1}+\Omega_7{\mathrm{lnRAIN}}_{\mathrm t-1}+\sum\limits_{i=1}^n\gamma_{\mathrm i}{\Delta\mathrm{lnRICE}}_{\mathrm t-1}+\sum\limits_{i=1}^n\gamma_{\mathrm i}{\Delta\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^n\gamma_{\mathrm i}{\Delta\mathrm{lnAGIL}}_{\mathrm t-1}+\\\sum\limits_{i=1}^n\gamma_{\mathrm i}{\Delta\mathrm{lnCARI}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^n\gamma_i{\Delta\mathrm{lnCO}}_{2\mathrm t-1}+\sum\limits_{i=1}^{\mathrm n}\gamma_{\mathrm i}{\Delta\mathrm{lnTEM}}_{\mathrm t-1}+\sum\limits_{\mathrm i=1}^n\gamma_{\mathrm i}{\Delta\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\varepsilon_{\mathrm t}\end{array}$$
(3)
$$\begin{array}{l}{\Delta\mathrm{lnMAIZ}}_{\mathrm t}=\lambda_{\mathrm o}+\varphi_1{\mathrm{lnMAIZ}}_{\mathrm t-1}+\varphi_2{\mathrm{lnFERZ}}_{\mathrm t-1}+\varphi_3{\mathrm{lnAGIL}}_{\mathrm t-1}+\varphi_4{\mathrm{lnCAMA}}_{\mathrm t-1}+\varphi_5{\mathrm{lnCO}}_{2\mathrm t-1}+\\\varphi_6{\mathrm{lnTEM}}_{\mathrm t-1}+\varphi_7{\mathrm{lnRAIN}}_{\mathrm t-1}+\sum\limits_{i=1}^{\mathrm n}\phi_{\mathrm i}{\Delta\mathrm{lnMAIZ}}_{\mathrm t-1}+\sum\limits_{i=1}^n\phi_{\mathrm i}{\Delta\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}\sum\limits_{i=1}^n\phi_i{\Delta\mathrm{lnAGIL}}_{\mathrm t-1}+\\\sum\limits_{i=1}^n\phi_{\mathrm i}{\Delta\mathrm{lnCAMA}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^n\phi_{\mathrm i}{\Delta\mathrm{lnCO}}_{2\mathrm t-1}+\sum\limits_{i=1}^n\phi_{\mathrm i}{\Delta\mathrm{lnTEM}}_{\mathrm t-1}+\sum\limits_{i=1}^n\phi_{\mathrm i}{\Delta\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\varepsilon_{\mathrm t}\end{array}+$$
(4)

where \(\Delta\) and \({\varepsilon }_{\mathrm{t}}\) represent the first difference operator and the error term, respectively. The parameters \({\Omega }_{\mathrm{i}}\) and \({\varphi }_{\mathrm{i}}\) connote the long-run elasticities for rice production and maize production models, correspondingly. The parameters \({\upgamma }_{\mathrm{i}}\) and \({\phi }_{\mathrm{i}}\) denote the short-run elasticities for rice production and maize production models, respectively. In Eqs. (3) and (4), \({\delta }_{0}\) and \({\lambda }_{0}\) stand for the drifts for the rice production and maize production models, respectively.

The ARDL bounds test procedure involves the estimations of Eqs. (3) and (4) using the ordinary least squares (OLS) method, and performing the Fisher test is to examine the joint significance of all the variables in the respective models. The null propositions of no long-term co-movement among the variables in the two models are given as follows:

$${H}_{\mathrm{Rice}}:{\Omega }_{1}={\Omega }_{2}={\Omega }_{3}={\Omega }_{4}={\Omega }_{5}={\Omega }_{6}={\Omega }_{7}=0$$
$${H}_{\mathrm{Maize}}:{\varphi }_{1}={\varphi }_{2}={\varphi }_{3}={\varphi }_{4}={\varphi }_{5}={\varphi }_{6}={\varphi }_{7}=0$$

The Fisher test, which normalizes rice and maize production, is defined as follows:

$$F\mathrm{y}(\mathrm{lnrice}/\mathrm{lnatem},\mathrm{ lnrain},\mathrm{ lnco}2,\mathrm{ lnagil},\mathrm{ lncari},\mathrm{ lnferz})$$
$$\mathrm{Fy}(\mathrm{lnmaiz}/\mathrm{lnatem},\mathrm{ lnrain},\mathrm{ lnco}2,\mathrm{ lnagil},\mathrm{ lncama},\mathrm{ lnferz})$$

The computed F-statistic is compared to the two sets of critical values (Pesaran et al. 2001). Correspondingly, the lower and upper bound critical values are ascribed to the I(0) and I(1) regressors. If the F-statistic falls below the lower limit, it suggests that there is no cointegration among the variables. If it falls between the upper and lower limits, there is an inconclusive cointegration among the variables. However, there is a cointegration among the variables if the F-statistic exceeds the upper limit critical value.

After cointegration has been proven, the next step is to estimate the predicted values of the long-run relations and draw conclusions regarding their predicted values (Pesaran and Pesaran 1997). The following specifications are the long-term ARDL (\(p\), \({q}_{1},{q}_{2},{q}_{3},{q}_{4}, {q}_{5},{q}_{6}\)) models for rice production and maize production in Eqs. (5) and (6), respectively.

$$\begin{array}{l}{\mathrm{lnRICE}}_{\mathrm t}=\delta_o+\sum\limits_{i=1}^p\Omega_{1\mathrm i}{\mathrm{lnRICE}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_1}\Omega_{2\mathrm i}{\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_2}\Omega_{3\mathrm i}{\mathrm{lnAGIL}}_{\mathrm t-\mathrm i}+\\\sum\limits_{i=1}^{q_3}\Omega_{4\mathrm i}{\mathrm{lnCARI}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_4}\Omega_{5\mathrm i}{\mathrm{lnCO}}_{2\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_5}\Omega_{6\mathrm i}{\mathrm{lnTEM}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_6}\Omega_{6\mathrm i}{\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\varepsilon_{\mathrm t}\end{array}$$
(5)
$$\begin{array}{l}{\mathrm{lnMAIZ}}_{\mathrm t}=\lambda_o+\sum\limits_{i=1}^p\varphi_{1\mathrm i}{\mathrm{lnMAIZ}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_1}\varphi_{2\mathrm i}{\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_2}\varphi_{3\mathrm i}{\mathrm{lnAGIL}}_{\mathrm t-\mathrm i}\\+\sum\limits_{i=1}^{q_3}\varphi_{4\mathrm i}{\mathrm{lnCAMA}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_4}\varphi_{5\mathrm i}{\mathrm{lnCO}}_{2\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_5}\varphi_{6\mathrm i}{\mathrm{lnTEM}}_{\mathrm t-\mathrm i}+\sum\limits_{i=1}^{q_6}\varphi_{6\mathrm i}{\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\varepsilon_{\mathrm t}\end{array}$$
(6)

The study used the Akaike information criterion (AIC) to determine the lag length of the variables in the long-run ARDL models for rice and maize production. The error correction models, which account for the short-run dynamics of rice and maize production, are denoted as follows:

$$\begin{array}{l}{\Delta\mathrm{lnRICE}}_{\mathrm t}=\delta_o+\sum\limits_{1=i}^p\gamma_{1\mathrm i}{\Delta\mathrm{lnRICE}}_{\mathrm t-1}+\sum\limits_{1=i}^{q_1}\gamma_{2\mathrm i}{\Delta\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}+\sum\limits_{1=i}^{q_2}\gamma_{3\mathrm i}{\Delta\mathrm{lnAGIL}}_{\mathrm t-1}+\\\sum\limits_{1=i}^{q_3}\gamma_{4\mathrm i}{\Delta\mathrm{lnCARI}}_{\mathrm t-\mathrm i}+\sum\limits_{1=i}^{q_4}\gamma_{5\mathrm i}{\Delta\mathrm{lnCO}2}_{\mathrm t-1}+\sum\limits_{1=i}^{q_5}\gamma_{6\mathrm i}{\Delta\mathrm{lnTEM}}_{\mathrm t-1}+\sum\limits_{1=i}^{q_6}\gamma_{7\mathrm i}{\Delta\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\psi{\mathrm{ECM}}_{\mathrm t-1}+\varepsilon_{\mathrm t}\end{array}$$
(7)
$$\begin{array}{l}{\Delta\mathrm{lnMAIZ}}_{\mathrm t}=\lambda_o+\sum\limits_{1=i}^p\phi_{1\mathrm i}{\Delta\mathrm{lnMAIZ}}_{t-1}+\sum\limits_{1=i}^{q_1}\phi_{2\mathrm i}{\Delta\mathrm{lnFERZ}}_{\mathrm t-\mathrm i}+\sum\limits_{1=i}^{q_2}\phi_{3\mathrm i}{\Delta\mathrm{lnAGIL}}_{\mathrm t-1}+\sum\limits_{1=i}^{q_3}\phi_{4\mathrm i}{\Delta\mathrm{lnCAMA}}_{\mathrm t-\mathrm i}+\\\sum\limits_{1=i}^{q_4}\phi_{5\mathrm i}{\Delta\mathrm{lnCO}2}_{\mathrm t-1}+\sum\limits_{1=i}^{q_5}\phi_{6\mathrm i}{\Delta\mathrm{lnTEM}}_{\mathrm t-1}+\sum\limits_{1=i}^{q_6}\phi_{7\mathrm i}{\Delta\mathrm{lnRAIN}}_{\mathrm t-\mathrm i}+\psi{\mathrm{ECM}}_{\mathrm t-1}+\varepsilon_{\mathrm t}\end{array}$$
(8)

Here, \(\psi\) connotes the speed of adjustment to attain equilibrium in a dynamic model after a disruption to the system.

Quantile regression

For robustness checks, this study employed the quantile regression (QR) model that accounts for distributional asymmetry in the association between the explained variables (rice and maize production) and explanatory variables (fertilizer consumption, agricultural irrigated lands, cultivated lands, carbon emissions, temperature, and rainfall). The QR approach is impervious to outliers and skewed distributions while estimating the gradient values at different proportion points (quantiles) of pertinent distribution. The QR methodology assists in comprehending (and thus forecasting) the connection between cereal crops and climatic variables based on the propensity of rainfall, temperature, and carbon emissions in the low, medium, and upper quantiles. The quantile regression model is best suited because it offers the said benefits and solves the drawbacks of the linear regression approaches like the ARDL model. The quantile regression is formulated as follows:

$${y}_{\mathrm{it}}={\sigma }_{\mathrm{i}}+{\eta \left(q\right)x}_{\mathrm{it}}+{\varepsilon }_{\mathrm{it}}$$
(9)

Here, \(y\) denotes the vector of explained variable (rice production or maize production), \(x\) stands for the vector of all explanatory variables (fertilizer consumption, agricultural irrigated lands, cultivated lands, carbon emissions, temperature, and rainfall), \(t\) represents time dimension, \(\sigma\) signifies the evidence of fixed effects, and \(q\) indicates the quantile \((0<q<1)\) of the conditional distribution. The impacts of \(x\) factors hinge on the quantile \(q\). According to Koenker (2004), the preceding minimization problem is solved to obtain the results of Eqs. (1) and (2) for many quantiles simultaneously:

$${\mathrm{min}}_{\mathrm{\sigma \eta }}\sum_{k=1}^{\tau }\sum_{j=1}^{n}\sum_{i=1}^{m}{w}_{k}{\rho }_{qk}\left({y}_{ij}-{\sigma }_{i}-{\eta \left({q}_{k}\right)x}_{ij}\right)$$
(10)

For \({\rho }_{qk}=u(q-I\left(u<0\right))\) is the piece-wise linear quantile loss expression presented by Koenker and Bassette (1978). The \(\tau\) quantiles \(({q}_{1}, \dots , {q}_{\tau })\), which are regulated by the weights \({w}_{k}\), have a relative effect on estimating the \({\sigma }_{i}\) parameters. Koenker (2004) proposed a method that involves regularizing or shrinking the effects of a penalty on an expected value. This method, which is known as penalized quantile regression, can be specified as follows:

$${\mathrm{min}}_{\mathrm{\sigma \eta }}\sum_{k=1}^{\tau }\sum_{j=1}^{n}\sum_{i=1}^{m}{w}_{k}{\rho }_{qk}\left({y}_{ij}-{\sigma }_{i}-{\eta \left({q}_{k}\right)x}_{ij}\right)+\lambda p\left(\sigma \right)$$
(11)

where \(p\left(\sigma \right)= \sum_{i=1}^{n}\left|{\sigma }_{i}\right|\) connotes the penalty anticipated.

Empirical results and discussions

Trend analysis of climate variability

This section presents the trend analysis of climatic variability to determine whether there are specific trends in temperature and precipitation patterns in China over the period under study. Panel A of Table 2 presents the statistical test results for monthly, seasonal, and annual temperature trends from 1978 to 2015. The results showed a significant positive trend in average temperature from February to October in China, given the Z statistics of 2.33, 2.34, 3.35, 2.59, 2.07, 2.01, 3.44, 2.55, and 2.66, respectively, which exceed the critical Z value of 1.96. On the other hand, the country experienced a negligible trend in temperature in January, November, and December over the period 1978–2015, considering the Z values of 1.59, 1.82, and 0.41, congruently, which are lesser than the critical Z value of 1.96. Seasonally, the results revealed a profound increasing trend in temperature in spring, summer, and autumn, with the estimated Z statistics of 2.86, 2.29, and 3.13, respectively, which are greater than the critical Z value of 1.96. Nevertheless, there was no significant trend in average temperature in winter from 1978 to 2015. The statistics further indicated a negligible trend in yearly average temperature over the period.

Table 2 Results of statistical tests for temperature and rainfall trends (1978–2015)
Full size table

Examining the trend in rainfall patterns in China is decisive since most Chinese farmers moderately depend on rainfall for crop production (Pickson et al. 2020). Panel B of Table 2 presents the results of statistical tests for monthly, seasonal, and annual rainfall trends spanning the period 1978–2015. Monthly precipitation patterns showed a negative but insignificant trend in January, February, March, April, August, September, and October in China, given the Z values of − 0.42, 0.30, − 1.24, − 0.45, − 0.82, − 0.36, and − 0.20, respectively, which are lesser than the critical Z value of 1.96. Also, monthly precipitation patterns exhibited a positive but negligible trend in May, June, July, November, and December, considering the Z values of 1.77, 1.13, 0.16, 1.71, and 1.72, in that order, which are lesser than the critical Z value of 1.96. Besides, the statistics indicated a positive but insubstantial trend in China’s annual precipitation patterns from 1978 to 2015. In spring, summer, autumn, and winter, there was a negligible trend in precipitation patterns. The results showed that China experienced no significant trend in monthly, seasonal, and annual rainfall patterns over the period under study.

Trend analysis of rice and maize production in China

This section presents the trend analysis of rice and maize production in China (see Fig. 1). Rice production increased from 93,395,906 tons in 1978 to 120,790,867 tons in 1984 but fell to 114,269,687 tons in 1985 and thereafter rose to 127,806,992 tons in 1990. Also, rice production decreased by 3,661,858 tons in 1991, but it later upsurged from 124,145,134 tons to 135,248,819 tons in 1997. Between 1997 and 2003, rice production dwindled by 26,991,864 tons but sharply improved from 108,256,955 tons in 2003 to 142,553,729 tons in 2015. On the other hand, maize production surged from 56,057,169 tons in 1978 to 133,197,612 tons in 1998, but it declined to 106,178,315 tons in 2000. Since then, maize production rose to 265,157,307 tons in 2015. From Fig. 1, maize production between 1978 and 2001 was dismal in contrast with rice production since greater volumes of rice were produced over the same period. However, no similar observations were made between 2002 and 2015. During 2002 and 2015, China experienced more volumes of maize produced than the amount of rice produced.

Fig. 1
figure 1

Trends in rice and maize production in China (1978–2015). Source: Authors’ construction

Full size image

Descriptive statistics

Table 3 presents the descriptive statistics of the concerned variables. The statistics showed that rice production, maize production, and wheat production averaged around 18.6%, 18.5%, and 18.4%, correspondingly, over the period 1978Q1–2015Q4. Rice production recorded a maximum of 18.8% and a minimum of 18.4% over the same period. Maize production showed a minimum of 17.8% and a maximum of 19.4%. Average cultivated areas for rice and maize were 10.3% and 10.2%, as against maximum cropped areas of 10.5% and 10.3% correspondingly. However, the minimum cropped areas stood at 10.2% for rice production and 10% for maize production. Average temperature and rainfall were 1.9% and 3.9%, respectively, during 1978Q1–2015Q4. Maximum rates of average temperature and rainfall were 2.1% and 4%, respectively. The minimum levels of average temperature and rainfall were 1.8% and 3.8% in that order. Carbon emissions averaged around 15.1%, with a maximum of 16.2% and a minimum of 14.2%. Considering fertilizer consumption, the statistics revealed an average of 5.6%, against a maximum rate of 6.1% and a minimum rate of 4.7% over 1978Q1–2015Q4. Eventually, agricultural irrigated lands recorded an average of 10.9%, against a maximum value of 11.1% and a minimum of 10.7%, as presented in Table 3.

Table 3 Descriptive statistics
Full size table

Stationarity and cointegration test results

The Ng-Perron unit root test was used to determine the integration levels of all the variables of interest (i.e., average temperature, average rainfall, carbon emissions, agricultural irrigated lands, fertilizer consumption, maize production, rice production, cropped lands under rice production, and cultivated areas under maize production) in the specified cereal crop models. This unit root test was performed to ensure that no I(2) variables were included in the cereal crop models to avoid misleading regression results.

Table 4 presents the outcomes of the Ng-Perron unit root test. From the stationarity test outcomes using only intercept and trend, almost all the concerned variables did not attain stationarity at the levels apart from maize production, which was stationary at the 1% threshold. Nonetheless, average temperature, average rainfall, carbon emissions, agricultural irrigated lands, fertilizer consumption, rice production, cropped lands under rice production, and cultivated areas under maize production attained stationarity after they were differenced once. The stationarity test outcomes implied that the variables were integrated of zero and one (i.e., I(0) and I(1)). Accordingly, the study proceeded to apply the ARDL technique for the cointegration test coupled with the estimations for short-run and long-run relationships.

Table 4 Results of the Ng-Perron unit root test
Full size table

Table 5 depicts the results of the bounds testing technique for cointegration analysis between the explained variables, say rice and maize production, and their explanatory variables. From the results displayed in Table 5, the estimated F-statistic of the rice production model is 4.970, which exceeds the upper bound critical values of 3.61 and 4.43 at the 5% and 1% significance levels, respectively. This result suggests a long-run relationship between rice production and their explanatory variables. For the maize production model, the estimated Fisher statistic (6.093) exceeds the upper bound critical values of 3.61 and 4.43, implying the rejection of the null proposition of no long-run association at the 5% and 1% thresholds. This outcome means a cointegrating relationship between maize production and its predictor variables. Finally, the study established a long-run association between the explained variables, say rice and maize production, and their explanatory variables.

Table 5 Results of the ARDL bounds test for cointegration
Full size table

Long-run and short-run estimations of the rice production model

This section presents the estimated results for the role of climate change in rice production. The cointegration results are presented in Table 5, whereas Table 6 presents the estimated long-run and short-run ARDL rice production models in China.

Table 6 Estimated long-run and short-run ARDL models of rice production in China
Full size table

Fertilizer consumption positively affected rice production in China in both the long and short runs, which was statistically significant at the 1% level. Thus, as Chinese farmers apply more fertilizers, the quantities of rice produced are expected to increase. Specifically, an increase in fertilizer consumption will cause long-run and short-run rice production to increase by 0.43% and 0.03%, respectively. These results are consistent with Chandio et al. (2021a) and Chandio et al. (2020b), who indicated that increases in fertilizer consumption significantly improve rice production in Pakistan. However, Pickson et al. (2021) showed that fertilizer consumption directly supports long-run rice production, although it has no impact on short-run rice production in China. They further revealed that there exists a varied relationship between fertilizer consumption and short-run rice production across 30 provinces of China.

The results reveal that agricultural irrigated lands exerta positive and significant effect on rice production in China. Thus, as the Chinese government improves irrigation infrastructure, the quantities of rice produced are expected to increase by 1.67% and 0.13% in the long and short run, respectively. This effect of agricultural irrigated lands on rice production is essential for ensuring food self-sufficiency in China. Therefore, given the changing climate with its attendant problem of increasing water scarcity, irrigation systems for agriculture have become more prevalent.

As expected, the cropped areas had a favorable impact on rice production in China, which was statistically significant at the 1% threshold. This outcome implies that an upsurge in the cultivated lands under rice production will cause long-run and short-run rice production to increase by 2.22% and 0.17%, respectively. The study agrees with the studies conducted by Hussain (2012), Chandio et al. (2021a), Ahsan et al. (2020), Sial et al. (2011), Chandio et al. (2020b), Pickson et al. (2020), Chandio et al. (2021b), Abbas (2021), Warsame et al. (2021), and Pickson et al. (2021), who established that cropped lands under cereal production directly affected the cultivation of cereal crops.

Carbon emissions had a significant positive association with long-run and short-run rice production. This result implies that an increase in carbon emissions by 1% causes rice production to rise by 0.29% and 0.02% in the long and short run, respectively. Higher levels of atmospheric carbon dioxide can boost the growth of crops by accelerating the process of photosynthesis and reducing the amount of water lost through transpiration. As the concentration of carbon dioxide increases, the pores of plants close, which lowers the amount of transpiration (Wu and Wang 2000; Hui et al. 2001; Kimball et al. 2002; Srivastava et al. 2002; Prior et al. 2010). While carbon dioxide may positively impact crop production, other studies have warned that excess emanation of carbon dioxide is detrimental. In this study, our objective is not to estimate the quantity of carbon dioxide required for crop growth. As a result, this finding should be interpreted with caution. This study is in line with Janjua et al. (2014), Casemir and Diaw (2018), Ahsan et al. (2020), Chandio et al. (2018), and Chandio et al. (2020b). Nonetheless, Pickson et al. (2020) found that carbon emissions have a detrimental effect on cereal production in China. Besides, Chandio et al. (2021a), Sossou et al. (2019), and Chandio et al. (2021b) found no association between the emanation of carbon dioxide and the production of cereal crops.

As evident from the outcomes displayed in Table 6, rice production in China was not substantially affected by a rise in average temperatures in the long run and short run. If all other factors remain constant, a 1% increase in average temperature increases rice production by 0.39% in the long run and 0.03% in the short run. However, the effect of the average temperature on rice production was not statistically significant in the long run and short run. The results are consistent with Chandio et al. (2020b) and Chandio et al. (2021b), who showed that average temperature has no association with cereal production. Nevertheless, other studies have shown that the average temperature has a positive and significant impact on crop production (Sossou et al. 2019; Pickson et al. 2020; Warsame et al. 2021). According to Pickson et al. (2021) and Abbas (2021), average temperature contributes to cereal production significantly and negatively in the long run but positively in the short run.

The findings further revealed no confirmation that average rainfall improves rice production—the approximate values of average rainfall in the long run and short run were negated and not statistically significant at any of the conventional thresholds. With its coefficients of − 0.2245 and − 0.0170, a 1% increase in average rainfall will influence long-run and short-run rice production to decline by 0.23% and 0.02%, correspondingly, in China. However, the impact of rainfall patterns on rice production was statistically insignificant. The main reason rice production in China was not significantly affected by the rainfall was that the country did not experience significant changes in the annual average rainfall patterns over the period under study. This study gives further credence to Chandio et al. (2021a), Casemir and Diaw(2018), and Chandio et al. (2021b), who found no significant interaction between rainfall and cereal crop cultivation. Contrary to the previous studies by Sossou et al. (2019), Attiaoui and Boufateh (2019), and Pickson et al. (2020), there is a significant and positive relationship between average rainfall and the cultivation of cereal crops in Burkina Faso, Tunisia, and China, correspondingly. Besides, Pickson et al. (2021) indicated that precipitation has a significant and direct association with rice production in the long run.

The speed of adjustment (error correction) coefficient shows the rate at which specific sets of variables converge to equilibrium in the presence of shocks. It should show a significant negative coefficient. This study found that the error correction coefficient (− 0.0757) was statistically significant with its anticipated sign. This result helps reaffirm the existence of the long-run association among the considered variables in the rice production model. The R-squared (0.8956), which is the explanatory power of the rice production model, indicates that about 89.56% of the changes in rice production in China are explained by the changes in temperature, rainfall, carbon emissions, agricultural irrigated lands, cropped lands under rice production, and fertilizer consumption. Furthermore, the study rejected the Fisher test of the joint null proposition that all the predicted values are zero at 1% statistical significance.

Long-run and short-run estimations of the maize production model

This section focuses on the impacts of climate change on maize production. The cointegration results are displayed in Table 5, whereas the estimated long-run and short-run coefficients are shown in Table 7.

Table 7 Estimated long-run and short-run ARDL models of maize production in China
Full size table

Fertilizer consumption positively affects maize production, as chemical fertilizers enhance soil fertility to improve crop production. The results indicated that fertilizer consumption positively impacted maize production in China. It is, therefore, expected that a 1% increase in fertilizer consumption will lead to 0.27% and 0.30% increase in long-run and short-run maize production, respectively. Previous studies by Ammani et al. (2012) and Wu et al. (2021) suggested that an increment in fertilizer consumption significantly improves the quantities derived from maize farming in Nigeria (Kaduna State) and China, respectively. Interestingly, Abbas (2021) indicated that fertilizer consumption directly correlates with long-run cereal production, but it has no impact on short-run cereal production in Pakistan.

Agricultural irrigated lands showed a positive sign, suggesting that developing and improving irrigation infrastructure will significantly help enhance long-run and short-run maize production in China. This study indicates that maize production will increase by 0.83% in the long run and 0.16% in the short run, resulting from a 1% increase in agricultural irrigated lands in China. The implication of the results is that, to maintain continuous and productive agricultural production, irrigation systems are needed. These systems can help enhance crop water use efficiency (Abdoulaye et al. 2021). Although maize can withstand high temperatures (Wu et al. 2021), improving irrigation is an efficient adaptation strategy to reduce the susceptibility of maize to the changing climate (Zhou and Turvey 2014; Kang et al. 2017).

The study established that cropped lands under maize production positively contributed to maize production in both the long and short run at the 1% significance level. The outcome means that increased cultivated areas can increase maize production in China. The results indicated that a 1% change in cropped lands would result in a 0.61% and 1.10% change in long-run and short-run maize production, respectively, in China, ceteris paribus. These observations support the findings from Ahsan et al. (2020), Sial et al. (2011), Chandio et al. (2020b), Pickson et al. (2020), Chandio et al. (2021b), Abbas (2021), Warsame et al. (2021), Pickson et al. (2021), and Chandio et al. (2021c). However, Khan et al. (2019) showed that cultivated lands under cereal production did not influence the cultivation of cereal crops in Pakistan.

The results indicated that carbon emissions had a significant positive impact on maize production in China. Thus, an increase in carbon emissions significantly improves long-run and short-run maize production in China. With coefficients of 0.3736 and 0.0703, a 1% increase in carbon emissions leads to 0.37% and 0.07% upsurge in long-run and short-run maize production, respectively, in China. These results agree with Ammani et al. (2012), who indicated that carbon emissions positively impacted the quantities of maize produced in Nigeria (Kaduna State).

However, the results contrast with the finding obtained by Pickson et al. (2020) for China, where carbon emissions had a significant adverse effect on cereal production. Meanwhile, Ahsan et al. (2020) and Chandio et al. (2021c) reported that carbon emissions had no substantial impact on the cultivation of cereal crops in Pakistan and India, respectively.

The study unearthed that average temperature has a substantial positive impact on maize production in China in both the long run and short run. This result implies that a 10% rise in average temperature raises the long-run and short-run maize production by 3.95% and 3.77%, correspondingly, in China. Thus, the average temperature has a favorable effect on maize production in China. Conversely, empirical evidence suggests that average temperature harms maize production (Xu et al. 2016; Luhunga 2017). Also, the findings differ from Zhou and Turvey (2014), Pickson et al. (2020), Chen et al. (2016), Sossou et al. (2019), Wei et al. (2014), Liu et al. (2020a, b), Warsame et al. (2021), Wu et al. (2021), and Chandio et al. (2021c). They observed a negative impact of temperature on the cultivation of cereal crops.

Additionally, the results showed that average rainfall had a significant positive relationship with maize production in China in the long and short run. The estimated coefficients associated with average rainfall (0.7004 and 0.3933) for maize production in China suggest that a 10% increase in rainfall causes maize production to increase by 7.0% and 3.93% in the long and short run, correspondingly. All the other things being equal, the statistical significance of average rainfall means that rainfall significantly influences the long-run and short-run maize production in China. These results confirm the studies conducted by Ammani et al. (2012), Khan et al. (2019), and Wu et al. (2021), which found that average rainfall had a significant favorable influence on the quantities of maize produced in Nigeria (Kaduna State), Pakistan, and China, correspondingly. The results are contrary to the empirical evidence obtained by Chandio et al. (2021c), which suggested that average rainfall had a negligible impact on cereal production in India.

The speed of adjustment coefficient is a statistical indicator that shows how quickly various factors can converge to equilibrium in the case of any disruptions. It should have a negative sign if it suggests a statistically significant coefficient. As indicated in Table 7, the speed of adjustment coefficient is statistically significant at the 1% threshold, implying a long-run association among the considered variables in the maize production model. With its coefficient of 0.1882, maize production disruptions caused by climatic variables (temperature, rainfall, and carbon emissions), agricultural irrigated lands, cropped lands under maize production, and fertilizer consumption are corrected within 5 years (1/0.1882). The R-squared is 0.7741, suggesting that changes in the explanatory variables (i.e., temperature, rainfall, carbon emissions, agricultural irrigated lands, cropped lands under maize production, and fertilizer consumption) explain approximately 77.41% of the variations in the dependent variable (maize production). The Fisher statistic of 17.1908 indicates that all the predicted values are jointly and statistically significant at the 1% threshold.

Results of the diagnostic and stability tests

The ARDL models were subjected to various statistical tests to establish their correctness. The tests include diagnostic tests (i.e., normality, autocorrelation, heteroscedasticity, and functionality tests) and stability tests (i.e., the cumulative sum (CUSUM) and the cumulative sum of squares (CUSUMSQ) tests). The CUSUM and CUSUMSQ tests examine the stability of the long-run coefficients and short-run dynamics. Thus, the CUSUM and CUSUMSQ tests determine the strength of the regression coefficients (Bahmani-Oskooee 2001). Figures 2, 3, 4, and 5 illustrate that both the CUSUM and CUSUMSQ plots are within the 95% confidence interval, indicating that the parameters of the rice and maize production models do not exhibit structural instability over the research period. The CUSUM and CUSUMSQ tests affirm the steadiness of the long-run parameters of the rice and maize production models. Table 8 presents the diagnostic and stability tests. The diagnostic tests revealed that the rice and maize production models are homoscedastic, not serially correlated, regularly distributed, and correctly specified, as none of the tests yielded statistical significance.

Fig. 2
figure 2

CUSUM test (rice production model)

Full size image
Fig. 3
figure 3

CUSUMSQ test (rice production model

Full size image
Fig. 4
figure 4

CUSUM test (maize production model)

Full size image
Fig. 5
figure 5

CUSUMSQ test (maize production model)

Full size image
Table 8 Results of the ARDL model diagnostic and stability tests
Full size table

Robustness checks using quantile regression

The baseline ARDL results in Tables 6 and 7 show that the climatic variables have varied associations with China’s cereal crop production (i.e., rice and maize). The ARDL model only accounts for linear associations between the explained and explanatory variables. For robustness checks, this study employed the quantile regression (QR) model that accounts for distributional asymmetry in the association between the explained variables (rice and maize production) and explanatory variables (fertilizer consumption, agricultural irrigated lands, cultivated lands, carbon emissions, temperature, and rainfall). The QR approach is impervious to outliers and skewed distributions while estimating the gradient values at different proportion points (quantiles) of pertinent distribution. The QR methodology assists in comprehending (and thus forecasting) the connection between cereal crops and climatic variables based on the propensity of rainfall and temperature in the low, medium, and upper quantiles. The QR results in Table 9 indicate the robustness of the results against the acute value of the explained variable. The effect of fertilizer consumption on rice production was significant, with a positive sign in all quantiles. The findings suggest that an increase in fertilizer consumption has a favorable effect on rice production in China at the 0.25, 0.50, 0.75, and 0.90 quantiles of the conditional distribution.

Table 9 Quantile regression (QR) estimations
Full size table

Similarly, fertilizer consumption positively affected maize production at the 0.25, 0.50, 0.75, and 0.90 quantiles. Besides, this study established that the effect of agricultural irrigated lands on rice production was positive and significant in China at the 0.25, 0.75, and 0.90 quantiles. Likewise, agricultural irrigated lands were observed to positively influence maize production in China across all quantiles (\(\tau =0.25, \tau =0.50, \tau =0.75, \mathrm{and} \tau =0.90)\). The favorable influence of agricultural irrigated lands on rice and maize production implies that the Chinese government should improve the irrigation infrastructure across the various provinces and autonomous regions to intensify the production of cereal crops, say rice, and maize. The study further showed that the cultivated lands under rice production have a significant positive association with rice production in China at all quantiles. In the case of the maize production model, the effect of the cropped lands under maize production on maize production was significantly positive in China in all quantiles (\(\tau =0.25, \tau =0.50, \tau =0.75, \mathrm{and} \tau =0.90)\).

The results revealed that the effect of carbon emissions on rice production was significantly deleterious in China for \(\tau =0.75 \mathrm{and }\tau =0.90\). However, there was a palpable and positive relationship between carbon emissions and maize production in China in all quantiles (\(\tau =0.25, \tau =0.50, \tau =0.75, \mathrm{and} \tau =0.90)\). The study found that average temperature and rainfall had no palpable influence on rice cultivation in China across any of the quantiles. The findings imply that temperature and rainfall patterns have no impact on rice production. Conversely, the average temperature had a significant effect on maize production in China at the lower (\(\tau =0.25\)) and middle (\(\tau =0.50\)) quantiles. The outcome implies how maize production responds to low and moderate temperatures in China. Also, average rainfall was positively related to maize production at the lower, middle, and upper quantiles. This result indicates how maize production responds to increasing rainfall patterns in China. Figures 6 and 7 depict how the effects of the explanatory variables on rice and maize production differ considerably from the linear regression estimates, with a 95% confidence interval. These figures show the dynamic behavior of the parameters (i.e., estimate coefficients) as a function of quantiles for rice and maize production in China.

Fig. 6
figure 6

Plots of quantile parameter estimates for the rice production model

Full size image
Fig. 7
figure 7

Plots of quantile parameter estimates for the maize production model

Full size image

Conclusions and policy implications

This study assessed the impacts of climate change on rice and maize production in China over 1978Q1–2015Q4 using different analytical techniques. It first conducted a nonparametric analysis using the Mann–Kendall test and Sen’s slope estimator to ascertain the actual trend and the degree of the observed trend in terms of climatic variables. The results showed a significant positive trend in average temperature from February to October in China. Seasonally, the results revealed a profound increasing trend in temperature in spring, summer, and autumn. However, China experienced no significant trend in monthly, seasonal, and annual rainfall patterns over the study period. The estimated results of the autoregressive distributed lag (ARDL) model showed that fertilizer consumption, agricultural irrigated lands, cultivated areas, and carbon emissions positively affected maize and rice production in China in both the long and short run. The findings also indicated that average temperature and rainfall had no significant impact on rice production in the long and short run. But in both periods, average temperature and rainfall positively and substantially improved maize production. For robustness checks, this study further employed the quantile regression (QR) model that accounts for distributional asymmetry in the association between the explained variables and explanatory variables. The QR results were not different from the ARDL results.

This paper has some policy implications. The Chinese government should promote the cultivation of more lands and increase the areas under rice and maize cultivation to boost food production and self-reliance. Increasing cropped lands helps achieve higher crop productivity and self-reliance in food production. Given the scarcity of freshwater resources in China, the government should encourage rainwater harvesting and construct water towers to improve the management of the water resources for agricultural purposes. Besides, advanced irrigation systems should be developed to allow farmers to use water efficiently and minimize the effects of climate change on crop production. This strategy will assist in mitigating the impact of high temperatures on crops by ensuring that water is available when needed. Climate change-related information should also be disseminated by establishing effective extension services so that farmers can adapt to the likely effects of climate change on crop production.