Assessing the long- and short-run asymmetrical effects of climate change on rice production: empirical evidence from India

In recent years, environmental change has arisen as a ubiquitous problem and gained environmentalist’s attention across the globe due to its long-term harmful effects on agricultural production, food supply, water supply, and livelihoods of rural households. The present study aims to explore the asymmetrical dynamic relationship between climate change and rice production with other explanatory variables. Based on the time series data of India, covering the period 1991–2018, the current study applied the nonlinear autoregressive distributed lag (NARDL) model and Granger causality approach. The results of the NARDL reveal that mean temperature negatively affects rice production in the long run while positively affecting it in the short run. Furthermore, positive shocks in rainfall and carbon emission have negative and significant impacts on rice production in the long and short run. In comparison, negative rainfall shocks significantly affect rice production in the long and short run. Wald test confirms the asymmetrical relationship between climate change and rice production. The Granger causality test shows feedback effect among mean temperature, decreasing rainfall, increasing carbon emission, and rice production. While no causal relationship between increasing temperature and decreasing carbon emission. Based on the empirical investigations, some critical policy implications emerged. Toward sustainable rice production in India, there is a need to improve irrigation infrastructure through increasing public investment and to develop climate-resilient seeds varieties to cope with climate change. Along with, at the district level government should provide proper training to farmers regarding the usage of pesticides, the proper amount of fertilizers, and irrigation systems.


Introduction
Due to the long-term adverse effects on agricultural productivity, food production, water availability, and rural lives, climate change has garnered environmentalist and policymaker attention across the globe since the 1990s (Chavas et al. 2009;Mohorji et al.2017). Changes in the long-term trends in mean temperature and shifting rainfall patterns, increasing variability, and greater prevalence of extreme events are the facet of climate change. Shifting rainfall patterns may exert a more substantial effect on rice production (Abbas 2021). However, frequent floods due to heavy rainfall may result in higher rice yield losses under climate change (Wassmann et al. 2009;Ozdemir 2021). Climate change results from increasing human activities on the land, including deforestation, land use, urbanization, increasing population, production, and consumption activities to fulfill people's demand for food supply. Climate steadily changes due to global temperature, precipitation, and carbon emission, significantly impacting cereal production and agricultural productivity (Chandio et al. 2021a, b;Klutse et al. 2021;Kumar et al. 2021a).
Agricultural productivity has decreased due to climate change's main drivers, such as precipitation and warmer temperature (Haile et al. 2017). However, increase in temperature, variation in rainfall, and frequent floods and droughts are mostly faced by the developing nations, situated in the tropical regions, and relies heavily on the agricultural sector (Janjua et al. 2014). Agriculture and its allied activities are sensitive to climate change, and on the other hand, it is also contributed to carbon emission (Swaminathan and Kesavan 2012). Climate change is harmful to agriculture production and enhances the vulnerability among small and medium farmers whose livelihoods are mainly dependent on agricultural and allied activities (Zakaria et al. 2020). Climate change's impact may vary from region to region based on geographic allocation. In the case of a developing nation, climate change deteriorates the performance of the agricultural sector (Abbas 2020;Janjua et al.2014;Nath and Behera 2011). Likewise,  revealed that climate change has significantly affected crop production and food security in South Asia in the long run. Swaminathan and Kesavan (2012) stated that climate change adversely affected food production. The developing nations are more vulnerable than developed countries due to more extensive dependence on the agricultural sector for livelihood, lack of technological advancement, and lack of adaptation policies of climate change (Dogan and Inglesi-Lotz 2020;Praveen and Sharma 2020;Warsame et al. 2021). However, Ahmad et al. (2020) stated that technological advancement and sustainable investment play a decisive role in coping with climate change. According to Chandio et al. (2021a, b), temperature negatively affected cereal production, whereas financial development positively improved cereal production in the case of Pakistan. Likewise, Ahsan et al. (2020) demonstrated that energy consumption, agricultural labor force, cultivated area, and CO 2 are the main determinants of agriculture productivity. Similarly, Warsame et al. (2021) explained that mean temperature and CO 2 negatively influenced agriculture productivity in the context of Somalia. In addition, Coulibaly et al. (2020) concluded that temperature and drought are the main factors that negatively affect agriculture productivity. Increasing carbon emission leads to a cascade of impact mechanisms that have harmful and beneficial effects on rice production.
The world's total rice production about 90% is mainly produced by Asian countries (FAO 2019). However, India is the first largest exportable country of rice in the world counted 9.8 million tonnes, followed by Thailand (7.5 million tonnes), Vietnam (6.5 million tonnes), Pakistan (4.6 million tonnes), and the USA (3.1 million tonnes). India is the second rice producer country in Asia after China, followed by Indonesia, Bangladesh, and Vietnam (Fig. 1). The Indian agricultural sector is the most sensitive and exposed area to climate change due to its less adaptive capacity to cope with the adverse effects of climate change (Guntukula 2020). Investigating the impacts of climate change on agricultural productivity is of immense importance because more than 50% of India's population primarily depends on agricultural activities for their livelihoods (Pattanayak and Kumar 2013). Changes in environmental factors such as temperature, precipitation, CO 2 , and rainfall pattern directly affects agricultural productivity (Res et al. 1998). Increasing carbon emission and global warming created challenges for the countries to cope with it through different strategies and policies (Alharthi et al. 2021). Therefore, it is indispensable to examine the effect of changes in climatic conditions on rice production. More than 60% of the population in India mainly depends on agriculture and its allied sectors (Baig et al. 2021). The trend of rice production and area under crop is shown in Fig. 2. Rice output grew from 74.7 million tonne (MT) in 1991 to 116.5 (MT) in 2018. Simultaneously, the cultivated rice area in India has increased from 42.7 million hectare (MH) in 1991 to 44.2 (MH) in 2018. The area under rice has risen by around 1.5 times, but rice production has increased by more than five times. Climate change may be the effect of food security by hampering agricultural productivity from one-way and multiple ways.
Climate change, on the other hand, has a global impact, and its negative consequences are projected to be more severe in India's agro-ecological zones. Climate models predict the severe impacts of climate change on the agricultural sector (Bahl 2015). Climate change significantly affected agricultural productivity and food supply, threatening food security (Moses et al. 2015). Because rice is more vulnerable to fluctuation due to climate change and its associated components, the rising negative effects of climatic change would put pressure on agricultural yield (Bahl 2015). Given rice's vulnerability to environmental change, particularly those connected to temperature increases and extended drought spells, meeting future global rice demand appears to be a difficult undertaking. Temperature-related changes in the duration of the growing season will reduce rice yield and shift farming frameworks away from rice and toward crops with greater temperature optimums (Korres et al. 2017).
The present study explores the nonlinear long-term and short-term effects of climate change on rice production in the context of India, spanning from 1991 to 2018. Previous studies applied crop simulation models (Gupta and Mishra 2019;Kumar 2011;Kumar et al. 2011;Lal et al. 1998;Mishra and Chandra 2016;Mukherjee and Huda 2018), linear econometric models (Baig et al. 2020;Bhanumurthy and Kumar 2018;Birthal et al. 2014;Guntukula 2020;Kumar et al. 2020;Nath and Mandal 2018;Praveen and Sharma 2020;Kumar et al. 2021b), and nonlinear models (Mitra 2014;Pal and Mitra 2018) to assess the impacts of climate change on the Indian agricultural production. Numerous studies examined the effects of climate change on rice yield by using linear regression analysis. As a result, these studies have produced only linear effects of climate change on rice production that might lack nonlinear effects. The current study adds to the existing literature by addressing the asymmetric long-term and short-term impacts of climate change on rice production in the case of India rather than sticking to a linear approach. Furthermore, in this study, we incorporated other important variables such as agricultural rural labor, agricultural credit, usage of fertilizers, and cultivated land in the model to examine the long-term and short-term impacts of these factors on rice production. Previous studies concluded that average precipitation, mean temperature, and carbon emission are the main climatic factors Ahsan et al. 2020) and agricultural credit, fertilizer usage, cultivated area, and rural population are non-climatic factors Rahman et al., 2017) of rice production. The dynamic nexus between study variables are represented in Fig. 3.
It is essential to investigate the asymmetrical implications, as it helps to understand whether positive and negative shocks dominate rice production in India. In this manner, the current work adopts a more comprehensive understanding. Also, it provides the main factors of rice production for India, which will help to formulate economic policies to cope with climate change and enhance rice production in India and other countries with the same agriculture profile.
The remainder of the paper is framed as follows: "Literature review" section deals with the existing literature. The data and technique are discussed in "Data and methodology" section. "Results and discussion" section presents the empirical findings and comments, while "Conclusion and policy implications" section concludes with policy implications.

Literature review
Numerous studies have examined the impacts of climate change on crop production and agricultural productivity in different regions of the world. There is growing consensus among environmentalists and researchers that a negative relationship exists between climate change and agricultural productivity in developing nations (Khanal et al. 2018). South Asia is the most susceptible terrain to climate change globally, with the largest population growth, poverty, and insecurity. Climate change, Fig. 3 The dynamic nexus between the study variables such as extreme weather, unexpected rainfall, and temperature fluctuations, severely affected agricultural production in developing nations (Masud et al. 2014;Shabbir et al. 2020). However, it is the primary concern to frame a suitable policy to tackle climate change problems for policymakers, researchers, and government organizations. At the global, regional level, researchers have undertaken numerous studies to assess the impacts of climate change on the agriculture sector (Chandio et al. 2020a, b, c, d, e;Praveen and Sharma 2020;Warsame et al. 2021).
Among previous studies conducted by Gupta and Mishra (2019) at the country level and Kumar et al. (2020) at the states level, i.e., Uttar Pradesh and Haryana respectively applied the Crop Simulation Model (CSM) and Ricardian regression approach to assess the nature of the relationship between climate change and rice productivity. According to Gupta and Mishra (2019), the multi-Global Climate Model predicts an increase in rice productivity in the most agro-ecological zones in representative concentration pathways (RCP) 2.6. Guiteras (2009) explained that major crop yield would harmfully be affected by 4.5 to 9% due to climate variation from 2010 to 2039 in India. In the same order, the crop would reduce up to 25% in the absence of adaptation productivity. Kumar et al. (2020) found that any large deviation in the rainfall harms rice and wheat production in Uttar Pradesh.
On the other hand, maximum temperature has a negative impact on rice and wheat in Uttar Pradesh and Haryana. While rising temperatures have a positive effect on rice production, they have a detrimental effect on grain. Abbas and Mayo (2021) reported that maximum temperature harms rice plants. Rice crop at the replantation stage during the vegetative phase has benefited from a decrease in the number of plants in the plantation stage and a lower minimum temperature. During the heading and flowering periods, rain has a harmful impact on rice crop. Likewise, Auffhammer et al. (2012) point out that heavy rainfall and drought have a negative effect on rice yield in the rain-fed areas during the 1966-2002 period, and lower rainfall and warmer night would not occur then rice yield would increase by 4% in India. In contrast, Rayamajhee et al. (2020) stated that there is no direct relationship between rainfall and rice production in Nepal. Likewise, Chandio et al. (2021a, b) employed the ARDL cointegration approach to investigate the impacts of climatic factors (CO 2 , average temperature, and precipitation), technological advancement (consumption of fertilizer used as a proxy variable), and other controlled variables such as the area under cultivated land, improves seed, and agriculture credit on rice production. They stated that average temperature and precipitation positively influenced rice production, while CO 2 has a significant and negative impact on rice production in Nepal. Furthermore, agriculture credit and area under cultivated land have a positive effect on rice production.
In the context of China, Pickson et al. (2021) explored the relationship between climate change and rice production using panel data spanning 1998-2017. The long-run and short-run effects of climate change on rice production were investigated using pooled mean group methodologies. Rice production positively influenced by average rainfall, while rice production negatively influenced by average temperature. Furthermore, findings reveal that in the long run, rice production positively influenced by cultivated area and fertilizer consumption. Additionally, the causality test shows that cultivated land and rice production have a bidirectional connection.
Similarly, Jan et al. (2021) investigated the impact of climate change on cereal crops, namely wheat and maize, in the Khyber Pakhtunkhwa (KP) province of Pakistan using panel data from 1986 to 2015. The results indicated that precipitation has a significant and positive impact on wheat and maize yield in the long and short run. In the short run, minimum temperature has a large beneficial effect on maize yield but has no effect on wheat output, according to the estimated results. Maximum temperature, on the other hand, has a detrimental impact on wheat and maize yields while having a beneficial impact on crop output in the short term.
Attiaoui and Boufateh (2019) and Abbas (2020) find a linear long-run dynamic relationship between climate change and agriculture productivity. Empirical results reveal that deficiency of rainfall and high temperature respectively negatively and positively affected agriculture productivity. Baig et al. (2020) also employ a linear dynamic ARDL model to assess the impact of climate change on the yield of major crops, including rice, wheat, coarse cereals, and pulse in India. Findings showed that temperature positively impacts wheat, coarse grains, and pulse except for rice. At the same time, rainfall has a positive impact on rice, coarse cereals, and pulse, except for wheat in India. In contrast, Mitra (2014) and Pal and Mitra (2018) investigated the nonlinear relationship between climate change and crop productivity in India. Mitra (2014) found no asymmetric relationship between rainfall and food grain in India and observed that average rainfall has a greater impact on food grain production than below-average rain. In contrast, Pal and Mitra (2018) explained that rainfall has a greater effect on food grain production up to 75 th quantile and reduces after that in India. While Nsabimana and Habimana (2017) conducted a study in Rwanda's context, they stated that rainfall has an asymmetric impact on crop prices in the short and long run.
Furthermore, the price of food crops has decreased during the harvest season and then increased. Likewise, Moore et al. (2017) used database yield to compare results from processbased and empirical studies in order to comprehensively investigate the influence of climate change on agricultural production and welfare. He claims that the asymmetric impacts of climate change on welfare and agricultural yield show a high possibility of severe welfare losses with warming of 2-3 °C, even after accounting for the CO 2 fertilization effect. Fezzi and Bateman (2016) and Kabubo-mariara and Karanja (2007) observed a nonlinear relationship between climate change and the revenue of agriculture crops. Therefore, it is challenging to cope with it due to the complex asymmetrical association between climate change and agriculture production. Table 1 shows a summary of the review of the literature.

Data and methodology
In this study, we explore the asymmetrical causal relationship between climate change and rice production in the case of India using times series data from 1991 to 2018. The data is obtained from different sources, including Reserve Bank of India (RBI), World Development Indicators (WDI), and the Climate Change Knowledge Portal (CCKP) ( Table 2). Figure 4 represents the trend of the scrutinized variables used in the analysis.
This study undertakes rice production (MT) as a dependent variable, mean temperature (°C), average rainfall (mm), carbon emission (kt), rural population (% of the total population), consumption of fertilizer (kg/ha), agriculture credit (billions), and area under crops (MH) used as independents variables. Annual mean temperature, annual average rainfall, and carbon emission are the main factors of climate change (Chandio et al. 2020a, b, c, d, e;Kumar et al. 2021a;Pickson et al. 2021). Chandio et al. (2021a, b), Pickson et al. (2021), and Warsame et al. (2021) also incorporated agriculture credit, consumption of fertilizer, rural population, and area under crops as non-climate factors of agriculture production. All the variables were transformed into logarithmic.

NARDL bound test for cointegration
This study employs the recently developed and advanced technique NARDL to investigate the asymmetrical effect of climate change on the production of rice. The ARDL technique ignored nonlinearity and the asymmetrical association between the underlying variables (Kumar et al. (2021b). The ARDL model is expanded to an asymmetric ARDL or NARDL by Shin et al. (2014) to assess the pattern of dynamic adjustment and asymmetries relationship in the short and long run between the variables. To explore the relationship between the variables following model can be specified as: We can rewrite Eq. (1) as follows: (1) (2) lnPR t = 0 + 1 lnAT t + 2 lnRF t + 3 lnCO 2t + 4 RP t + 5 lnAC t + 6 lnF t + 7 lnAUR t + t where lnPR is the natural log of rice production, lnAT is the natural log of mean temperature, lnRF is the natural log of average rainfall, lnCO 2 is the natural log of carbon emission, RP is rural population, lnAC is the natural log of agricultural credit, lnF is the natural log of consumption of fertilizer, and lnAUR indicates natural log of the area under rice crop. Before presenting a full depiction of the NARDL model, general forms of long-run asymmetry relationships are given as follows: where lnPR t is a k × 1 vector of rice production at time t, and ( 0 , + 1 , − 2 , + 3 , − 4 , + 5 , − 6 , + 7 , − 8 , 9 , 10 and 11 ) are the associated asymmetric long-run parameters. Here, lnAT t , lnRF t , lnCO 2t , and RP t, as k × 1 vector of regressors is subdivided as; RP − t are partial sum processes of positive ( +) and negative (-) changes in lnAT t , lnRF t , lnCO 2t , RP t respectively. Equation shows partial decomposition of lnAT, lnRF, lnCO 2 , and RP. Shin et al. (2014) prolong ARDL model adopted (Pesaran et al. 2001) by utilizing the concept of cumulative positive (3)  (2019) 1968-2014 Pakistan ARDL + CO 2 , Avg. Temperature, Area under cultivation--> + Rice production both in short and long run + Fertilizers--> + Rice production in long run but -Rice production in short run 2 Chandio et al. (2021a, b) 1980-2016 Turkey ARDL + CO 2 -> -Rice Production both in short and long run + Temperature, Precipitation, Area harvested of rice--> + Rice production both in short and long run + Domestic Credit--> -Rice production in long run but + Rice Production in short run 3 Yuliawan and Handoko (2016) 1973-2018 South Korea ARDL + CO 2 , Mean Temperature, Area under rice--> + Rice production both in long and short run + Rainfall--> -Rice production both in long and short run + Fertilizer--> + Rice production in long run but has no impact in short run 12 Chandio et al. (2020a, b, c, d, e)  + Temperature--> + Rice production initially but harmful beyond a certain optimal temperature + Precipitation does not harm rice productivity and negative partials sums. In this manner, the NARDL model proposed by Shin et al. (2014) represents asymmetric error correction form is specified as:  In the above equation, ( i ) , indicates long-run coefficients, while ( i ), ( i ), ( i ), ( i ), and( i ) are the short-run coefficients. The NARDL's estimation method is the same as linear ARDL. The null hypothesis of asymmetrical long-run relationship, = + = − = 0 between the variables. Null hypotheses have been tested by computing the general F-statistics F PSS or t-statistics ( t BDM ) proposed by Banerjee et al. (1998) determined these values by comparing them to the two critical bounds (lower and upper bound), which define a band including all conceivable classifications of the regressors as solely I (0), I (1), or mutually cointegrated. We accept the null hypothesis if the F-statistics are less than the lower bound value, i.e., I(0). We can infer that there is no long-run association between the variables. If the F-statistics are in the range I(0) to I(1), the outcome is inconclusive. If the F value is greater than the I(1) bound value, the null hypothesis can be rejected, indicating that variables are long-run cointegrated. ECT -1 is the error correction term, and is the rate at which the asymmetrical long-run equilibrium relationship is restored following a disruption.
T h e l o n g -r u n ( + = − ) a n d s h o r t -r u n ( + 1 = − 2 , + 1 = − 2 , + i = − i , + 1 = − 2 ) asymmetries estimates through the Wald test for mean temperature (lnAT), average rainfall (lnRF), carbon emission (lnCO 2 ), and rural population (RP) variables. Where p and q are representing optimal lags order of dependent and independent variables, respectively. The Akaike and Schwarz information criteria have been used to find out the optimal lag selection in the model. The long-term asymmetric coefficients are calculated based on L mi + = + ∕ andL mi − = − ∕ . These longrun coefficients measure the connection between variables in the long-run equilibrium with respect to independent variable shocks. By utilizing the cumulative dynamic multiplier effect, these long-run and short-run asymmetry trajectories can be described in the following ways: a unit percentage change in X + t andX − t on Y t are obtained through the following equation: where if h → ∞ , then m + h → L mi + andm − h → L mi − . The adequacy and stability of the specified NARDL models are also checked with various diagnostic tests. The results of correlation analysis are reported in Table 4, which indicate that all the variables are positively correlated

Results and discussion
with production of rice except rural population which is negatively correlated. The next step is to check the stationarity of the underlying variables to guarantee that none of them are integrated at order I(2). Because the NARDL model requires that variables be integrated at the order I(0) or I(1) to investigate cointegration among variables, a unit root test must be performed. We used the augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests in this order, and the results are shown in Table 5. We can infer from Table 5 that mean temperature, average rainfall, rural population, and  land area under rice crop are I(0), while rice production, carbon emission and agriculture credit series are I(1). By neglecting structural breakdowns in the data, common unit root tests such as the ADF and PP allow results to be misled. To address this issue, we employ the Zivot and Andrews (2002) test. The results of the Zivot and Andrews (2002) test are shown in Table 6, which reveal that rice output, mean temperature, average rainfall, fertilizer usage, and area under rice crop are integrated at order I(0). In contrast, carbon emission, agricultural credit, and rural population are stationary after being first differenced with different structural breaks in the series. Due to the drought in 2002 in India, agricultural productivity sharply declined (Gulati et al. 2013). Hence, the structural break has arisen in the data of rice production. Due to the presence of structural breaks in the data, the variables may have nonlinearity. As a result, to check for nonlinearity, we use the BDS independence test, which checks for the presence of linear dependency in the dependent variable in the model. The BDS test for nonlinearity in the residual of the dynamic relationship is performed. The results of the BDS are reported in Table 7, which indicate that all the variables are not identically and independently distributed (iid) except mean temperature and average rainfall. The BDS statistics show the null hypothesis of residual of being independent and identically residual also is rejected at 1% level of significance of rice production at all the dimension. After confirming the nonlinearity in the series, we move toward the estimation of the NARDL model.

NARDL cointegration results
In the present study, Schwarz (1978) information criterion is used to choose the optimal lag length of the NARDL (p,q). Then we applied general to specific approach by ignoring all insignificant regressors since their inclusion may produce imprecise estimation results. Table 8 delineated the asymmetric impacts of climate change and other controlled agriculture inputs on rice production. Two operational testings are used for the existence of an asymmetrical cointegration relationship based on the NARDL. We find that the F-statistics are greater than the critical upper bound value at the 1% level of significance, confirming the presence of cointegration between mean temperature, average rainfall, carbon emission, rural population, agricultural credit, fertilizer consumption, area under rice crop, and rice production from 1991 to 2018. The Wald test highlights the importance of asymmetry in both the short and long run, implying that nonlinearity must be considered when researching the relationship between climate change and rice output. At a 1% level of significance, the t-statistics support the cointegration among the variables. It means that in India, mean temperature, average rainfall, carbon emissions, agricultural finance, fertilizer usage, rice crop area, and rice production have a long-term asymmetric relationship.

Long-and short-run asymmetric estimates
A positive and negative component in mean temperature has negative and significant impacts on rice production, which represent that any positive and negative shock in mean temperature deteriorates rice production. However, the sign of both coefficients are the same but different in magnitude, which indicates that mean temperature has a significant asymmetric impact on rice production. This study is in line with previous studies (Chandio et al. 2020a, b, c, d, e;Haris et al. 2013;lal et al. 1998;Matthews et al. 1997;Warsame et al. 2021;Yuliawan and Handoko 2016), which corroborates the same findings. Chandio et al. (2020a, b, c, d, e), Matthews et al. (1997), and Warsame et al. (2021) reported that temperature has an adverse effect on rice production both in the short and long run. For instance, increases  (decreases) of 1% in temperature reduce rice production by 9.23 (10.32) percent in the long run in India. Several reasons can support this finding; increasing mean temperature is beneficial for rice production initially. However, beyond a certain optimal temperature, further temperature increases become harmful for rice production. Second, temperature rise would make the age of rice shorter and decrease the rice yield (Kumar et al. 2021a). Higher temperature increases the sea level; consequently, highly productive rice cultivation areas will be more exposed to inundation and salinity intrusion. Moreover, the increased mean temperature has adversely impacted rice production in various parts of South Asia such as India, Bangladesh, Sri Lanka, and Pakistan, which results in reduced average yields by 4% (Matthews et al. 1997). Table 8 reported the results of the long-and short-run asymmetrical impacts on rice production. Estimated outcomes in the long-run indicate that positive shock in the rainfall has negative and significant effects on rice production at a 1% level in India. The estimated coefficients of positive shock in average rainfall indicate that a 1% rise in average rainfall leads to a decrease of 1.24% of rice production in India. These findings are supported by the previous study Nasrullah et al. 2021), which stated that excess rainfall negatively influenced rice production in rain-fed areas. Rice production has tremendous pressure due to the high variability of rainfall in rain-fed regions of India (Pal and Mitra 2018). However, heavy rainfall, i.e., the floodlike situation, adversely affected rice production in India (Pal and Mitra 2018). Some previous studies Chandio et al. 2021a, b;Siddiq et al. 2012;Warsame et al. 2021) contradicted this result and stated that excess rainfall had enhanced rice production in rain-fed areas. In contrast, coefficients of negative shocks in the rainfall have a positive and significant impact on rice production at a 1% level in the long run. This study is in line with Mitra 2014), they found that any negative shock in the rainfall has positively affected rice production in India. Pal and Mitra (2018) stated that scanty rainfall and drought have reduced food grain production in India. We can infer from the estimated results that 1% increases (decreases) in average rainfall has reduced (boosts) rice production by approximately 1.24 (2.87) percent in India. Any positive shock in the carbon emission has a negative impact on rice production at the 1% significance level in India. The estimated outcomes indicate a rise in carbon emission in the atmosphere by 1%, which reduces rice production by 1.95% approximately. This outcome is in line with Chandio et al. (2021a, b), who found that in the short and long run, carbon emissions negatively affected rice production in the case of Turkey. In contrast, carbon emission negative shocks have an insignificant positive impact on rice production. The coefficient of the negative component of carbon emission indicates that it increases rice production by 0.4% when 1%reduces the carbon emission. We can infer from the estimated results that rice production has been boosted by the reduction of carbon emission in the atmosphere in India. Global warming results from increasing carbon emissions in the atmosphere, which is critical in reducing crop production in developing countries (Jan et al. 2021). The positive components have a dominant effect over negative shock on rice production, which implies that increasing carbon emission has harmful for rice production.
Furthermore, a positive shock in the rural population has a statistically insignificant impact on rice production with a coefficient of 0.49 in the long run. Interpretively, rice production is growing by 0.49% due to a 1% increase in the rural population. The coefficients indicate that rice production increases with increase in rural population. Whereas a negative shock in the rural population has negatively influenced rice production by 0.39% in the long run at a 1% level of significance. This study is in line with previous studies (Kumar et al. 2021a;Warsame et al. 2021), who found that the rural population has a negative impact on cereals production. It is because the marginal productivity of agriculture labor is zero due to working surplus labor in the same piece of land (Thirlwall 1994). Agriculture labor productivity has decreased because land cannot produce more than its capacity (Kumar et al. 2021a). Table 8 reported the results of the short-run asymmetrical impacts on rice output. The positive and negative shocks in mean temperature have positively influenced rice production. Estimated coefficients indicate that a 1% increase and decrease in mean temperature can lead to increases the rice production by 17.23% and 2.60%, respectively, which implies that positive shocks have a more dominant effect than the negative shock on rice production in the short run. Results advocated that rice production has more affected by the increasing temperature rather than decreasing temperature. Moreover, rainfall positive shock has a negative and significant effect on rice production at a 1% level of significance. It is found that rice production reduced by 0.74% when 1% increase in positive shock of rainfall. In contrast, coefficients of negative shocks in the rainfall have a positive and significant impact on rice production at a 1% level of significance in the short run. We can infer from the estimated results that 1% decreases in average rainfall has boosted rice production by approximately 0.64%. Furthermore, any positive shock in the carbon emission has a negative and significant impact on rice production at the 1% level of significance. The estimated outcomes indicate a rise in carbon emission in the atmosphere by 1%, which reduces rice production by 6.16% approximately. In comparison, carbon emission negative shocks positively impact rice production at the 1% significance level. The coefficient of the negative component of carbon emission indicates that it increases rice production by 1.69% when there is 1% reduction in carbon emission. We can infer from the estimated results that rice production has been boosted by reducing carbon emissions in the atmosphere in the short run. Likewise, the impact of positive shock in the rural population has a negative and insignificant effect on rice production in the short run. Interpretively, a 1% increase in rural population leads to a decrease in rice production by 0.50%. The estimated coefficients indicate that rice production decreases when increasing rural population. In comparison, negative shock in the rural population has positively influenced rice production by 1.82% in the short-run at a 1% level of significance.
Moving on to other controlled variables such as fertilizer consumption (lnF), agricultural credit (lnAC), and area under rice crop (lnAUR), these are three core elements of rice production (Chandio et al. 2021a, b). Our findings show that a 1% increase in fertilizer consumption, agricultural credit, and area under crop enhance rice production by 1 3 0.70%, 0.04%, and 2.34%, respectively. These findings are consistent with previous studies (Chandio et al. 2021a, b;Chandio et al. 2020a, b, c, d, e;Janjua et al. 2014;Nasrullah et al. 2021;Omoregie et al. 2018;Zakaria et al. 2020). In the context of India, agricultural credit plays a significant role to boost agricultural production and farm income (Mohan 2006). Chandio et al. (2021a, b) found that agriculture credit has a positive and significant impact on rice production in the case of Nepal. Baig et al. (2020) reported that fertilizer positively influenced rice production in India. Due to might be the reason that fertilizer enhances soil fertility and nutrition, which create a considerable positive impact on rice production (Janjua et al. 2014). Chandio et al. (2021a, b) stated that the area under crop positively impacts rice production in Turkey. The area under rice has the largest share in India, which positively contribute to rice production. The negative and significant ECT value shows that all the variables move toward long-run stability at a medium annual speed of adjustment of 70.97%.
Finally, we performed several dynamic adjustments, the results of which are given in Fig. 4, which depicts the cumulative dynamic multipliers. These multipliers depict the pattern of rice production adjustment toward its new long-term equilibrium as a result of a negative or positive unitary shock in rainfall, mean temperature, carbon emissions, and rural population, respectively. The dynamic multipliers are computed using the Akaike information criterion (AIC) best-fit NARDL model. A particular prediction horizon's rice production adjustment to positive (green line) and negative (red line) shocks is captured by the positive and negative curves. As seen in the graph, the asymmetric curve (dashed red line) represents the difference between the dynamic multipliers for positive and negative shocks, respectively. There is a 95% confidence interval between the lower and upper bands (dotted red lines) of this curve. Figure 5 confirms a negative association between rainfall and rice output. A negative shock in rainfall outperforms a positive shock over the horizon. There is also a large asymmetric reaction to rainfall shocks. As with mean temperature, Fig. 5 Dynamic multiplier adjustment graph rice production is negatively correlated. This confirms the results in Table 8 that a negative shock in mean temperature dominates a positive shock in the long term. Furthermore, positive carbon emission shocks must outweigh beneficial effects on rice production for there to be a negative correlation. However, a negative shock in rural population outweighs a positive one. Table 9 displays the results of different diagnostic tests used to assess the model's reliability (normality, autocorrelation, heteroscedasticity, and Ramsey RESET model). The NARDL model does not suffer from any diagnostic problem. CUSUM and CUSUMQ tests were used to assess model stability. In Fig. 6a and b, the predicted line is within the crucial values at the 5% level of significance, indicating that the model is highly stable.

Granger causality results
Asymmetrical causality between dependent and independent variables are reported in Table 10. We observed a bidirectional impact between a negative shock in rainfall and rice production. In contrast, one-way causality running from positive shock in rainfall to rice production. In addition, we found bi-direction asymmetrical causality among mean temperature and rice production. Furthermore, a two-way causal relationship exists between carbon emission (positive and negative shock) and rice production. Similarly, we found bidirectional asymmetrical causality running among the rural population and rice production. However, bidirectional impact between fertilizer consumption and rice production while one-way causal nexus between area under crop and rice production. Meanwhile, no causal relation runs from agricultural credit to rice production. It implies that positive and negative shocks in mean temperature, carbon emission, and rural population will influence rice production and vice-versa. This work is in line with Chandio et al. (2021a, b), who stated that average rainfall, consumption of fertilizer, and agriculture credit has positively influenced production of rice in Nepal. This study contradicts Warsame et al. (2021), who argued that there is no causal relationship between average rainfall, mean temperature carbon emission, and cereals crop production in Somalia. While negative shock in rainfall, fertilizer consumption, and area under crop has granger causes rice production and vice versa.
Moreover, one-way causality flows from rainfall positive shock toward the area under crop to rice production. Furthermore, unidirectional causality also running from rice production to increasing carbon emission and agricultural credit, which indicates that increasing rice production will increase carbon emission and agricultural credit. In contrast, there is no asymmetrical causality running from average rainfall positive shock, a negative shock in carbon emissions, and a positive shock in agricultural credit to rice production. It indicates that increasing rainfall, decreasing carbon emissions, and increasing agricultural credit has no significant impact on rice production.
Similarly, two-way causality exists between variables such as lnRF + < = > lnRF − , lnRF + < = > lnAT + ,   l n R F + < = > l n C O 2 + , l n R F + < = > l n C O 2 − , l n R F − < = > l n AT + , l n R F + < = > l n C O 2 + , l n R F − < = > l n C O 2 − , l n R F − < = > R P + , lnRF − < = > RP − , lnRF − < = > lnF, lnRF − < = > lnAC, and lnRF − < = > lnAUR. While unidirectional causality running from positive and negative shock in rural population, agricultural credit to increasing rainfall. Furthermore, two-way directional causality running between lnAT + < = > lnAT − , lnAT + < = > lnCO2 + , l n A T + < = > R P − , l n A T + < = > l n A C , l n AT + < = > l n AU R , l n AT − < = > l n C O 2 + , lnAT − < = > lnCO2 − , and lnAT − < = > lnAUR. This finding is consistent with Warsame et al. (2021), who stated that area under crop has positively influenced mean temperature in the atmosphere. Likewise, one-way causality running from increasing and decreasing temperature to increasing rural population, which indicates that increasing and decreasing temperature will positively influenced rural population. Rahman et al. (2021) contradict this finding, and they argued that mean temperature has no significant impact on population growth. Furthermore, there is also evidence that decreasing temperature (lnAT − ) will increase fertilizer consumption (lnF) and agricultural credit (lnAC). Moreover, at 1% significance level, asymmetrical causality between decreasing carbon emission and increasing rural population which indicates reducing carbon emission leads to the increase in rural population. Apart from, one-way directional causality running from increasing rural population to increasing carbon emission means that increasing population leads to decrease environmental quality in the atmosphere. Population increase in rural areas leads to increase deforestation, which plays a key role in deteriorating environmental quality. Researchers stated that the rising population is a dominant cause of environmental degradation ).
However, evidence shows that causality runs from increasing and decreasing carbon emissions toward fertilizer consumption and agricultural credit at the 1% level of significance. The outcome indicates that increasing and decreasing carbon emissions has influenced fertilizer consumption. The causal relationship between agricultural credit and decreasing carbon emission demonstrates that unidirectional causality running from agricultural credit toward decreasing carbon emission at 5 levels of significance, which indicates ≠ > , -> , and < -> indicate "does not Granger cause," "unidirectional causality," and "bidirectional causality," respectively that increasing agricultural credit leads to increase environmental quality in the atmosphere. Asymmetrical causality exists between increasing carbon emission and area under crop, which suggests that increasing carbon emission leads to the increasing area under crop and vice-versa. Unidirectional asymmetrical causality also running from decreasing carbon emission toward the area under crop at the 1 level of significance.

Conclusion and policy implications
Rice production significantly contributed to agricultural growth and food security. Rice is a staple food, and more than 50% of the Indian population consumes it in their daily life. Rice crop has widely grown, followed by wheat, coarse cereals, and pulse in India. The present study examines the asymmetrical relationship and granger causality between climate change and rice production through nonlinear ARDL using time series data spanning from 1991 to 2018 in India. The estimated outcomes confirm the presence of asymmetric relationships among selected variables in the short and long run. The findings of this study reveal that increasing and decreasing temperature influenced rice production adversely in the long run while positively affected in the short run by different magnitude. However, excess rainfall adversely affected rice production, while a decrease in rainfall has no evidence of an adverse effect on rice production in the long and short run. Furthermore, in the long and short run, increased carbon emission levels in the atmosphere had impeded rice production. In contrast, decreased carbon emissions had no adverse impact on rice production. In the long and short run, positive shock in the rural population positively affected rice production, while negative shock adversely affected rice production. The estimated outcomes indicate that other controlled variables, for instance, fertilizer consumption, agricultural credit, and area under crop positively contributed to rice production. The results from asymmetrical causality divulge a feedback effect between negative shock rainfall and rice production. At the same time, a one-way direction causal relationship runs from positive shock in rainfall toward rice production. Furthermore, there is a two-way directional causal relationship between a positive and negative shock in mean temperature and rice production. At the same time, there is no causal relationship between mean temperature and decreasing carbon emission. Moreover, there is a feedback effect between increasing carbon emission and rice production, while a one-way causal relationship runs from rice production to decreasing carbon emission. However, we observed the two-way directional causal relationship among a positive and negative shock in rural population and rice production. Likewise, a two-way causal relationship runs between fertilizer consumption and rice production, while a one-way causal relationship runs from rice production to agricultural credit and from the area under crop to rice production.
Based on the empirical investigations, some key policy implications emerged. Specifically, the government should promote mechanisms of research and development to meet the demand of the population. In this regard, the new fertilizers are required to produce and provided at a subsidized rate to the farmers. Toward sustainable rice production in the context of India, improve irrigation infrastructure through increasing public investment, and develop climate-resilient seeds varieties to cope with or adapt to climate change. Along with, at the district level government should provide proper training to farmers regarding the usage of pesticides, a proper amount of fertilizer, and irrigation systems. The present study was conducted at the national level and undertaken only on rice production, which cannot explain the main influence of climate change or unlike the agro-environment region. However, to tackle regional disparities and season-wise production (Rabi or Kharif) into consideration, area-specific and season-specific research for better insight should be performed. In addition, the effects of technological advancement and subsidy on rice production can also be examined in the case of India by using the ARDL model under the presence of climate change.