Introduction

Climate change has gained global interest within the past decades due to their long-term effect on humanity. Real-world changes and growing scientific evidence prove that humanity is driving global environmental change leading the world into a new geological era—the Anthropocene (Steffen et al. 2011). Thus, a change in the Earth’s global mean near-surface temperature shows a steady state due to doubling of anthropogenic greenhouse gas (GHG) concentrations in the atmosphere (Mohiuddin et al. 2016a; Owusu and Asumadu-Sarkodie 2016; Owusu et al. 2016; Zeebe 2013). Additional human-pressure stands a chance of causing a possible irreversible damage to planet earth resulting in extreme weather changes, food insecurity, and loss of biodiversity, which could hamper development and trigger global humanitarian crisis (Griggs et al. 2013). Possibility of halting human carbon dioxide today poses climate risks that would persist for millennia, leading to increasing global mean temperatures and sea-level rise (Davis et al. 2010; Friedlingstein and Solomon 2005; Keith 2009; Ramanathan and Feng 2008; Wigley 2005). Scientific studies suggest that the cumulative emissions of long-existing greenhouse gases are strongly related with the extent of global warming which by stabilizing global temperature via mitigating global greenhouse gas emissions will likely reduce the impact of climate change (Allen et al. 2009; Huntingford et al. 2012; Meinshausen et al. 2009).

The UN 2030 Agenda for Sustainable Development is a “plan for action for people, planet, and prosperity.” The Sustainable Development Goal (SDG) 2 seeks to end hunger and achieve long-term food security, including better nutrition through sustainable systems of production, distribution, and consumption. These essential benefits for human development led to the establishment of the SDGs by the United Nations (Asumadu-Sarkodie and Owusu 2016d, g). Thus, agriculture is a significant component to achieving the SDG 2 and improving quality of life. However, the quest to increase global food supply to meet the increasing population has led to poor agricultural practices that are detrimental to the environment (Asumadu-Sarkodie and Owusu 2016h; Tilman et al. 2002).

According to IPCC report, agriculture, forestry, and other land use are the second contributing sector of global GHG (24 % of 49 Gt CO2-eq) after electricity and heat production sector (IPCC 2014). Agricultural practices have doubled the natural rate of terrestrial nitrogen (N) fixation over hundreds of years leading to increased atmospheric nitrous oxide (N2O), nitric oxide (NO), methane (CH4), carbon dioxide (CO2), and ammonia (NH3) concentrations which contribute immensely to global climate change (Li and Lin 2000; Sanford et al. 2012). Animal and agricultural crop production systems including ruminants, animal waste, flooded rice fields, and biomass burning is the main sources and sinks for atmospheric methane (CH4) which amount to one third of all global emissions (Mosier et al. 1998).

A number of scientific research (see Table 1) have studied the impact of agricultural ecosystem towards climate change, through the emission of gases such as nitrous oxide, nitric oxide, methane, carbon dioxide, and ammonia. The intensification of emissions is dependent on the type of crops, the indigenous factors related to soil type, weather patterns, and agricultural practices (Dobbie et al. 1999). Agricultural practices like fertilizer application have been found as the main contributor of atmospheric nitrous oxide emissions while livestock production and animal waste are the main contributors of ammonia (Li 2000; Parton et al. 2015). Farming practices like continuous cropping of food security crops like cassava, maize, and sorghum play a critical role in the exacerbation of emissions in Ghana (Asumadu-Sarkodie and Owusu 2016h). Studies show that agricultural soils, rice paddies, and crop residue burning have a potential to escalate global warming through the emission of nitrous oxide and methane (Bhatia et al. 2013). However, ammonia emissions have a seasonal pattern and a strong correlation with temperature, planting time, and agricultural practices (Xu et al. 2015).

Table 1 Compilation of research on the impact of agricultural ecosystem on GHG emissions
Full size table

Table 1 shows a compilation of scientific research on the impact of the agricultural ecosystem on greenhouse gas emissions. Previous studies (Table 1) have been limited to the traditional method of quantification and estimation, meta-analysis, linear mixing model, and multiple linear regression analysis. Nevertheless, the use of modern econometric methods to analyze the impact of the agricultural ecosystem on GHG emissions is limited in the scope of the study. Modern econometric methods like vector error correction model and autoregression distribution lag have been employed in several studies to examine the causal relationship between environmental pollution, energy consumption, and macroeconomic variables in different countries (Asumadu-Sarkodie and Owusu 2016a, b, c, f; Mohiuddin et al. 2016a). However, the assessment of carbon dioxide emissions from developing countries like Ghana has been ignored for so long a time due to their low impact towards global climate change. It is estimated that 43 % of Ghana’s GDP, 50 % of its export earnings, and 70 % of its total employment depends on agriculture and forestry (FAO 2016). As a result, climate change vulnerability is sensitive among population and economies that depend on agriculture and forestry. A recent study that examined the nexus between agricultural production and carbon dioxide emissions found a weak equilibrium relationship which dies over time (Asumadu-Sarkodie and Owusu 2016h). Due to sporadic and limited research in this area, little is known about the empirical relationship between the agricultural ecosystem and carbon dioxide emissions in Ghana.

Against the backdrop, the study makes an attempt to examine the causal nexus between carbon dioxide emissions and agricultural ecosystem using an econometric approach. Though similar studies have been done in developed and some developing countries, however, due to the diversity in socio-economic framework, institutional and political environment in Ghana, the strengths, weaknesses, opportunities, and threats in this environment may affect the outcome of the study. In addition, the study will serve as a reference tool for integrating climate change measures into national policies, strategies, and planning by the Government of Ghana. Given the immense contribution of agricultural ecosystem towards greenhouse gas emissions, the study examines the causal nexus between agricultural ecosystem and carbon dioxide emissions in Ghana. The study employs a multivariate cointegration analysis (Mohiuddin et al. 2016a), vector error correction model (Asumadu-Sarkodie and Owusu 2016b), and a Granger causality test (Granger 1988), to examine the cointegration, long-run effect, and the direction of causality of the area of rice paddy harvested, agricultural machinery (tractors), agriculture value added, biomass-burned crop residues, cereal production, enteric emissions of methane, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, emissions of nitrous oxide from manure application, and stock of livestock towards carbon dioxide emissions.

Methodology

In order to examine the causal nexus between carbon dioxide emissions and agricultural ecosystem, the study employs a time series data from World Bank (2014) and FAO (2015) from 1961 to 2012. Ten variables are employed in the study, with carbon dioxide emissions (CO2) as the dependent variable while biomass-burned crop residues, agriculture value added, enteric emissions of methane, emissions of nitrous oxide from manure application, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, stock of livestock, agricultural machinery, area of rice paddy harvested, and cereal production are the independent variables (see Table 2).

Table 2 Data and variable definition
Full size table

Table 3 presents the descriptive statistical analysis of the study variables. Analysis from 52 observations shows that almost all the variables with the exception of agriculture value added and agricultural machinery have a long right tail (positive skewness); however, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers has the longest right tail while agricultural machinery has the longest left tail (negative skewness). Emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers and agricultural machinery show a leptokurtic distribution to the normal while the remaining variables show a platykurtic distribution to the normal. The Jarque-Bera test statistic operates under the null hypothesis that a series is normally distributed at a p value greater than 5 %. Since CO2, stock, and agricultural machinery are smaller than 5 % p value, the null hypothesis of the normal distribution is rejected. In other words, with the exception of CO2, stock, and agricultural machinery, the remaining variables are normally distributed (see Table 3). With the exception of agricultural machinery and agriculture value added in the Pearson’s correlation, almost all the variables show a positive monotonical relationship with CO2 emissions; however, the strength of association is stronger between CO2 and enteric emissions of methane as rho (ρ) approaches 1. Figure 1 shows the trend of variables over time. With the exception of agricultural machinery and agriculture value added, almost all the variables increase periodically but is, however, inconclusive without making an inferential analysis.

Table 3 Descriptive statistical analysis
Full size table
Fig. 1
figure 1

Trend of variables over time

Full size image

Model estimation

The study employs the vector error correction (VEC) model, which has been used frequently in existing literature at macroeconomic levels. This modern econometric approach has been utilized in global research (Asumadu-Sarkodie and Owusu 2016b, f). Granger causality is employed to ascertain the direction of the causal relationship between the variables under investigation (Granger 1969). For brevity, the model estimation is available in the Appendix.

Results and discussion

This section presents the results and the discussion of the study outcome. After estimating the VEC model of cointegrated variables, Granger causality, diagnostic, and stability tests were performed to ascertain the direction of the causal relationship between study variables, to verify and validate the estimated model.

Unit root test

Performing unit root test is a requirement before estimating the Johansen method of cointegration. Table 4 presents the results of augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests. The null hypothesis cannot be rejected by both ADF and PP unit root tests at 5 % level but rejected at first difference. Hence, the variables are integrated at I(1).

Table 4 ADF and PP unit root tests
Full size table

Johansen’s test of cointegration

This method is used in order to determine the existence of a cointegration among the variables using the selected optimal lags. Table 5 presents the results of unrestricted cointegration rank test for the variables. The trace test statistic indicates six cointegrating equations while Max-eigenvalue test statistic indicates five cointegrating equations at the 5 % level.

Table 5 Unrestricted cointegration rank test (trace and maximum Eigenvalue)
Full size table

Long-run equilibrium relationship

The study employs six cointegrating equations and two optimal lags in the VEC model. Table 6 presents the results of the long-run equilibrium relationship based on VEC model. Results from Table 6 shows that the speed of adjustment (CointEq1 = −0.96) in the VEC model is negative and statistically significant at the 5 % level, showing the existence of a long-run equilibrium relationship running from biomass-burned crop residues, agriculture value added, enteric emissions of methane, emissions of nitrous oxide from manure application, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, stock of livestock, agricultural machinery, area of rice paddy harvested, and cereal production to carbon dioxide emissions. Thus, the variables can converge in the long-run regardless of possible shocks in the short-run. Ninety-six percent of the past year’s disequilibria are corrected in the present year, signifying a good speed of adjustment in the relationship between agricultural ecosystem and carbon dioxide emissions following a shock.

Table 6 Vector error correction model estimates
Full size table

Long-run elasticity

In addition, a 1 % increase in the area of rice paddy harvested will increase carbon dioxide emissions by 1.49 % in the long run, a 1 % increase in biomass-burned crop residues will increase carbon dioxide emissions by 1.00 % in the long run, a 1 % increase in cereal production will increase carbon dioxide emissions by 1.38 % in the long run, while a 1 % increase in agricultural machinery will decrease carbon dioxide emissions by 0.09 % in the long run.

Short-run equilibrium relationship

A Wald’s test of coefficient restriction was used to check the short-run equilibrium relationship between the dependent variable and the independent variables. Results from Table 7 show no evidence of short-run equilibrium relationship. The results suggest that the short-run effect of the area of rice paddy harvested, agricultural machinery (tractors), agriculture value added, biomass-burned crop residues, cereal production, enteric emissions of methane, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, emissions of nitrous oxide from manure application, and stock of livestock may not occur in the short-term but will rather increase carbon dioxide emissions in the long run.

Table 7 Short-run equilibrium relationship based on VEC model
Full size table

Granger causality tests

After estimating the cointegration between the variables using the Johansen’s test of cointegration and establishing a long-run equilibrium relationship with the vector error correction model, the direction of causality among the variables is uncertain. From this regard, the study examines the direction of causality among the variables using the Granger causality test. Table 8 presents the results of pairwise Granger causality test. Table 8 shows a bidirectional causality between carbon dioxide emissions ↔ cereal production and carbon dioxide emissions ↔ biomass-burned crop residues and a unidirectional causality between: carbon dioxide emissions → area of rice paddy harvested, carbon dioxide emissions → agricultural machinery, carbon dioxide emissions → agriculture value added, carbon dioxide emissions → enteric emissions of methane, and carbon dioxide emissions → emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers.

Table 8 Pairwise Granger causality tests
Full size table

Diagnostics and robustness of the VEC model

In order to check the robustness of the model, the VEC model was subjected to a series of diagnostic tests. Table 9 presents the diagnostic tests of the VEC model. The results from Table 9 shows that the residuals are normally distributed (Jarque-Bera test), no residual serial correlation exists (Lagrange multiplier test), and no conditional heteroskedasticity exists in the model (Breusch-Pagan-Godfrey test). Figure 2 depicts the roots of linear recurrence relation in the VEC model. Figure 2 shows that the VEC model specification imposes only one unit root without a single root outside the circle, meaning that the VEC model obeys the VAR stability conditions thus, statistically valid and robust to make unbiased inferences.

Table 9 VECM diagnostics test
Full size table
Fig. 2
figure 2

Inverse roots of AR polynomials

Full size image

Discussion

The bidirectional causality between carbon dioxide emissions and cereal production shows that increasing or decreasing carbon dioxide emissions in Ghana would result in either positive or negative outcome. An elevated level of carbon dioxide emissions in the atmosphere leads to climate change which results in increasing temperature and changing rainfall patterns affecting cereal production. A study by Fitzgerald et al. (2010) suggests that elevated carbon dioxide emissions alter crop physiology, grain quality, yield, growth, pest and disease dynamics, and soil processes, which might change future human nutrition and cropping systems and patterns. In the reverse option, cereals absorb carbon dioxide to prepare its food through a process called photosynthesis. A recent study showed that some cereals’ biomass increased significantly by 25–30 % due to carbon dioxide emissions (Fitzgerald et al. 2010). In contrast, the long-run elasticity in Table 6 shows that an increasing rate of cereal production will increase carbon dioxide emissions by 1.38 %, which may be due to poor agricultural practices in the production of cereals in Ghana.

The bidirectional causality between carbon dioxide emissions and biomass-burned crop residues is multifaceted. Some studies (Awasthi et al. 2010; Kludze et al. 2013; Viana et al. 2013) suggest that biomass-burned crop residues lead to economic loss, health impacts, and are not sustainable due to its environmental impact like greenhouse gas emissions, disruption of the physical, chemical, and biological composition of soil quality, and its effect on crop yield. On the contrary, some studies (Kutcher and Malhi 2010; Vasilica et al. 2014) suggest that burning biomass of crop residues in the field has been proven to control the emergence of invasive weed species, insects, crop diseases, repairs damaged soil structure, improves soil water retention, improves soil fertility, increase the total organic carbon, increase the total soil nitrogen, improve soil-aggregated stability, etc. which directly affect agricultural productivity. However, evidence from the long-run elasticity in Table 6 shows that an increasing rate of biomass-burned crop residues will increase carbon dioxide emissions by 1.00 % in Ghana. Burning biomass of crop residues is one of the unsustainable farming practices in Ghana which has received public health concern as it underpins environmental and health problems like air pollution. Changes in farming practices will undoubtedly improve Ghana’s agricultural sector while curbing environmental pollution and pollution-related health problems.

Conclusion

In this study, the causal nexus between carbon dioxide emissions and agricultural ecosystem was examined by employing a data spanning from 1961 to 2012. Evidence from the study shows that the variables are integrated at I(1) and cointegrated. There was an indication of a long-run equilibrium relationship running from the area of rice paddy harvested, agricultural machinery (tractors), agriculture value added, biomass-burned crop residues, cereal production, enteric emissions of methane, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, emissions of nitrous oxide from manure application, and stock of livestock to carbon dioxide emissions.

The study shows that a 1 % increase in the area of rice paddy harvested will increase carbon dioxide emissions by 1.49 % in the long run, a 1 % increase in biomass-burned crop residues will increase carbon dioxide emissions by 1.00 % in the long run, a 1 % increase in cereal production will increase carbon dioxide emissions by 1.38 % in the long run, while a 1 % increase in agricultural machinery (tractors) will decrease carbon dioxide emissions by 0.09 % in the long run.

The Granger causality shows a bidirectional causality between carbon dioxide emissions and cereal production and carbon dioxide emissions and biomass-burned crop residues. In addition, there was a unidirectional causality running from carbon dioxide emissions to the area of rice paddy harvested, carbon dioxide emissions to agricultural machinery, carbon dioxide emissions to agriculture value added, carbon dioxide emissions to enteric emissions, and carbon dioxide emissions to emissions of nitrous oxide from manure application.

Results from the vector error correction model shows that the combined effect of biomass-burned crop residues, agriculture value added, enteric emissions of methane, emissions of nitrous oxide from manure application, emissions of carbon dioxide equivalent of nitrous oxide from synthetic fertilizers, stock of livestock, agricultural machinery, area of rice paddy harvested, and cereal production affect carbon dioxide emissions in the long run. A visual inspection of the regression line in Fig. 1 throws more light on the vector error correction model. It is evidential that an increase in value-added agricultural products and agricultural machinery like tractors and wheelers reduce the rate of carbon dioxide emissions. Harvest and post-harvest losses have become the major burden of Ghana’s agricultural sector. The majority of Ghana’s farming communities are located in rural areas with low harvest and post-harvest management systems which affect the carbon footprint as agricultural products deteriorate. Poor value addition to agricultural products has recently affected the food commodity market leading to an extreme food price volatility. Improving the use of energy-saving agricultural machinery in Ghana would help reduce carbon dioxide emissions in the long run. Improving manure management by employing carbonization technologies for livestock excretion thus, utilization of manure for energy generation would reduce enteric emissions of methane thereby reducing carbon dioxide emissions. Employing high-speed puddling machines in the area of rice paddy harvested would play a role in the reduction of methane emissions from rice harvesting. Cycle usage of biomass residues into effective use rather than burning it in open fields would help reduce carbon dioxide emissions. A recent study shows that biomass residue can be used to generate electricity while providing economic advantages (Asumadu-Sarkodie and Owusu 2016e; Mohiuddin et al. 2016b). Finally, reducing the use and input of fertilizer on farm land during crop production would help reduce nitrous oxide emissions thereby reducing environmental pollution.

Future studies should employ multivariate analysis methods like non-linear iterative partial least squares or statistically inspired modification of partial least squares that are capable of dealing with problems of multicollinearity which falls short in econometric methods like vector error correction model which limits the use of important explanatory variables.