Economic Growth, CH 4 and N 2 O Emissions in Sudan: Where Should the Policy Focus Be?

This study aimed to investigate the effect of economic growth, agricultural growth and energy use on methane (CH 4 ) and nitrous oxide (N 2 O) emissions in Sudan. Within the context of the EKC, the study applies the OLS, cointegration, vector error correction modelling (VECM) and Granger causality methods. The study has established a long run equilibrium relationship for both CH 4 and N 2 O in their relation to economic growth, agricultural growth and energy use in presence of trade openness (TOP) and inflows of foreign direct investments (FDI). The estimated VECM shows that emissions of CH 4 are significantly affected by economic growth, TOP, and FDI with no effect of agricultural growth in the short run while CH 4 is found to be significantly affected by economic growth, agricultural growth, TOP and FDI in the long run. The estimated VECM for N 2 O shows that N 2 O emissions are more significantly affected by energy use, agricultural growth and FDI with no effect of economic growth in the short run, while N 2 O is found to be significantly affected by economic growth, agricultural growth, TOP and FDI in the long run. Consistently, findings from the estimated OLS and VECMs show that the EKC does not hold for either CH 4 , or N 2 O emissions, and that N 2 O emissions are more significantly affected by economic growth, agricultural growth and energy use than emissions of CH 4 . Findings from impulse response and variance decomposition analysis confirm that emissions of N 2 O are more responsive to economic growth, agricultural growth and energy use than emissions of CH 4 . Granger causality analysis shows existence of bidirectional relationship between CH 4 and agricultural growth, but a unidirectional relationship from CH 4 to FDI. For N 2 O, the study finds a unidirectional relationship running from agricultural growth to N 2 O, while N 2 O emissions are found to cause GDP per capita, the squared GDP per capita, OIL consumption and FDI. In terms of causality, these results suggest that emissions of CH 4 and N 2 O have been generated more by agricultural activities than by overall eco-How nomic activity, and that activities generating N 2 O emissions in particular have been contributing significantly to economic growth. Within the context of the country’s intended nationally determined contributions, the findings of this study suggest that policies should be directed cautiously but more effectively to control N 2 O than CH 4 emissions. Economic growth could be pursued without significant environmental harm from both CH 4 and N 2 O emissions. However, Sudan should expand adoption of energy efficiency measures, expansion of renewable energy use, place restrictions on production and use of fuel woods and charcoal for low carbon economy and green growth.


Introduction
Global warming and climate changes are international environmental problems caused by emissions and concentration of greenhouse gases (GHGs) chief among them is carbon dioxide (CO 2 ). Globally, in 2016, about 72% of GHG emissions consist of CO 2 , but non-CO 2 emissions, namely methane (CH 4 ) and nitrous oxides (N 2 O) make up shares of 19% and 6% and that CH 4 , N 2 O and Chlorofluorocarbons (CFCs) are the main GHGs contributing to global warming potential (WGP) (Olivier et al., 2017). CH 4 was the largest contributor to non-CO 2 GHG emissions, from non-dairy and dairy cattle and cattle manure management, while collectively coal mining, oil and natural gas production and distribution, and rice cultivation accounted for 25% and 10% of CH 4 emissions respectively (Olivier et al., 2017). N 2 O emissions both anthropogenic and natural originate from soils which account for over half of total atmospheric N 2 O inputs, (Parton et al., 2001). Emissions of N 2 O are directly and indirectly related to land use and land cover changes (LULCC) and thus to agricultural growth and development. Sources emitting N 2 O directly include cultivated soils and fertilized and/or grazed grassland systems, while indirect emissions result from transport of nitrous from agricultural systems into ground and surface waters, or emission as ammonia or nitrogen oxides (Smith et al., 1999). N 2 O emissions stem also from agricultural biomass burning, industrial activities. According to the Intergovernmental Panel on Climate Change, IPCC, Climate Change (1995), compared with CO 2 heat trapping, a kilogram of CH 4 is 21 times as effective at trapping heat in the earth's atmosphere as a kilogram (kg) of carbon dioxide within 100 years, while the per kg global warming potential (GWP) of N 2 O is nearly 310 times that of CO 2 within 100 years. The same report (IPCC, Climate Change, 1995) attributes growth of CO 2 , CH 4 and N 2 O to fossil fuel use, land use change and agriculture. Moreover, it has been documented that one pound of N 2 O warms the atmosphere almost 300 times more than one pound of CO 2 (United emissions are related to LULCC, and intensive statistical analysis of these emissions can be found in Parton et al. (2001) and Houghton et al. (2012). Importantly, it has been recognized that policy targeting reduction of CH 4 and N 2 O emissions result in higher social benefits compared with reductions of CO 2 emissions based on the GWP of these trace gases (Marten & Newbold, 2014). Earlier studies such as of Grossman & Krueger (1991) found that concentrations of sulfur dioxide and smoke increase with per capita gross domestic product (GDP) at low levels of national income, but decrease with GDP growth at higher levels of income in 42 countries. Then, Panayotou (1993)  Agreement. Yet, where climate change policy should be directed needs to be grounded on sound findings from rigorous empirical studies. Both mitigation and adaptation policies to climate change have costs, and thus only the most cost effective should be designed and implemented to control GHG emissions. In general, cross-sectional studies of agriculture using the Ricardian method suggest that efficient adaptation will reduce damages from climate change (Mendelsohn et al., 1994(Mendelsohn et al., , 1996Mendelsohn & Dinar, 1999 Sudan is a low-income country; member of the UNFCCC, signed and ratified the KP and submitted its INDCs to Paris 2015 Agreement. Although mitigation is not a priority and it's not binding with specific emission cut targets for Sudan, adaptation measures are extremely important since Sudan is vulnerable to different impacts of climate change including drought, flooding and reduced agricultural products (Nimir & Ismail, 2013). Nonetheless, for effective climate change adaptation, the driving forces of CH 4 and N 2 O emissions as the main GHGs in Sudan need to be identified and their relative importance needs to be well understood. Emissions of these gases in Sudan stem mainly from agriculture, deforestation, energy use and municipal wastes. Yet, commercial energy in terms of oil production and consumption has been steadily increasing since the late 1990s and contributing increasingly to GHGs in general and CO 2 emissions in particular. In Sudan, N 2 O increased by 439.29% and CH 4 increased by 242.44% over the period 1970-2017 amounting to an annual rate of increase of 9.15% and 5.05% for N 2 O and CH 4 respectively. Meanwhile, CO 2 emissions have increased by 301.33% at an annual rate of 6.28%. Emissions from agriculture account for an average of 84% and 91% form total emissions for CH 4 and N 2 O respectively (calculated from World Bank, 2017). The second source of CH 4 and N 2 O emissions in Sudan is the energy sector. Since agriculture also accounts for more than 35% of GDP over the period 1970-2017, GDP separately and its agricultural component jointly could be major determinants of CH 4 and N 2 O emissions in Sudan.
Despite their importance to global warming and climate change, relatively less analytical and empirical attention has been paid to determinants of CH 4 and N 2 O emissions, globally and at country level. For Sudan, no such study on determinants of CH 4 and N 2 O exists, and the country's INDCs have not been grounded on findings of rigorous empirical studies. Since the early 1990s, Toth (1995) notes that although CO 2 continues to be the single most important GHG, contributions of other trace gases to the aggregated global warming potential are sufficiently significant to justify their more equitable treatment in climate policy analysis. Even more recently, Sinha & Sengupta (2019) note that there are a limited number of studies on the estimation of EKC hypothesis for N 2 O emissions, though it is one of the most harmful GHGs present in ambient atmosphere.
Upon this introduction and background information, the aim of this study is to contribute to this literature and knowledge twofold. First, how differently economic growth, agricultural growth and energy use affect emissions of CH 4 and N 2 O in Sudan. Secondly, the study comparatively addresses the two most important non-CO 2 emissions with the same explanatory variables, in a context of a single country. Findings from such comparative analysis guide which and where policies can be directed effectively toward control of these emissions, and thus serves as evidence-based framework to climate change policy in Sudan in the context of its INDCs.

Literature Review
Huge empirical literature has been developing to investigate what development factors that lead to increases in GHGs emissions at the global and national levels.
Studies from the field of environmental economics in the early 1990s the focus was on determining a level of per capita income at which GHGs generally and CO 2 emissions in particular start a reversion from upward to downward slop-  (Panayotou, 1993). Empirical tests of the EKC were earlier undertaken by Selden & Song (1994) and Grossman & Krueger (1995), with focus on economic growth, trade liberalization and GHGs emissions. But, over time and across countries CO 2 emissions per capita have been found to strongly correlate with GDP per capita (Stern, 2011). The focus has shifted to investigate probable determinants of GHGs emissions but focusing mostly on CO 2 emissions, so as to design environmental and climate change policies compatible with economic development objectives.
In empirical literature, identified factors affecting GHGs emissions in general include the scale and rate of economic growth, energy use, openness to trade and flows of capital, urbanization and the contribution of the services sector to total economic output. Technological changes in the energy sector play an important role in GHGs emissions (Wheeler, 2012;Wang et al., 2012). Population density which is important factor in energy distribution and use by sectors has also been identified as important factor affecting GHGs emissions. Empirical literature on CO 2 determinants of emissions can be found in Omri (2009) Their results show that the coefficient of electricity production from renewable sources with respect to CO 2 emissions is negative and significant in the long run and found a U-shaped (EKC) relationship between per capita GHGs and income, with a peak point of GDP per capita of $9920 which according to them was outside the observed sample period.
Vavrek & Chovancova (2016)  The authors state that in order to meet their 2050 objectives to reduce GHG emissions, these countries need to accelerate restructuring the ways how they meet their demand for energy, food, transport and housing. Utilizing time series data between 1970 and 2012 and the ARDL approach, Manuel & Mario (2017) analyze the relationship between N 2 O emissions, economic growth, agricultural land used and exports in Germany. Their results show a quadratic long run relationship between N 2 O emissions and economic growth, confirming the existence of an EKC for Germany. They estimated a turning point at $27,880 which according to them was within the sample and implies that Germany was in the decreasing part of the curve of environmental degradation. They also find that agricultural land area affects N 2 O emissions positively, whereas exports affect emissions negatively. The paper shows that, contrary to testing the EKC in less developed countries, mitigation of N 2 O emissions does not negatively affect growth in Germany and as such, it is feasible to undertake any conservative policy in order to reduce emissions without major consequences on economic sectors. Using panel unit root and cointegration tests, Cho et al. (2014) analyze determinants of total GHG, CH 4 and N 2 O emissions. They found a quadratic relationship in the long run for OECD countries. They also found effects of trade openness on GHG emissions. Although exports may not affect N 2 O emissions according to Kearsley & Riddel (2010) the existence of an EKC for exports may be because of the pollution haven hypothesis (PHH) asserting transfer of pollution across countries instead of mitigating it which could be described as an inverse U-shaped relationship between exports and economic growth. For the European countries, Gielen & Kram (1998) find evidences that CH 4 , N 2 O and CFCs play a significant role in meeting Kyoto targets by EU member states, and that their emission reductions were forecast to contribute one quarter to the total emission reduction in 2010, given the emission reduction goals of individual European countries. Furthermore, Gambhir et al. (2017) find non-CO 2 mitigation measures are less costly than CO 2 mitigation measures, with the majority of their abatement potential achievable at US2005$100/tCO2e or less throughout the 21st century compared to a marginal CO 2 mitigation cost which was estimated to be already greater than this by 2030 in the most stringent mitigation scenario. Benavides et al. (2017) use the ARDL method to investigate the relationship between emissions of CH 4 , GDP, electricity production from renewable energy sources (excluding hydro) and trade openness in Austria. They find an inverted U-shaped between GDP per capita and CH 4 emissions as well as a unidirectional causality between CH 4 , GDP per capita and electricity production. Applying econometric methods on annual data for 1981-2012, Maralgua (2017) investigates the EKC hypothesis for N 2 O emissions, income, exports, urbanization, and growth in different sectors of the economy for Mongolia. The author finds a highly significant and robust long-run U-shaped relationship between N 2 O emissions and income. Furthermore, exports, urbanization, and growth in the industrial and services sectors were found to decrease N 2 O, while growth in the agricultural sector was found to increase N 2 O emissions. The author found significant short-and long-run Granger causal relationships amongst the variables. However, until recently, Sinha & Sengupta (2019) argue for better understanding of the impact of energy consumption pattern on N 2 O emissions and revision of energy policies. They analyzed the impact of renewable and fossil fuel energy consumption on N 2 O emissions for Asia Pacific Economic Cooperation (APEC) countries over the period of 1990-2015 in the context of the EKC hypothesis. They indicate that the efficacy of the renewable energy solutions help to reduce In conclusions, the findings of the above reviewed studies indicate that policy measures could be effective and less costly to be directed to non-CO 2 emissions that directly targeting reduction of CO 2 emissions.

Methodology
This study investigates the effects of economic growth, agricultural growth and energy use with separate inclusion of oil consumption (OIL) on emissions of CH 4 and N 2 O in Sudan in presence of trade openness (TOP) and inflows of foreign direct investments (FDI). The study utilizes annual time series data over the period 1970-2016. Within the context of the EKC, the study applies the ordinary least squares (OLS), cointegration, vector error correction modelling (VECM) and Granger causality methods. The study departs with a graphical analysis of total GHGs, CH 4 and N 2 O emissions over time. Figure 1 shows some type of co-movements of CH 4 and N 2 O emissions in Sudan, but recently emissions of N 2 O started to overtake CH 4 . The figure also depicts that total GHG emissions have been steadily increasing over the study period. Over this period, Sudan has also experienced positive economic growth rates with some years of slowdown, as well as agricultural expansion with increased production and consumption of modern energy sources, namely oil and electricity but also biomass energies.

OLS Modelling
General two OLS models to analyze how differently economic growth, agricultural growth and energy use affect CH 4 and N 2 O emissions in presence of (TOP) and (FDI) in natural logarithms (L) are written as follows: The variables are defined as follows: CH 4 is Methane emissions measured in Kiloton (Kt) of CO 2 equivalent. N 2 O is Nitrous Oxide emissions measured in thousand metric tons of CO 2 equivalent.
GDPP is GDP per capita measured in current US Dollars. EUP is the energy use per capita measured in kg of oil equivalent per capita.
Oil is oil consumption measured in kg per capita. CPIN is crop production index which shows agricultural production for each year relative to the base period [2004][2005][2006]. It includes all crops except fodder crops.
TOP is trade openness measured as exports plus imports as percentage of GDP. FDI is the inflow of foreign direct investments to Sudan, measured in current US Dollars.

E. S. E. Mohamed
The study departs with descriptive statistical analysis for both CH 4 and N 2 O emissions together with the set of explanatory variables, as presented in Table 1.
Based on Jarque-Bera (J-B) statistics, CH 4 , N 2 O and GDPP are not normally distributed.
The correlation matrix in Table 2 shows that CH 4 is positively correlated with GDPP, CPIN and FDI. N 2 O is highly positively correlated with GDPP and the squared value of GDPP, CPIN and FDI. The squared value of GDPP is highly positively correlated with GDPP and of the other independent variables only FDI and CPIN, OIL and FDI are highly positively correlated.
The correlation matrices give an initial idea that N 2 O emissions are more affected by economic growth and agricultural growth than CH 4 emissions.

Dynamic Econometric Modelling
As some of the main variables of the study, i.e., CH 4 , N 2 O and GDPP are found to follow non-normal distribution, and the likelihood that the variables may not follow an autoregressive (AR), and a first differenced random walk I(1), which in fact were reflected by the unreliability of the above estimated OLS models, the study proceeds to dynamic econometric methods of cointegration, VECM and Granger causality analysis.

Stationarity and Cointegration 1) Stationarity Analysis
For meaningful estimation of time series econometric models and for reliable results, the stationarity of the variables must be established otherwise, estimations may be spurious (Granger, 2001). The non-stationarity and presence of a unit root in the series implies that and any shock will have a permanent effect on The null hypothesis (non-stationarity) is that H 0 : γ = 0, and the alterative (sta- With a linear trend in order to test whether the series is I(1) against the alternative that it is stationary around a fixed linear trend, a trend term component needs to be added along with the constant, i.e., The lag p should be large enough to make the error term ε white noise. The estimated ADF statistics is compared with the simulated MacKinnon (2010) critical values, which employ a set of simulations to derive asymptotic results and to simulate critical values for arbitrary sample sizes. The ADF statistics must be larger than critical values in absolute value and have a minus sign. For CH 4 , using the ADF and PP unit root tests, with the assumption of intercept only, and with trend and intercept, all variables included in the study are found to be nonstationary at level I(0), but they all turn to be stationary at first difference I(1) at 5 percent level of significance as presented in Table 3. For N 2 O, using the ADF and PP unit root tests the study variables are found to be nonstationary at level but they all turn to be stationary at first difference I(1) at 5 percent level of significant as presented in Table 4.

2) Cointegration
Given that all variables are found to be integrated at the first order, the long run nature of the relationships between CH 4 , N 2 O and their explanatory variables included in the model are tested by the Johansen cointegration method. Applying this method, for CH 4 with the assumption of intercept only, the maximum Eigen value statistics indicate existence of at least three cointegrating vectors. With the assumption of intercept and trend in data, the maximum Eigen value statistics indicate existence of at least two cointegrating vectors as summarized in Table 5. For N 2 O with the assumption of intercept only, the trace statistics indicates existence of seven cointegrating vectors while the maximum Eigen value statistics indicate existence of at least four cointegrating vectors. With the assumption of intercept and trend in data, the maximum Eigen value statistics indicate that N 2 O is cointegrated with four vectors as summarized in Table 5.
Thus, stationarity of the study variables is established by the unit root tests and a long run equilibrium relationship is confirmed for both CH 4 and N 2 O with economic growth, agricultural growth and energy use in presence of trade openness and inflows of FDI. Moreover, it appears that emissions of N 2 O have more and persistent co-movements with economic growth, agricultural growth and energy use than emissions of CH 4 . Upon establishment of stationarity at I(1) and the cointegration of the variables, the study proceeds to estimate an unrestricted vector autoregressive VAR model upon which an optimal lag length is selected for estimation of VECM. Results of the optimal lag length are summarized in Table 6.

VECM Model Estimations
A VECM version of Equation (1) is estimated for CH 4 emissions and of Equation (2) for N 2 O emissions, both on lag length of 2 according to LR criterion as in Table 6. Summary results of the estimated VECM for CH 4 and N 2 O emissions are reported in Table 7. The VECM models estimates show that the coefficients of N 2 O emissions with respect to economic growth and agricultural growth are much higher and more significant than the coefficients of CH 4 with respect to economic growth and agricultural growth. Results of the two models for both types of emissions show that low level of economic growth is also associated with low level of emissions, but while it is found that emissions of CH 4 increase significantly with the squared value of the GDP per capita, this does not hold for emissions of N 2 O in the long run. Positive coefficients for both CH 4 and N 2 O emissions with respect to the squared value of GDP per capita contradict predications of the EKC in the case of Sudan, more strongly and clearly for CH 4 emissions. Also, N 2 O emissions are better correctly adjusting back to equilibrium than CH 4 emissions in response to shocks in the system. The VECM models for both types of emissions proved to be stable and not miss-specified as depicted by Figure 6 for CH 4 and in Figure 7 for N 2 O.   As presented in Table 8     Granger non-causality or exogeneity of variables is also tested through the Wald test in order to judge which variables lead and which lag the others, for both CH 4 and N 2 O emissions. As in Table 9, CH 4 , EUP and CPIN are found to be lagging (endogenous variables) and all other variables of interest are leading highly exogenous, which is consistent with the results of cointegration and VECM results. For N 2 O, all variables are found leading highly exogenous and only N 2 O is found to be lagging (endogenous variable).
In accordance with the VECM, impacts on CH 4 and N 2 O and their volatility with economic and agricultural growth shocks are also assessed through the impulse response function (IRF). As in Table 10 CH 4 is largely explained by its own lagged shocks, followed by TOP, FDI, and the squared value of GDPP, with the least contribution coming from GDPP and CPIN. Table 10 also shows that N 2 O is largely explained by its own lagged shocks, followed by TOP, FDI, and the GDPP, with the least contribution coming from squared value of GDPP, OIL and CPIN.
The method of variance decomposition shows that variations in CH 4 are much explained by CH 4 itself, TOP and FDI, than by EUP and CPIN. GDPP contributes the least in explaining variations in CH 4 as reported in Table 11. Also, the method of variance decomposition shows that variations in N 2 O are much explained by N 2 O itself, TOP and FDI, than by GDPP, OIL. CPIN, the squared value of GDPP and EUP contribute the least in explaining variations in N 2 O as reported in Table 11.    Table 11. Variance decomposition of L(CH 4 ) and N 2 O.

Discussions and Conclusion
This study comparatively investigated the effects of economic growth, agricultural growth and energy use on methane and nitrous oxide emissions in Sudan emissions, the estimated VECM shows that emissions of this trace gas are significantly affected by economic growth, TOP and FDI with no effect of agricultural growth in the short run, while its emissions are found to be significantly affected by economic growth, agricultural growth, TOP and FDI in the long run.
It also shows nonexistence of an EKC in the long run which is consistent with the OLS finding. For N 2 O, the VECM model results show that emissions of this trace gas are more significantly than CH 4 affected by energy use, agricultural growth and FDI in the short run, while significantly affected by economic growth, agricultural growth, TOP and FDI in the long run. Also, no EKC is found for N 2 O emissions. Impulse response and variance decomposition analysis confirm that emissions of N 2 O are more responsive to economic growth, agricultural growth and energy use than emissions of CH 4 . Granger causality test shows existence of only one bidirectional relationship between CH 4 and agricultural growth and only one unidirectional relationship running from CH 4 to FDI.
For N 2 O there exists only one unidirectional relationship running from agricultural growth to N 2 O while there is unidirectional relationship running from N 2 O emissions to GDPP, the squared value of GDPP, OIL consumption and FDI with no sign of significant feedback effects. These results suggest that policies toward control of N 2 O emissions will have significant negative effects on economic growth, and inflows of FDI may likely be discouraged with stringent environmental policies toward nitrous oxide emissions.
The study concludes that N 2 O emissions are more responsive to changes in economic and agricultural growth compared with CH 4 emissions. Furthermore, energy use only affects N 2 O emissions with no effect on CH 4 emissions. Oil consumption has no effect on emissions of both CH 4 and N 2 O. Results from Granger causality analysis suggest that economic growth could be pursued without significant environmental harm from both CH 4 and N 2 O emissions.
However, policies toward control of N 2 O emissions in particular should be set and implemented with caution as their effects on emissions will be transmitted to negatively affecting economic growth, and inflows of FDI. Furthermore, the findings suggest that Sudan should adopt energy efficiency measures, expansion of production and use of liquefied petroleum gas and place restrictions on production and use of fuel woods and charcoal for low carbon economy and green growth. Again, such policy measures should more effective if cautiously directed to control of N 2 O emissions within the country INDCs for the purpose of dealing with climate change obligations, more through than to CH 4 emissions control. As a least developed country (LDC) Sudan is not obliged to pursue a GHG emission reduction target. However, the country has set plans to reduce GHG emissions and pursue low-carbon development, promoting sustainable resource management in balancing national economic objectives and sustainable development requirements. Notwithstanding that Sudan INDCs give priority to energy, forestry and waste sectors in mitigation of CO 2 , CH 4 and N 2 O emissions, Sudan Intended Nationally Determined Contributions (Republic of Sudan, 2015). Also, within the INDCs, the energy intensity of the economy defined as total primary energy use per unit of GDP can also be reduced by relocation of resources from energy intensive sectors to labour and capital intensive sectors.
The emission intensity of energy, represented by CO 2 per unit of energy can also be decreased by substitution of fuels (with lower emission factors) and through increases of renewable energies in the country's energy mix.