Forecasting of CO₂ emissions in Iran based on time series and regression analysis

Hosseini, Seyed Mohsen; Saifoddin, Amirali; Shirmohammadi, Reza; Aslani, Alireza Article Forecasting of CO₂ emissions in Iran based on time series and regression analysis Energy Reports Provided in Cooperation with: Elsevier Suggested Citation: Hosseini, Seyed Mohsen; Saifoddin, Amirali; Shirmohammadi, Reza; Aslani, Alireza (2019) : Forecasting of CO₂ emissions in Iran based on time series and regression analysis, Energy Reports, ISSN 2352-4847, Elsevier, Amsterdam, Vol. 5, pp. 619-631, https://doi.org/10.1016/j.egyr.2019.05.004 This Version is available at: https://hdl.handle.net/10419/243616 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Energy Reports 5 (2019) 619–631 Contents lists available at ScienceDirect Energy Reports journal homepage: www.elsevier.com/locate/egyr Research paper Forecasting of CO2emissions in Iran based on time series and regression analysis Seyed Mohsen Hosseini, Amirali Saifoddin∗, Reza Shirmohammadi, Alireza Aslani Department of Renewable Energies and Environment, Faculty of New Sciences & Technologies, University of Tehran, Tehran, Iran article info Article history: Received 27 February 2019 Received in revised form 21 May 2019 Accepted 22 May 2019 Available online xxxx Keywords: Regression Paris agreement CO2emission Energy Scenario abstract Iran has become one of the most CO2emitting countries during the last decades. The country ranks after Japan and Germany in terms of CO2emissions. However, from an economic viewpoint, the gross domestic product (GDP) of Iran is lower than the summation of Berlin and Tokyo GDP. Moreover, a large proportion of Iran’s revenue comes from the crude oil export; therefore, this level of CO2emission cannot be economically driven and is as a result of high energy intensity in this country. This is while the government also has not a clear program in this regard. The Sixth Five-year Development Plan of Iran, in addition, sets a number of ambitious targets mostly regarding the energy intensity, GDP growth, and renewable energies, but does not mention to CO2emission issue. Therefore, prospects for an early settlement of the dispute are seemingly dim. Our aim is to predict Iran’s CO2emissions in 2030 under assumptions of two scenarios, i.e. business as usual (BAU) and the Sixth Development Plan (SDP), using multiple linear regression (MLR) and multiple polynomial regression (MPR) analysis. Findings suggest that Iran most likely will not meet its commitment to the Paris Agreement under the BAU’s assumptions; however, full implementation of the ambitiously shaped SDP could have met the target by end 2018. ©2019 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction The continuous rise of carbon dioxide emissions is a major cause of global warming. Global warming is one of the biggest and probably most difficult environmental, social and economic threats that the world has faced so far in the recent century (Shirmohammadi et al.,2018). Increasing carbon dioxide emission into the atmosphere is one of the main reasons for global warming with adverse environmental effects such as sea level rise, floods, droughts, etc. During the 20th century, the average temperature of the earth has increased by 0.6 degrees, and it is estimated that it will increase 1–5 more degrees in the next century (Bistline and Rai,2010). According to the studies conducted by the world meteorological organization, the average global temperature in 2015 has been the highest one ever recorded. Based on research, a combination of El Nino streams and global warming resulted by human activities has led to the highest recorded average global temperature in 2011 to 2015. All nations nowadays have faced a thorny problem of finding a way to lessen the adverse effects of climate change. This is a worldwide issue and not limited to a particular area. All of the people living from the Arctic Ocean to the Antarctic Ocean ∗Corresponding author. E-mail address: [email protected] (A. Saifoddin). are in danger due to the unprecedented rise in the earth’s average temperature. Melting permafrost in Alaska and Russia, as well as, polar ice cap indeed will influence all the living things. Unfortunately, excessive consumption of fossil fuels and deforestation in the last two decades also have exacerbated the situation (Razmjoo and Davarpanah,2019). The atmosphere contains about 44% more CO2molecules from the 1750 level at the present time (Tarasova et al.,2018). The issue has led to a 0.8 ◦C global warming since 1880 with an increase rate of approximately 0.15– 0.20 ◦C per decade. About two-thirds of the figure is related to the post-1975 years (Carlowicz,2010). Continuation of the anthropogenic CO2emissions reportedly could have easily increased the figure to 2 ◦C in the coming years (Allen et al.,2009). Therefore, the CO2emission is an urgent problem and needs tackling straight away. Environmentalists continuously have urged the international community to take decisive actions on this disastrous problem. COP21 conference in Paris could be regarded as a sign of solidarity and amenability for tackling the problem of climate change. The primary aim of COP21 is to prevent the increase of global temperature under 2 ◦C and also to limit this increase under 1.5 ◦C compared to the time before industrialization (Armin Razmjoo et al.,2019). Most of the developed countries, particularly the EU members, like the Nordic Countries have shaped and adopted coherent policies on renewable energies and low-carbon technologies over the last two decades to satisfy the strong desire https://doi.org/10.1016/j.egyr.2019.05.004 2352-4847/©2019 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 620 S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 of carbon-free societies (Ardakani et al.,2018;Hosseini et al., 2017). This is while most of the third world countries such as Iran still have not put forward a strong action plan in order to tackle the climate change problem (Armin Razmjoo et al.,2019). Table 1 defines Iran’s policies on sustainable development and low-carbon technologies in summary. Because failed to put forth applicatory action plans to follow these policies, the country still has not achieved the desired results from the formulation of development policies. After Venezuela, Saudi Arabia, and Canada, Iran has the largest proven oil resources in the world (Davarpanah and Mirshekari, 2018). The country also advantageously have the biggest natural gas reservoirs after Russia. The abundance of fossil resources in this country, and deficiency in the formulation of integrated energy policies, as well as, inconsistency in the implementation of the existing plans have vastly deteriorated the climate situation. This is why the energy use per capita in this country is 57% higher than the global average (Bank,2019b). Moreover, Iran has the seventh position among the highest CO2emitting countries as shown in Fig. 1 (ATLAS,2016). Iran’s CO2emission has had a jump between 1960 and 2014. The figure has risen by more than 17 times in this period (Bank,2019a). After all, Iran is one of the signatories of the Paris Agreement and has a commitment to reduce 4% of the overall CO2 emissions by 2030. However, Iran’s deputy environment chief has stated that there is a potential for a 12% reduction in the case of receiving international support and no further sanction exposure. The agreement has its particular opponents and proponents in the government body. Opponents mainly point to the significant expenses of this fulfillment and remark that the developed countries must incur the relevant expenses. The other reason is fear of further sanctions and embargo due to environmental reasons. But Iran, either way, is going to the expenses of buying the CO2share of the other countries in the case of nonperformance. As a result, controversy has been arising over the Iran retraction after the US withdrawal from the Paris Agreement and the Joint Comprehensive Plan of Action. Along with the serious administrative problems in decision-making, cultural and behavioral problems also have been considered as the other drivers of climate change in this country. Indeed, findings suggest that social and economic issues had been the first priorities of the Tehran residents in the 2000s and the environmental issues were of little importance (Calabrese et al.,2008); however, serious concerns have been arising over the air and water quality of Iran in the recent years. The evidence against environmental problems in this country is overwhelming and the relevant authorities need to formulate comprehensive policy action plans in order to tackle the abovementioned problems (Amirnekooei et al.,2012; Lotfalipour et al.,2013). This paper aims to make a prediction about the environmental state of Iran in 2030 and its commitment to the Paris agreement. For this, a literature review over the potential driving forces of CO2emissions will be conducted in order to determine the prominent determinants of CO2emissions in Iran. With the acquisition of historical data from valid databases and analysis of their trends over time, the identified determinants then are used in two statistical approaches, multiple linear regression (MLP) and multiple polynomial regression (MPR) models, in order to develop prediction models for estimation of CO2emissions between 1971 and 2014. From the regression analysis, a linear equation and a polynomial equation will be derived which relate the annual CO2emission to the identified determinants. Through a variety of forecasting methods, including expert neural net, hybrid forecasting method, quadratic growth, linear exponential smoothing, single exponential smoothing, damped exponential smoothing, triple exponential smoothing, linear growth, moving average, polynomial growth, and double moving average, the predictors will be forecasted by 2030. Predictive models with the lowest mean absolute percentage error (MAPE) will be selected for forecasting of the determinants. Finally, with putting the forecasted predictors into the derived equations, annual CO2emission values will be projected by 2030. Two scenarios also have been designed in this study to scrutinize the possible influence of adaptation of different policies over the state of Iran in terms of CO2emission in 2030. The first scenario considers insignificant changes in the current policies of the government; while, the second one considers the targets of ‘‘Law on the Sixth Five-Year Economic, Cultural, and Social Development Plan for 1396–1400 (2016–2021)’’ — ‘‘The Sixth Development Plan’’. 2. Literature review 2.1. Driving forces of CO2emission in Iran At the early 2000s, the International Energy Agency (IEA) strongly urged the country members, particularly, the developing countries to reconsider the policy over energy subsidy. In most cases, this action results in a small increase in economic growth and a great reduction in energy consumption and CO2 emissions(1999). Despite the IEA’s notifications in about the last two decades, Iran still has not changed its policy and allocates a large proportion of the GDP to the energy subsidies in all sectors. Actually, China is in the face of stiff competition from Iran on this matter. This is a legitimate reason why per capita CO2emission and per capita energy use are 66% and 57% higher than the global average (Bank,2019a,b). Population is considered as a major driving force of CO2 emissions in the majority of regions (O’Mahony,2013;O’Mahony et al., 2013;Xu et al.,2014). Although Begum et al. (2015) failed to find a meaningful relationship between CO2 emissions and population in Malaysia, a number of studies (Ardakani et al.,2018;Hosseini et al.,2017;Jorgenson and Clark,2010;Xu et al.,2014;Yang et al.,2015) feature a positive correlation between population and CO2emissions in both developed and developing societies. Iran gradually became an energy-intensive society after the revolution in 1979. In fact, steep population growth between 1977 and 1990, particularly, during the Iran–Iraq War triggered serious problems during the post-war years. There is a growing alarm at population growth and per capita energy consumption and CO2emission in this country. Ministry of Health and Medical Education predict that Iran’s population will reach the top in 2050 and fall into a decline afterward. Therefore, Iran needs to adopt flexible policies toward sustainable energy systems to secure the energy supply and environmental health in the coming years. Fig. 2 shown Iran’s population growth during the past years. Globally and regionally, energy consumption is the main driving force of anthropogenic CO2emissions. Since CO2is a product of fossil fuel consumption, therefore, energy consumption bears a positive relationship with CO2emission. As a result, CO2emission is drastically high in the major fossil fuel consuming countries, particularly, China and the U.S. (Apergis and Payne,2010;Heidari et al.,2015;Kais and Sami,2016;Kaivo-oja et al.,2016;Soytas et al.,2007). Fig. 3 shows the growth trend of per capita energy use in Iran between 1971 and 2014. As the population, energy use also has been on increase during this period. Fossil fuels are the main sources of energy in this country. Actually, a significant proportion of Iran’s CO2emissions is related to the consumption of vehicle fuels in the transportation sector and natural gas in the building sector in this country. Fig. 4 shows the composition of Iran’s carbon footprint in 2014. Despite the emissions from coal consumption, the per capita CO2emissions from the other sources are significantly higher than Europe’s average. That is S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 621 Table 1 Iran’s policies on sustainable development in the last decade. Policy Year Policy sector Agency Targets Liquid fuel exchange purchase 2013 Electricity, Framework policy Ministry of Energy — Renewable Energy Organization of Iran (SUNA) •Improvement of energy efficiency in governmental and private power plants; •Development of renewable energies; •Optimization of energy consumption. Supply of 20% of electricity consumed by governmental and public organizations from renewables 2016 Electricity, Framework policy Renewable Energy and Energy Efficiency Organization (SATBA) •Supply of 20% of the electricity demand of governmental and public buildings from renewable energies for two years; •Set of electricity consumption tariff based upon the renewable feed-in tariff. Renewable portfolio standards — The Sixth Development Plan 2016 Electricity, Framework policy Renewable Energy and Energy Efficiency Organization (SATBA) •Enhancement of share of carbon-free energies in the energy mix to 5% by the end of 2021. Renewable electricity compliance 2015 Framework policy Renewable Energy and Energy Efficiency Organization (SATBA) •Introduction of tariff as renewable electricity duty in electricity bills to develop and maintain rural electricity grids. Payment of benefit of conserving fossil fuels 2015 – Renewable Energy and Energy Efficiency Organization (SATBA) •Improvement of energy efficiency in all sectors; •Development of renewable energies; •Increase in natural gas consumption in large cities; •Production of hybrid, fuel efficient, and electric cars; •Reduction of cost of freight; •Improvement of energy efficiency through construction of combined heat and power facilities and small scale generation power plants; •Reduction of generation and transmission losses; •Electrification of agricultural wells using renewable technologies. •Involvement of electricity instead of fossil fuels in the wealth-producing activities; •Production of electricity from waste plants. Renewable energy development fund 2013 Electricity, Framework policy, Multi-sectoral policy Ministry of Energy — Renewable Energy Organization of Iran •Introduction of tariff by an amount of 30 Rials per kilowatt-hour as electricity duties to develop and maintain rural electricity grids. Promotion of awareness about renewable technologies 2011 Framework policy, Multi-sectoral policy Science and technology deputy of presidency — Renewable energies technology development headquarters •Acquisition of technical know-how on production of renewable technologies and construction of renewable power plants; •Allocation of research and development funds for renewable projects; •Concentration on realization and localization of renewable technologies in the country; •Introduction and implementation of pilot projects on renewable energies. Financial support for the feed-in tariff 2005 Electricity Ministry of Energy — Renewable Energy Organization of Iran (SUNA) •Undertaking a definite proportion of electricity generation cost via renewable technologies by Iran Department of Environment; •Guaranteed purchase of electricity generated via renewable technologies by subsidiary companies of Ministry of Energy. Financial supports 2011 – Ministry of Energy — Renewable Energy Organization of Iran (SUNA) •Persuasion of investment in renewable energies. Renewable portfolio standards 2012 Electricity, Framework policy, Multi-sectoral policy Ministry of Energy — Renewable Energy Organization of Iran (SUNA) •Installation of 5000 MW solar and wind energy by 2020. mainly because Iran has limited is agenda to crude oil, while it has 1.15 billion tons of proven coal reserves according to the remarks of deputy head of Middle East Mines Industries Development Holding Company (MIDHCO). Natural gas production projects also have not sufficiently financed by the government, despite the huge potential in this country. Increase in the share of renewables and nuclear energy in energy mix also can exert a negative influence over CO2emissions in both short-term and long-term (Apergis et al.,2010;Baek, 2016;Cansino et al.,2015). For those trend-setters of climate change, China and the U.S., who failed to involve renewable energies in the wealth production activities, GDP still bears a close relationship with CO2emissions. This is while in some part of the world such as European Union (EU) Member States who efficiently raised the share of carbon-free energies in their energy mix, such relationships between CO2emissions and GDP are not observable (Kaivo-oja et al.,2016). Technological development in the field of nuclear energy may have a negative impact on CO2 622 S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 Fig. 1. Top CO2emitting countries in 2016 (ATLAS,2016). Fig. 2. Iran’s population growth between 1960 and 2017. Fig. 3. Growth trend of per capita energy use over the years in Iran. S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 623 Fig. 4. Composition of CO2emission in Iran and European countries. emissions in the long-term (Apergis et al.,2010;Baek,2016). Development of renewable energies also can produce the same result, however, only in the short-term (Baek,2016). About 79% of Iran electricity production in 2015 was from natural gas power plants according to the Ministry of Energy’s data. This is while the share of nuclear power plants and renewable was only 1.7% of the total. As per Fig. 5, the share of fossil fuels in power generation has increased in the past three decades and reached the top in 2007. About 97% of Iran power generation was from fossil fuels in this year. However, the figure shows a slight decline trend afterward. Industry and building sectors are the major electricity consumers in this country. Moreover, Iran exports electricity to the neighboring countries, particularly, Iraq. Therefore, Iran needs practical policies to cut CO2emission intensity in power generation to meet the Paris agreement’s target. Trade level refers to the level of imports and exports per year in a country. Since it bears a close relationship with Gross Domestic Product (GDP), the factor generally forges a strong relationship with CO2emissions. As a result, any rise in the trade level can simply result in a higher level of CO2emission in some particular countries, such as China, India, Turkey, and Malaysia (Halicioglu, 2009;Jayanthakumaran et al.,2012;Ozturk and Acaravci,2013; Shahbaz et al.,2016). Technological progress also could have a major impact on the amount of CO2emission. Development of energy infrastructure has brought about lower CO2emissions in both developed and developing societies (Liou and Wu,2011). Energy Intensity (EI) has been suggested as a measure of technological progress in the literature which directly influences the amount of CO2emission. As a result, a low EI implies a high level of energy-efficient and innovative energy systems in society (Donglan et al.,2010;Yue et al.,2013;Zhang and Tan,2016). Iran also suffers from a high energy intensity. The energy intensity level of primary energy in Iran was 7.794 Mj per PPP GDP in 2015 according to the World Bank Data. Considering that a large proportion of Iran’s GDP is supplied from crude oil export, and Iran has not an energy-intensive industry, the energy intensity is drastically high in this country. Moreover, due to the reduction in crude oil price and the US sanctions against Iran after 2011, Iran has faced a reduction in oil export revenue, but the energy intensity was in increase between 2011 and 2015. This is mainly because of a dramatical increase in the energy consumption of Iran during this period. Fig. 6 shows the per capita GDP of Iran between 1971 and 2014. In fact, the increase in energy use between 2011 and 2014, alongside the promotion of fuels quality standards have slightly reduced the CO2intensity per energy use in this country. Fig. 7 illustrates the trend of CO2intensity in Iran between 1971 and 2014. Adoption of constructivism policies like electrification, supply natural gas to the majority of residential buildings, utilization of electric water pumps instead of diesel engines, and production of vehicles fuels in compliance to the Euro Standards have triggered a slight decline trend in the CO2intensity during the post-war years. However, as per Fig. 7, these policies already became inefficient and only may stabilize the level of CO2intensity per energy consumption in this country. Actually, Iran needs to adopt more flexible policies regarding the renewable energies and CO2capture facilities if have a strong desire to lower the CO2 intensity and meet the Paris Agreement’s target. Determinants of CO2 emissions are various and sundry and encompass a wide range of socioeconomic and technological factors. Inasmuch as there is a consensus among the researchers on these driving forces, determinants of CO2emissions may vary from region to region because of their particular economic structure and energy system. With reviewing the possible driving forces of CO2emissions in the literature and scrutiny of their historical trends in Iran, this paper finds population, CO2intensity, GDP per capita, share of fossil fuels in electricity production, and per capita energy use as the major driving forces of CO2emissions in this country. Through regression analysis, these factors will be used as predictors for the development of forecasting models to estimate CO2emission by 2030. The next section provides a brief overview over the possible approaches adopted for forecasting CO2emission in the last two decades. 2.2. Common approaches for forecasting CO2 emissions Numerous approaches have been adopted for forecasting CO2 emissions in the short-term and long-term. These approaches mainly forecast the future trends based on the historical data and encompass a wide range of regression models and machine learning tools combined with scenario analysis. A number of studies have applied decomposition-based approaches to forecasting CO2emissions. These papers suggest a variety of forecasting models, including the STIRPAT model (Noorpoor and Kudahi,2015;Wen and Liu,2016), Kaya model (O’Mahony,2013), and the IPAT model (Qiang et al.,2015). However, given the major improvements in environmental policies and considerable development in modern technologies, in addition, due to the complex nature of CO2production from various resources and abundance of the driving forces, traditional approaches seemingly fail to accurately forecast CO2emissions. 624 S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 Fig. 5. Growth trend of electricity production from fossil fuels in Iran. Fig. 6. Iran’s per capita GDP between 1971 and 2014. Fig. 7. Iran’s CO2intensity between 1971 and 2014. Forecasted results from the traditional models may exhibit different trends from the historical data. As a result, novel and innovative methodologies are needed to efficiently address the issue (Guo et al.,2018). Scenario analysis has received considerable attention for this purpose. With the consideration of possible solutions and alternatives for the future, scenario analysis reduces the uncertainty of outcomes and make a better projection about the future trends S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 625 of CO2emissions. A large number of studies have applied scenario analysis to forecasting CO2emissions. As an illustration, a combination of the scenario analysis and a system dynamic model was applied to forecasting CO2emissions in China by 2020 (Xiao et al.,2016). Scenario analysis also has been applied to scrutinize the long-term solutions for reducing energy utilization and CO2 emissions of the steel industry in China (Karali et al.,2016). Another study has adopted the methodology in a combination with a linear programming model to prioritize alternative solutions to the problem of global warming (Tokimatsu et al.,2017). Machine learning tools, in particular, artificial neural network (ANN) models have been widely used in the forecasting processes. These methodologies best suit to the cases which there are nonlinear relationships between target and response variables (Guo et al.,2018). Combining the ANN and a bees algorithm, Behrang et al. (2011) have speculated about the state of global CO2emissions in 2040. Using a Back-Propagation Neural Networks model combined with a genetic algorithm (GA), Sun et al. (2016) also forecast future values of CO2emissions for a province of China. Fang et al. (2018) offer a novel methodology for forecasting of CO2emissions in China by 2020. The approach is a compound of Gaussian Process Regression (GPR) and genetic algorithm and has a good performance in terms of accuracy. A number of researchers also put efforts into the forecasting of CO2emissions in Iran. Davoudpour and Ahadi (Davoudpour and Ahadi,2006) have proposed a scenario-based econometric model to forecast annual energy demand and CO2emissions of Iran between 2000 and 2011. From scenario analysis, possible impacts of pursuing energy efficiency measures and price adjustment policy on the total energy demand and CO2emissions have been apprised during the 2000s. Under a business-as-usual (BAU) scenario, the yearly growth rate of CO2emissions has been obtained 6.8%; while, with the implementation of a management scenario, the country most likely could have lowered the rate to a half extent. Köne and Büke (2010) have applied regression analysis to forecast CO2emissions from the top 25 CO2emitting countries by 2015 and 2030. Based on the analysis, CO2emission in Iran almost has followed a linear increasing trend in the majority of times. With the application of a grey box model and an Autoregressive Integrated Moving Average (ARIMA) model, Lotfalipour et al. (2013) have forecasted Iran’s CO2emissions by 2020. With superior performance, the grey box model has estimated that Iran’s CO2emissions will cross the border of 925 million tons by 2020. Azadeh et al. (2017) have performed a regression analysis in an effort to forecast CO2emissions from Iran’s industrial sector based on the socioeconomic indicators. Heydari et al. (2019) have proposed a novel methodology, combining a General Regression Neural Network model with a Grey Wolf Optimization approach, for forecasting of CO2emissions from fossil fuel consumption in Iran, Italy, and Canada. 3. Methodology This research applies multiple regression (MR) models (Ostertagová,2012) in order to forecast Iran’s CO2emissions based on the historical data derived from World Bank Open Data. Two models have been developed using Matlab: MLR and MPR. Moreover, using proForecaster 2.1, this paper adopts various forecasting approaches, i.e. expert neural net, hybrid forecasting method, quadratic growth, linear exponential smoothing, single exponential smoothing, damped exponential smoothing, triple exponential smoothing, linear growth, moving average, polynomial growth, and double moving average, in order to forecast the predictors’ trend between 2015 and 2030. It is worth to mention that the model selection is based on the mean absolute percentage error (MAPE) of the forecasting approaches. This paper splits the dataset according to 70/30 rule. That means 80% of the dataset is used for the training set and 20% for the validation set. Fig. 8 elaborates the forecasting steps in this paper. The next section explains the development procedure of regression models in summary. 3.1. Regression analysis Multiple regression (MR) model is a type of regression models when there are more than one predictors for a response variable. As one of the extrapolation methods, regression analysis assumes that the past is a proxy for the future. Therefore, this paper assumes that future values of CO2emission most likely follow the pattern of historical data. The general form of a multiple regression model is: Y=Xβ+e(1) In this equation, the response vector Yand error vector eare two n-length column vector, βis a column vector of length k+1, and Xis a nby k+1 matrix. The MR estimates the regression parameters using the least square error (Tabasi et al.,2016). The approach firstly computes the sum of the squared errors, then seek for the best estimators minimizing the sum as follow: e=Y−Xβ(2) eTe=(Y−Xβ)T(Y−Xβ) (3) Normal equations will be derived by making Eq. (3) equal to 0, XTX⌢ β=XTY(4) By multiplying both sides of Eq. (4) by (XTX)−1, an unbiased estimator of βwill be calculated as below: ⌢ β=(XTX)−1XTY(5) Therefore, the mean estimated values of Yare calculated by the following equation: ⌢ Y=X⌢ β=X(XTX)−1XTY=HY (6) The symmetric matrix of H(hat matrix) converts Yto ⌢ Y. Finally, the following equation computes the error terms: ⌢ e=Y−⌢ Y=(I−H)Y(7) Eqs. (8) and (9) represents the general form of MLR and second-degree MLP equations respectively. yi=β0+β1xi1+β2xi2+ · · · + βkxik +ui,for i =1,2,3, . . ., n (8) yi=β0+β1x2 i1+β2x2 2+ · · · + βkx2 ik +ui,for i =1,2,3, . . ., n (9) Coefficient of determination is defined as follow: R2=1−∑n i=1(yi−ˆ y)2 ∑n i=1(yi−y)2(10) As well, residual sum of squares (RSS) is calculated from the following equation: RSS = n ∑ i=1 (yi−ˆ y)2(11) 626 S.M. Hosseini, A. Saifoddin, R. Shirmohammadi et al. / Energy Reports 5 (2019) 619–631 Fig. 8. Forecasting process flowchart. 4. Prediction scenarios Iran’s CO2emissions have been in a substantial increase after 1980. As per Fig. 9, between 1975 and 1980, the figure has followed a declining trend, though considerable increased afterward and reached the top in 2015. Additionally, there is a great deal of uncertainty about the future of Iran’s CO2emissions, as the government still have not brought forward specific proposals to tackle the problem of increasing CO2emissions. Further, the abovementioned issues also will expose the future of Iran’s CO2 emissions to considerable uncertainties. Thus, any possible scenario should take these uncertainties into consideration in order to provide an accurate prediction from upcoming CO2emissions. This paper presents two scenarios in order to forecast Iran’s CO2 emissions between 2015 and 2030: business as usual (BAU) and Sixth Development Plan (SDP). 4.1. BAU scenario The BAU scenario considers an insignificant change in the currently available composition of the indicators. In fact, the indicators represent the same trends with their historical data in this scenario. Table 2 contains the relevant information and assumptions regarding the indicators in the BAU scenario. 4.2. SDP scenario On March 19th, 2017, the Iranian Parliament approved the Sixth Development Plan which describes the structural policies of Iran between 2016 and 2021. However, despite the Paris Agreement, Iran has not set any target regarding CO2emissions in the next 5 years. Instead, the Parliament has set three high targets regarding renewables, per capita GDP, and energy intensity as below: •Increase of renewable shares in primary energy consumption by 1% each year; •Increase of per capita GDP (in comparison to constant prices in 2004) by 7.6% each year; and •Reduction of energy intensity by 3% each year. In addition to assumptions of the BAU scenario, the STP takes the three targets into consideration in order to forecast the CO2 emissions by 2030. Though the two first targets influence the MLP model straightforwardly, the third target makes an alteration in both per capita GDP and per capita energy use. The scenario applies the relevant impacts to both determinants and accordingly estimates the amount of yearly CO2emissions. 5. Results Using the MPR, a polynomial function has been derived from the historical data, which represents a strong correlation between the predictors and response variable: y= −1.4e−11x12+1.7e−3x1x2+1.2e−7x1x3(12) +9.4e−7x1x4+4.1e−6x1x5+1341x22 −4x2x3−325.3x2x4+50x2x5+1.7e−4x32 +3.2e−2x3x4−2.2e−3x3x5−29.2x42+1.9e −1x4x5−1.7e−2x52−5.6e−3x1−51051.3x2 +3.7x3+5649.9x4−170.9x5−17426.4 where, x1is the total population, x2represents CO2intensity in kg per kg of oil equivalent energy use, x3is per capita GDP in US$, x4is the share of electricity production of fossil fuel sources, and x5represents per capita energy use in kg of oil equivalent. R2for the training set is 0.99 and RSS is 13927877.94. For the validation set, R2is 0.98 and the RSS is 14865647.54. This implies that the model best fits the predictors and correlations between the estimated and actual data concentrate in close proximity of the ideal fit. Eq. (13) represent the function derived from the MLR methodology. y=3.9e−3x1+42256.6x2+6.7x3−274.7x4