Driving Forces of Air Pollution in Ulaanbaatar City Between 2005 and 2015: An Index Decomposition Analysis

: Ulaanbaatar, the capital city of Mongolia, is facing serious air pollution challenges— especially during the cold and long winter months—mainly due to fossil fuel combustion. This study investigates the socioeconomic drivers of the sulfur dioxide (SO 2 ), nitrogen dioxide (NO 2 ), and particulate matter (PM 2.5 ) concentration changes in Ulaanbaatar between 2005 and 2015 by applying the index decomposition analysis (IDA) method. Five socio-economic driving forces are considered in the decomposition analysis. All the driving forces contributed to more air pollution concentration changes in 2015 than in 2005, despite the decreasing trends of decomposition results for the period of 2010–2015. In general, economic growth, pollution intensity, and energy intensity signiﬁcantly contributed to the changes of air pollutant concentrations, while energy structure and population growth had marginal e ﬀ ects. Finally, appropriate policy recommendations are proposed to the local government so that they can initiate feasible policies to e ﬀ ectively reduce air pollution, protect human health, and respond to climate change in Ulaanbaatar.


Introduction
Migration to Ulaanbaatar, the capital city of Mongolia, has increased the population by 55% over the past fifteen years. Now, it is estimated that 1.4 million residents are living in Ulaanbaatar [1]. With its increasing population, this city has experienced rapid economic growth [2]. In particular, this city relies on fossil fuels (see Appendix A, Figures A1 and A2) to achieve its gross domestic product (GDP) growth and to support its rapid urbanization, leading to concerns of both air quality deterioration and climate change. Located in the dry North Asia, Ulaanbaatar has a very cold winter, with the minimum temperature below −35 • C in January [3]. The heating season lasts seven months, from mid-October to mid-May [4]. Unfortunately, Ulaanbaatar has very limited access to petroleum or natural gas and has to rely on coal for both power generation and heat supply due to its abundant coal reserve. Over 790,000 people reside in settlements without water, sanitation, or basic infrastructure-often in traditional Mongolian felt tents, known as gers, where these people rely on wood or coal burning
Such studies can help to trace the economic, environmental, and social trends [31]. Major research methods include structural decomposition analysis (SDA) and index decomposition analysis (IDA) [32][33][34], which have been widely applied to analyze the impacts of economic growth,  [21]).
Under such circumstances, the purpose of this study is to uncover the key drivers of SO 2 , NO 2 , and PM 2.5 concentrations in Ulaanbaatar, including pollution intensity, energy structure, energy intensity, Sustainability 2020, 12, 3185 3 of 17 economic growth and population growth, by employing the Logarithmic Mean Divisa Index (LMDI) method over the period of 2005-2015. We expect that valuable recommendations can be offered to the local government so that they can prepare effective mitigation policies. We also hope that the policy implications gained from this paper can be disseminated to other cities with similar challenges. The remaining part of this paper is structured as the following: Section 2 reviews relevant literature; Section 3 details research methods and data; Section 4 presents research results; Section 5 discusses policy implications; finally, Section 6 summarizes the research conclusions.

Literature Review
Globally, there are numerous studies on the socioeconomic driving forces of different emissions, although the majority of these studies focus on carbon emissions [22][23][24], and on PM 2.5 , SO 2 , and NO 2 [25][26][27][28]. In particular, several studies on the changes of PM 2.5 air concentrations are studied [29,30].
Such studies can help to trace the economic, environmental, and social trends [31]. Major research methods include structural decomposition analysis (SDA) and index decomposition analysis (IDA) [32][33][34], which have been widely applied to analyze the impacts of economic growth, technology changes, and sectorial shifts by using different indicators [31].
By applying an input-output model to the SDA method, it decomposes relevant changes into different indicators. This method uses an additive form and a multiplicative form in order to complete the decomposition analyses [35]. One of the advantages of this method is that it takes the advantages of the sector details from the input-output tables [31]. For example, with the application of SDA, Rose and Chen [36] analyzed the changes in sectorial energy consumption in USA. Similarly, Cansino et al. [37] studied Spanish CO 2 emissions from structural changes. In addition, Yamakawa and Peters [38] uncovered the key driving forces on greenhouse gas emissions in Norway by applying this method.
The IDA method can uncover the process of energy supply, energy conversion, and energy consumption so that key driving factors, such as end-use energy structure, energy mix in electricity generation, and electricity generation efficiency, can be quantitatively identified. Ang and Zhang [39] conducted a global review on this method and found that the IDA method is more suitable for analyzing energy systems, since it pays more attention to technical details. Due to this advantage, this method has been applied in analyzing energy-related carbon emissions-in particular, in the manufacturing and heavy industrial sectors [40,41]. Furthermore, several studies applied the IDA method to uncover the driving forces of air pollutant emissions and concentrations. For instance, at a national level, the main driving factors inducing changes in China's air pollutant emissions between 1997 and 2002 were investigated. The major findings of this study showed that the main drivers on inducing more PM 2.5 , SO 2 , and NO x were economic growth and energy intensity [28]. Another study conducted by Zhang et al. [30] focused on investigating the key drivers of PM 2.5 concentrations at a municipal level so that useful policy recommendations were raised to the municipal authorities to mitigate their PM 2.5 pollutions.
From the above introduction, it is clear that most existing studies focus on one specific air pollutant concentration without combining these different air pollutant concentrations together. In fact, the combustion of fossil fuels (in particular, coal) can lead to simultaneous concentrations of PM 2.5 , SO 2 , and NO x . However, few studies focus on other pollutants than PM 2.5 . Therefore, it is crucial to initiate integrated studies on all three air pollutants so that more effective policies can be prepared to improve the overall air quality. With regard to Ulaanbaatar, the existing relevant studies focus on PM 2.5 concentrations, chemical components, and its human health impact [6, 16,42,43], without examining the driving forces of air pollutants. Consequently, the literature review provides rationale to conduct this research.

Methods
IDA can not only identify the processes of energy supply, energy conversion, and energy consumption, but can also uncover critical driving forces, including end-use energy structure, energy mix in electricity generation, and electricity generation efficiency. Unlike IDA, SDA relies on the input-output table released by the official statistical agencies. Such a table is not available every year and its accuracy is also not reliable in many countries where statistic infrastructure is still weak and data are not accountable. On the contrary, IDA only uses aggregated sector information and has the advantage of a reduced data requirement and a simpler presentation of results. LMDI is one of the most widely used IDA methods [44,45]. This method is path-independent and provides a perfect decomposition, which can resolve the problem of remaining unexplained residual terms in the decomposition results [45]. Moreover, this method can solve zero-value problems [46]. Owing to these reasons, this paper applies the LMDI method in order to investigate the driving forces of air pollutants in Ulaanbaatar City.
According to Ang [33], V is an energy or environmentally-related aggregate. Assuming that there are n factors driving the changes in V over time, each factor is associated with a quantitative variable, whereby there are n variables, including X 1 , X 2 . . . . X n . Let subscript i be a sub-category of the aggregate, and V i expressed as the product of X 1,i , X 2,i , . . . and X n,i . Therefore, the general IDA identity is shown in Equation (1): The difference of the aggregate during the period 0 to T can be further decomposed, as shown in Equation (2): The sum of absolute change driven by each variable should be equal to the total absolute change of the aggregate. The term on the right side of Equation (2) is the effect associated with the respective factors in Equation (1).
Five key driving forces are investigated for uncovering the SO 2 , NO 2 , and PM 2.5 concentration changes in this study, including pollution intensity, energy structure, energy intensity, economic growth, and population growth. The change in each factor helps to quantify all three air pollutant concentration changes from fossil fuels. The IDA identity in Equation (1) can be rewritten as Equation (3): where C i represents different air pollutant concentrations; i refers to SO 2 , NO 2 , and PM 2.5 concentrations, respectively; F is fossil fuel consumption; E represents total energy consumption; GDP represents gross domestic product (GDP); P represents population growth; CI = C/F represents pollution intensity; ES = F/E represents energy structure; EI = E/GDP represents energy intensity; Y = GDP/P represents economic growth; P represents population growth (see Table 1 for the definition of variables).
The additive form gives direct information about the magnitude of concentration changes by different decomposed factors. This form is adopted in this study so that the results can be calculated from different aspects. Equation (4) shows how to calculate the changes of SO 2 , NO 2 , or PM 2.5 concentrations.
where ∆C i represents the pollution changes of different air pollutants; i represents SO 2 , NO 2 , or PM 2.5 , respectively; ∆C CI represents pollution intensity effect; ∆C ES represents energy structure effect; ∆C EI Sustainability 2020, 12, 3185 5 of 17 represents energy intensity effect; ∆C Y represents economic scale effect; ∆C P represents the population scale effect. These five effects can be further calculated based on Equations (5)-(9), respectively: The pollution intensity effect reflects the impact of the change in fossil fuels mix. The energy structure effect captures the influence of the change of fossil fuel shares in total energy consumption. The energy intensity effect denotes the effect of the change of economic dependence on energy use. The economic growth effect indicates the impact of the change of GDP per capita. The population growth effect denotes the effect of the population change living in the city. By applying this decomposition method, the total changes in SO 2 , NO 2 , or PM 2.5 concentrations in Ulaanbaatar City can be obtained by the sum of the five effects.

Data Collection
Due to data accessibility, the related data between 2005 and 2015 were collected from different sources. Ulaanbaatar's primary SO 2 , NO 2 , and PM 2.5 concentration data were gathered from the National Statistical Information Center [21]. The final energy consumption data were collected from the IEA World Energy Balances database [47]. The GDP data were collected from the World Bank national accounts data [48] and the population data were acquired from the Statistic Department of Ulaanbaatar [49].

Results
Using Equations (5)-(9), we calculated the contribution rates of pollution intensity, energy structure, energy intensity, economic growth, and population growth to the changes of SO 2 , NO 2 , and PM 2.  Tables A1-A4 respectively (See Appendix B, Tables A1-A4). Figure 2 shows the decomposition results of SO 2 concentration. We can see that the most important factor that contributed to the increase of SO 2 Figure A3). Normally, economic growth can result in high energy consumption [50]. In this regard, the total energy consumption of Ulaanbaatar increased from 2960 ktoe in 2005 to 4714 ktoe in 2015, which followed this relationship. On the contrary, pollution intensity played a crucial role in reducing SO 2 concentration in most of the years during 2005-2015. The pollution intensity effect of the SO 2 concentration reduction increased by 6.8 times, ranging from 2.96 µg/m 3 in 2005 to 23.11 µg/m 3 concentration in 2015. Interestingly, the pollution intensity effect of SO 2 concentration decreased from 2011 to 2015, coupled with the decrease of the economic growth effect during the same period. In 2010, the SO 2 concentration level increased by 85%, and this surge affected the pollution intensity effect, which grew from 6.5 µg/m 3 in 2009 to 28.9 µg/m 3 in 2010. In contrast, the pollution intensity effect of SO 2 decreased to −21.1 µg/m 3 in 2013 due to the reduced concentration level of SO 2 , which was around 23% compared to the previous year. The pollution intensity of SO 2 concentration was -4.2 µg/m 3 in 2014 and decreased to −23.1 µg/m 3 in 2015. As a result, the SO 2 concentration level reduced by 53%.
Sustainability 2020, 12, x FOR PEER REVIEW 6 of 17 national accounts data [48] and the population data were acquired from the Statistic Department of Ulaanbaatar [49].

Results
Using Equations (5)-(9), we calculated the contribution rates of pollution intensity, energy structure, energy intensity, economic growth, and population growth to the changes of SO2, NO2, and PM2. 5         Finally, with regard to PM 2.5 concentration, pollution intensity was the dominant factor, with 17.47 µg/m 3 of PM 2.5 concentration growth during the first sub-period (2005-2010), followed by economic growth of 15.95 µg/m 3 . In contrast, population growth and energy structure had marginal contributions to the PM 2.5 concentration increase during the first sub-period. Energy intensity remained the largest mitigation factor of PM 2.5 concentration during the two sub-periods. The concentration−reduction effect of energy intensity had little increase, from −11.63 µg/m 3 during 2005-2010 to −12.14 µg/m 3 during 2010-2015. Pollution intensity was the second driving factor of the PM 2.5 concentration reduction, while energy structure exerted the least mitigating effect on PM 2.5 concentration during the second sub-period (2010-2015). On the contrary, economic growth was the largest contributor to the PM 2.5 concentration increase during 2010-2015, followed by population growth. As a result, the PM 2.5 concentration increased by 170% during the first sub-period and decreased by 7% during the second sub-period. SO 2 and PM 2.5 are pollutants that can primarily be associated with the types and amounts of combusted fuels. Therefore, it can be concluded that the contents of the sulfur and ash were responsible for the concentrations of SO 2 and PM 2.5 , given the fact that fossil fuel combustion in ger districts is a major contributor to air pollution in Ulaanbaatar.

Discussions
Despite the fact that the SO 2 , NO 2 , and PM 2.5 concentrations in Ulaanbaatar decreased during the second sub-period (2010-2015), the concentration levels remain higher than the standard level set by the World Health Organization [10]. This decrease could be associated with the actions taken by the local government's effort to subsidize electricity costs in ger districts to combat the air pollution. However, the combustion of fossil fuel remains the largest source of air pollution in the city and it can be concluded that the contents of the sulfur and ash are responsible for the concentrations of air pollutants. Hence, more efforts should be made to improve the local air quality in Ulaanbaatar.
According to the decomposition results, the effect of economic growth was the key factor for the increase of all three air pollutants in Ulaanbaatar, especially during the rapid economic boom period that happened between 2009 and 2011. While this economic growth effect played a significant role in the increase of air pollutants, pollution intensity and energy intensity had significant effects on the decreases of SO 2 , NO 2 , and PM 2.5 concentrations in most years between 2005 and 2015.
Throughout the two sub-periods, the contributions of pollution intensity to the SO 2 , NO 2 , and PM 2.5 concentrations fluctuated. However, its overall effect on those concentrations decreased in 2015, compared to 2005. This negative contribution was due to the increased use of electricity, leading to the pollution intensity effect on PM 2.5 concentrations becoming positive in 2015 compared to its negative effect in 2005. Although the energy structure factor led to the increase in all three pollutant concentrations, these changes were not significant enough to make any differences. Interestingly, its effects on air pollutant concentrations started to decrease during 2012-2015.
Lastly, the contribution of population growth reduced for all three air pollutants in the second sub-period compared to the first sub-period. As per the data released by the Statistic Department of Ulaanbaatar [51], the number of people living in the ger districts increased between 2010 and 2018, in line with the overall population growth of Ulaanbaatar. Thus, there was no direct link between population growth in the ger districts and the population growth factor.
Sustainable urbanization requires that urban development processes should not deteriorate the local environment, while still meeting the demands of the local population [13]. According to the Green Development Strategic Action Plan, Ulaanbaatar should consider effecting the environmentally sustainable policies, which will achieve inclusive economic growth, active public participation, and a safe and healthy living environment for its citizens [52].
Developing countries often face a dilemma with their pollution control, as their efforts may undermine economic growth. It is crucial for developing countries to provide cheap electricity to support their industries and citizens. Renewable energy requires significant investment despite the low operating cost. Therefore, it is not easy to widely promote renewable energy applications in the short term due to the lack of financial resources. However, innovative mitigation policies should be prepared to reduce air pollution and protect human health, even without the wide promotion of renewable energy. For instance, district heating is one effective measure to reduce the overall air pollutants. Furthermore, the local government may set up pollution limits from a range of sources and ban or limit the used motor vehicle imports. Financial subsidies should be provided to those factories that produce semi-coke briquettes and use energy efficient stoves. Moreover, long term action plans should focus on renewable energy production. For instance, there is an opportunity to promote solar water heating systems for new housing complexes because, annually, there are 250 sunny days in this city [53].
Practically, "the 2017 Mongolia National Program" states to reduce air pollutants by 80% in 2025 [54] and calls for phasing out the practice of unprocessed coal, excluding three power plants in Ulaanbaatar. In order to achieve this target, the local government initiated several efforts. For example, in 2016, the local government offered subsidies for more efficient wood and coal burning stoves and provided free electricity to several ger districts during the night. Another measure was to promote the use of electrical heat generators and the application of central heating systems in the ger districts. However, such efforts were not sufficient, since only 3.6% of total final energy consumption in this city in 2015 was based upon renewable energy sources [55]. Consequently, it is critical for the local government to actively encourage the promotion of renewable energy, such as solar power and wind power. In addition, geothermal power, such as the use of ground source heat pumps, can significantly reduce the overall use of fossil fuels and corresponding air pollutants and greenhouse gas emissions [56]. Therefore, the local government should support the potential application of such innovative technologies.
Improving coal quality while reducing coal consumption is another measure that should be further applied in Ulaanbaatar, especially for controlling energy intensity-driven concentrations. According to the World Bank study [13], coal is expected to be the primary energy source for the next 5-10 years in Ulaanbaatar. Facing such a challenge, the local government is currently promoting the use of improved coal, such as semi-coke briquettes. Also, as one measure for cleaning up this city's air quality, the local government banned the usage of raw coal in May 2019. However, the real implementation of this policy is not effective, since raw coal is cheaper and can be easily obtained by those in the ger districts. Under such a condition, the initiation of more capacity-building activities has been suggested in order to educate different stakeholders so that they can realize the potential damages to their health and local air quality. Similar efforts should also be initiated to grassroot enforcement officials so that they can better enforce related policies. Moreover, the local government should support more research and development (R&D) activities so that more energy efficient technologies will be available, such as energy cascading technology, energy efficient stoves, and energy audits.
Finally, accurate and timely data are essential for effective air quality management. As a developing country, Mongolia is lacking advanced information and communication technologies. Local decision makers find it difficult to obtain relevant information. Thus, the local government may consider investing to build a dynamic information platform so that all the related information and data can be integrated. Such an information platform can serve as one decision support system so that different stakeholders can share the data and information. It can also be beneficial for the decision makers to identify the key areas with the application of a geographical information system (GIS) and predict the potential impacts of different policies. Besides this information platform, it is crucial for the local government to install on-site sensors in the key pollution sites (such as the ger districts and coal burning power plants) so that real data can be collected and local air quality can be monitored. In order to make sure that such an integrated system can function well, it is recommended that the local government should collaborate with the local research institutes or universities so that the whole system can be operated and maintained more effectively.

Conclusions
The serious air pollution in Ulaanbaatar receives increasing attention. The reliance on coal has further exacerbated this problem. Under such conditions, this study investigates the changes of three air pollutant concentrations (including SO 2 , NO 2 , and PM 2.5 ) for the period of 2005-2015 in Ulaanbaatar by applying the LMDI method so that the driving forces for the changes of these air pollutant concentrations can be identified. Five socio-economic driving forces are considered in the decomposition analysis: i.e., pollution intensity, energy structure, energy intensity, economic growth, and population growth. The results showed that all the driving forces contributed to more air pollution concentration changes in 2015 than in 2005. Economic growth, pollution intensity, and energy intensity significantly contributed to the changes of air pollutant concentrations, while energy structure and population growth had marginal effects. Economic growth was the major factor that contributed to the increase of SO 2 , NO 2 , and PM 2. The effects of pollution intensity on SO 2 , NO 2 , and PM 2.5 concentrations were significant, but with volatile intervals of pollution increase and decrease.
The effects of all factors, excluding energy intensity, decreased during the second sub-period compared to the first sub-period. Economic growth maintained to be the largest contributor to the increase of three air pollutant concentrations in both of the two sub-periods. In contrast, the energy intensity factor was the mitigating factor during the two sub-periods. The effect of pollution intensity transformed from positive to negative on SO 2 , NO 2 , and PM 2.5 concentrations from the first sub-period (2005)(2006)(2007)(2008)(2009)(2010) to the second sub-period (2010-2015). Population growth and energy structure had relatively small effects on the three pollutants.
In order to respond to this air pollution issue, the Ulaanbaatar government has taken several measures, such as the promotion of semi-coke briquettes and subsidies to local communities for using electric stoves. However, these efforts are far from achieving their proposed target. By considering the local realities, policy recommendations, including energy structure optimization, energy efficiency improvements, and capacity-building efforts, are provided. We expect that the key findings from this study can help other cities with similar challenges to better prepare their mitigation policies by considering local realities.   Appendix B