Sustainable Development – East Asia and Paci�c Region 66082 v1 D I S C U S S I O N P A P E R S M O N G O L I A Air Quality Analysis of Ulaanbaatar Improving Air Quality to Reduce Health Impacts December 2011 THE WORLD BANK Air Quality Analysis of Ulaanbaatar Improving Air Quality to Reduce Health Impacts December 2011 THE WORLD BANK © 2011 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, NW Washington, DC 20433 USA December 2011 All rights reserved. This study was prepared by the Sustainable Development Department (EASSD) of the East Asia and Paci�c Region, and was mainly funded by the Bank-Korea Environmental Partnership (BKEP), the government of the Netherlands (through the “Netherlands–Mongolia Trust Fund for Environmental Reform�) and the Norwegian Development Organization (NORAD). Sustainable development issues are an integral part of the development challenge in the East Asia and Paci�c (EAP) Region. The World Bank’s Sustainable Development Strategy for the region provides the conceptual framework for setting priorities, strengthening the policy and institutional frameworks for sustainable development, and addressing key environmental, social and rural development challenges through projects, programs, policy dialogue, non-lending services, and partnerships. The EASSD Discussion Paper series provides a forum for discussion on good practices and policy issues within the development community and with client countries. The publication and primary data are available online at www.worldbank.eapenvironment and www.worldbank.org.mn This volume is a product of the staff of the International Bank for Reconstruction and Development / The World Bank. The �ndings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Of�ce of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202- 522-2422, e-mail pubrights@worldbank.org. Table of Contents FOREWORD...................................................................................................................................vii ACRONYMS ....................................................................................................................................ix ACKNOWLEDGMENTS ......................................................................................................................xi EXECUTIVE SUMMARY .................................................................................................................. xiii 1. INTRODUCTION ....................................................................................................................... Previous studies of air pollution in Ulaanbaatar and the AMHIB air pollution study ........................ 1 Objectives of this report .................................................................................................................... 2 Scope of the project ........................................................................................................................... 2 2. AIR QUALITY MANAGEMENT APPROACH ..................................................................................... Air Quality Management Concept .................................................................................................... 5 Air quality assessment in Ulaanbaatar by the AMHIB monitoring program ...................................... 6 Assessment of the spatial distribution of PM concentrations and the contributions from different emission sources .................................................................................................................. 7 Air pollution and population exposure assessment ............................................................................. 7 Methods for health effects assessment ................................................................................................ 8 Methods for assessment of air-pollution-related health costs .............................................................. 9 Methods for assessment of costs of abatement options ....................................................................... 9 The analytical scheme for cost-benefit evaluations ........................................................................... 10 3. ASSESSMENT OF THE PRESENT AIR POLLUTION EXPOSURE AND SOURCE CONTRIBUTIONS INULAANBAATAR ................................................................................................................... Assessment of air pollution concentrations by monitoring ............................................................... 14 Monitoring program ............................................................................................................... 14 Measured PM concentrations and their variability in Ulaanbaatar, June 08–May 09 ............... 16 Summary of PM levels in Ulaanbaatar ..................................................................................... 20 Ger stove firing practices and variability in the PM concentrations in Ulaanbaatar .................. 24 Quantification of contributions to PM air pollution from various source categories ........................ 24 Emission inventory.................................................................................................................. 24 Contributions to ground-level concentrations ......................................................................... 25 Source apportionment by receptor modeling from AMHIB data analysis ........................................ 26 Assessment of air pollution exposure through air pollution modeling .............................................. 30 The spatial distribution of PM concentrations in UB and the contribution from main sources .................................................................................................................................... 30 The exposure of the UB population to PM concentrations, and its source contributions ......... 36 Implications of air pollution mapping for the monitoring network in UB ....................................... 36 iii Air Quality Analysis of Ulaanbaatar 4. ESTIMATING THE EFFECTS OF AIR POLLUTION ON MORTALITY AND HOSPITALIZATION IN ULAANBAATAR ................................................................................................................... Introduction .................................................................................................................................... 37 Data and Methods........................................................................................................................... 38 Results............................................................................................................................................. 42 Summary and Discussion ................................................................................................................ 45 Implications .................................................................................................................................... 46 5. WILLINGNESS TO PAY TO REDUCE AIR POLLUTION IN ULAANBAATAR ........................................... Introduction .................................................................................................................................... 49 Literature Review ............................................................................................................................ 49 Survey design and adaptation .......................................................................................................... 50 Data preparation and sample characteristics .................................................................................... 53 Results............................................................................................................................................. 54 Discussion ....................................................................................................................................... 56 6. CURRENT HEALTH COSTS OF AIR POLLUTION IN ULAANBAATAR AND BENEFITS FROM MANAGEMENT SCENARIOS .............................................................................................. Population-weighted exposure (PWE) at present and for 30 percent/50 percent/80 percent scenarios .......................................................................................................................................... 59 Avoided premature deaths, cases of chronic bronchitis and hospital admissions ............................... 62 Monetized health benefits ............................................................................................................... 65 7. AIR POLLUTION ABATEMENT OPTIONS AND THEIR COSTS IN ULAANBAATAR.................................. Main findings .................................................................................................................................. 69 Ger area emissions reduction options and their costs ....................................................................... 69 Introduction ............................................................................................................................ 69 Description of the scenarios..................................................................................................... 71 Results—cost implications ...................................................................................................... 74 Results—emission reductions .................................................................................................. 74 Discussion ............................................................................................................................... 76 HOB emissions reduction option and costs (Scenario 6) ......................................................... 77 Soil suspension reduction options and their costs .................................................................... 78 Summary......................................................................................................................................... 80 8. HEALTH BENEFITS OF THE AIR POLLUTION MANAGEMENT SCENARIOS .......................................... 9. COST-BENEFIT CONSIDERATIONS ............................................................................................. Cost-benefit considerations for each of the scenarios ....................................................................... 87 Summary......................................................................................................................................... 90 Conclusions .................................................................................................................................... 90 REFERENCES ................................................................................................................................. INTRODUCTION TO ATTACHED CD ROM .....................................................................................  Annex A: Baseline PM Concentrations/Baseline in Ulaanbaatar June 2008 – May 2009 ............... 101 Annex B: Identification and Apportionment of PM Pollution Sources by Receptor Modeling ....... 102 Annex C: Air Pollution Dispersion Modeling for Ulaanbaatar and Assessment of Source Contributions ............................................................................................................................... 102 Annex D: Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar ............................................................................................................................... 102 Annex E: The Willingness to Pay for Mortality Risk Reductions in Ulaanbaatar, Mongolia ........... 104 Annex F: Air Pollution Abatement Options and Their Costs in Ulaanbaatar ................................. 104 Annex G: Data Annex ................................................................................................................... 104 iv Table of Contents FIGURES Figure 2-1 General concept for development of cost-effective air quality management strategies .... 6 Figure 2-2 Detailed analytical scheme to arrive at the cost-benefit evaluation............................... 11 Figure 3-1 Location of AMHIB, MUB, and French sampling sites in Ulaanbaatar....................... 15 Figure 3-2 Monthly average concentrations of PM10 in Ulaanbaatar, June 2008–May 2009......... 17 Figure 3-3 Monthly average concentrations of PM2.5 in Ulaanbaatar, June 2008–May 2009. ....... 17 Figure 3-4 Examples of individual daily measured concentrations at station 6 (3 Kholoolol, in ger area) .................................................................................................................. 18 Figure 3-5 Examples of individual daily measured concentrations at station 2 (NRC, East of city center, outside ger areas)................................................................................... 18 Figure 3-6 Annual average PM10 concentrations as measured at AMHIB stations in Ulaanbaatar, June 2008–May 2009 ........................................................................ 19 Figure 3-7 Annual average PM2.5 concentrations as measured at AMHIB stations in Ulaanbaatar, June 2008–May 2009 ........................................................................ 19 Figure 3-8 Comparison of UB PM10 concentrations (2008–09) with Chinese cities (2008) and other capitals in the world (2004) ........................................................................ 21 Figure 3-9 PM10 (total), PM10-2.5 (coarse), and PM2.5 (fine) concentrations at the NRC station (AMHIB site 2) for 2006–09 .......................................................................... 22 Figure 3-10 Monthly averaged daily variation of PM10 and PM2.5 at the MUB Western Cross Station during work days ............................................................................................ 23 Figure 3-11 Example of time series of PM for one day at the beginning of winter, at Takhilt meteorological site, November 19–20, 2009 ............................................................... 25 Figure 3-12 Concentration contributions (μg/m3) to PM in air from main sources in UB ............. 28 Figure 3-13 Relative contributions (percent) to PM in air from main sources in UB. ..................... 29 Figure 3-14 Spatial distribution of emissions from ger stoves, tons/km2/year.................................. 32 Figure 3-15 Spatial distribution of PM10 and PM2.5, annual average (μg/m3) .................................. 33 Figure 3-16 Spatial distribution of source contributions, PM10, annual average (μg/m3) ................. 34 Figure 3-17 Spatial distribution of source contributions, PM2.5, annual average (μg/m3) ................ 35 Figure 4-1 The size of PM2.5 and PM10 relative to human hair and beach sand............................. 38 Figure 4-2 Health effects of particulate matter ............................................................................. 39 Figure 4-3 Administrative zones and AMHIB stations, Ulaanbaatar, Mongolia ............................ 40 Figure 5-1 Depiction of risk change (American version used to develop Mongolian survey)......... 51 Figure 5-2 Payment card elicitation of willingness to pay for a latent risk reduction (Mongolian version) ................................................................................................... 52 Figure 6-1 How much is air pollution reduced if emissions are reduced by 30%/50%/80%? PWE reductions for given emission reductions of PM10 .............................................. 61 Figure 6-2 Relative long-term mortality risk associated with different levels of PM10, estimated using three different functions..................................................................... 65 Figure 7-1 Estimated emission reduction from ger area interventions........................................... 76 Figure 8-1 Annual health benefit from abatement scenarios (mill USD) (discounted values), base case ..................................................................................................................... 82 Figure 8-2 Annual health benefit from abatement scenarios (mill USD) (discounted values), high case assuming an annual population growth rate of 8% and an income elasticity of VSL of 2................................................................................................... 83 Figure 8-3 Annual health benefit from abatement scenarios (mill USD) (discounted values), low case assuming an annual population growth rate of 5% and an income elasticity of VSL of 1................................................................................................... 83 Figure 8-4 Difference in health benefits ($ million) between five short to medium term measures and a long-term scenario (relocation into apartments) ................................................ 85 v Air Quality Analysis of Ulaanbaatar TABLES Table 3-1 Annual average PM concentration of Ulaanbaatar city, corrected according to results from sampler comparisons ........................................................................... 20 Table 3-2 Indicated ranges for annual average PM concentrations in Ulaanbaatar, June 08–May 09, linked to the areas where the measurements were carried out .......... 21 Table 3-3 Summary of the emissions inventory for Ulaanbaatar, 2008 (tons/year) ...................... 25 Table 3-4 Population-weighted exposure (PWE) to PM in Ulaanbaatar as calculated by the air pollution model, contributions from main sources (μg/m3) .............................. 36 Table 4-1 PM monitor stations and associated hospital districts ................................................. 40 Table 4-2 Administrative district with associated PM monitor and population ........................... 41 Table 4-3 Descriptive statistics of data used in mortality analysis ................................................ 43 Table 4-4 Descriptive statistics of data used in hospital analysis .................................................. 43 Table 4-5 Descriptive statistics for mortality and hospitalization data used in analysis ................ 44 Table 4-6 Statistically significant or near significant mortality effect estimates for particulate matter and nitrogen dioxide, June 2008–May 2009 .................................................... 44 Table 4-7 Selected statistically significant results of logistic regressions for hospitalization and particulate matter, June 2008–May 2009 .................................................................... 45 Table 5-1 Study Design (cumulative probabilities over a ten-year period) ................................... 52 Table 5-2 Descriptive statistics for cleaning criteria variables....................................................... 53 Table 5-3 Comparison of demographic variables by cleaning and place of birth ......................... 54 Table 5-4 External scope tests under alternative data cleaning approaches (respondents age 40–65). ................................................................................................................. 55 Table 5-5 Construct validity of WTP for the current and latent risk reductions (using a Weibull distribution) ................................................................................................ 56 Table 6-1 Population-weighted average PM concentrations (PWE) in Ulaanbaatar, and reductions from abatement scenarios, μg/m3 ........................................................ 60 Table 6-2 Exposure-response coefficients (% change in incidence of health effect per mg/m3 PM10), baseline annual incidence rates, willingness-to-pay (WTP) for avoiding premature death (long-term effect) and new cases of chronic bronchitis, and cost of illness (COI) of hospital admissions........................................................................ 66 Table 6-3 Estimated current health damage due to PM pollution in Ulaanbaatar (base case), number of cases avoided due to interventions, and monetized current cost and benefit from interventions (in mill USD).................................................................... 67 Table 7-1 Cost elements that play a role in the different scenarios .............................................. 74 Table 7-2 Net present value of the costs and direct benefits ........................................................ 75 Table 7-3 Average and maximum emission reductions ................................................................ 75 Table 7-4 Relative cost-effectiveness of the different options ....................................................... 77 Table 7-5 HOB baseline data ..................................................................................................... 78 Table 7-6 Summary performance of the various options ............................................................. 80 Table 8-1 Present value of health benefits ($ million), accumulated for the period 2010–23 ....... 84 Table 9-1 Comparison of present value (PV) of health benefits (base case) with net present value (NVP) of implementing costs, and net benefit (PV minus NPV) for the eight abatement scenarios, 2010 (mill USD) ......................................................... 88 vi Foreword A ir pollution has major health impacts presents the results of the entire study including on people living in Ulaanbaatar. The estimates of the massive health costs associated excessively high particulate matter with air pollution and analyses of short-, medium- concentrations, especially in the winter and long-term intervention options. and in the ger areas, increase the incidence of heart and lung diseases, and lead to premature Despite the grave picture presented by this deaths. Improving air quality management study regarding air pollution in Ulaanbaatar, it in Ulaanbaatar and reducing pollution also allows for a positive outlook—the health concentrations would prevent illnesses, save lives burden due to air pollution in Ulaanbaatar can and avoid enormous health costs. be significantly reduced through measures that have a favorable cost-benefit ratio. The AMHIB In order to get a sound information basis for study has been carried out in close collaboration a strategy to improve air quality in Ulaanbaatar, with our Mongolian partners, and we are looking the World Bank in partnership with Mongolian forward to continuing this partnership with the counterparts launched an �Air Monitoring and aim of making a contribution towards a cleaner, Health Impact Baseline� (AMHIB) study in greener and healthier Ulaanbaatar. 2008. This was the first time that air pollution was monitored year round in the surrounding ger Ede Jorge Ijjasz-Vasquez areas of Ulaanbaatar and that estimations of the Sector Manager health impacts from the monitored air quality China & Mongolia Sustainable levels were undertaken. The AMHIB study also Development Unit includes analyses of the sources of the pollution East Asia and Pacific Region concentrations, and cost-benefit analyses of The World Bank measures to reduce these levels. Coralie Gevers After the first AMHIB study phase, an initial Country Manager assessment focusing on pollution concentrations Mongolia and sources was published in December 2009. The World Bank The preliminary results of the initial assessment indicated severely high particulate matter Magda Lovei concentrations. After the year-round monitoring Sector Manager data, including measurements from the winter Social, Environment and Rural months, became available the full extent of Development Unit particulate matter pollution was revealed, and East Asia and Pacific Region even more alarming annual average concentrations The World Bank and peak levels were found. This final report vii Acronyms AMHIB Air Monitoring and Health Impact Baseline Study ADB Asian Development Bank AirQUIS Air quality Information System (developed by NILU) AQ Air quality AQS Air quality standards ASTAE Asia Sustainable and Alternative Energy Program BC Black carbon CBDICFP (JICA) Capacity building for development and implementation of carbon finance Project. CHP Combined heat and power station CLEM Central Laboratory for Environmental Monitoring COI Cost of illness CP Coarse particles (particles between 2.5 and 10 microns) CVD Cardiovascular diseases EBRD European Bank for Reconstruction and Development EF Emissions factors EI Emissions inventory FRM Reference method GDP Gross domestic product GENT Air monitoring equipment GIS Geographic Information System GIZ Deutsche Gesellschaft für Internationale Zusammenarbeit (formerly GTZ) GRIMM Air monitoring equipment GTZ Deutsche Gesellschaft für Technische Zusammenarbeit GW Gigawatts HOB Heat-only boiler ICD International Classification of Diseases IPCC Intergovernmental Panel on Climate Change IQR Interquartile range IT Interim targets JICA Japan International Cooperation Agency LPB Low-pressure boiler LPG Liquefied Petroleum Gas LV Limit values MCA Millennium Challenge Account, USA MJ Megajoules MMRE Ministry of Mineral Resources and Energy MNS Mongolian standards MNT Mongolian currency (Tugrik) ix Air Quality Analysis of Ulaanbaatar MUB MS owned by the Air Quality Department of Ulaanbaatar Municipality MW Megawatts NAMHEM National Agency of Meteorology, Hydrology and Environmental Monitoring of Mongolia MS Monitoring Station NILU Norwegian Institute of Air Research NOX Nitrogen oxide (combine nitric oxide/NO and nitrogen dioxide/NO2) NPV Net present values NRC Nuclear Research Center NRC- NUM Nuclear Research Center (of the National University of Mongolia) PM10 Particulate matter (particles of diameter 10 micrometers or less) PM2.5 Particulate matter (particles of diameter 2.5 micrometers or less) PMF Positive matrix factorization POP Total population PV Present values PWE Population-weighted average exposure RFF Resources for the Future RR Relative risk SA Source apportionment SCC Semi-coked coal SO2 Sulfur dioxide UB Ulaanbaatar UB CAP UB Clean Area Project UK United Kingdom UNEP United Nations Environment Programme US United States USEPA United States Environmental Protection Agency VSL Value of statistical life WB The World Bank WHO World Health Organization WTP Willingness-to-pay Note: Unless otherwise noted, all dollars are U.S. dollars. x Acknowledgments T he implementation of the Air Batnyam, Sarangerel Enkhmaa (NAMHEM) and Monitoring and Health Impact Baseline Ganzorig Ayush (Department of Finance, NUM). (AMHIB) study has brought together In addition, a number of Mongolian researchers, Mongolian and international air students and officials have been engaged in the quality experts as well as public health experts data collection processes performed as part of this and economists who have taken an synergetic study. approach of linking public health, air quality and economic issues. This report builds upon the International team members engaged in discussion paper Air Pollution in Ulaanbaatar: drafting the main report and assisting the national Initial Assessment of Current Situation and Effects team in the preparation of technical background of Abatement Measures that was published in reports on air quality monitoring, modeling December 2009, and reflects the final results and work, health impact assessment and , abatement recommendations from the AMHIB project. option scenarios and cost benefit analyses, included Steinar Larssen (World Bank consultan In order to provide overall guidance to the and chief technical advisor, former Norwegian project, a steering committee was set up and Institute of Air Research/NILU), Bruce Denby headed by Jargalsaikhan Ch, Vice Minister and Liu Li (NILU), Bart Ostro (Office of of the Ministry of Nature, Environment and Environmental Health Hazard, California Tourism (MNET). The committee included EPA), Alan Krupnick (Resources for the Future/ representatives from the Ministry of Health, RFF), Sandra Hoffmann (formerly RFF, now Ulaanbaatar Government, the National Agency the Food Economics Division, US Department of Meteorology, Hydrology and Environmental of Agriculture), Robert van der Plas (World Monitoring of Mongolia (NAMHEM), Bank consultant) and Kristin Aunan (Center the Central Laboratory for Environmental for International Climate and Environmental Monitoring (CLEM) and the World Bank. Research/CICERO). Technical assistance has also been provided by Qin Ping (Fudan University), The Mongolian study team responsible for Dam Vu Than (NILU) and Stephen Rauch data collection, initial data analyses and initial (California EPA). drafting of technical background report, was implemented under a contract arrangement with The project was co-managed by Ede the National University of Mongolia (NUM). Ijjasz (Sector Manager, China & Mongolia The team was led by Sereeter Lodoysamba Sustainable Development Unit), Magda Lovei and included Dagva Shagjjamba and Gunchin (Sector Manager, Social, Environment and Rural Gerelmaa (Nuclear Research Centre, NUM), Development Unit), Arshad Sayed and Coralie Altangerel Enkhjargal and Batbaatar Suvd Gevers (Country Managers for Mongolia), (Mongolian Public Health Institute), Burmaajav Jostein Nygard (task team leader) and Gailius Burmaa (Mongolian Ministry of Health),Lamzav Draugelis (co-task team leader). World Bank xi Air Quality Analysis of Ulaanbaatar team members included Stefan Csordas, Natsuko cooperating closely with the Asian Development Toba, Neelesh Shrestha, Erdene Ochir Badarch, Bank (ADB), the European Bank for Tumentsogt Tsevegmid, Tijen Arin and Shawna Reconstruction and Development (EBRD), Fei Lee. Administrative assistance was provided Deutsche Gesellschaft für Internationale by Minhnguyet Le Khorami and Nomuuntugs Zusammenarbeit (GIZ), JICA, the Millennium Tsevegmid. Challenge Corporation (MAC), and many other partners in Mongolia to contribute to reduced air The report drafts have been peer reviewed by pollution in Ulaanbaatar. Dick van den Hout (Netherlands Organization for Applied Research/TNO), Jordi Sunyer We would like to express our gratitude to the Deu (Center for Research in Environmental various donors, who provided funds to carry out Epidemiology), Mohammed Khaliquzzaman the study. This includes the Korean Environment (World Bank consultant), Taizo Yamada (Japan Management Corporation (KEMC) (through International Cooperation Agency/JICA), Carlos the Bank-Korea Environmental Partnership), Marcelo Bortman, Sameer Akbar, Hua Wang the government of the Netherlands (through the and Giovanna Ruta (World Bank). Additional World Bank administered “Netherlands-Mongolia comments were received from Dejan Ostojic, Trust Fund for Environmental Reform�) and Robert W. Bacon, Jan Bojo and Paul Procee the government of Japan (through the Bank (World Bank). administered trust fund project “Capacity Building for the Development of Carbon The report has been edited by Robert Financing Projects in Mongolia�). The Norwegian Livernash and Charles Warwick, and designed Development Organization (NORAD) funded and typeset by Shepherd Inc. awareness raising through the Together for a Green and Clean Ulaanbaatar Event in June 2011, and In addition to the outlined AMHIB dissemination seminars through its Eco-town donors, and by the request by the Mongolian program in Ulaanbaatar. Without this generous Government, the World Bank has been support, this study would not have been possible. xii Executive Summary A mbient annual average particulate homes of the poor and the most vulnerable of matter (PM) concentrations in UB’s population. PM2.5 concentrations in the ger Ulaanbaatar (UB) are 10–25 times areas are much higher than in the center, with an greater than Mongolian air quality annual average concentration in the range of 200 standards (AQS) and are among the highest to 350 µg/m3. The main sources of the particles recorded measurements in any world capital. in the ger areas are coal burning for heating and PM is the main pollutant in Ulaanbaatar, and cooking during the winter, and the suspension one that is of greatest concern due to its broader of dust by wind action throughout the year, but health impacts. This study thus focuses on especially in the warm season. Much of the ger monitoring and analyses of PM10 and PM2.5 areas were not included in the epidemiological concentrations, particulates below 10 and 2.5 study because area coverage of the monitors and micrograms per cubic meter (µg/m3), respectively, the sample size was too small. Nonetheless, the which have known health impacts. Among health effects in the ger areas are likely to be all measurements taken at the study’s eight substantially more severe than in the city center monitoring stations, the worst recorded annual due to the high ambient PM concentrations. average concentration was more than 10 times Assuming that the toxicity of PM2.5 in the ger higher than the Mongolian AQS for PM10 and 25 areas is generally similar to that in the city times higher than the Mongolian AQS for PM2.5.1 center and that the dose-response function (the Compared to other cities for which data are calculation of the incidence of a health event due available in global data bases, and also compared to inhaling excessive particulate matter of a given to Chinese cities with high PM concentrations, size) remains linear at the higher levels of PM2.5, UB appears to be the most PM-polluted capital there is a 1.38 percent increase in daily mortality and is among the cities with the worst air quality per every 10 µg/m3 in PM2.5 in the ger areas. This in the world. suggests that the current PM2.5 concentrations in the ger areas led to a 24 to 45 percent Ger households are both the main source higher mortality than would be the case at the and the main casualties of air pollution. Prior Mongolian air quality standard of 25 µg/m3. This to this study, no systematic air quality monitoring constitutes a significant public health problem. had been undertaken in the ger areas of UB and continued data collection and analysis is The main sources of ground-level air recommended. The highest PM concentrations pollution are coal and wood burning for are measured in the ger areas, the location of the heating of individual residences in ger areas and the suspension of dry dust from open soil surfaces and roads, representing 75–95 percent of PM concentrations. Other significant sources 1 The Mongolian annual ambient air quality standards are of ground-level PM concentrations are emissions 50 µg/m3 and 25 µg/m3 for PM10 and PM2.5, respectively, while the WHO interim targets for developing countries are from power plants, heat-only boilers (HOBs), and 70 µg/m3 for PM10 and 35 µg/m3 for PM2.5. car and vehicle exhaust. xiii Air Quality Analysis of Ulaanbaatar Figure ES-1: Comparison of UB PM10 concentrations (2008–09) with Chinese cities (2008) and other world capitals (2004) Source: Authors’ illustration based on data from the China Environment Yearbook 2009 for Chinese cities, AMHIB study for UB, and WHO Air Quality Guidelines - Global Update 2005 for other cities. Table ES-1: Indicated ranges for annual average PM concentrations in Ulaanbaatar, June 08–May 09 Area PM10 PM2.5 Exceedance: Ra o to AQSs 3 3 μg/m μg/m Mongolian: WHO Central city areas 150–250 75–150 3–6 7–15 Ger areas 350–700 200–350 7–14 17–35 Source: AMHIB data Wintertime PM measures show due to increased winter ger area heating. For alarmingly high levels. PM concentrations example, monthly average PM10 concentration during summer are much lower than during measured at the Zuun railway station in a ger winter, therefore annual average figures mask area was 1,850 µg/m3 in January 2009, while alarmingly high monthly average and daily the four highest daily average measurements at maximum PM concentrations during peak the same station were in the range of 3,612– heating periods. Due to the dominance of ger 4,360 µg/m3. The highest PM2.5 concentrations stoves as emissions sources, PM concentrations were measured at the Bayanhoshuu station, in UB show a strong seasonal variation with also in a ger area, where the monthly average the winter being much worse than the summer figure was about 1,500 µg/m3 and the five xiv Executive Summary Figure ES-2: Health Effects of particulate matter Source: Authors’ illustration. highest daily concentrations were in the range costs of between $ 177 and $ 727 million (mean of 2,310–4,060 µg/m3.2 value $463 million). The health impacts were estimated, first, by setting a threshold value for There are also intra-day peaks when population-weighted exposure, below which the cooking takes place in the ger areas. Hourly health effect is very small. Second, modeling measurements of PM correlate strongly with techniques were used to calculate exposure to the daily routine firing practices of heating excessive levels of ambient PM concentrations. stoves, illustrating the importance of addressing The exposure of the population to pollution was emissions from cold-start lighting and refueling estimated by calculating the physical dispersion of stoves. Importantly, measurements generally of ambient concentrations of PM and then corroborate the results from dispersion superimposing a population distribution. Third, modelling, which is based on an emissions the value of a statistical life (VSL), which is a inventory and not the air monitoring station function of the willingness to pay (WTP) for a data, and on which abatement impacts are reduction in mortality risk, was estimated based estimated. on a survey administered to a random sample of the population. The results show that the people The magnitude of the estimated negative of UB attach a high value to reducing mortality health impacts is large. The alarming PM risk compared to cities in China as well as in the concentrations in UB lead to significant health range typically measured in developed countries. impacts, amounting to estimated annual health The epidemiological study shows statistically significant associations between 2 The concentrations measured on some days at individual monitoring stations are very high. However, no specific reasons cardiovascular mortality and coarse particles have been found to reject these measurements. Although errors with a one-day lag: Every 10 µg/m3 change in cannot be ruled out, they are regarded as examples of the very coarse-particle concentrations would increase high exposures to PM that can occur due to a combination of very large emissions at low heights locally near the stations and daily cardiovascular mortality on the next day adverse dispersion conditions. by 0.25 percent. This means that a reduction of xv Air Quality Analysis of Ulaanbaatar coarse-particle concentrations by 200 µg/m3 in emissions of key pollution sources. Simulations would lead to a decrease in cardiovascular were made for of 30, 50, and 80 percent emissions mortality by 5 percent. Likewise, for PM2.5 reductions for the main sources. These results are each 10 µg/m3 increase relates to a subsequent shown in Figure ES-3. For example, an 80 percent 1.4 percent increase in daily all-cause mortality emission reduction from coal combustion in ger in the warm season so that a reduction of PM2.5 areas, irrespective of the abatement measures concentrations by 200 µg/m3 would lead to used, yields an estimated 48 percent reduction in a decrease in all-cause mortality of about ambient PM2.5 concentrations, or approximately 28 percent. In the cold season, a reduction of $66 million in annual avoided health costs. coarse particles concentration by 200 µg/m3 would lead to a reduction of mortality the next The results of this study show that a solution day by 5-6 percent. The strongest associations will require a wide range of pollution abatement between particulate matter and hospitalization options—there is no one magic bullet. An were between PM2.5 and PM10 and cardiovascular 80 percent emission reduction from ger area disease. For each 10 µg/m3 change in PM2.5, a heating, heat-only boilers, and suspended soils 3-day lag was associated with a 0.82 percent (dust) yields a significant 69 percent reduction of increase in admissions for cardiovascular disease ambient concentrations but is still not sufficient to and a 0.24 percent increase for respiratory meet Mongolian air quality standards. To achieve disease. For a 10 µg/m3 change in PM10, a 3-day Mongolian AQS, an emission reduction of about lag was associated with a 0.21 percent increase 94 percent from these three source sectors would be in cardiovascular admissions, while a 2-day lag needed. was associated with a 0.04 percent increase in respiratory admissions. Costs and benefits vary strongly between different pollution abatement options. Cost- Abatement measures should demonstrate benefit analysis helps to prioritize options that their emission reduction potential to achieve emissions reduction targets. The study ensure significant impacts on ambient PM compared eight options, taking into account concentrations. The study simulates reductions direct benefits (e.g. fuel savings), health benefits, of ambient PM concentrations from reductions and investment costs. The options are not Figure ES-3: Air pollution concentration reductions due to emissions reduction of 30%/50%/80% (PWE of PM10 and PM2.5 reduction) Estimated population weighted average concentration for given emission reductions PM10 PM2.5 Note: The projected emission reduction scenarios (-30%, -50%, and -80%) do not result in equal reductions in concentration values. Source: Authors’ illustration based on AMHIB data. xvi Executive Summary Table ES-2: Estimated current health damage due to PM and benefits and costs from air pollution abatement options in Ulaanbaatar Annual number of cases Monetized (mill USD) Chronic bronchi s / Chronic bronchi s / All-cause All-cause Share of Hospital admissions Hospital admissions SUM mortality mortality GDP in UB (Respiratory disease) / (Respiratory disease) / (mill USD) (chronic) (chronic) (2008) Hospital admissions (CVD) Hospital admissions (CVD) 1591 1411 / 4465 / 4063 352 100 / 4.71 / 6.92 463 18.8 % Current health (385 - (1219 - 1516)* / (1828 - (85 - (86 - 107)* / (1.9 – 8.5)* / (177 – (7.2 – damage 2721)* 8083)* / (2290 - 6122)* 601)* (3.9 – 10.4)* 727)* 29.5)* 30% reduc on of Ger stoves 63 74 / 619 / 528 14 5 / 0.65 / 0.90 21 0.8 % 80% reduc on of Ger stoves 198 253 / 1663 / 1444 44 18 / 1.75 / 2.46 66 2.7 % 30% reduc on of HOBs 3 3 / 28 / 24 1 0 / 0.03 / 0.04 1 0.0 % 80% reduc on of HOBs 7 8 / 75 /64 2 1 / 0.08 / 0.11 2 0.1 % 30% reduc on of suspended dust 53 60 / 520 / 443 12 4 / 0.55 / 0.75 17 0.7 % 80% reduc on of suspended dust 159 199 / 1395 / 1205 35 14 / 1.47 / 2.05 53 2.1 % 30% reduc on of all 3 sectors 129 159 / 1172 / 1009 29 11 / 1.24 / 1.72 43 1.7 % 80% reduc on of all 3 sectors 522 707 / 3169 /2822 115 50 / 3.34 / 4.81 174 7.0 % * These intervals were calculated based on the 95% confidence interval of the estimated dose-response coefficient and a +/– 30% interval of the PWE values. The lower (higher) value represents the estimate using the lower (higher) ends of both the 95% confidence interval of the dose response coefficient and PWE concentration. Similar confidence intervals apply to the estimated health damage reductions in the various scenarios but are not displayed here for the sake of clarity of presentation. Source: AMHIB study exhaustive of the abatement measures in UB, but Emissions and Efficiency Laboratory in Mongolia represent some of the most discussed. has not yet been asked to test any fuel alternatives to Nailakh coal, the dominant traditional fuel, The costs of pollution management with cleaner stove technologies. intervention include the costs of the program to promote the intervention (publicity campaign, A strategy combining short- and medium- training, possible subsidies, cost of testing and term pollution abatement measures is certification), cost of additional equipment recommended. The sheer magnitude of the needed (stoves, electric heaters, apartment air pollution-induced health damage calls for buildings), cost of additional infrastructure immediate action, while medium- to long-term needed (additional production capacity for clean solutions are sought in parallel. Figure ES-5 fuels, electricity and/or heat, electricity/heat illustrates the estimated cost of delaying short distribution network), and cost of incremental term actions, based on available information. heating energy. Five of the scenarios generate a The available options to manage air pollution net benefit (avoided health costs minus abatement and reduce health damage could take effect over costs) in the range of $393–$1,635 million over the short (1–5 years) and long term (+5 years), a 15-year period. This suggests air pollution depending on the nature of the intervention. management can be carried out with a substantial Medium- and long-term solutions require economic gain when health costs are taken preparation time during which people continue into account. The analysis requires continuous to breathe unhealthy air. Given the relative costs updating as new analysis on costs and emissions and benefits of various interventions analyzed in becomes available. For example, the Stove this report, fiscal and affordability constraints, xvii Air Quality Analysis of Ulaanbaatar Figure ES-4: Annual health benefit from abatement scenarios (mill USD) (discounted values) Source: Authors’ illustration. Figure ES-5: The cost of delaying short-term measures—the difference in health benefits between a short-term scenario (certified stoves) and a long-term scenario (relocation into apartments) Note: Total estimated “avoided cases� by taking scenario 2 (certified stoves) compared to scenario 5 (relocation into apartments) in the 2010–23 period: 2,052 premature deaths, 2,666 cases of chronic bronchitis, 16,775 cases of respiratory diseases (HA), 14,615 cases of cardiovascular diseases (HA). Source: Authors’ illustration. xviii Executive Summary and the time needed for adequate preparation, the f. Introduce an interagency coordination following strategy is recommended: mechanism with the UB city and central government with a working level a. Set targets that would reach Mongolian secretariat that will be responsible for AQS for PM10 as soon as possible and year-round monitoring, gathering and PM2.5 by 2020; given socioeconomic processing information on all abatement constraints, set interim targets and a programs (government, donor and reduction time line for PM pollution private initiatives), providing research reduction. and support to this coordination body, b. Ensure that abatement measures have and supplying the government and UB reasonable, demonstrated emission city with up-to-date information on exposure reduction potential that is program progress, latest news, as well sufficient to contribute significantly to as private sector for information and achieving targets. coordination. The GoM and UB city have c. To achieve targets, all main PM sources already implemented elements of this need to be tackled, but these should be strategy, which is supported by the donor prioritized based on the potential of their community, including the United States emissions exposure reduction, costs, Millennium Challenge Corporation, affordability, and the level of preparation Asian Development Bank, European Bank and time required for implementation, for Reconstruction and Development, ease of implementation, and governance Japan International Cooperation Agency, issues. Reductions are needed first from United Nations Environment Program, ger heating systems (stoves, wall stoves, the World Bank, and bilateral programs and low-pressure boilers) and from dust from the Netherlands, Korea, Germany suspension from uncovered soil surfaces, and France. as roads, and near-road surfaces are the g. Strengthen air quality monitoring and second most important PM10 source. have sufficient operating budgets in place d. Maintain an open and candid discussion to evaluate the results of various efforts to of actual costs and benefits of abatement reduce ambient concentrations of PM and measures by (i) ensuring the abatement the effectiveness of key strategic programs, measures are technically feasible and such as the replacement of ger stoves. their emission reduction benefits are h. Strengthen air quality emissions justified with sufficient evidence; inventories for dispersion modeling. (ii) appraising the full costs of abatement i. Maintain and strengthen the Stove measures and their contribution to Emissions and Efficiency Testing Laboratory, overall improvements in ambient PM set up by ADB under the auspices of the concentration reductions, so they can be Ministry of Mineral Resources and Energy, compared to the health cost reductions; which can test emissions factors for new (iii) if subsidies are provided, avoid over stove models and support clean stove subsidization to ensure sustainability; technology development. and (iv) keep the public informed of j. Continue studies of health impacts. plans, manage expectations and find k. Maintain public awareness through a opportunities to mobilize citizen action. professional public awareness campaign e. Implement concerted actions that that can adequately distill technical combine (i) short-term measures that are messages, and monitor results and other well-prepared and can be carried out in a information to build knowledge and relatively short period of time (1–3 years) understanding. The secretariat of the to have a visible effect, (ii) medium-term coordination mechanism could be a focal measures that require more resources and point for this effort with the support of a time to prepare and roll out. professional public relations agency. xix Air Quality Analysis of Ulaanbaatar This study concludes that it is important Environment Monitoring, under the guidance to establish a well-coordinated framework of the AMHIB Steering Committee chaired for pollution abatement in Ulaanbaatar— by the Ministry of Nature, Environment and one that relies on best available scientific and Tourism and including the participation of the economic data, that involves its citizens and Ulaanbaatar Municipality’s Air Quality Agency. shares information with them, and continuously The previous National Coordination Committee updates urban air quality analysis, such as the one on the Reduction of Air Pollution in Ulaanbaatar presented in this study, based on new information (NCC), chaired by the Minister of Mineral as it becomes available. Concerted efforts, over Resources and Energy and vice chaired by the several years, are needed. Table ES-3 shows air UB Municipality, were informed of the interim pollution, health effects, and abatement options results. This committee has now been replaced in UB at a glance. The study’s interim findings by a similar NCC headed by the president of have been shared with UB citizens, MUB, GoM, Mongolia. Mongolian experts involved in this and the donor community since its inception. study have participated in the drafting of new The study was largely conducted by Mongolian programs, policies, and legislation. This process institutions, especially the National University should continue. Ulaanbaatar is already the of Mongolia, the Ministry of Health’s Public world’s coldest capital city; it need not be the Health Institute, and the Mongolian National most polluted. Agency for Meteorology, Hydrology, and xx Table ES-3: Ulaanbaatar air pollution burden at a glance, 2008/09 Air Pollu on Emissions Source PM10 PM2.5 Spa al distribu on (tons/year) Ger households 19,731 15,785 Throughout Ger areas HOBs 1,077 646 Dispersed over UB surroundings CHPs 18,589 7,436 3 point sources to the west of UB centre Vehicle exhaust 1,161 1,161 Mainly throughout the central city areas Dust from paved roads 9,954 771 Mainly throughout the central city areas Dust from unpaved roads 4,812 722 Mainly throughout the Ger areas Concentra on (μg/m3) Central city areas 150–250 75–150 Ger areas show much higher concentra on levels Ger areas 350–700 200–350 Exposure (μg/m3 ) Popula on weighted average 427 260 Ger households are exposed to higher levels of air pollu on Health Effects Mortality and ~4.2% in cardiovascular admissions 200 Hospitaliza on ~0.8% admissions for respiratory disease 3 200 μg/m reduc on ~5–6% all-cause mortality in cold season 3 Reduc on in μg/m and associated health in coarse par cles ~5% in cardiovascular admissions bene�ts ~28% all-cause mortality in warm season ~16.4% admissions for cardiovascular disease 200 ~4.8% admissions for respiratory disease ~27.6% in all-cause mortality in Ger areas Willingness to Pay Value of sta s cal life: US$ 221,000 Health Costs Current health damage corresponds to 19% of GDP in Ulaanbaatar and 9.1% of GDP in Mongolia Cost Bene�t Emission Control NPV costs US$ per t PV health bene�t Net Bene�t (million US$) reduced (million US$) (million US$) Analysis Scenarios Reduce start-up emissions through -39% -53 -463 865.8 918.8 backligh ng the �re and Reduce start-up emissions through slight -58% -35.8 -208 1,599.0 1,634.8 modi�ca ons of the stove Health Bene�ts Replace exis ng stoves with cleaner coal -26% -0.3 -4 1,605.0 1,605.3 stoves, without changing the fuel Replace exis ng stoves and fuels with -22% 36.7 550 1,028.5 991.8 cleaner stoves and SCC Electric hea ng in exis ng Ger homes -63% 1,410.0 7,523 1,802.9 392.9 Reloca on of Ger households into -31% 4,094.0 44,925 597.1 -3,496.9 apartments Heat-only Boilers -32% 5.9 49 19.7 13.8 Road dust -1.7% 66.7 39,332 66.9 0.2 Greening -0.2% 2.5 18,016 58.1 55.6 xxi Executive Summary Source: AMHIB data. 1. Introduction Previous studies of air pollution in Ulaanbaatar on an initial source inventory made available and the AMHIB air pollution study to the Bank consultants. This contributed both to the estimates of PM concentrations and to a Ulaanbaatar’s population has likely been facing first understanding of the PM source structure air pollution challenges for several decades— (Guttikunda 2007). through the development of Ulaanbaatar (UB) as a growing industrial city with increased Through further development of the UB- construction, traffic, and power generation Clean Air Program, it was discovered that the facilities located in a valley with a relatively dry Mongolian authorities already had established climate. But it was the rapid expansion of the some PM monitoring on a research basis, surrounding ger3 areas that started to indicate that estimating both actual PM concentrations UB was facing severe air pollution challenges. levels as well as chemical composition, through The Mongolian government had established which an initial understanding of the PM an air monitoring system within the traditional source structure emerged. This information was city center, but it largely monitored SO2, NO2, also circulated within the UB government and temperature, and wind direction. Nevertheless, environmental authorities. knowledge about air pollution—especially particulate matter (PM), the pollutant with the UB covers a relatively large geographical most destructive impact on human health—was area. It has a population of about 1 million with limited. At the National University of Mongolia diversified living conditions in the traditional city (NUM), sporadic PM10 and PM2.5 monitoring center and the emerging surrounding ger areas. began on a very limited scale in 2004–05 with To monitor these areas, the World Bank, in close incomplete yearly collected data sets. However, cooperation with their Mongolian counterparts, it represented a start, indicating the growing established an Air Monitoring and Health Impact concern in the city’s air quality situation. Baseline (AMHIB) study. The study was intended to create an air quality baseline from June 2008 to The World Bank first started to examine May 2009 which would then be used to compare the air quality situation in UB with the initial changes in the city’s air quality. Simultaneously, a development of the UB Clean Air Program. Prior health impact assessment was also undertaken in to the preparation of the program, the Bank approximately the same locations as the AMHIB decided to take a closer look at the PM levels air quality monitoring network. The system in UB by estimating PM concentrations based included eight monitored locations distributed among traditional city and ger area locations. 3 Gers are round removable wooden homes traditionally used In December 2009, the preliminary results from by Mongolian (and other Central Asian) nomads. Today, UB’s city center is surrounded by gers with non-populated rural the AMHIB work were presented (World Bank areas beyond the ger areas. 2009) reflecting results from the first few months 1 Air Quality Analysis of Ulaanbaatar of operations. In addition, the emissions inventory health, both mortality and morbidity, through used for the work was presented along with initial chronic bronchitis and cardiovascular and modeling results, based upon the initial knowledge respiratory system diseases. Finally, the objective is of PM concentration levels and source structure. to present abatement scenarios to cost-effectively reduce the high PM concentrations. This report presents the final results of the AMHIB study. The main audience for this Scope of the project study is the government of Mongolia (GoM) and Ulaanbaatar decision makers, as well as The aim of this study is to inform decision those actively engaged in the improvement makers about the sources of air pollution in of the city’s air quality, including air quality UB, the health damage caused by PM, and the professionals, health experts, and development options for abatement measures. The results and partners. The study contains (a) the results of air recommendations of this study neither represent quality monitoring from June 1, 2008, to a full and final treatment of air pollution in UB, May 31, 2009 in eight locations, including nor do they provide a complete treatment of all for the first time in the city’s ger areas; (b) the available options for intervention or prescribe outcome of air quality modeling to predict the a specific course of action. Rather, this report effects of various pollution abatement measures; intends to highlight the nature and scale of air (c) a comparison of over 50,000 hospital records pollution and associated health costs, and outline in six of nine administrative districts in UB to some possible interventions to address this issue. correlate the incidence of cardiovascular and Certain aspects of air pollution management such respiratory disease, and premature death, caused as capacity constraints, institutional issues and by exposure to excessive particulate matter (PM) social acceptability have not been included and pollution; and (d) a willingness-to-pay survey the statistical analyses underlying this report are of 629 respondents in Ulaanbaatar to estimate subject to a number of sources of uncertainties. the value placed by UB residents on avoiding The report has to be read with these caveats premature diseases and mortality, thus providing in mind, and should therefore be viewed as an a basis to quantify the health costs of air pollution information source for designing policies aimed and the benefits of reducing air pollution. The at addressing air pollution in UB, rather than a report shares the findings of these efforts and detailed air pollution management strategy. recommends an air pollution reduction strategy. We hope this study marks the beginning of a In many developing countries indoor continuous assessment of issues and options as air pollution is more severe than outdoor air new information and data is learned on solutions pollution, particularly because of indoor heating to this severe public health crisis. and cooking practices and the use of charcoal, wood etc as fuel. However, in UB the design of Objectives of this report ger stoves means that PM is absorbed through the stove chimneys, thereby resulting in relatively less By including the first components in an air severe indoor air pollution. Indications show that quality management concept, the AMHIB study outdoor air pollution is substantially graver than intends to establish a baseline for air quality and indoor air pollution (see box below). As a result, health impact measures upon which future air the report is focused on outdoor pollution. pollution control activities can be compared. By showing the PM concentration levels at the The report focuses on local air pollution various AMHIB monitoring stations in the impact on health caused by PM from main 2008–09 period, it is possible to compare the pollution sources. Therefore, the report identifies development over the future years in both ger the main sources that contributes to PM areas and traditional city centers. concentrations in Ulaanbaatar, the health and economic effect of these high concentrations and The objective of the report is also to estimate how to reduce the emissions and contributions the effects of PM concentrations on human from these sources. In that context, it should 2 Introduction be noted that certain sources may in general This does not mean that CHPs are not important contribute to high air pollution concentrations, sources to focus on in an overall air pollution for example combined heat and power (CHP) management plan; they certainly are and stations would be critical contributors to overall initiatives to reduce emissions from CHPs play an sulfur dioxide (SO2), PM, carbon dioxide (CO2) important role in controlling overall air pollution and greenhouse gas (GHG) emissions and emissions (e.g. SO2 contributions in a regional concentrations. However, since the focus of this context and CO2 in a global context). study is on PM emissions and concentrations in Ulaanbaatar, we have restricted the report mainly Similarly, other analyses, such as on to a focus on the main sources of the high PM greenhouse gases (GHGs) quantification and concentrations in which CHPs seem to be a more Syngas potential have not been considered as they limited source (e.g. due to their high stacks). lie beyond the scope of this study. Comparison of indoor air pollution with the outdoor air pollution concentrations In 2008, a comprehensive Indoor Air Pollution (IAP) study, “Indoor Air Quality Survey� was undertaken in UB by the Ministry of Health (MoH), Public Health Instititue (PHI) and WHO. The annual average concentration values of PM10 and PM2.5 from the study are much lower, particularly in gers, than the measured values in the AMHIB study for outdoor air pollution. Indoor air pollution concentration vs outdoor air pollution concentration (PM10 and PM 2.5) PM 10 (ug/m3) PM 2.5 (ug/m3) Indoor Air Quality Survey 117.91 55.16 (2008)* AMHIB Outdoor air pollu on 350-700 150-250 study (2011)** (year round) * Study conducted during March-June 2007 ** Study conducted throughout the year of 2009 3 2. Air Quality Management Approach Air Quality Management Concept The AMHIB report provides analytical inputs for integrated air quality management. In Given the complex nature of air pollution addition to specific findings, based on available problems in Ulaanbaatar, integrated air data, for current air pollution abatement strategy, quality management is recommended. Used in its analysis should be replicated with updated and successful air pollution abatement programs better quality data, as it becomes available and world-wide, this is a structured approach to lessons learned are collected. a continuous cycle of planning, implementing, evaluating, and adjusting abatement strategies To some extent, the integrated air quality and measures. This includes (a) an assessment management concept contains many common- of air quality and its distribution and variations sense elements. To successfully combat urban in order to understand pollution levels, and the air pollution in Ulaanbaatar, it is important to contributions from various sources to ground understand the characteristics of the various level concentrations; (b) an assessment of the pollutants prevalent in the city, their sources, environmental damage caused by air pollution and their effects. This understanding is the basis and the effects of exposure to air pollution on for effective management of the air pollution the population; (c) identification and assessment problem. of feasible abatement options; (d) a cost-benefit analysis to compare costs of abatement options Figure 2-1 illustrates the integrated air quality with decreases in costs in pollution-related management concept. The outer circle shows how damages; (e) prioritization of abatement measures polluting activities—such as energy use, industrial based on selection criteria that should include activities, and transport—cause emissions that technical feasibility, ease of implementation (e.g. lead to polluted air, which in turn leads to adverse does the locality have the capacity to implement), health effects. This emphasizes the need reduce and the rate of return of the abatement measure the pollution, which usually reduces the emissions in terms of the value of reduced damage costs; and hence improves air quality. The boxes inside (f ) design of a pollution abatement strategy; the circle indicate how the analytical process finds (g) implementation; and (h) monitoring and the most effective solutions. These are linked to evaluation to provide feedback for continuous the outer circle through the starting point—the readjustment and improvement.4 emissions—and through the end point—the solution strategies. In between these points, the analytical process undergoes monitoring data, modeling, health effects assessment, selection of 4 The integrated air quality management concept is described abatement options. These are analyzed to find in detail in the Urban Air Quality Management Strategy, Asia Guidebook: http://books.google.com/books?id=9G0c7d_ the most effective options in terms of comparison nQcEC&pg=PA8&lpg=PA8&dq=urbair+guidebook&source= of control costs with reduced damage costs. bl&ots=9pUtYHlKot&sig=ZlwtkH1m3KqY2OfO3AByz6Zat b0&hl=en&ei=B34SSu2SEI-ysgaus5mFDg&sa=X&oi=book_ The continuous cycle of improvement is a key to result&ct=result&resnum=1#PPP1,M1 success factor. For example, spatial distribution 5 Air Quality Analysis of Ulaanbaatar Figure 2-1: General concept for development of cost-effective air quality management strategies Pollu on drivers / emissions Dispersion Monitoring modeling of air pollu on Solu ons to reduce Emissions Air pollu on Air pollu on concentra ons pollu on Abatement op ons Op mized Exposure abatement costs strategies cost and bene�t value comparison Damage – Effects Impact / health damage Source: Authors’ illustration. of ground-level air pollution (the pollution both these fractions of PM are relevant as a basis people are exposed to) must be overlaid with for estimating the health effects of PM on the spatial distribution of the population. These population. The time period for the monitoring two distributions can change for many reasons covered a full year, from June 2008 to May 2009, including, changes in weather, housing choices, in order to cover all the seasons and to provide increased/decreased mobility, migration, new the annual average concentrations needed for the sources of pollution, new and better quality data, health effects assessment. During the course of etc. Thus, strategies that make sense today may the AMHIB monitoring period, four additional need to be adjusted over time. Figure 2-2 provides monitoring stations were established through a further detailed illustration of how this general donor program.5 The data from this program concept has been applied by this report. provided additional basis for the PM assessment in UB. The PM monitoring programs are summarized Air quality assessment in Ulaanbaatar by the in Chapter 3 and described in detail in Annex A. AMHIB monitoring program There are large spatial variations in the PM The AMHIB air pollution baseline monitoring concentration levels in UB as a result of the program in UB established eight monitoring large spatial variations of the main PM emission stations for PM. Sites for the stations were selected sources. Although the spatial coverage of the on the basis of the following criteria: (a) use of monitoring network for each of the PM fractions existing sites and equipment operated by local (PM10 and PM2.5) was limited the network was environmental agencies and research institutes designed so that the large span of concentration as much as possible; (b) monitor the PM near levels across the city areas could be shown. the local hospitals participating in the health study; and (c) cover the various areas in UB 5 Refurbished air quality monitoring stations were donated by affected by different main sources. The program Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ), included monitoring of PM10 and PM2.5 because an international enterprise for international cooperation. 6 Air Quality Management Approach Assessment of the spatial distribution desulfurization devices, the following equation of PM concentrations and the contributions was used: from different emission sources Emissions Activity Emission Factor Two methods are available for estimating the (1 Efficiency in %) contributions of the various main sources in UB to ground-level PM concentrations in different Emission factors for specific sources are either parts of the city. developed empirically through source testing or cited from publications by authorities such The receptor modeling approach (or known as the United States Environmental Protection as the “top-down� approach) collects monitoring Agency, the European Environment Agency, the data from air pollution monitoring stations and International Energy Association, and accredited uses sophisticated filter analysis to determine the academic sources. The selection of good emission sources of pollution. For the AMHIB study, PM factors is crucial to compiling an accurate and filters from three of the sites were used for this dependable emissions inventory. Otherwise, an analysis. Methods, description, and results are unreliable emission factor could translate into summarized in Chapter 3 and in more detail in large discrepancies in the total emission estimates, Annex B. and even larger discrepancies in the assessment of concentrations and contributions. The dispersion modeling approach (known as the “bottom-up approach�) builds on an A preliminary emissions inventory for UB emissions inventory (EI) of the main sources was first developed in 2007 (Guttikunda 2007). of air pollution. The EI data are used as input This inventory was updated in 2009 (World Bank data to a dispersion model that, together with 2009), and further updated—with improved meteorological and other statistical data, can population distribution data—in 2010 as part estimate the effects of sources on the hourly and of the AMHIB study. The details of the previous thus daily and annual concentrations of major emissions inventories are given in the references pollutants such as particulate matter (PM), SO2, above, while the resulting inventory after the NOX, ozone (03), and others. The dispersion AMHIB updating is summarized in chapter 4 model produces a map of the concentrations in with more details in annex C. The methodologies the model area, giving their spatial distribution. In used in the AMHIB data for UB have provided a Ulaanbaatar the accuracy of the predictions of the basis for assessing the spatial distribution of PM model was evaluated by comparing the modeled contributions and the contributions from the results with the actual monitoring data from the different main sources. PM monitoring stations. The dispersion model and its evaluation are described in Chapter 4 Air pollution and population exposure and in further detail in Annex C. assessment In the development of an emissions inventory In air pollution analysis worldwide, much for Ulaanbaatar, the basic methodology for attention is given to “population exposure�, or estimating emissions from industrial or household the actual pollution concentration level that fuel consumption, as well as traffic activity, used a people are exposed to. It is important to assess standard formula regardless of emission type: the specific contributions of each key pollution source to population exposure—also known as the “source-sector-specific contribution to population Emissions Activity Emission Factor exposure.� This indicator reveals the importance of each source to the health effects of the For industrial or household sources with population and is more meaningful than simply emission reduction and control technology comparing gross emissions amounts per source, such as scrubbers, electrostatic precipitators, or even the average ground-level concentration 7 Air Quality Analysis of Ulaanbaatar contributed by each source. The population heating systems, heat-only-boilers, power exposure should ideally be calculated based upon plants, road traffic, soil suspension etc. data from each person’s movements within the 3. Establish air pollution modeling. A Eulerian various parts of the city day-by-day, or even grid model was used, which has embedded hour-by-hour. However, this detail of population sub-grid models for calculation of pollutant exposure has not been carried out anywhere in concentrations resulting from different types the world and such data are also not available in of sources (area-, line- and point sources). The Ulaanbaatar. model solves the time-dependent advection/ diffusion equation in a three-dimensional grid. The assessment of population exposure in The model grid used for the UB modeling this report is constrained by the data available in is 30 30 km, and the grid cell size used is Ulaanbaatar. Because pollution and people are 1 1 km (see details in Annex C). Input to the unevenly distributed across the city, their exposure model includes the emission inventory and levels are different, depending on where they live meteorological and population data. The and where and when they move about the city. emissions are preprocessed to provide hourly The exposure of the population is summarized emissions in each of the grid cells to the as one number, Population-weighted Exposure model. (PWE), which represents the whole population. The model generates output hourly The PWE sums up the average pollution concentrations throughout the modeled concentrations within one km2 cells on a period (for this report: one full year) in each distribution map (a grid of one km2 cells covering of the grid cells, as well as in specific points, the six central districts of UB) multiplied by the such as the locations of the monitoring total number of people in each cell and divided by stations the total population. This PWE exposure number 4. Calculate the annual average population- is based on the outdoor concentration in the grid weighted exposure, total and contribution cell where each individual lives, and does not, as from each of the source categories, by mentioned above, take account of the difference combining the air pollution and population in exposure that people are subjected to when distributions in the grid. moving outside their area and going to work and school, etc. A time-activity pattern would need to Methods for health effects assessment be established, which is unavailable. In UB this may be less important than in other cities, since For this epidemiological analysis, a method the highest concentrations occur during early similar to those of previous researchers was morning and late afternoon/evening hours when used to examine the relationship between daily people tend to be mostly at home. exposure to air pollution and subsequent health effects. In these studies, statistical associations The PWEs are then used to calculate health are determined between air pollution, including effects (see following section). particulate matter, and several adverse health outcomes, including premature mortality (death) The steps to assess the exposure and the and hospital admissions for heart and lung effects of abatement of sources on the exposure diseases. The method, called “case-crossover,� are as follows: is similar to the case-control study often used in epidemiology. In this method, pollution on 1. Analysis of all available monitoring data. the date of the health event (case) is compared This gives an overview of the current air to pollution on several other reference days pollution problem, and it also provides (controls) occurring within the same month and a basis for evaluating the air pollution year. As a result, each individual in the study dispersion model. serves as his or her own control and there is 2. Establish an emissions inventory, which matching on a wide range of individual-level includes all main sources including ger characteristics. Thus, most factors that can impact 8 Air Quality Management Approach mortality, and which can also vary on a daily response functions. This gives the estimated basis with air pollution, are implicitly taken into health impacts of exposure to excessive levels of account by the study design. ambient concentrations of PM. The AMHIB study includes a time-series study using data This method significantly reduces the from a broad range of hospitals in Ulaanbaatar likelihood that some unmeasured factors will affect to evaluate the exposure-response relationship the results. By limiting cases and controls to the between incidences of hospital admittances for same month, factors that vary over a longer time various respiratory illnesses and cardiovascular scale (e.g., smoking, diet, exercise, age) will not diseases (CVD) and ambient concentrations change. Thus these factors will not have an impact of PM in the city (see Chapter 4). The results on the statistical analysis. This method also allows from that study are applied in terms of exposure- for the examination of the impact of exposures that response functions for hospitalization, as well as occur several days before the actual event. The basic for current hospitalization rates for respiratory analysis involves the full year of data collected in and cardiovascular disease in Ulaanbaatar. UB. However, given the large seasonal differences in concentrations and sources of particulate matter, Third, a unit cost value is estimated for each additional analysis for mortality was conducted health end-point. For premature deaths and by dividing the year into cool and warm seasons. chronic bronchitis, the study team relies on results Ultimately, the study team determined the from the survey on the WTP of Ulaanbaatar expected change in the health effect per every unit residents (see Chapter 5) to reduce risk of death increase in the pollutant under investigation. associated with air pollution. Hospitalization cost estimates are transferred from a previous study in Methods for assessment of air-pollution-related China (World Bank 2007). health costs Methods for assessment of costs of abatement Health impacts are estimated in physical and options monetary terms in the following way: Scenarios analyses were conducted to examine First, a threshold value is set for annual potential costs and PM10 reduction impacts average population-weighted exposure (PWE) of selected PM10 reductions measures. These below which no health risks are assumed. The scenarios include the following. threshold value chosen could be the guideline value set by WHO or other international guideline Baseline; business as usual values or local standards. WHO guideline Scenario 1a: Reduce start-up emissions values are typically determined by the level of through backlighting the fire pollution concentrations that are identified in Scenario 1b: Reduce start-up emissions epidemiological studies as “threshold� levels for through slight modifications of the stove observable effects. Thus, they are a metric for the Scenario 2: Replace existing stoves with actual physical impacts rather than what may cleaner coal stoves, without changing be defined as the target or acceptable level in a the fuel specific setting. For PM, there is no threshold Scenario 3: Replace existing stoves and level below which there are no effects. This study fuels with cleaner stoves and semi-coked uses the annual average PM10 concentration of Coal (SCC) 15 µg/m3 as the lower threshold level for the Scenario 4: Electric heating in existing ger effects, as was done by WHO in the Global homes Burden of Disease assessment (Cohen et al. 2004). Scenario 5: Relocation of ger households into apartments Second, the PWE values are combined Scenario 6: Heat-only boilers with estimated baseline rates of the given health Scenarios 7 and 8: Fugitive dust (paving end-point in the population and exposure- roads and greening the city). 9 Air Quality Analysis of Ulaanbaatar In each alternative scenario, relative changes estimate of the costs of the health effects of compared to the baseline scenario in terms of PM in Ulaanbaatar, and their reduction for a cost and PM10 reductions are examined. Detailed selected set of emission reduction scenarios. assumptions and methodologies are presented in 3. Cost-benefit calculations. This includes a study Chapter 7 and Annex F. on pollution management scenarios and their costs for three main sources (ger stoves, The analytical scheme for cost-benefit soil suspension, heat-only boilers (HOBs), evaluations and the estimated health benefits and their monetary value for each option. The control Figure 2-2 shows the more detailed analytical scenarios cover a 15-year period, 2010–23. scheme followed in the work for this report. This This also includes results in a cost-benefit is based upon the methodologies described above, evaluation, comparing the control costs with and the air quality management concept shown in the value of the health benefit. overview form in Figure 2-1. Figure 2-2 shows the type of values that are The analytical scheme has three parts. obtained from each of the studies and activities (each of the boxes in the figure), the relationship 1. Air pollution assessment. This includes the air between them, and how the values are used in the pollution monitoring, emissions inventory, analytical line. and the modeling (dispersion and receptor) activities. It also includes results in a modeled A strategy for air pollution abatement in air pollution (PM) distribution (spatial and Ulaanbaatar can be developed on the basis of the temporal) for the AMHIB year (June 2008– cost-benefit evaluations of abatement options May 2009). developed in this report, together with other 2. Health effects assessment. This includes the considerations, such as the technical, economic, assessment of the population exposure and political feasibility of the various options. An (population weighted average exposure, air pollution abatement strategy will typically be PWE) and the studies on health effects step-wise, with various options implemented in and willingness-to-pay to avoid effects in the short, medium, and long term. Ulaanbaatar. It also includes results in an 10 Air Quality Management Approach Figure 2-2: Detailed analytical scheme to arrive at the cost-benefit evaluation Source: Authors’ illustration. 11 3. Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar T he objective of this chapter is to the pollution originating from suspension of soil develop an estimate of the exposure dust are the largest contributors to PM pollution of the population in Ulaanbaatar to in UB, and together the two sources dominate particulate matter—the dominant PM concentrations. air pollutant in UB affecting the health of its population. The exposure of the population to PM was calculated by combining the spatial PM The chapter explains the process of concentration distribution across the city developing the exposure assessment, with and ger areas with the similar population references to Annexes A,B,C, and G, where the distribution giving the “population weighted various parts are described in detail. Chapter 3 average exposure,� or PWE. This is very high in describes the monitoring program in UB under Ulaanbaatar—about 430 µg/m3 for PM10 and the AMHIB umbrella, and uses the these results 260 µg/m3 for PM2.5—as an annual average for to estimate the annual average PM concentration the AMHIB year (June 2008–May 2009). This in the two main types of areas in UB—the ger average exposure is about 10 times higher than the areas and the central urban areas—based upon Mongolian air quality standards, and 6–7 times average and maximum measured concentrations. higher than the most lax WHO target values (see The PM concentrations are extremely high box below). It should also be noted that, naturally, (see table 3-2), especially in the winter period. the average exposure value implies that about half They are probably the highest in any capital city the population is exposed to higher values than globally. In addition, the concentrations in the ger the average, and about half to lower values. The areas are much higher than in the central areas. most highly exposed segment of the population experiences annual average concentrations more The chapter assesses the contributions to than two times the average exposure level. The the PM concentrations from the various sources population, especially in the ger areas, is exposed in UB. Two different methods were used. The to extremely high daily PM levels, reaching several first was based upon a combination of chemical thousand µg/m3 during many days of the winter. and statistical analysis of PM samples taken over the year at three locations (called “source The PWE results are used as an important apportionment,� or SA) through receptor input to the calculations of the health effects on modeling. The second was based upon dispersion the population, described in chapters 6 and 8. modeling using emission inventories as its main input. The dispersion model was evaluated The results of this mapping of PM through comparison with the measured results, concentrations in UB should have certain and with strong correlation between measured implications for the development of a long-term and modeled PM. There was also good agreement PM monitoring program in UB, not least that between the two SA methods, with both showing the ger areas should be covered with a number of that the pollution originating from ger stoves and modern air pollution monitoring stations. 13 Air Quality Analysis of Ulaanbaatar Air Quality Standards and Guidelines for Particulate Matter (PM) Air quality standards and guidelines are set to protect the public against risks of negative health impacts caused by the air pollution. The table below summarizes Mongolian air quality standards (AQS) as well as WHO guidelines, USEPA standards, and EU limit values (LV) for PM10 and PM2.5. WHO guidelines are the lowest. They represent the levels where effects are very small, and should be considered as goals for the future. WHO has established interim targets (IT-1-3), realizing that in many developing countries, the WHO guidelines cannot be met in the short term. USEPA standards and EU limit values differ. The EU LV is stricter than the US AQS for PM10, while it is more lax for PM2.5. They represent to some extent what is politically and technically feasible. 3 Guidelines, Standards, Limit values (µg/m ) PM2.5 PM10 Annual 24 hour Annual 24 hour average average average average (daily) (daily) Mongolian Standards, 2007,MNS 4585:2007 25 50 50 100 WHO Guidelines, 2005 10 25 20 50 WHO Interim Targets (IT) IT-1 35 75 70 150 IT-2 25 50 50 100 IT-3 15 37.5 30 75 1 USEPA AQS, 2006 15 35 - 150 3 EU LV 25 4 - 2 20 40 50 1) 7 days above 35 per year is allowed (98th percentile) 2) 35 days above 50 per year is allowed (90th percentile) 3) To be met by 2010 4) To be met by 2020 1Source: WHO 2005. Assessment of air pollution concentrations by health effects of PM. It has also been established monitoring that PM concentrations are much higher in the areas dominated by traditional households Monitoring program (the ger areas) than in the central parts of the cities dominated by apartment houses, as well as more modern single-family houses. Thus, a PM It has been established that particulate matter monitoring program has to cover both the central (PM) constitutes the dominant air pollution areas of the city, as well as ger areas. problem contributing to health effects on the population in Ulaanbaatar (World Bank 2009). As a result, the AMHIB monitoring program Monitoring sites for the AMHIB project concentrates exclusively on monitoring PM are shown in Figure 3-1. They were selected concentrations. The program includes PM10 based upon the knowledge of the spatial air and PM2.5, the two fraction sizes of PM in air pollution distribution in Ulaanbaatar using data that are most relevant for the evaluation of the from stationary monitoring stations that existed 14 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar prior to the start of the AMHIB (operated by The measurement program ran from June 1, NAMHEM, CLEM-NAMHEM, NRC-NUM); 2008, to May 31, 2009. Measurements were the geography and topography of UB; population carried out two days per week (24 hour sampling density representations; and the locations of on Wednesdays and Sundays). In addition, from family hospitals participating in the health October to April measurements were carried out study part of the project. Before the AMHIB every day during one week per month. monitoring started, PM measurements had already been conducted for a short period at the In addition to the AMHIB stations, CLEM station (PM10), at the NAMHEM site four stations provided by a German donor (PM2.5 and PM10), and at the NRC site (PM2.5 and operated by GTZ (now known as GIZ)6 and PM10). The other sites (3, 4, 6, 7, and 8) were came into operation in January 2009, using chosen because they were near the participating automatic monitoring equipment. Two of the family hospitals. The stations were located such stations were located near traffic crossings/ that they represent the general air pollution level streets (termed “Western� and “Eastern Cross� in the surrounding area, and were not affected in figure 3-1). The TV station location is near a significantly by any one single nearby source. ger area, while the airport station is located in a Stations 3, 4, 6, 7, and 8 are located in the more fairly clear area near the airport. The data from polluted ger areas, while 1, 2, and 5 are in more the MUB station provides important additional centrally located populated areas. PM10 was measured at stations no. 1, 2 (central areas), and at 3, 5, and 6 (ger areas). PM2.5 was measured at 6 The GTZ-funded and originally GTZ-operated monitoring stations (MS) have been transferred to the Urban Air Quality stations 1, 2 (central), and at 3, 4, 7, and 8 (ger Department of the Ulaanbaatar Municipality (MUB), now areas). referred to as “MUB stations�. Figure 3-1: Location of AMHIB (yellow), MUB (red), and French (green) sampling sites in Ulaanbaatar Note: Blue points mark meteorological stations. Source: Authors’ illustration. 15 Air Quality Analysis of Ulaanbaatar insight that supports the conclusions regarding very cold days with low winds and ground-level the PM assessment from the AMHIB stations, inversion trapping the pollution emitted at low as well as regarding the importance of the heights within a shallow layer of air near the various source categories in the different areas ground. The very high concentrations measured of Ulaanbaatar. on some days at individual monitoring stations (as shown in the example figures 3-4 and 3-5 Annex A provides details about the below) can represent as much as 50-100 µg/ monitoring program, the stations, the m3, or up to about 25% of the annual average resulting monitoring data, and an assessment concentrations at some stations and are thus of their quality. The annex includes a detailed important for determining the annual PM presentation and visualization of the main features pollution level.7 of PM air pollution in Ulaanbaatar. A summary of the results is provided as follows. Figures 3-2 to 3-5 show the different typical PM pollution features at different stations Measurement equipment at the AMHIB representing the two different types of areas. stations varied, depending upon which At station 6 within the ger areas, the seasonal equipment was already available from the variation and episodic peaks are much more different agencies involved in the AMHIB pronounced than at the NRC station outside the monitoring. The data quality assessment, ger areas. partly based upon parallel sampling with all equipment at the same place carried out during Figure 3-6 and 3-7 show the annual average three different weeks, revealed sampling artifacts PM concentrations as measured at the AMHIB and problems with most of the samplers (see stations. The relevant air quality standards annex A). The data from the parallel sampling, (AQS) are indicated in the figure (as also in together with additional information, gave a figures 3-2–3-5). This shows clearly that the basis for adjusting the output from some of PM concentrations in Ulaanbaatar during the instruments and a final assessment of the the AMHIB measurement period far exceed PM concentrations at the AMHIB stations (see Mongolia’s own AQSs. The concentrations annex G). were also much higher than the most moderate WHO interim targets for PM10 (70 µg/m3) and Measured PM concentrations and their PM2.5 (35 µg/m3). The annual average PM10 variability in Ulaanbaatar, June 08–May 09 concentration at the most affected station was more than 10 times higher than the The most significant feature of PM levels in Mongolian AQS, while the annual average PM2.5 UB is the very strong seasonal variation, with concentration at the most affected station was extremely high concentrations during the more than 25 times higher. winter period, from November to March, as shown in figure 3-2. The ger area stations have It should be noted that the PM2.5 and PM10 much higher concentrations than the NRC levels measured at the stations cannot be directly station, which is located near, but shielded from compared, as they are not measured at the same ger area emissions by houses. This is almost sites in general, and some of the PM2.5 stations are certainly caused by extensive coal burning for located in more polluted areas than some of the heat and power generation during the cold PM10 stations. season, although suspension of dry soil dust is also a significant PM source in UB. Figures 3-4 and 3-5 show the episodic nature of the PM 7 No specific reasons have been found to reject these pollution, with extremely high concentrations measurements. Although errors cannot be ruled out, they are on some days, especially in the winter. Very high regarded as examples of the very high exposures to PM that can occur due to a combination of very large emissions at daily episodes are typically due to a particular low heights locally near the stations and adverse dispersion type of meteorological conditions, typically conditions. 16 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-2: Monthly average concentrations of PM10 in Ulaanbaatar, June 2008–May 2009 Monthly average concentra on PM10 2000 1900 1800 1700 NRC (2) 1600 1500 1400 Zuun ail (3) 1300 1200 3 khoroolol (6) 1100 1000 900 800 700 600 500 400 300 200 100 0 Note: Red line in Figure shows the Mongolian Air Quality Standard. Source: Authors’ illustration based on AMHIB monitoring. Figure 3-3: Monthly average concentrations of PM2.5 in Ulaanbaatar, June 2008–May 2009 1600 Monthly average concentra on PM2.5 1500 1400 NRC (2) 1300 1200 1100 100 ail (3) 1000 900 6 buudal (4) 800 700 600 500 400 300 200 100 0 Note: Red line in Figure shows the Mongolian Air Quality Standard. Source: Authors’ illustration based on AMHIB monitoring. 17 Air Quality Analysis of Ulaanbaatar Figure 3-4: Examples of individual daily measured concentrations at station 6 (3 Kholoolol, in ger area) Timeseries PM 10 (III khoroolol) 1600 1500 3800 2985.7 1400 1300 2725.0 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 6/4/08 6/14/08 6/24/08 7/4/08 7/14/08 7/24/08 8/3/08 8/13/08 8/23/08 9/2/08 9/12/08 9/22/08 10/2/08 10/12/08 10/22/08 11/1/08 11/11/08 11/21/08 12/1/08 12/11/08 12/21/08 12/31/08 1/10/09 1/20/09 1/30/09 2/9/09 2/19/09 3/1/09 3/11/09 3/21/09 3/31/09 4/10/09 4/20/09 4/30/09 5/10/09 5/20/09 5/30/09 Four peaks in the range 2,725-3,800 µg /m3 extend above the top of the vertical axis. Source: Authors’ illustration based on AMHIB monitoring. Figure 3-5: Examples of individual daily measured concentrations at station 2 (NRC, East of city center, outside ger areas) Timeseries PM 10 (NRC) 1000 900 1118 1403 800 700 600 500 400 300 200 100 0 6/4/08 7/4/08 8/3/08 9/2/08 2/9/09 3/1/09 6/14/08 6/24/08 7/14/08 7/24/08 8/13/08 8/23/08 9/12/08 9/22/08 10/2/08 10/12/08 10/22/08 11/1/08 11/11/08 11/21/08 12/1/08 12/11/08 12/21/08 12/31/08 1/10/09 1/20/09 1/30/09 2/19/09 3/11/09 3/21/09 3/31/09 4/10/09 4/20/09 4/30/09 5/10/09 5/20/09 5/30/09 Two peaks of 1,118 and 1,403 µg /m3 extend above the top of the vertical axis. Source: Authors’ illustration based on AMHIB monitoring. 18 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-6: Annual average PM10 concentrations as measured at AMHIB stations in Ulaanbaatar, June 2008–May 2009 Annual average concentration PM10 (in µg/m3) 600 500 400 300 200 100 0 NAMHEM(1) NRC (2) Zuun ail (3) CLEM (5) 3 khoroolol (6) Note: The instruments at stations 1 and 5 measure too low concentrations, and are thus indicated with a different color in the figure. Source: Authors’ illustration based on AMHIB monitoring. Figure 3-7: Annual average PM2.5 concentrations as measured at AMHIB stations in Ulaanbaatar, June 2008–May 2009 Annual average concentration PM2.5 (in µg/m3) 700 600 500 400 300 200 100 0 NRC (2) 100 ail (3) 6 buudal (4) Bayan Airport(8) hoshuu(7) Note: The instrument at stations 4, 7 and 8 measure exceptionally higher concentrations during the winter months, due to influence from high relative humidity. Source: Authors’ illustration based on AMHIB monitoring. 19 Air Quality Analysis of Ulaanbaatar Figures 3-2–3-7 present the concentrations Summary of PM levels in Ulaanbaatar measured by the various equipment. The data quality assessment based upon inter-comparison The UB data show the large difference in between the various instruments revealed concentration between the AMHIB ger area sampling artifacts and problems with most of stations and the NRC station outside the ger the samplers (see annex A). Data correction areas. The GTZ stations provide additional procedures were developed based upon the inter- data. They are located mostly outside the ger comparisons and other information. areas, in central UB locations as well as in a thinly populated ger area near the airport. The Applying these corrections to the GTZ station at the TV antenna is different, measured concentrations gives the corrected although it is located near ger areas. The PM concentrations presented in table 3-1. concentrations reported for the MUB stations The uncertainty of the corrected numbers are significantly lower than the AMHIB ger is indicated with indices 1–3, with 1 being area concentrations (see annex A, table A.6). It the least uncertain and 3 the most uncertain. is not possible simply from the measurements Statistics for quantifying the degree of at the stations to establish an average overall uncertainty are not available. It is judged it to PM concentration number for Ulaanbaatar. Air be of the order of ±10–15 percent for index 1, pollution modeling is required to achieve that. ± 15–20 percent for index 2, and ± 40–50 percent However, from the measurements the annual for index 3. average concentration levels in the two types of areas, the ger areas and the central UB areas, can Compared to the data from other cities be indicated separately. The line of reasoning in global data bases, Ulaanbaatar could be is described in chapter A5 in annex A. Table the most PM polluted capital in the world, 3-2 shows the result of this assessment. The with higher concentrations than in Chinese results from the PM measurements at the four cities. Figure 3-8 shows examples of PM MUB stations that provided data for the period concentrations in other highly polluted cities December 2008–May 2009 confirm the ranges (World Bank 2009). indicated in table 3-2. Table 3-1: Annual average PM concentration of Ulaanbaatar city, corrected according to results from sampler comparisons Site Site name PM2.5 PM10 No 2 Nuclear Research Center 78 3 253 1 3 Zuun ail 236 3 558 1 4 6 Buudal 225 2 6 III Khoroolol >700 3 7 Bayankhoshuu 338 2 8 Airport 190 2 1 NAMHEM 59 4 76 4 5 C LEM 67 4 Note: Indexes 1-3 indicate the uncertainty of the numbers, with 1 the least uncertain and 3 the most uncertain. Index 4: The instruments at these stations were shown to give too low values. They are included here for completeness. Source: AMHIB data. 20 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-8: Comparison of UB PM10 concentrations (2008–09) with Chinese cities (2008) and other capitals in the world (2004) Source: Authors’ illustration based on data from the China Environment Yearbook 2009 for Chinese cities, AMHIB study for UB, and WHO Air Quality Guidelines - Global Update 2005 for other cities. Table 3.2: Indicated ranges for annual average PM concentrations in Ulaanbaatar, June 08-May 09, linked to the areas where the measurements were carried out Area PM10 PM2.5 Exceedance: Ra o to AQSs 3 3 µg/m µg/m Mongolian: WHO Central city areas 150–250 75–150 3–6 7–15 Ger areas 350–700 200–350 7–14 17–35 Note: These indicated PM levels represent the situation in the AMHIB period, June 2008-May 2009. They are linked to the areas where the measurements were carried out. They do not cover all polluted areas in Ulaanbaatar. Source: AMHIB data. PM has been measured at the NRC station during the last (AMHIB) winter were also higher since 2006. Figure 3-9 shows monthly time series than in the first two winters of measuring. For of PM concentration at the NRC site (AMHIB PM2.5, the highest levels were measured during site 2) for 2006–09. This time series shows a the last winter, and these were significantly higher trend toward increased concentrations over the than in earlier years. Possible explanations are period, both during the summer and winter. The increased polluting activities resulting in higher highest winter concentrations of PM10 occurred emissions and/or in changes in meteorological during winter 2007–08, while concentrations conditions. 21 Air Quality Analysis of Ulaanbaatar Figure 3-9: PM10 (total), PM10-2.5 (coarse), and PM2.5 (fine) concentrations at the NRC station (AMHIB site 2) for 2006–09 700 NRC 2006-2009 600 500 400 �ne Coarse 300 Total 200 100 0 Jan/06 Mar/06 May/06 Jan/07 Mar/07 Jan/08 Jan/09 May/07 Mar/08 May/08 Mar/09 May/09 Sep/06 Sep/08 Jul/06 Nov/06 Sep/07 Nov/07 Nov/08 Jul/07 Jul/08 Source: National University of Mongolia and authors illustration based on AMHIB monitoring Contributions to PM concentrations from main emissions sources. The main sources of PM concentrations contribution from vehicle exhaust near the in Ulaanbaatar have previously been assessed as heavily trafficked road crossing. At that time of emissions from the burning of coal and wood for the year, concentrations from ger heating sources heating of individual residences in ger areas (ger and other combustion sources (except the power stoves) and suspension of dry dust from roads plants) are very low. Concentrations during the and other surfaces (World Bank 2009). Emissions morning hours, as well as during the afternoon from power plants, heat-only boilers (HOB), rush hour and evening hours, are increased, and car and vehicle exhaust are also important although this is not very apparent in the scale sources, although they were assessed as smaller used in figure 3-10. This increase is the result contributors to ground-level PM concentrations. of the combined influences of the typical daily wind speed variation (increased day-time wind The elemental analysis of sampled PM on speed), the daily variation in traffic volume with filters from the AMHIB stations supported the rush hours, as well as a small influence from dominance of coal combustion and suspended soil evening cooking in gers. These variations and particles. source influences are very small compared to the very much higher winter time concentrations. An additional supporting indication of The winter curves show strong morning and the extent of contributions from main sources evening increases at the Western Cross station, can be extracted from the AMHIB and MUB which is located in a central UB area but fairly measurements. Starting with the PM situation close to ger areas to its north and west. The in central UB areas, the hourly measurements of periods of increase correspond with periods of PM10 and PM2.5 at the MUB station “Western ger heating. The prolonged evening/night peaks Cross� can be used as a descriptive illustration (see can be partly explained by the normally reduced figure 3-10). wind speed at night, combined with often strong ground-level temperature inversions causing PM2.5 concentrations during summer reduced dispersion and increased concentrations. (given as June in the figure) represent the In addition, the afternoon traffic rush hours 22 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar are not as pronounced in the winter months A similar analysis from the MUB TV station indicating that the vehicle exhaust contribution located close to ger areas shows similar variations to the PM2.5 concentrations is limited when to the Western Cross station, although these compared to other sources, even at this traffic- are more pronounced, indicating the exposed site. importance of ger area emissions (see annex A, figure A.31). The much higher PM10 concentrations indicate a large influence from sources of the The measurements presented in this sub- coarse PM fraction (particle sizes between 2.5 chapter indicate that ger heating sources are a and 10 micrometer) even in the summer (June). major contributor to PM2.5 concentrations in Sources of coarse PM are suspension of dry the ger areas. This also contributes to PM in the dust from surfaces, such as open dry ground, central areas of UB. Dust suspension from open as well as roads and their surroundings. The soil surfaces and roads is another major source combustion of coal and wood in gers and other of coarse PM. The influence from road vehicle combustion sources also contribute to some exhaust is limited, even in central traffic areas. extent to coarse PM. The average daily variation The contribution from power plants and HOBs shown in figure 3-10—increases significantly in to PM concentrations in UB cannot be quantified the morning, increases slightly less during mid- by means of the data presented in this section. day, and has a strong evening peak—indicating that road dust and other coarse PM sources are Such evaluations, based upon analysis of involved. The coarse fraction is especially large measured time series, can only be used as an during spring time (March–May) providing a indication of PM source contributions and their clear indication that suspension of dust from importance. Quantitative analysis for source dry surfaces (open soil surfaces and roads) is an contributions are described in the following important source. sections. Figure 3-10: Monthly averaged daily variation of PM10 and PM2.5 at the MUB Western Cross Station during work days continued 23 Air Quality Analysis of Ulaanbaatar Figure 3-10 continued Source: Authors’ illustration based on AMHIB monitoring. Ger stove firing practices and variability in day and night as is the practice in the colder mid- the PM concentrations in Ulaanbaatar and late winter. Hourly measurements of PM show typical daily By an approximate calculation, about 50 variations that correlate with the daily routine firing percent of PM concentrations corresponds to the practices of the typical ger stove. Figure 3-11 shows, morning ignition phase (cold start) (8.30–9.30), as a typical example, PM hourly concentrations the evening phase (18.30–19.30), and during the on November 19–20 at the Takhilt meteorological reloading of stoves (20.30–21.30). site (Meteo site no. 3) in a ger area to the west of UB center. Clear peaks are shown in the morning Quantification of contributions to PM air (the right-most peak in the figure) and two-three pollution from various source categories successive peaks in the afternoon/evening. These correspond with ger stove firing periods. The Emission inventory evening peak is longer, corresponding to prolonged firing, while the morning firing is shorter, for The amount of emissions from the various heating and cooking before going to work. The source categories provides an initial estimate of evening peaks correspond to stove loadings. their contribution to air pollution in an area. The emissions inventory presented in table 3-3 It is notable that smoke pollution tends to is based upon a preliminary inventory prepared be higher in December early in the winter than for the World Bank in 2007 (Guttikunda later (see figure A.36 and figure A.37 in annex 2007), updated in 2008–2009 (World Bank A). This occurred despite the daily average 2009) and further updated in the preparation of temperature being slightly in lower in January this final report. The inventory shows that ger 2009 than December 2008. This could be households are the largest source of PM emissions because at the beginning of the winter people in total, while power plants (CHP) and road keep the stoves cold, and fire them two or more dust suspension also contribute large amounts times a day. This makes more pollution than of emissions. The CHP emissions used here are when the stoves are kept hotter throughout the based upon emission measurements carried out by 24 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-11: Example of time series of PM for one day at the beginning of winter, at Takhilt meteoro- logical site, November 19–20, 2008 Note: Measurement using a GRIMM 107 PM monitor. Source: Authors’ illustration based on AMHIB monitoring. Table 3-3: Summary of the emissions inventory for Ulaanbaatar, 2008 (tons/year) Source PM10 PM2.5 SO2 Height of Spa al distribu on emissions, meters Ger 19,731 15,785 8,784 3–5 Throughout ger areas households HOBs 1,077 646 4,360 10–20 Dispersed over UB surroundings CHPs 18,589 7,436 33,600 100–200 3 point sources to the west of UB center Vehicle 1,161 1,161 1,354 <1 Dispersed along main road exhaust system mainly throughout the central city areas Dry dust from roads Paved 5,142 771 <1 Mainly throughout the central city areas Unpaved 4,812 722 <1 Mainly throughout the ger areas Source: AMHIB data. JICA (see annex C, which includes an uncertainty Contributions to ground-level evaluation). It is important to note that soil dust concentrations suspension from surfaces other than roads is not included in this overview of emissions, although Table 3-3 also shows the height of the source it is included as a source in the air pollution emission points and their spatial distribution. modeling. The higher the emission point, the lower tends 25 Air Quality Analysis of Ulaanbaatar to be the general contribution from the source Source apportionment by receptor modeling to ground-level concentrations, due to the from AMHIB data analysis increased dispersion of the emissions before it reaches the ground. Thus, the contribution Samples and methods of the CHPs to ground-level concentrations is much less than indicated by the emission The samples collected at three AMHIB sites were amount in the table. The spatial distribution used for the analysis: (1) NRC (site 2); (2) Zuun of emissions is obviously also important, Ail (site 3), where two fractions, PM2.5 (fine) and and causes the source contributions to vary PM10-2.5 (coarse) were collected; and (3) at III across different areas of the city. The large Khoroolol (site 6) where only one fraction, PM10 amount of emissions from the ger areas at a was collected. Figure 3-1 shows their location. low emission height, dispersed and at the same The three sites represent different locations and time concentrated among the population in the positions relative to main PM sources in UB, and Ger areas, account for a potentially significant thus give a degree of spatial understanding of the contribution in those areas. Road dust and source contributions to PM in the air: vehicle exhaust account for a significant contribution in the areas with large amounts ■ Site 2, NRC, is located a few kilometers to of traffic. The contribution from HOBs can be the east of central UB away from the main significant near and downwind of areas with ger areas and main roads, in an area with many HOB stacks. This is also the case for open soil surfaces in the neighborhood. It is power plant emissions, although it is limited expected that the soil sources will contribute due to the height of their stacks. significantly to this site. ■ Site 3, Zuun Ail, is located a few kilometers Assessment of the source contributions to northeast of central UB, near small valleys to ground-level concentrations can be carried out the north (Chingeltei, Hailaast, and Selbe). using one of the following two methods: Extensive ger areas are located in those valleys to the north and expose the site to emissions ■ Source apportionment of PM sources from ger heating systems when air flows through receptor modeling. This method down the valleys toward the south. Few open requires extensive analysis of a large number soil surfaces surround the site, and the site is of PM samples collected on filters. The also removed from the main roads, although AMHIB data collection provided the basis there are unpaved roads with light traffic. for such an analysis. This method provides Sampling height at the site 3 is at 6m, which estimates of source contributions at the sites might slightly reduce the influence of local where the samples were taken. dust re-suspension. It is expected that ger area ■ Source apportionment through dispersion emissions will contribute significantly to this modeling. This requires an emission site. inventory, meteorological and other data, ■ Site 6, III Khoroolol, is located well within a and a tested dispersion model evaluated for ger area surrounded by local unpaved roads. the area. This method can potentially assess The sampler was placed at a height of 4m the contributions spatially throughout the height on a balcony on the second floor of a modeling area. residential house. Again, ger area emissions along with soil and unpaved road dust, are Results from receptor modeling of expected to contribute significantly to this source contributions are given in the sections site. immediately below. Results from dispersion modeling on average source contributions and its A total of 545 particulate matter samples spatial variation are given in the last section of this were included in the receptor modeling analyses chapter. of AMHIB data, from June 2008 to May 2009. 26 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Annex B gives details on the methods, analysis, Combustion: In some cases, it has been and results from the receptor modeling. A possible to identify two distinct combustion summary is provided below. source types with differing combustion characteristics. One contains black carbon and a Identification of PM sources significant sulfur component (called Combustion 1); while the other also has black carbon, but Output from the statistical analysis, using the with relatively lower sulfur content and higher Positive Matrix Factorization method (PMF) soil elements (called Combustion 2). The high- (see annex B), identifies a number of “factors� sulfur profile could be associated with high- which are defined by their specific and separate temperature coal combustion (such as in power element composition, or profile. From this plants and boilers), while the low-sulfur could elemental profile, it is often possible to allocate be associated with low-temperature combustion, the “factor� to a certain pollution source, such as in small-scale residential stoves. The based upon knowledge of tracer elements or combustion sources contribute mostly to the fine the elemental composition of the emissions PM fraction (PM2.5), while they also have smaller from the source type. From the statistical contributions to the coarse PM fraction. analysis of the AMHIB data, it was possible to allocate factors to sources associated with coal Motor vehicles/road traffic: The profile combustion, motor vehicles, road dust, and soil. associated with a local motor vehicle and road dust component contains BC, zinc, and lead, The following sources could be identified along with elements typical of crustal matter. through the factors derived from the PMF analysis: Mongolia has recently phased out leaded petrol, although there is likely to be residual lead in local Soil: Airborne soil originates from crustal road dust. This is a mixed profile consisting of matter. It has been possible to identify two exhaust particles in the fine fraction (PM2.5) and different sources of airborne soil in UB: (i) a suspended road dust in the coarse PM fraction source identified by soil elements (named Soil (PM10-2.5). These two parts of the profile are highly 1) and (ii) a Soil 2 source with a significantly correlated in time since they both originate from higher black carbon (BC) component. The road traffic with its specific time variation, and difference between the two crustal matter sources thus they appear in the analysis as one source. is most likely the location, with the Soil 2 source originating more locally in Ulaanbaatar, Biomass burning: The profile associated where there is likely to be a greater concentration with biomass burning contains black carbon and of settled combustion particles and coal dusts potassium in the samples. In UB, biomass burning mixed into the crustal matter, hence the higher contributes mostly to the fine PM fraction (PM2.5). presence of BC in the source profile. Soil 1 is likely to represent the general crustal matter in These profiles/sources contribute differently the area around Ulaanbaatar. The soil sources to the coarse and fine fractions of PM at the contribute to PM in air through the well-known various sites. action of wind and turbulence to suspend the particles in the soil surface in the air. The soil Results of the source apportionment using sources in UB contribute mostly to the coarse the PMF method PM fraction (PM10-2.5), but it also accounts for a significant portion of the fine PM fraction Contributions from the various sources were (PM2.5). The very dry and cold climate in calculated for the three sites (NRC, Zuun Ail, and Mongolia probably causes a distribution of the III Khoroolol). For the NRC site, a five-year data top-soil particles consisting of a higher fraction of series (2004–09) was also available, and source very fine particles, compared to areas with more apportionment (SA) analysis was also carried out humid and mild climates. for this five-year period. 27 Air Quality Analysis of Ulaanbaatar The SA results are summarized in figures accounts for 35–92 percent of the total PM2.5 3-12 and 3-13. concentration at the three sites (see figure 3-12). The suspended soil particles dominate the coarse This indicates that the combustion source fraction (PM10-2.5), accounting for 70–90 percent is the largest contributor to the fine PM fraction of the total. This source also contributes to PM2.5, (PM2.5) concentrations, both in concentration particularly at the NRC site 2. When these two and percentage, especially at site 3. Combustion fractions are combined into PM10, both soil and combustion contribute significantly to PM10, but there is variation among the sites. Soil dominates Figure 3-12: Concentration contributions at site 2 (NRC), combustion dominates at site 6 (μg/m3) to PM in air from main sources in UB (III Khoroolol), while they contribute equally at Top: PM2.5; Mid: PM10-2.5; Bottom: PM10 site 3 (Zuun ail). The motor vehicles/road dust source accounts for about 3-12 percent of PM2.5 at the three sites, and 5–20 percent of the coarse PM fraction, due to suspended road dust. The biomass contribution is very small except for PM10 at site 6, which is located in a ger area. Discussion of source contributions The source apportionment analysis indicates that coal combustion and soil suspension dominate as sources of PM at stations 2, 3, and 6, and that motor vehicles and biomass burning account for smaller contributions. There are significant differences between the stations in the amount of contributions to the different PM size fractions. PM2.5 (fine particle fraction) Soil suspension and coal combustion both make substantial contributions to PM2.5 at site 2 (NRC), while coal combustion dominates completely at site 3 (Zuun Ail). This can be explained by the differences in site locations. NRC is located away from main ger areas but within an area with open and often dry soil surfaces, while Zuun Ail is exposed to ger area emissions with no extensive dry soil surfaces close to the station. Both are located well away from paved main roads and while there is traffic on smaller unpaved roads near the sites the motor vehicle contribution to PM2.5 is low at both sites. At NRC, there is some discrepancy between the AMHIB period and the three and a half-year period of analysis (2006-09). There was relatively Source: Authors’ illustration based on AMHIB monitoring. more soil in the AMHIB period at the expense of 28 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-13: Relative contributions (percent) to PM in air from main sources in UB Top: PM2.5; Mid: PM10-2.5; Bottom: PM10. Source: Authors’ illustration based on AMHIB monitoring. 29 Air Quality Analysis of Ulaanbaatar the coal combustion source, while the opposite ground. At site 3, which is exposed to ger area holds for the three and a half-year period. Figure emissions, the source of combustion in Ger stoves 3-9 shows that the PM2.5 fraction was very high in dominates completely. At site 6, located within a the AMHIB period compared to earlier winters. Ger area where the combustion factor dominates The source apportionment analysis indicates that completely, the analysis could not distinguish the this increase is linked to a larger contribution two profiles. from the soil source. It is possible that drier conditions in the AMHIB period led to increased Soil 1 vs. Soil 2 The two soil profiles could fine particle suspension. be separated in the analysis only for the coarse fraction at site 2 (NRC) where they contribute PM10-2.5 (coarse particle fraction) Soil roughly equally to coarse PM at the sites. Thus, suspension dominates at both sites 2 and 3. the general (Soil 1) and the more local (Soil 2) The motor vehicle source provides a measurable dust suspension sources appear to be equally contribution to both sites, due to road dust important. This indicates that control of local soil suspension. Although the distance to main roads suspension, e.g. by covering open surfaces, will is considerable, there is traffic on smaller unpaved only address part of the soil suspension problem. roads near the sites. also In addition, traffic in the entire area causes road dust concentrations in the Assessment of air pollution exposure through whole UB air shed. air pollution modeling PM10 PM10 is the sum of PM2.5 and PM10- The spatial distribution of PM 2.5. The PM10-2.5 (coarse fraction) concentration concentrations in UB and the contribution is generally substantially higher than the PM2.5 concentration (see figure 3-9). As a consequence of from main sources this and the above, soil dust suspension accounts for the largest contribution of PM10 at sites 2 and Air pollution dispersion models applied and 3. Coal combustion is a very large source a site 3, compared with measurements A state-of-the- while road dust makes a significant contribution to art Eulerian grid model established for urban site 2. scale applications (Slørdal et al. 2003) was used to model the spatial and temporal air pollution concentrations in Ulaanbaatar for the AMHIB For site 6, only PM10 apportionment has study. A diagnostic wind field model was used been carried out. Site 6 is located well within a to provide the hourly meteorological data fields ger area with small unpaved roads surrounding it. needed as input to the dispersion model. The Hence coal combustion dominates this area and emission inventory for Ulaanbaatar was used as road dust gives a significant contribution, while input to an emissions model that provides hourly also biomass (wood) burning and soil suspension gridded emissions as input to the dispersion are also noticeable as sources. calculations. The models are described in annex C. The model calculations for Ulaanbaatar have Combustion 1 (high-temp) vs. combustion concentrated on particulate matter, PM. 2 (low-temp) These two combustion-related profiles could be distinguished in the analysis The software system platform AirQUIS was of the fine PM fraction (PM2.5) both at sites used to perform the modeling work for UB. The 2 (NRC) and 3 (Zuun ail). At site 2 (NRC), GIS-based AirQUIS system (AirQUIS 2008) the two sources contribute about equally, is an integrated air quality management system corresponding to the location of site 2, which is that contains different modules for treating away from main ger areas, but in an region with and combining various data such as emissions heat-only boilers. Also, the power plant emissions inventory data, geographical information data, might contribute to this site, as it is located measurement data, as well as the various models downwind of the power plant stacks at a distance needed to perform dispersion and exposure where the power plant plumes could reach the calculations. 30 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar The dispersion model was evaluated by Main results from the update of the comparing the results from the modeling emissions inventory The emissions inventory with those from the measurements and source resulting from the updated AMHIB study, as apportionment analysis, as described in the shown in table 4-3, is described in detail in previous two sections. Details of the evaluation annex C. Updated population data resulted in are in annex C. an increase in the calculated PM emissions from the ger households by 20.5 percent. The total The comparison with concentration emissions from the power plants (CHPs) and measurements shows that the modeled and the heat-only boilers (HOBs) have also been measured annual PM2.5 concentrations at modified, based upon new data and information. stations 3, 4, 7, and 8 (located in ger areas) The road traffic (vehicle exhaust and road dust agree within 10 percent. The modeled suspension) emissions are unchanged from concentration is also within the uncertainty the discussion paper (World Bank 2009). The range of the PM10 measured at station 3. even more significant change resulting from This is also the case for PM10 at station 6, the updated population data is in the spatial although the measured concentration is quite distribution of the ger household emissions. The uncertain. Station 2, NRC, is located a couple new population data resulted in a distribution of kilometers east of the UB central area. much more concentrated in the ger areas The model substantially overestimates the closer to the UB central areas , a result that concentrations at this location, for PM10 and is important for the calculation of the spatial more so for PM2.5. The model grids surrounding ground-level PM10 concentration distribution. station 2 are, according to the population Figure 3-14 shows the large change in ger area distribution, located within the ger areas with emissions distribution as a result of the new a high emission density. The emissions might population distribution. In addition, the spatial also be overestimated for this area. Another distribution of HOB emissions changed to some explanatory factor could be that the model degree. underestimates the windier conditions along the river valley area, where station 2 is located. Modeled present PM concentrations in More detailed wind measurements are needed to UB, its spatial distribution and contributions examine this issue further. from pollution sources Figure 3-15 shows the modeled spatial distribution of concentrations A comparison with the statistical source of PM10 and PM2.5, as an annual average. The apportionment (SA) indicates that on the model calculates the average concentration in average for the three PM10 cases, the dispersion each of the km2 grids. The concentrations vary model gives about 10 percent (absolute) higher considerably between different areas within combustion contribution and about 14 percent UB, with the highest concentrations in the ger lower soil and road dust contribution than the areas just north of the city’s center. There is a measured SA. The opposite applies for the two very steep gradient toward lower concentrations PM2.5 cases, where the dispersion model gives when moving away from the ger area and south approximately 17 percent lower combustion into the city. The PM10 concentrations reach contribution and about 9 percent higher soil as high as 900 µg/m3, and the PM2.5 as high as and road dust contribution than the measured 550–600 µg/m3, as an annual average. These are SA. extremely high concentrations by any standard, including compared to Mongolia’s own Air Therefore, it can be concluded that Quality Standards, which are 50 µg/m3 for PM10 the dispersion model compares with the and 25 µg/m3 for PM2.5 (see earlier part of this measurements of PM and its elemental chapter). composition sufficiently well to be used as a basis for cost effectiveness and cost-benefit analysis of The calculated spatial distribution maps are abatement measures. in accordance with the general assessment of the 31 Air Quality Analysis of Ulaanbaatar Figure 3-14: Spatial distribution of emissions from ger stoves, tons/km2/year Note: The updated distribution (top) compared to the preliminary distribution used previously (bottom). Source: World Bank 2009. ranges of PM concentrations indicated from the The contributions from the various categories results of the measurements and as summarized in of sources to the concentrations has been assessed table 3-1. From the evaluation of the model (see the both by measurements, and by the dispersion section above and annex C), the results from the air model calculations. Both methods indicate that pollution modeling provide an acceptable basis for combustion in ger heating systems and suspension assessing the PM concentrations and its distribution, of dry dust from open soil surfaces and roads contributions from sources, and subsequent are the main contributors to PM in UB. These assessments of the effects of various control scenarios two sources represent 75–95 percent of the PM on reducing PM pollution in Ulaanbaatar. concentrations, varying between locations, on the 32 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar Figure 3-15: Spatial distribution of PM10 (top) and PM2.5 (bottom), annual average (μg/m3) Note: Model calculations for the period June 2008 to May 2009. Source: Authors’ illustration based on AMHIB data. higher end for PM2.5 and on the lower end for and 3-17 show the spatial contributions from PM10. Vehicle exhaust from traffic accounts for three of the less dominant source categories: only a minor contribution, except areas near the HOBs, road traffic (including exhaust particles most heavily trafficked roads. plus suspended road dust), and the power plants (CHPs), for PM10 and PM2.5. The HOB Maps of the spatial distribution of the distribution reflects the HOB locations and gives contributions from the gers and from the soil an estimated annual average PM concentration suspension would be very similar to the total up to 20 µg/m3 for PM10 and 12 µg/m3 for distribution shown in figure 3-15. Figures 3-16 PM2.5 in the most affected areas. The traffic 33 Air Quality Analysis of Ulaanbaatar contribution is concentrated in the city center contribution is considerably lower in the ger areas and its contribution is quite substantial in this in general, although it is substantial near main area, with an estimated annual average up to and roads, especially because of road dust suspension. above 40 µg/m3 for PM10 and 24 µg/m3 for PM2.5 The CHP distribution is mostly downwind of in the most affected areas. The influence of traffic the power plants in the direction of the prevailing Figure 3-16: Spatial distribution of source contributions, PM10, annual average (μg/m3) Note: Model calculations for the period June 2008 to May 2009. Source: Authors’ illustration based on AMHIB data. 34 Assessment of the Present Air Pollution Exposure and Source Contributions in Ulaanbaatar westerly wind. Its calculated contribution meteorological conditions allow the plumes to annual average PM concentrations is from the chimneys to reach the ground level, comparatively low. CHP emissions and resulting particularly if this weather coincides with periods ground-level concentrations can rise in shorter when the cleaning equipment of the power plants periods (less than several hours) when suitable cleaning equipment is not operating properly. Figure 3-17: Spatial distribution of source contributions, PM2.5, annual average (μg/m3) Note: Model calculations for the period June 2008 to May 2009. Source: Authors’ illustration based on AMHIB data. 35 Air Quality Analysis of Ulaanbaatar The exposure of the UB population to PM calculations, influence ground level in the more concentrations, and its source contributions populated areas most of the time. Vehicle exhaust also makes only a small contribution, since total The exposure of the population of Ulaanbaatar PM emissions from road traffic is relatively small, to PM in the air is calculated in terms of the although the suspension of road dust makes a population-weighted averaged concentration, the noticeable contribution to the levels of PM10. PWE, as described in chapter 2. Table 3-4 gives the calculated PWE for PM10 and PM2.5 for UB Implications of air pollution mapping for the for the AMHIB period June 2008–May 2009 monitoring network in UB and present the contributions to PWE, calculated from the various main sources. The mapping of the PM concentration distribution in Ulaanbaatar, both from the Emissions from the ger heating systems monitoring results and from the modeling, (burning coal and wood in stoves and small provided a basis for recommendations on boilers in ger housing units and kiosks in the improving the stationary air pollution monitoring ger areas) account for the largest contribution to network in UB. For example, the mapping shows the PWE (46 percent for PM10 and 60 percent that parts of the ger areas have much higher for PM2.5). Wind-blown dust from open surfaces concentrations than the central districts. This in the city represents the second largest source leads to the conclusion that the stationary air of PWE (31 percent for PM10 and 16 percent pollution monitoring network in UB should for PM2.5). Combustion residues, essentially ash include automatic stations in the ger districts. and soot accumulating in ger stoves, which are In chapter A8 in annex A, recommendations dumped in open space and dispersed by the wind, regarding the design and content of a stationary contribute 11 percent and 15 percent to PM10 and long-term monitoring network in UB have PM2.5 concentrations, respectively. Power plants, been developed. The recommendations were although representing a large emission source, based on an overview of the existing AMHIB provide only a relatively small contribution and the availability of modern monitoring because their tall chimneys lift the power plant equipment, both within the Mongolian and UB plumes and do not, according to the model institutions as well as through the German and French donor programs. (see annex A) One of the main recommendations is to also establish modern monitoring stations in the ger districts. Table 3-4: Population-weighted exposure (PWE) A suggested monitoring network is included in to PM in Ulaanbaatar as calculated by the air figure A.34. pollution model, contributions from main sources (μg/m3) The suggested minimum network includes: ■ 3 urban background stations in the most polluted ger areas ■ 1 to 3 urban background stations in central areas ■ 1 station between ger and central areas ■ 3 traffic stations (located near main road links) ■ 1 reference station in a relatively clean area ■ some additional specialized sites. The final design of a long-term monitoring network should result from discussions among the Source: AMHIB data. stakeholders. 36 4. Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar T his chapter summarizes the statistical Introduction analysis of the association between daily exposure to air pollution and This chapter summarizes the results of an analysis both premature death (mortality) of the association between air pollution and both and hospitalization (morbidity) using one year premature death and hospitalization in UB. The of data from Ulaanbaatar. The analysis involved study team compared data on daily readings from consideration of daily concentrations of four particulate matter (PM) monitors with concurrent pollutants: PM10 and PM2.5 (particulate matter data from health records from UB. Specifically, less than 10 microns and 2.5 microns in diameter, the team examined whether concentrations respectively), coarse particles (particles between of PM2.5, PM10, and coarse particles (CP) are 2.5 and 10 microns), and nitrogen dioxide. statistically related to either premature death While analysis of this nature generally requires (mortality) or additional hospital admissions several years of data to ensure that the data are (morbidity) from either heart or lung disease. sufficiently detailed to detect an effect, several Fine particles more easily evade the natural important results were obtained. The study team defense mechanisms in the human airways than observed that when using the full year of data coarse particles and are more likely to penetrate collected, daily exposure to coarse particles was deeply into the lungs. The coarse particles will associated with increases in mortality. In addition, more likely be filtered out by the nose and a strong effect was observed between PM2.5 and upper airways. However, if the concentration of mortality during the warm season, when people coarse particles (and hence PM10) is particularly have greater exposure to outdoor air. Strong high, there is evidence that they also will cause associations were observed for hospital admissions significant health effects (Malig and Ostro 2009). for both heart and lung disease with PM2.5, Figure 4-1 indicates the relative size of particles. PM10, and nitrogen dioxide. The effect estimates for mortality and morbidity, measured as the The ambient concentrations of both PM2.5 percent change in the health outcome per every and PM10 in UB are extremely high. Therefore, it unit change in the pollutant concentration, were is important to determine the magnitude of their generally similar in magnitude to those reported effects on public health. Besides PM, the impacts in studies in North America, Europe, and other of nitrogen dioxide (NO2)—a product of fuel parts of Asia. Thus, despite the relative sparseness combustion—are also examined in this analysis. of data, the findings indicate that particulate As is the case for PM, associations between NO2 matter and other pollutants in UB have serious and both death and disease have been observed in negative effects on public health. Additional years epidemiologic studies (U.S. EPA 2008). Accordingly, of data would help confirm these findings and the study team analyzed the relationship between also aid in determining the sources of pollution daily exposures to PM and NO2 with both mortality that have most significant impact on health. and hospital admissions in UB. 37 Air Quality Analysis of Ulaanbaatar Figure 4-1: The size of PM2.5 and PM10 relative to human hair and beach sand Source: Courtesy of U.S. EPA. For this epidemiological analysis, the study presented below, particularly for mortality, should team used methods similar to those used over the be viewed as preliminary, pending additional years last decade to examine the relationship between of analysis. Details on the complete analysis and a daily exposure to air pollution and subsequent discussion of the results are in annex D. health effects. These studies, conducted in dozens of cities around the world, have demonstrated Data and Methods statistical associations between several adverse health outcomes and daily, multiday, and long- Exposure data term (one year to several years) changes in outdoor air pollutants, including particulate All pollution data were obtained from the matter (PM) (Brook et al. 2010). As summarized Air Monitoring and Health Impacts Baseline in Figure 4-2, there is a wide range in the severity (AMHIB) study, sponsored by the World of health outcomes associated with PM, including Bank and based on recommendations from minor respiratory symptoms, asthma attacks, several Mongolian representatives. Data cover emergency room visits, heart attacks, hospital the period from June 2008 to May 2009. The admissions, and death. PM10 and PM2.5 data come from a network of eight monitoring stations (see figure 3-1 and In the following sections, the methodology table 4-1). Figure 4-3 shows the administrative used and the subsequent analytic results of effects districts in UB and the approximate location of due to daily changes in PM10, PM2.5, CP, and the monitoring stations relative to the district NO2 are presented using the one year of currently borders. Pollution data included 24-hour available data in UB. The estimated effects of averages of PM10, PM2.5, CP, and NO2. CP was exposure to these pollutants on specific diseases calculated by subtracting PM2.5 from PM10. Data resulting in either death or hospital admissions were sampled twice a week, on Wednesdays and have been summarized. Most studies of this kind Sundays. In addition, there were six weeks, mostly use several years of data, and therefore the results during the winter, when data were collected every 38 Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar Figure 4-2: Health effects of particulate matter Source: Guttikunda (2007) and authors’ illustration based on AMHIB monitoring. day. Unfortunately, wide variations in sampling Mortality and hospitalization data methods were used for PM2.5 and PM10 and comparisons among instruments show significant These six districts had a total population of discrepancies (see annex C). In addition, humidity approximately 930,000 in 2008. The other has a differential impact on the accuracy of the three districts were very small ( total population measurements of the different monitors. These under 60,000), were spatially distinct from the errors in measuring actual pollution levels may rest of the urban area of UB, and the population reduce the likelihood of observing a health effect. exposures are not well represented by the The study team attempted to minimize this monitors. Therefore, the analysis was restricted to problem by selecting only a subset of the existing the more populated and contiguous six districts. monitors. It is still likely that the health effects For mortality, both “natural� and cardiovascular will be underestimated, although it is difficult deaths were examined. Natural mortality is to determine from the available data how much defined as total mortality minus deaths due to higher the ¨true¨ effect would be. Data for NO2 injuries, accidents, violence, and suicides (and were provided as an average of the four NO2 labeled, in this analysis, as “all-cause mortality�) monitors in UB. since these latter causes are not likely to be 39 Air Quality Analysis of Ulaanbaatar Figure 4-3: Administrative Zones and AMHIB Stations, Ulaanbaatar, Mongolia Source: Authors’ illustration based on AMHIB monitoring. Table 4-1: PM monitor stations and associated hospital districts – – – – – – – – Source: AMHIB study. 40 Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar Table 4-2: Administrative district with associated PM monitor and population Note: Districts 7, 8 and 9 did not have a nearby PM monitor and were excluded from the analysis. Source of Population data: World Bank. 2009. Air Pollution in Ulaanbaatar: Initial Assessment of Current Situation and Effects of Abatement Measures. Washington, DC: World Bank. related to pollution exposure. For hospitalization, ambient concentrations. The level of accuracy both cardiovascular and respiratory admissions depends on resources and data availability. were examined. The coding of cause of death and hospitalization was based on the current To assign exposure for the particle international standard—the tenth version of the metrics, the administrative district of residence International Classification of Diseases, (ICD- (and presumed location of the deaths and 10), codes S through Z. In addition, for both hospitalizations) was matched with the closest PM mortality and morbidity, cases due specifically monitor (see figure 4-3). This was undertaken to either cardiovascular or respiratory disease in order to minimize the misclassification of were examined. Cardiovascular deaths and pollution exposure, given the distinct spatial hospitalizations were assigned ICD codes starting orientation of the city and surrounding ger areas. with “I,� while ICD codes for respiratory cases Table 4-2 summarizes the matches between event began with “J.� location and monitor. Of the eight monitors that were initially available for the analysis, only Assigning Exposure monitor #8 (Airport) and #5 (CLEM) were not used. The former was not used since it is located In such a study it is crucial to utilize the pollutant at the airport, a significant distance away from measurements that most accurately reflect the the center city, and the latter was not used since it actual exposures. Statistical studies have shown only measured PM10 and these measurements had that errors in the measurement of exposure are unexplainable variations—they were much lower likely to underestimate the true effect of pollution than all other monitors, and even lower than the and make it more difficult to find a “statistically PM2.5 measurements for the same district (from significant� effect of pollution. Conceptually, a different monitor). It is important to note that statistical significance is an estimated pollution the monitor corresponding to Administration effect that is not due to chance alone and one District 1 (AMHIB station 6, III khoroolol) only that would be observed repeatedly in replicated measures PM10. However, AMHIB station #1 also studies. In order to estimate the exposure of the matches up with Administrative District 1, so population to ambient concentrations of studied District 1 residents were assigned PM2.5 data from pollutants, the population distribution needs that monitor. Likewise, station #6 only measures to be superimposed on the distribution of these PM10 for District 3, so PM2.5 data were taken 41 Air Quality Analysis of Ulaanbaatar from nearby station #7. Similar substitutions were of the occurrence of death (or hospitalization) is made in other districts using nearby monitors. examined to determine whether it is more likely The study team also explored other options for to occur on days of higher air pollution. assigning exposure, but the subsequent results were not quantitatively different from those using In the analysis, both the same day of exposure this method. (i.e., without a lag in the response time or lag0) and previous days of exposure are tested to see Method if they have an impact on mortality. This is important since it is likely that an adverse health As in many previous studies (Malig and Ostro effect will occur several days after the actual 2008), a time-stratified case-crossover study exposure occurs. Thus, besides lag0, the study design was used. This design is similar to the team examined exposures up to three days before case-control study often used in epidemiology. In the event (e.g., lag1, lag2 and lag3). Results are the case-crossover method, pollution on the date presented in terms of the percent increase in of the death (case) is compared to pollution on the health effect over the existing conditions, several other referent days (controls) occurring per every 10 micrograms per cubic meter (µg/ within the same month and year. As a result, m3) increase in the air pollutant being measured. each individual in the study serves as his or her The latter is a standard measure of the amount own control and there is matching on a wide of particles found in the air. Besides this central range of individual-level characteristics. This estimate, a 95 percent confidence interval is also method significantly reduces the likelihood that provided. This provides a range within which the some unmeasured factors will affect the results. true value is likely to fall 95 percent of the time if As a result, most factors that can have an impact other studies were conducted with similar data. on mortality and that vary on a daily basis with pollution are implicitly taken into account by Results the study design. In addition, by limiting cases and controls to the same month, factors that The distributions (i.e., mean and median) of the vary over a longer time scale (e.g., smoking, pollution data used in the analysis of premature diet, exercise, age) will not change. Thus, these mortality are summarized in table 4-3. For factors will not have an impact on the statistical example, for the mortality analysis, based on all of analysis. The basic analysis involves the full year the districts, the average temperature was –2 Co, of data collected in UB. However, given the large and the mean pollution concentrations of NO2, differences in concentrations and sources of PM PM10, PM2.5 and CP (in µg/m3) were 29, 486, by season, a sensitivity analysis for mortality 301 and 206, respectively. Median PM10, PM2.5 was conducted by dividing the year into a cold and CP are much lower at 227, 95, and 112 µg/ and warm season. The latter was defined as May m3, respectively, due to the skewed distribution through September. and the very high concentrations that occur in the winter. As expected, PM concentrations vary Logistic regression analysis was used to widely among districts. For example, median estimate the impact of pollution exposure on PM2.5 ranges from 42 µg/m3 in District 6 to health (morbidity and mortality) and the accuracy 256 µg/m3 in District 3. (predictive value) of the model’s estimate. The goal of the logistic regression is to determine the Table 4-4 summarizes the pollution data best fitting model to describe the relationship, used in the analysis of hospital admissions and in this case, between death on a given day (the table 4-5 summarizes the mortality and hospital dependent variable) and pollution (the predictor). data. The mean age at death was 56.3, with 1,425 A logistic regression will generate a coefficient (an cases of all-cause mortality, 568 cardiovascular effect per unit of the pollutant) and significance (40 percent of the total), and 53 respiratory level (an indication of the likelihood of the (4 percent of the total). Males made up about observed effect). In this model, the probability 57 percent of all mortality. Due to the small 42 Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar Table 4-3: Descriptive statistics of data used in mortality analysis µ µ µ µ Source: AMHIB data. Table 4-4: Descriptive statistics of data used in hospital analysis µ µ µ µ Source: AMHIB data. 43 Air Quality Analysis of Ulaanbaatar Table 4-5: Descriptive statistics for mortality and hospitalization data used in analysis Source: AMHIB data. Table 4-6: Statistically significant or near significant mortality effect estimates for particulate matter and nit-ogen dioxide, June 2008–May 2009 µ µ Source: AMHIB data. amount of respiratory-specific mortality, it was daily mortality. In the cold season, coarse particles dropped from subsequent analysis. Regarding were associated with cardiovascular mortality. hospitalizations, there were 6,291 cardiovascular- In this case, a one-day lag in these particles specific and 21,991 respiratory-specific was associated with a 0.28 percent increase in admissions. Males made up 44.7 percent of the mortality on the following day. Full-year analysis cardiovascular and respiratory hospital admissions. of NO2 showed modest associations with both all-cause and cardiovascular mortality based on Table 4-6 summarizes the statistically the previous day’s exposure (i.e., a one-day lag). significant regression results (i.e., associations Exposures to NO2 were not stratified by season between pollution and health that are unlikely since the concentrations were reasonably similar to be due to chance alone) for the pollution data in both the warm and cold season. collected between June 2008 and May 2009 (for the full set of results, see annex D). Using the Table 4-7 summarizes selected results for full year of data on particulate matter, the only hospital admissions for both cardiovascular and associations observed were between cardiovascular respiratory disease. Unlike the case for mortality, mortality and coarse particles with a one-day lag. there were many strong associations between The results suggest that every 10 µg/m3 change in particulate matter and hospitalizations (the full coarse particles would increase daily cardiovascular set of results are presented in annex D). For mortality on the next day by 0.25 percent (95 many of the outcomes, the three-day lag between percent confidence interval = 0, 0.51). Additional exposure and hospital admissions generated the associations were observed when the data were best fit of the data and the largest impacts. For stratified by season. For example, in the warm example, for each 10 µg/m3 change in PM2.5, season there were associations between PM2.5 and a 3-day lag was associated with a 0.82 percent all-cause mortality, with each 10 µg/m3 increase increase in admissions for cardiovascular disease relating to a subsequent 1.4 percent increase in and a 0.24 percent increase for respiratory 44 Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar Table 4-7: Selected statistically significant results of logistic regressions for hospitalization and particulate matter, June 2008–May 2009 µ µ Note: PM10, PM2.5, CP = particulate matter less than 10 microns, less than 2.5 microns, and between 2.5 and microns, respectively; Lag = number of days prior to death that the exposure occurs; confidence interval = likely range within which the true value is likely to fall 95 percent of the time if many other studies were conducted with similar data. Source: AMHIB study. disease. For a 10 µg/m3 change PM10, a 3-day data also could lead to inaccurate or unstable lag was associated with a 0.21 percent increase results. in cardiovascular admissions, while a 2-day lag was associated with a 0.04 percent increase For mortality, the strongest associations in respiratory admissions. Coarse particles using the full year of particle data were observed did not appear to have a significant effect on for coarse particles. Specifically, coarse- hospitalizations in this sample and only a small particle concentrations were associated with effect on respiratory admissions was observed. cardiovascular mortality on the following day. Effects were also observed between NO2 and Exposure to coarse particles was associated both cardiovascular and respiratory admissions, with an increase of approximately 0.3 percent with a 3-day lag (i.e., exposure three days prior to per every 10 µg/m3 change. However, given the effect) being the most important. the very high exposures to these pollutants in UB, exposures equal to the interquartile range Summary and Discussion (a metric commonly used in air pollution epidemiology studies signifying the amount UB’s population is exposed to high between the 75th and 25th percentile of the concentrations of PM from different sources, distribution) would generate a daily increase each with different chemical constituents and in mortality of about seven percent above the size fractions. The findings from the limited one normal levels. In addition, a strong effect was year of available data support the conclusion that observed between PM2.5 and mortality during there are significant public health implications the warm season. The magnitude of this effect is related to air pollution in UB. However, because generally similar to, but at the higher end of, the most similar studies involve several years of range of effects observed in previous studies of data, it would be important to repeat this study PM2.5 in other parts of the world. Based on the once more data are available. Nevertheless, interquartile range observed during the warm several important associations were observed season in UB, this would relate to about a four between daily changes in PM and NO2 and both percent increase in daily mortality over baseline. mortality and morbidity. This is particularly Finally, more modest associations (i.e., somewhat noteworthy given the scarcity of data, i.e. only less statistically significant) were observed in the one year of data generally collected on an every analysis of the full-year of data on NO2 with both third or sixth day basis. The sparseness of the all-cause and cardiovascular mortality. 45 Air Quality Analysis of Ulaanbaatar For hospital admissions, the strongest additional year of data, the signal from PM2.5 will associations were observed for PM2.5 and PM10 become stronger and the public health burden with cardiovascular disease. Using the IQR, will become more obvious. exposures to PM2.5 and PM10 could increase daily hospital admissions for cardiovascular In summary, the population of UB is exposed disease by approximately nine percent over the to high concentrations of PM from different normal level of hospitalizations. Associations sources and with different chemistry and size were also observed for NO2, particularly with fractions. There is existing scientific literature hospitalizations for cardiovascular disease. to suggest that many of these sources—coal combustion, local boilers, motor vehicles, There are several possible explanations for the biomass, and even crustal material—can lack of an effect of PM2.5 in the full-year analysis, contribute to adverse health effects, especially but an observed effect during the warm season. at the concentrations experienced in UB. The The first factor might be the differential seasonal findings of this analysis of the limited available pattern of sources and exposures during the year. data support the conclusion that there is a It is possible that during the winter, exposure to significant public health burden related to PM2.5 is easier to avert for the majority of the exposure to air pollution in UB. population living in houses and apartments in the central city. About half of the population of UB Implications lives in apartments located in the central areas of the city. About 80 percent of these apartments are The following is a simple example to illustrate supplied with central heating and hot water from the quantitative health implications of the above three power plants located to the west of the city. results for short-term changes in PM2.5. For this The remainder of the apartments is supplied by exercise, a hypothetical change from the current local boilers. It is likely that during the warmer annual concentration of 260 µg/m3 to the current months, when an effect of PM2.5 on mortality was annual standard of 25 µg/m3 is assumed. Recall observed, residents will spend more time outdoors that the analysis indicates that during the warm (or indoors with windows open) and incur season, every 10 µg/m3 change in PM2.5 would greater exposures. Several previous studies have result in a 1.38 percent (95 percent CI = 0.42, reported effects of PM2.5 in particular seasons. 2.44) increase in daily mortality and that over For example, in a study of 100 U.S. cities from the full year, a similar change in PM2.5 results in a 1987 to 2000 (see references in World Bank 0.82 percent (95 percent CI = 0.64, 1.00) change 2009), found strong effects of PM2.5 on mortality in hospital admissions for cardiovascular disease during the warmer months, particularly in the and a 0.24 percent (95 percent CI = 0.14, 0.34) northern sections of the U.S. which experience change in hospital admissions for respiratory significant seasonality in climate. Second, the disease. It is assumed that the estimated toxicity of the components of PM2.5 may be relationship between the percent change in health different in the summer than the winter. Third, and PM2.5 is linear (i.e. there is a constant slope the full-year mortality results may also be due to or effect per µg/m3 over the entire range). At very low statistical power (i.e., the ability to detect an high concentrations, this relationship would have effect) in the study. While strong and consistent to become less than linear; that is, the slope would associations were observed between PM2.5 and be lower. It is also assumed for simplicity that the hospital admissions for cardiovascular disease annual change consists of 365 daily changes of a with many more cases, the low number of cases similar amount. This assumption does not alter of daily mortality makes it much more difficult the overall findings. There are approximately 11 to statistically detect an effect. Therefore, given deaths per day during the warm season and 55.5 the strong and consistent effects observed for and 190 cardiovascular and respiratory admissions hospitalization and the likely similar biological per day throughout the year, respectively. These mechanisms, an effect on mortality would averages are based on the data and sample used in be expected. Further, it is likely that with an the analysis; that is, using the period from June 46 Estimating the Effects of Air Pollution on Mortality and Hospitalization in Ulaanbaatar 2008 to May 2009, for people living in one of the study because the existing monitors were six administrative districts in the study, and using not necessarily representative of exposures in total mortality minus homicides, accidents, and those areas and the population sample size injuries. was small. Chapter 3 indicates that the PM2.5 concentrations in these areas are much higher The annual impacts, therefore, for mortality than those reported the central city with an (DM) can be calculated by: annual average PM2.5 in the range of 200 to 350 µg/m3. The source mix for the particles in DM change in PM2.5 % change mortality the ger areas will be somewhat different from per µg/m3 baseline daily mortality that in the central city since the sources are rate 146 days of the warm season dominated by extensive coal burning for heat and power generation (particularly during the (260 25) 0.138% 11 146 521 winter) and the blowing of dust. If it is assumed, deaths or 11% of all deaths however, that the toxicity of PM2.5 in these areas is generally similar to that in the urban area, we Thus, given these model assumptions, can provide a general quantitative assessment of meeting the standard would reduce premature the likely health impact. It is also necessary to mortality from short-term exposures by 521 assume that the concentration-response function deaths per year, with a 95 percent confidence remains linear (i.e., the slope or effect per µg/m3 interval of 160 to 920. This amounts to a 13 is constant) at the higher levels of PM2.5. If so, percent reduction in mortality with a 95 percent this suggests that the ger areas will have a 1.38 confidence interval of four percent to 22 percent. percent increase in daily mortality per every 10 Using the same calculations for hospital µg/m3 in PM2.5. Using the average of the likely admissions, the attainment of the PM2.5 standard annual range of 275 µg/m3 for PM2.5 in the ger would reduce annual cardiovascular hospital areas, a daily change from this concentration to admissions by 3,900 (95 percent CI 3,050 the current annual standard of 25 µg/m3 would 4,760) and respiratory hospital admissions by suggest a 34 percent increase in mortality above 3,910 (95 percent CI 2,280 5,540). the normal expected number of daily deaths in these areas. This predicted one-third increase As indicated above, many of the ger areas in daily mortality is clearly a significant and were not included in this epidemiological preventable public health burden. 47 5. Willingness to Pay to Reduce Air Pollution in Ulaanbaatar I n addition to understanding the severity air Air quality improvements reduce the risk of pollution in Ulaanbaatar and the extent to both illness and death. Past studies have found which it affects people’s health it is often that the WTP to reduce risk of death typically useful for policy makers to realize how accounts for 85 to 90 percent of total WTP strongly people feel about spending resources to for reductions in health risks associated with protect their health relative to other needs. This air pollution (U.S. EPA 2006). As a result, this chapter reports on a survey of Ulaanbaatar residents survey focused only on the WTP of Ulaanbaatar designed to estimate their willingness to pay to residents to reduce risk of death associated with reduce mortality risks. The study looks at mortality air pollution. Air pollution can increase the risk risk reductions of the magnitude and types typically of death immediately through its impact on those resulting from air pollution reduction policy. The in the population who have more vulnerable study asks residents about their willingness to pay respiratory and circulatory systems. It can also for a 5- and a 10-in-1,000 annual risk reduction increase the risk of death in the future to the over a 10-year period (e.g., 5 in 10,000 annually). extent air pollution exposure causes or exacerbates The study estimates that Ulaanbaatar residents diseases of the respiratory or circulatory system, are willing to pay approximately three percent of or cancer. The survey asked about residents’ WTP household income to achieve the 5-in-1,000 risk to reduce risks immediately and beginning at age reduction. This implies a value of statistical life 70. The size of the risk reductions were also of the estimate of 319 million tugrug, or $221,000 based magnitude typically seen in response to successful on the official exchange rate. air pollution policy. The survey was administered to a random sample of Ulaanbaatar residents in Introduction January and February 2010. People have many competing needs and desires. Literature Review Knowing how much people in Ulaanbaatar are willing to pay (WTP) to reduce the kinds of The study team is aware of no prior Mongolian health risks associated with air pollution can be stated preference surveys measuring willingness helpful in planning how to address air pollution to pay for reduction in mortality risk. It is also in the city. At the request of the AMHIB study aware of only five such studies conducted in Steering Committee, the World Bank study countries bordering Mongolia—all in China team conducted a survey estimating the WTP (World Bank 2009; Hammitt and Zhou 2006; Li of Ulaanbaatar residents for reductions in health et al. 2002; Wang et al. 2001; and Zhang 2002). risks of the size and type that have resulted from All of these studies converted WTP into a value air pollution policy improvements elsewhere.8 of statistical life (VSL) estimate. The value of a This chapter reports the results of this survey. statistical life (VSL) is the average willingness to pay for a small reduction in individual mortality 8 For further information about the WTP survey study, refer to risk expressed in terms of risk to a population. Annex E to this report. If a population of 10,000 people experiences a 49 Air Quality Analysis of Ulaanbaatar 1/10,000 excess risk of mortality over a ten-year et al., 2004) and more recently adapted for use in period, then statistically one extra person in the China (World Bank 2009). population would be expected to die during the ten-year period. This is what is meant by In adapting the survey for use in Mongolia, the loss of a statistical life. If the average person the study team had two fundamental goals. in a population of 10,000 people is willing to The primary goal was producing a survey that pay $600 for a 1/10,000 reduction in mortality Mongolian respondents would understand well risk, the VSL for the population is $6 million. and accept and that would have risk reduction Mathematically, VSL is the average WTP for levels appropriate to air quality policy and would a reduction in mortality risk divided by the permit estimation of WTP appropriate for reduction in the mortality risk. Estimates of VSL Ulaanbaatar. The secondary goal was to maintain are used by decision makers around the world to as much consistency as possible with versions of help set policy priorities affecting health. the survey that had been administered successfully in other countries. This consistency will provide In the above five Chinese studies, mean a stronger basis for judging the credibility of the WTP ranged from $32,000 to $64,000, implying Mongolian estimates. VSL estimates that are quite low in comparison to those found in similar studies conducted After translating the survey into Mongolian, elsewhere in the world. This likely reflects the focus groups were used to identify changes fact that each of these studies had methodological needed to accommodate Mongolian cultural, problems (World Bank 2009). sociological, and institutional conditions. These accommodations included accounting for the In two phases from 2004 to 2009, a World structure of health care delivery and adjusting Bank study team—made up of researchers from the list of primary causes of mortality and the Resources for the Future (RFF) of Washington, list of activities people engage in to protect their DC, Fudan University in Shanghai, and health. Ages, mortality rates, and WTP bids People’s University in Beijing—and the World were changed to reflect Mongolian vital statistics Bank conducted a stated-preference study (i.e. and economic data. As in the U.S., Canadian, a survey-based study where people are asked Chinese, and Japanese versions of the survey, about their willingness to pay) in Shanghai, respondents were offered a product that would Chongqing, Nanning, and Jiujiang (World Bank reduce their mortality risk if used over a 10-year 2009). The study used a survey instrument that period. The product was described as having had been successfully administered in six other been approved by the relevant country’s health industrialized countries prior to its administration authority. Based on focus group results, two in China (see annex E, table 12). This Chinese risk reductions, 5-in-1,000 and 10-in-1,000, study estimated a mean VSL of $440,000 ($US were chosen. Minor adjustments were needed 2008). Compared to other countries in which in the way information on probability and risk the survey had been administered, this estimate was communicated in the survey. For example, of Chinese WTP was lower in absolute terms the original survey used roulette to explain the than those in industrialized countries, but similar concept of chance. Mongolian respondents were as a percentage of household income. Like unfamiliar with roulette, but there is a widely respondents in Japan, Chinese respondents were viewed TV game that uses a similar wheel, so willing to pay relatively more for future death risk reference was made to that game rather than to reductions relative to reductions in immediate roulette. Finally a risk-attitude question referring death risks than did respondents in the western to air and train travel was changed to one related countries surveyed. to crossing busy streets. Survey design and adaptation The Questionnaire The survey used in this study is based on one first The questionnaire begins with demographic developed for use in the United States (Alberini questions and asks respondents about their 50 Willingness to Pay to Reduce Air Pollution in Ulaanbaatar health status and chronic disease history as well The basic form of the WTP question asks as that of their family members. The survey next how much the respondent would be willing to introduces the concept of probability and the pay for a product that, when used and paid for probability of dying or surviving. The survey over the next 10 years, will reduce baseline risk then shows respondents how probability of by x in 1,000 over the 10-year period. The latent dying will be represented in the survey and tests risk reduction questions ask whether they would their understanding of this representation and be willing to pay for a product that, when used the concept of probability in general (see figure and paid for over the next 10 years, will reduce 5-1). They are then presented information on their baseline risk by x in 1,000, for a ten-year age- and gender-specific leading causes of death period beginning at age 70 (see figure 5-2). The in Ulaanbaatar and on common risk-reducing latent WTP questions are preceded by a question behaviors. The survey then presents information asking the respondents their perceived chance of on the effectiveness and cost of common risk- surviving to age 70. This question encourages the reducing behaviors. Respondents are told the risk respondent to think about their future and can be of dying of someone of their age and gender living used to estimate discount rates. in Ulaanbaatar and are asked to accept this risk as their own for the purpose of the survey. They As shown in table 5-1, the sample design are then asked about their WTP for mortality has two treatments: in one, respondents receive risk reductions of a given magnitude, occurring a 5-in-1,000 risk reduction over 10 years as the at a specified time. The survey ends with initial question followed by a 10-in-1,000 risk further demographic questions and debriefing reduction as the second question. In the other, questions that can be used to evaluate whether respondents receive a 10-in-1,000 risk reduction the respondent understood the survey and took it and then a 5-in-1,000 risk reduction question. seriously. Respondents are asked two WTP questions. Figure 5-1: Depiction of risk change (American version used to develop Mongolian survey) Source: Authors’ illustration. 51 Air Quality Analysis of Ulaanbaatar Figure 5-2: Payment card elicitation of willingness to pay for a latent risk reduction (Mongolian version) Source: Authors’ illustration. Table 5-1: Study Design (cumulative probabilities over a ten-year period) – Source: AMHIB data. The initial question is a contemporaneous respondents with a matrix of ordered numbers risk reduction for all respondents. The second (possible bid amounts) and asks them to pick question is a latent risk reduction for respondents the one representing their maximum willingness age 40 to 65 and a contemporaneous risk to pay for the risk reduction. Range intervals reduction for respondents over age 65. were designed to be roughly constant in percentage terms. Psychometric studies and The survey uses a payment screen to elicit experimental economics suggest this as a means WTP (see figure 5-2). This screen presents of reducing response error (Rowe, Schulze, and 52 Willingness to Pay to Reduce Air Pollution in Ulaanbaatar Breffle 1996; Ready, Navrud and Dubourg the same training, including very explicit 2001). training regarding the level of interaction with respondents that was permissible to maintain The payment card elicitation approach (see consistency in administration across respondents. figure 5-2) presents respondents with a matrix A sample of 629 respondents completed the of ordered numbers and asks them to pick one survey. Participants were given a small gift for corresponding to their maximum willingness to participating. pay for the risk reduction. The chosen “stated� number can be thought of as (a) an appropriate Data preparation and sample characteristics estimate of WTP, (b) the top end of an interval between the chosen number and the next In any stated preference survey, some responses lowest number (which was estimated using a cannot be taken as accurately representing a Weibull distribution), or (c) the bottom end of respondent’s WTP. A respondent may fail tests an interval between this number and the next of whether they understand the concept of highest number (which was not estimated). In probability or the way probability is presented order of conservativeness, the Weibull estimate, in the survey. They may say they do not as called here, is more conservative (produces a understand the survey. They may also bid an lower WTP estimate) than the use of the stated unrealistically high percentage of their income. value as an estimate. Alternatively, WTP is set at Respondents who exhibit these criteria are the value in the matrix immediately below the typically eliminated from the sample. Table 5-2 chosen number, which in this context is similar reports observations targeted by these various to a Lower Turnbull estimate. This would be the cleaning criteria. The most common cleaning most conservative of the three estimates used in criteria met by respondents in this study (21 this analysis. percent of respondents) was a respondent saying that he or she did not understand probability well Administration of the survey (FLAG6). However, over 99 percent of the sample passed tests indicating that they had a basic The survey was administered in Ulaanbaatar, understanding of the graphical way probability the capital of Mongolia, using stratified random was presented in the survey. sampling by neighborhood. Respondents came to Neighborhood Family Health and Administrative Table 5-3 reports descriptive statistics of the full Centers to take the survey. The survey was sample. Mongolia has a young population and a low administered on weekends to accommodate life expectancy compared to other countries studied respondents’ work schedules. The survey was using this survey instrument. The sample similarly administered on laptop computers operated by is skewed toward a younger population despite an trained enumerators. Enumerators all received effort to oversample in the older age categories. Table 5-2: Descriptive statistics for cleaning criteria variables Note: *FLAG1 indicates a failure to pass tests of understanding of probability. **OVER80 identifies respondents who were over 80 years of age. ***FLAG6 indicates respondent said they did not understand the concept of probability. Source: AMHIB data. 53 Air Quality Analysis of Ulaanbaatar Table 5-3: Comparison of demographic variables by cleaning and place of birth – – – Note: *Cleaning approach C drops Flag1, Flag6, Over80, WTP/income <0.9 Source: AMHIB data. Eighty-three percent of the sample expects to live to who do not understand aspects of the survey are age 70. Twenty-three percent of respondents report addressed. For the preliminary analysis presented having problems with bronchitis and 44 percent in this chapter, the data using three alternative have heart disease. The study sample is also more cleaning specifications is analyzed. Further details female than male (38 percent male). 45 percent on estimation methods used in this study are of the sample have some college education. Mean provided in annex E. monthly household income is 390,063 tugrug, or 103,113 tugrug per capita. Results Estimation methodology Table 5-4 reports the mean WTP results based on the three cleaning approaches. Cleaning approach To estimate WTP and explain the factors that A drops respondents who failed tests of probability explain its variation, a model appropriate for understanding and who are over 80. Cleaning interval data is used. The underlying econometric approach B drops these respondents and those model is who say they do not understand probability well. Cleaning approach C drops these respondents log WTPi* = Xiβ + εI (1) as well as those who say they would be willing to pay more 90 percent of their income for the where WTP* is the underlying willingness to risk reduction. Table 5-4 contains information pay for a selected risk reduction; X denotes a on the WTP estimates in tugrug, both for vector of age, health, and other attributes; β is a contemporaneous and latent risk reduction, as vector of coefficients; and ε is an extreme value well as Wald test statistics testing for differences Type I error term (Alberini et al. 2004). WTP in mean values among the various estimates. The estimates can be sensitive to the manner in which preliminary analysis presented in this paper treats inconsistencies in responses and respondents the chosen bid as the true WTP of respondents. 54 Willingness to Pay to Reduce Air Pollution in Ulaanbaatar Table 5-4: External scope tests under alternative data cleaning approaches (respondents age 40–65) Note: Mean values are in tugrug. WTPx denotes an x in 1,000 reduction in risk of dying for the next ten years beginning immediately. WTPx_70 denotes an x in 1,000 reduction in risk of dying for ten years beginning at age 70. Source: AMHIB data. Certain regularities in response in WTP studies for a contemporaneous and latent reduction in are expected based on basic economic theory. If risk where the size of the risk reduction is held these regularities are not observed, it raises questions constant. In table 5-4 all comparisons show as about construct validity; that is, whether the survey statistically significant and positive discounting. instrument is effectively measuring respondents’ WTP. For example, WTP for risk reduction should Regression analysis provides further be increasing in the size of the risk reduction. information on whether the survey is performing Normally, one would also expect WTP for a risk well. Standard economic theory indicates reduction today to be higher than WTP for a latent that WTP would increase with income and risk reduction. This reflects a positive discount rate. risk reduction. If health is viewed as a scarce good, then a history of poor health might also This survey design provides two tests of contribute to an increase in WTP for reduction whether WTP responses increase with the size of of further health risks. In the case of latent risk the risk reduction. These are called scope tests. reduction that starts at age 70, one would also Does WTP increase with the scope of the benefit expect that WTP is increasing in the respondents’ provided? This means, are respondents who subjective probability of living until age 70. In received a 10-in-1,000 risk reduction over 10 years addition, several studies in the U.S. and China willing to pay more than those receiving a 5-in- have found an inverted U-shaped relationship 1,000 risk reduction. The study finds that for each between age and WTP (see Alberini et al. 2004; possible comparison, Ulaanbaatar respondents are Alberini et al. 2006).9 Table 5-5 presents results of willing to pay more for larger risk reductions. This regression analysis on WTP for contemporaneous effect is generally statistically significant for both and latent risk reductions respectively. contemporaneous and latent risk reduction. 9 The exception is Itaoka et al. (2005), who conducted a The survey design also provides two tests of contingent valuation survey in Japan, and found a positive discounting by looking at the difference in WTP linear effect from age, suggesting that WTP increases with age. 55 Air Quality Analysis of Ulaanbaatar Table 5-5: Construct validity of WTP for the current and latent risk reductions (using a Weibull distribution) Source: AMHIB data. Results for contemporaneous risk reduction positive, but very small and only significant for show all the expected signs on explanatory cleaning approach A. variables. The relationships most central to the theory are those between income and WTP and This study did not find strong evidence that risk reduction level and WTP. The coefficient people with bronchitis, high blood pressure, for income is of the correct sign and both and cancer are willing to pay more to reduce highly significant and large, and the one for risk mortality risk than those without. Theoretically, reduction is significant once people who bid more this relationship can be either negative or positive. than 90 percent of their income are dropped Krupnick et al. (2000) found that these measures from the sample. This is consistent with scope were either insignificant or had a positive effect test findings. The relationship between age and on WTP in Canada and the U.S. WTP for contemporaneous risk reduction shows the same inverted U-shape as found in some prior In our regressions gender did not turn out to studies, but its significance is weak. be statistically significant. Thus, the WTP is not affected by whether a person is male or female. Regressions on WTP for latent risk reduction also show expected results. The coefficient on Discussion the 5-in-1,000 risk reduction over 10 years is significant and negative, indicating that The overall conclusion from these tests and respondents said they would pay less for a smaller analyses is that respondents in Ulaanbaatar risk reduction. The coefficient on respondents generally understood the survey and appeared who thought they would be alive at 70 was also to be giving valid responses about their WTP to 56 Willingness to Pay to Reduce Air Pollution in Ulaanbaatar reduce mortality risk. That is, they appeared to exchange rate, or $493,000 based on a purchasing take the survey seriously and gave responses that power parity (PPP) exchange rate. behave in the way one would expect based on economic theory and past studies. PPP exchange rates are most often used in making cross-country comparisons. They are based This study produces WTP estimates for on the price of a similar basket of consumer goods contemporaneous and latent mortality risk in different countries rather than currency exchange reductions. It also uses a variety of sample rates. Official exchange rates may be affected cleaning approaches and three different by capital flows not closely related to consumer assumptions about how conservatively to interpret preferences and may be influenced by government WTP. For policy analysis, the authors recommend policy for macroeconomic reasons. Nevertheless, use of the Lower Turnbull estimate of WTP to PPP rates can also be problematic, particularly reduce contemporaneous mortality risk using for countries that are somewhat isolated and have sample cleaning approach C. They do so for three unique cultures. Here, the ratio of internationally reasons. First, only results for cleaning approach traded to non-traded goods would be low and it C are presented because only estimates using this might be difficult to find a market basket of goods approach passed the most stringent of the validity comparable to that used to compute PPP rates in tests. Second, proposed air pollution policy in other countries. Adjusting for quality differences of Ulaanbaatar is likely to focus on reduction in goods that appear to be in comparable categories particulate matter. can also be problematic. Therefore, dollar- denominated VSLs using the official exchange Finally, critics often claim that stated rate are also reported for Mongolia. For use in preference studies overstate WTP because Mongolian policy analysis, the authors of this study respondents do not actually have to expend their recommend using either the value of statistical own money and do not receive actual benefits. life estimates in tugrug or, if necessary, converted To address this concern, it is generally advisable to U.S. dollars using the official exchange rate— to consider whether policies will pass cost-benefit simply because conservatism in this measure is tests when the most conservative estimates of desirable, given that there are so many uncertainties WTP are used. The Lower Turnbull estimate in the VSL estimates. assumes that the WTP amount provided by the respondent is the lower bound of an interval This value should be interpreted as only between this amount and the next higher valid for Ulaanbaatar, not outlying rural areas amount on the payment card. As a result, the of Mongolia. Even based on conservative authors believe that a WTP of 159,000 tugrug assumptions, these results show that people living for a 5-in-10,000 annual contemporaneous risk in Ulaanbaatar place a high value relative to their reduction is a reasonable reflection of preferences income on reducing their risks of death, a value in this sample. This translates into a VSL of 319 comparable to that seen in other developed and million tugrug, or $221,000 based on the official emerging economies. 57 6. Current Health Costs of Air Pollution in Ulaanbaatar and Benefits from Management Scenarios T his chapter estimates the current Population-weighted exposure (PWE) at health damages attributable to air present and for 30 percent/50 percent/ pollution in Ulaanbaatar together with 80 percent scenarios the health benefits (i.e. the avoided health damage) that potentially can be obtained In order to estimate health costs associated with from alternative management scenarios. The the present Ulaanbaatar air pollution as well as scenarios show how interventions targeting their reduction as a result of pollution abatement main sources of PM pollution in the city may scenarios, this report uses the population-weighted reduce the population-weighted exposure, given average PM concentration, which can be referred alternative ambition levels of the interventions. to in its shortened form as the population- The main sources looked at are ger household weighted exposure (PWE). Compared to using stoves, suspended soil dust, and heat-only boilers the average annual concentration for the city, (HOBs). The scenarios reflect a 30 percent, the PWE more accurately reflects exposure of 50 percent, and 80 percent reduction in the UB’s population to air pollution by taking into emissions from each of these sources separately account the spatial distributions of the pollution and corresponding reductions for all sources and the population. together. Using exposure-response functions from epidemiological studies, the health effects This section calculates the present health associated with the different levels of pollution damage of air pollution, as well as the reduction (today and for future scenarios) are estimated of this damage as a result of some selected in terms of premature deaths, cases of chronic intervention scenarios. The costs related to bronchitis, and hospital admissions. Using the the health damage, and the avoided costs (i.e. estimated WTP for mortality risk reduction benefits) as a result of the interventions, are also from chapter 5, and estimated unit costs for calculated. the other health end-points (chronic bronchitis and hospital admissions), the estimated health The scenarios are selected to show the range effects are monetized. The current health damage of effects of interventions on the two main source attributable to air pollution is estimated at 19.4 groups contributing to the PM concentrations percent of Ulaanbaatar’s GDP, and an 80 percent and PWE—ger household heating systems and reduction in all main sources would yield a health suspended soil dust, as well as on the heat-only benefit of 7.5 percent of the city’s GDP. Achieving boilers (HOBs), another fairly easily controllable the air quality guidelines for Ulaanbaatar (50 µg/ source. The scenarios reflect 30 percent, 50 m3 for PM10 and 25 µg/m3 for PM2.5) is estimated percent and 80 percent reduction in the emissions to yield a health benefit of 13 percent of the city’s from each of these sources. GDP. 59 Air Quality Analysis of Ulaanbaatar In this type of assessment of avoided data in the table are presented in figure 6-1. health effects linked to an intervention, it is PWEs are calculated based upon the PM considered that the avoided effects and costs pollution maps presented in chapter 3. are annual for the year when the intervention has been fully implemented. These results are The reduction in the PWE shows that ger useful for comparing the effects situation before area interventions provide the largest reduction, intervention (present situation) with the situation while interventions in dust suspension also that will exist after full implementation of the provide significant reductions. The HOB interventions. In practice, an intervention is intervention provides only a minor PWE implemented over several years, with gradually reduction. reduced emissions, concentrations, and effects. When carrying out cost-benefit calculations to When interpreting figure 6-1, it is important compare different types of interventions, the to note that it represents the average PM time line of the interventions should be taken concentrations (PWE). Portions of the population into account, as well as integrated intervention will still remain exposed to levels above the AQ costs and benefits (avoided effects costs) over the standards even when the PWE does not exceed timeline, as is done in the cost-benefit analysis standard values. presented in chapter 9. Figure 6-1 shows the remaining PWE The PWE data in table 6-1 are used to for PM10 and PM2.5 under different emission estimate health benefits of the reductions. The reduction scenarios. The reductions drive the Table 6-1: Population-weighted average PM concentrations (PWE) in Ulaanbaatar, and reductions from abatement scenarios, μg/m3 Source: AMHIB data. 60 Current Health Costs of Air Pollution and Benefits from Management Scenarios health cost calculations given in the next section. Figure 6-1 shows that a larger reduction in The figures indicate the present relative health emissions is needed to comply with the various impacts, and also the reductions in health impacts AQ standards and guidelines. It should also of different interventions, based on the spatial be noted that even if the averaged population distribution of the population and of the PM exposure meets an AQ standard, a substantial sources and concentrations. For example, a 30 part of the population is still exposed to levels percent reduction in ger area emissions would above the standard, because PWE is an average result in a reduction in the population-weighted exposure. exposure of PM10 by about 13 percent and of PM2.5 by 18 percent. Similarly, an 80 percent Figure 6-1 shows that the WHO interim emission reduction in all the three sources in the targets IT-1, at 70 µg/m3 for PM10, can be met by figure would result in a PWE reduction of 69 reducing the three source sectors by more than percent for PM10 and 65 percent for PM2.5. The 95 percent. The IT-1 target for PM2.5 , at 35 µg/ total PWE is less than the emission reductions m3, can only be met by also reducing the “other� because the “other� sources (traffic, CHPs) are sources, which include road traffic and the power not reduced in this example. This PWE reduction plants (CHPs). does not, however, give a similar reduction in health costs due to the nonlinearity of health The Mongolian Air Quality Standards and effects as a function of PM concentration. the WHO guideline are more difficult to meet. For example, a 30 percent reduction in ger The present situation regarding PM pollution in area emissions yields a 5 percent reduction in Ulaanbaatar is so extreme that emissions have to calculated health costs, as shown in the next be reduced by about 95 percent, for all sources, to section. comply with these standards. Figure 6-1: How much is air pollution reduced if emissions are reduced by 30%/50%/80%? PWE reductions for given emission reductions of PM10 continued 61 Air Quality Analysis of Ulaanbaatar Figure 6-1 continued Source: Authors’ illustration based on AMHIB data. Avoided premature deaths, cases of chronic Bank (2007). The analysis includes the three bronchitis and hospital admissions major mortality and morbidity effect end-points associated with air pollution exposure: premature A large number of studies around the world deaths (i.e. deaths brought forward due to document a consistent association between air pollution exposure), new cases of chronic elevated ambient PM10 and PM2.5 levels and a bronchitis, and hospital admissions for respiratory variety of health end-points, such as respiratory and cardiovascular diseases.10 These are termed infections, the number and severity of asthma “health end-points.� We estimate both the current attacks, the number of hospital admissions, excess number of cases—the number of cases that school and work absenteeism, and mortality are attributable to current air pollution levels— rates (OECD 2000). The health impacts from and the number of cases that can be avoided by air pollution are determined by two main implementing the interventions described earlier. factors: (1) the concentrations of pollutants in the atmosphere, and (2) the number of people Premature deaths and enhanced rates of being exposed. A range of factors can modify chronic obstructive lung diseases (of which the extent to which a given air pollutant affects chronic bronchitis typically is the most prevalent) a population, as for instance the general health are two major health impacts associated with status and co-exposure to other pollutants. In long-term exposure to PM pollution. Exposure- Ulaanbaatar, especially in the winter season, very response relationships for these end-points high ambient PM concentrations occur in Ger are typically revealed by means of long-term areas, as evidenced by measurements, where the cohort studies or cross-sectional studies population densities are also high. (types of epidemiological studies). Enhanced hospitalization rates are found to be linked to air To calculate impacts of air pollution in Ulaanbaatar, this analysis applies the exposure- 10 The exposure-response functions for chronic bronchitis, hospital admissions for respiratory and cardiovascular diseases response functions used in the assessment of (CVD) applied in WB (2007) is based on a meta-analysis of costs of air pollution in China by the World several Chinese studies (Aunan and Pan 2004). 62 Current Health Costs of Air Pollution and Benefits from Management Scenarios pollution exposure on a shorter time-scale and are Based on exposure-response functions revealed in time-series studies. from Ulaanbaatar (for hospital admissions) and from the literature (for mortality and chronic Health impacts are estimated in physical and bronchitis), health impacts are derived using the monetary terms in the following way: following equation. First, a threshold value is set for PWE below E ((RR 1)/RR)* fp *POP which no health risks are assumed. The threshold value chosen could be the guideline value set where E is the number of cases of each health by WHO or other international guideline end-point attributed to air pollution (“excess values or local standards. WHO guideline cases�), RR is the relative risk of health effect values are typically determined by the level of between two levels of pollution (here the pollution concentrations that are identified in current level and a lower level obtained from an epidemiological studies as “threshold� levels for intervention or the lower threshold level), fp is observable effects. Thus, they are a metric for the current incidence rate of the health effect, the actual physical impacts rather than what may and POP is the exposed population considered. be defined as the target or acceptable level in a In the following, POP is assumed as the total specific setting. PM has no threshold level below population in Ulaanbaatar, which was 1.106 which there are no effects. This study uses the million in 2009. For hospital admissions fp*POP annual average PM10 concentration of 15 µg/m3 is replaced with the actual annual number of as the lower threshold level for the effects, as was hospital admissions. Except for the mortality used by WHO in the Global Burden of Disease function, where the study team relies on WB assessment (Cohen et al. 2004). (2007) (which assumes a 15 µg/m3 PM10 as a threshold level), RR is given by: Second, the PWE values are combined with estimated baseline rates of the given health RR exp(β*(C Ct)) end-point in the population and exposure- response functions. This gives the estimated where β is the exposure-response coefficient (see health impacts of exposure to excessive levels of table 6-2 where betas are given as percentage ambient concentrations of PM. The AMHIB values), C is the current pollution level, and study includes a time-series study using data Ct is the target pollution level obtained from from a broad range of hospitals in Ulaanbaatar an intervention or the assumed threshold to evaluate the exposure-response relationship value. We calculate the remaining number of between incidences of hospital admittances for cases attributable to air pollution after each various respiratory illnesses and cardiovascular intervention, and derive the number of cases diseases (CVD) and ambient concentrations of that can be avoided by subtracting these figures PM in the city (see chapter 5). The study team from the calculated excess cases in the current applied the results from that study in terms of situation (which is calculated by using the exposure-response functions for hospitalization threshold levels described above). To determine as well as for current hospitalization rates costs of air pollution, and avoided costs as for respiratory and cardiovascular disease in a result of an intervention (i.e., the benefits Ulaanbaatar. from reduced air pollution), this paper uses a willingness-to-pay methodology to monetize Third, a unit cost value was estimated for mortality and chronic bronchitis impacts and each health end-point. For premature deaths a cost-of-Illness methodology to estimate the and chronic bronchitis, the study team relied economic value of avoided hospitalization (see on results from the survey on the WTP of World Bank 2007). Ulaanbaatar residents (see chapter 6) to reduce risk of death associated with air pollution. For Due to the considerable attention these hospitalization cost estimates, a previous study in calculations may have when disseminated, it China was transferred (World Bank 2007). is necessary to provide some background in 63 Air Quality Analysis of Ulaanbaatar academic literature to disclose key assumptions so exposure-response function11. A direct application that others could use this work and improve it. of the exposure-response function in Pope et al. (2002) may lead to implausibly high damage An important reason for limiting the number estimates in polluted regions in Asia, and the U.S. of health end-points is the lack of background results were therefore calibrated toward the few data in Ulaanbaatar relating to prevalence rates cross-sectional studies on mortality rates that were for different diseases, absenteeism from work and available for high pollution cities in China (World school etc. Despite this lack of data, assumptions Bank 2007). The result is an exposure-response are made in the following about the prevalence of function that flattens toward higher PM10 levels. chronic bronchitis in Ulaanbaatar. According to However, there are particularly large uncertainties Lopez et al. (2006) the prevalence rate in China related to this adjustment. New findings from and Mongolia is around 3 percent in adults short-term studies in Asia find that the exposure– above 30 years of age, with large uncertainties response functions appear linear over a fairly in the estimate. The World Bank (2007) used a large range of ambient concentrations up to and prevalence rate of 3.4 percent and a corresponding sometimes exceeding 100 µg/m3. However, in a annual incidence rate (new cases per year) of recent study from the U.S. (Pope et al. 2009) a 0.15 percent for China. This incidence rate was comparison of exposure-response relationships assumed in this report. between urban ambient particulate matter, passive cigarette smoking, and active cigarette As mentioned earlier, the team used data for smoking demonstrates a logarithmic exposure- the total annual number of hospital admissions response relationship with ischemic heart disease, for cardiovascular and respiratory diseases in cardiovascular disease, and cardiopulmonary hospitals in Ulaanbaatar for the calculations disease.12 related to these end-points. According to the AMHIB time-series study in Ulaanbaatar, The study includes exposure levels far the current baseline hospitalization rates for exceeding the levels typically encountered cardiovascular diseases and respiratory diseases in urban settings—both in developed and are 0.0196 and 0.0685, respectively. This implies developing countries—but conversely includes the that a total of approximately 22,000 hospital levels encountered in Ulaanbaatar. The findings admissions for cardiovascular diseases and around in Pope et al. (2009) support the use of nonlinear 76,000 hospital admissions for respiratory diseases exposure-response functions. In addition to the in Ulaanbaatar are estimated on an annual estimated premature deaths resulting from the basis. It is important to note that the hospital adjusted exposure-response function in World admissions estimates may not represent the Bank (2007), this report provides an estimate of entire effect related to these end-points because the health effect using that exposure-response not all hospitals in Ulaanbaatar were included function from Pope et al. (2009). This will in the time-series study. However, some of the indicate to some extent the sensitivity to the final patients that were admitted to tertiary hospitals in Ulaanbaatar may not be residents of the city. This 11 Note, however, that in the analysis of the warm season in Ulaanbaatar in the time-series study described in chapter means that the number of avoided cases from 5 in this report, a relationship between PM2.5 and all-cause interventions could be overestimated. Given that mortality was found. Specifically, a 10 µg/m3 change in PM2.5 the patients at tertiary hospitals constitute only was associated with a 1.38 percent change in mortality. This estimate is similar to those reported from studies of PM2.5 12 percent of the total number, it is suggested that in the U.S. In our view, the finding on a short-term effect the net effect of the two uncertainty factors is not on mortality in Ulaanbaatar lends credence to a long-term exposure effect on mortality in the city. Thus, we suggest it is clear and could well cancel each other. reasonable to assume that the long-term exposure studies used here are relevant also to Mongolia. Because no long-term epidemiological cohort 12 The study by Pope et al. (2009) uses estimated daily dose as the exposure variable. Daily dose (DD) is estimated from studies on mortality rates and air pollution have concentration levels by assuming an inhalation rate (IR) of on been carried out in Asia, the well-known, large average 18 m3/day for adults, i.e. in the present calculation we assume DD = PWE×IR. The highest dose for passive smokers study in the United States by Pope et al. (2002) in the study corresponds to 50 mg/m3, while the lowest dose was used by the WB (2007) to establish an for active smokers in the study corresponds to 1000 mg/m3. 64 Current Health Costs of Air Pollution and Benefits from Management Scenarios results of the choice of function for the mortality As carried out in previous applied studies impact. (Mestl et al. 2004; Kan et al. 2004), the report uses the population-weighted exposure When applying the function from Pope et al. (PWE) estimates for the whole region (i.e. the (2009), a threshold level for PM10 is not inserted, concentration times population in each grid because the team considers it more correct to use averaged for the total population in all grids) as the original nonlinear function in the calculation. input to the health benefit analysis. Given that Again, it is necessary to note that the relationship the exposure-response functions for mortality are between the two estimates—derived from using nonlinear, the results probably deviate slightly the World Bank (2007) function and the Pope from the results that would have been obtained et al. (2009) function)—differs across abatement using geographically disaggregated PWE values. options. This is because the WB function uses This uncertainty, however, is regarded by the team PM10 values while the Pope et al. (2009) function to be minor compared to other uncertainties in uses PM2.5, and the ratio between PM10 and PM2.5 the analysis. varies across the different abatement measures. The relative risk estimates obtained by applying, Monetized health benefits respectively, the exposure-response function from WB (2007), Pope et al. (2009), and Pope et al. To assess the unit costs of a premature death, (2002) are shown in figure 6-2. RR estimates this report relies on the willingness-to-pay study obtained from using Pope et al. (2009) are higher carried out as part of the AMHIB study (see when compared to the estimates from WB chapter 5) which derived a value of statistical (2007). However, the slopes of the two curves are life (VSL) of $221,000 i.e. the value placed on not very different at high levels of PM10. avoiding premature death. As in WB (2007) Figure 6-2: Relative long-term mortality risk associated with different levels of PM10, estimated using three different functions Relative risk Ambient PM10 levels (mg/m3) Source: World Bank (2007), Pope et al. (2002), and Pope et al. (2009). 65 Air Quality Analysis of Ulaanbaatar this value is multiplied with 0.3213 to obtain an and the number of cases that can be avoided estimate of the WTP for avoiding a new case from implementing the interventions described of chronic bronchitis (see table 6-2). Unit cost earlier. It also shows the monetized health estimates for hospitalization are derived based on impacts. The current health damage corresponds WB (2007). The report uses the ratio between to 18.8 percent of GDP in Ulaanbaatar and the value placed on these end-points for China by 8.8 percent of GDP in Mongolia15 in 2008. In WB (2007) and the GDP/cap in China combined the sensitivity calculation using the exposure- with the GDP/cap of Mongolia in 2008 to response function from Pope et al. (2009), calculate the corresponding unit cost estimates for current damage corresponds to 27.9 percent of Mongolia.14 GDP in Ulaanbaatar and 13.1 percent of GDP in Mongolia. The difference in the estimated Table 6-3 shows the current estimated health damage derived from the use of two number of cases attributable to PM pollution different exposure-response functions indicates the substantial uncertainty in the health damage assessment. However, it is clear that health 13 This factor is derived from a study (Viscusi et al. 1999) indicating that people’s choices imply that the utility of living damage is also substantial. with chronic bronchitis is about 0.68 of the utility of living in good health (Viscusi, W.K., W. Magat, and J. Huber 1991. “Pricing environmental health risks: A survey assessment of 15 GDP in Mongolia was $5.258 billion in 2008, in current USD risk-risk and risk-dollar tradeoffs for chronic bronchitis.� (WB database, for 2008, available: http://econ.worldbank.org/ Journal of Environmental Economics and Management WBSITE/EXTERNAL/EXTDEC/0,,menuPK:476823~page 21:32–51.) PK:64165236~piPK:64165141~theSitePK:469372,00.html). 14 I.e. Unit cost (Mongolia) = [Unit cost (China)/GDP per cap We assume the ratio between GDP in Ulaanbaatar and GDP (China)]*GDP per cap (Mongolia). in Mongolia is the same in 2008 as in 2007, i.e. 0.47. Table 6-2: Exposure-response coefficients (% change in incidence of health effect per μg/m3 PM10), baseline annual incidence rates, willingness-to-pay (WTP) for avoiding premature death (long-term effect) and new cases of chronic bronchitis, and cost of illness (COI) of hospital admissions Source: AMHIB data. 66 Table 6-3: Estimated current health damage due to PM pollution in Ulaanbaatar (base case), number of cases avoided due to interventions, and monetized current cost and benefit from interventions (in mill USD) Note: Sensitivity estimates using Pope et al. (2009) for mortality impacts. * These intervals were calculated based on the 95% confidence interval of the estimated dose-response coefficient and a +/- 30% interval of the PWE values. The lower (higher) value represents the estimate using the lower (higher) ends of both the 95% confidence interval of the dose response coefficient and PWE concentration. Similar confidence intervals apply to the estimated health damage reductions in the various scenarios but are not displayed here for 67 Current Health Costs of Air Pollution and Benefits from Management Scenarios the sake of clarity of presentation. Source: AMHIB data. Air Quality Analysis of Ulaanbaatar The maximum achievable benefit (80 percent the intervention has been fully implemented. reduction in all 3 sectors) corresponds to 7.0 These results are useful for comparing the percent of GDP in Ulaanbaatar in 2008 (3.3 effects situation before intervention (present percent of GDP in Mongolia). In the sensitivity situation) with the situation that will exist after calculation this figure receives 6.2 percent of the interventions. In practice an intervention is GDP in Ulaanbaatar (2.9 percent of GDP implemented over several years, with gradually in Mongolia). These sensitivity estimates are reduced emissions, concentrations and effects. lower than the base case, while the total damage When carrying out cost-benefit calculations to sensitivity estimates are higher because of the compare different types of interventions, the shape of the two exposure-response functions that time line should account for the interventions, as are applied to calculate mortality impacts in the well as integrate intervention costs and benefits base case and in the sensitivity case, respectively, (avoided effects costs) over the timeline. In as well as the exact profile of the abatement chapter 8, such an assessment is carried out for options involved. This is because the PM2.5/PM10 some selected abatement options. ratio is not constant across abatement scenarios. If the PM2.5/PM10 ratio of the source that is subject While the estimated health effects from to abatement is high, the ratio of the calculated PM pollution in Ulaanbaatar presented here benefit in the sensitivity case versus the base case are significant, it should still be noted that the will be relatively higher than in a case when the calculations do not take into account the exposure PM2.5/PM10 ratio is low. to indoor air pollution in gers. Enhanced PM levels indoors in ger households dependent on It should be noted that in this type of solid fuels for cooking and/or heating (assumed assessment of avoided health effects linked to be high for the short periods of time of lighting to an intervention, it is considered that the phase) could imply that PWE values may in fact avoided effects are annual for the year when be even larger than estimated in this report. 68 7. Air Pollution Abatement Options and Their Costs in Ulaanbaatar Main findings Ger area emissions reduction options and their costs Comparing different scenarios can be useful in examining potential costs and PM10 reduction Introduction impacts of selected PM reduction measures. In addition to the baseline scenario, alternative The study team has explored the impact and scenarios include (1a) reducing start-up emissions cost-effectiveness of different options to reduce through backlighting the fire; (1b) reducing particulate emissions. The scenarios follow start-up emissions through slight modifications portions of the Mongolian Government’s of the stove; (2) replacing existing stoves with “Smokeless UB� proposal, albeit with some cleaner coal stoves, without changing the fuel; modifications based on the acquired experience (3) replacing existing stoves and fuels with during the preparation of the UB Clean Air cleaner stoves and semi-coked coal fuel (SCC); Project (UBCAP project, to be financed by (4) installing electric heating in existing ger the World Bank) and overall dialogue with homes; (5) relocating ger households into government agencies and donors since mid- apartments; (6) installing heat-only boilers; and 2007, when the World Bank started its work on (7 and 8) controlling fugitive dust (paving roads air pollution in Ulaanbaatar. One of the main and greening the city). differences is the increase in balanced short, medium, long term measures in the discussed A short-term measure (scenario 1a) of scenarios, while the Smokeless UB proposal had reducing the startup emission of stoves through a program with a shorter time-span. In addition, changing lighting methods has a negative the analysis and scenarios mainly focus on ger area NPV of $463 per ton of PM10 reduced. heating issues. Another short-term measure (scenario 1b) of reducing the startup emission of stoves through As this section may attract considerable modification of stoves in addition to a change attention in Mongolia, it is necessary to discuss in the lighting method, also has a negative the limitations of the estimates. At the time of NPV of $208 per ton of PM10 reduction. writing this final report, the capacity to measure These two scenarios also reduce a substantial the impact of emission reduction measures in amount of PM10 in an absolute term. The long- UB had only recently been established. Ambient term measures such as apartment relocation emission levels can easily be measured in UB, and electric heating, which require a large and have in fact been monitored for a number infrastructure investment, have much higher of years. However, measuring the impact of NPV per ton of PM10 reductions. mitigation measures at the level of the various 69 Air Quality Analysis of Ulaanbaatar individual sources of pollution (cars, CHP plants, measurements show that a well-operated ger stoves, HOB, etc) requires special expertise traditional stove emits more than 300 mg of and equipment. As a first step, a state-of-the- PM2.5 per MJ of heat produced. This contrasts art laboratory was established in UB in August with the measurements of a number of clean 2010 by the Ministry of Mineral Resources stoves that showed a range of 1–10 mg of PM2.5 and Energy with the financial support of the per MJ of heat produced. The PM2.5 emission Asian Development Bank and technical support reduction therefore could exceed 95 percent if from the World Bank. This laboratory is able these cleaner stoves are used on a large scale. to measure the absolute level of emissions in However, it must be noted that, although some the smokestack of a stove (the main source of stoves submitted for testing to the laboratory pollution), as well as the reduction in emissions were called “clean,� they performed equally from the various types of modifications (to the with traditional stoves leading to the conclusion stoves and the fuels). Before the establishment of that not all “clean� stoves are actually clean. this laboratory, such measurement capacity did This justifies the existence of the laboratory not exist and was one of the barriers to predicting as an unbiased method to determine the the effect of specific measures and, thus, the actual performance of the stoves as a first step, overall development of effective mitigation although the performance needs to be verified measures for ger heating systems. The laboratory when households use the stoves outside of the has only recently started to operate regularly. laboratory environment. In a second-order estimation, the addition of apartments will Some types of interventions have a increase the load for electricity and heat and relatively more straightforward impact. For because the three CHPs are currently close to the example, when ger area households move maximum capacity, additional capacity is needed. into apartments, under certain assumptions, A new plant means that more coal will be burned it is possible to reduce to nearly zero the in the valley, resulting in increased emissions. contribution of ger household heating to However, this level of analysis is beyond the harmful ambient concentrations of PM. This scope of the current exercise, and on a household assumes that the new CHP plant should basis the emissions from coal stoves far exceed the be constructed to accommodate the heat emissions from the CHP system.17 load, that it operates good emission control equipment and manages its waste (ash ponds) The impact of different stoves and appropriately—all of which would need to different fuels is more difficult to determine. be verified. It could be assumed, as an initial Combined stove-fuel combustion/emission estimation, that the end-use emission reductions testing is necessary to check the effects of would be nearly 100 percent.16 Households prevailing combinations. Particulate emissions would likely abandon their ger stove and would are the result of poorly performing heating use district heating from the CHP plant. systems (stove-fuel combinations and fuel consumption). Improving the combustion However, although the absolute levels of characteristics of the stove can reduce emissions. emissions from ger stoves have been measured, Such reduction requires changing or modifying more measurements are needed to chart the the stove. Behavioral changes, such as the way results with more confidence. The initial households light their stoves, can also lead to 17 In addition, the government of Mongolia has been initiating 16 This assumption does not consider a resulting increase of a 100,000 Solar Ger Program. The World Bank is supporting PM10 and other pollutants from power generation as a this program through the Renewable Energy and Rural first-order approach of the issue. CHP plants supply heat Electricity Access Project by assisting in increasing access to and electricity to industries and households. The effect of the electricity and improving the reliability of electricity service incremental number of apartments on the total emissions for among the herder population and in off-grid soum centers all users combined initially remains small. However, this would with solar home systems (SHSs). The project will contribute need to be studied in more detail to determine the incremental 65,400 of the 98,322 total SHSs supplied as a part of the emissions due to the increased electricity or DH load. GoM 100,000 GER Program. While the project focuses on Compared to the emission reduction from not using coal in a rural areas, this may partly, though not significantly, offset the heating stove, these can be ignored in the first-order approach. increased air pollution. 70 Air Pollution Abatement Options and Their Costs in Ulaanbaatar emission reductions.18 At the time this report, Scenario 6: Heat-only boilers the MMRE’s laboratory had tested about 10 different stove models, all with Nalaikh coal due Scenarios 7 and 8: Control fugitive dust to lack of availability of other types of fuels. The (paving roads and greening the city). more promising test results have been used in this scenario analysis. A baseline emission factor was Scenarios 6, 7, and 8 are discussed in the established for a traditional stove using traditional following two sections, as they do not deal fuel, and its operation simulated as closely as directly with ger household heating issues but do possible in a traditional manner.19 The following have an impact on emission levels in Ger areas. section discusses emission reductions relative to the baseline (i.e. relative emissions reductions). The first five scenarios reduce the emissions Due to the considerable attention paid to fuel from space heating in ger areas, the sixth reduces switching, especially to semi-coke coal products, emissions from HOBs, and the last scenarios estimates of combustion characteristics have been reduce emissions from fugitive dust. For each assumed. For the AMHIB analysis, estimated of the above intervention scenarios, the cost of relative emission reductions are used that will be realizing the action has been estimated as well as verified more precisely through absolute emission the incremental benefits, relative to the baseline level measurements when they are available from situation. The goal of this analysis is to indicate the UB laboratory. which of the intervention scenarios are the more cost-effective and faster. It is possible to analyze The analysis presents the following individual many combinations of different interventions options, with each leading to a scenario in which using this framework. The five scenarios below, the option is rolled out on a large scale. It is therefore, are intended to illustrate how various possible to combine the options to look for higher measures could be compared and prioritized for performing scenarios: government support and policy making. This analysis initially does not look at the resulting Baseline; business as usual benefits, only at the costs needed to realize each of the scenarios for reducing emissions. Benefits can Scenario 1a: Reduce start-up emissions initially be assumed to be approximately equal in through backlighting the fire terms of health benefits, and will be identified in detail later in other chapters. Scenario 1b: Reduce start-up emissions through slight modifications of the stove Description of the scenarios Scenario 2: Replace existing stoves with The five scenarios, particularly the expected cleaner coal stoves, without changing the fuel outcome and what is needed to achieve it, are described in more detail in the following. Scenario 3: Replace existing stoves and fuels with cleaner stoves and SCC ■ Baseline; business as usual. This describes the status quo as observed today with several Scenario 4: Install electric heating in existing trends expected to continue: (a) growth of ger homes the city through natural population increase and a large influx of migrant workers from Scenario 5: Relocation of ger households into the country side; (b) conversion of moveable apartments ger dwellings into fixed wooden and brick homes; and (c) relatively fast growth of low- 18 Measurements in a qualified laboratory in Johannesburg, pressure boilers and decreased use of simple South Africa, and recently (November 2010) confirmed in stoves (with or without heating wall), which MMRE’s Ulaanbaatar laboratory, indicate that it is possible to increases coal consumption. obtain emission reductions of 80 percent or more from such behavioral changes. ■ Scenario 1 (short-term). Reduce start-up 19 Testing protocols are expected to be available at the laboratory. emissions. This is a promising option that was 71 Air Quality Analysis of Ulaanbaatar first described in the World Bank’s Mongolia: ■ Scenario 2 (short-term). Replace existing Heating in Poor, Peri-Urban Ger Areas of stoves with new coal stoves (“cleaner stoves�) Ulaanbaatar (further discussed as “Ger Heating without changing the fuel. Although several Report�, World Bank’s Asia Sustainable models of emission reducing stoves exist, not and Alternative Energy Program, October all combinations have been tested. Of those 2009), the Air Pollution in Ulaanbaatar tested, measurements indicate that the right Initial Assessment of Current Situation and type of stove with a traditional fuel (nalaikh Effects of Abatement Measures (World Bank coal) can achieve relative reductions in excess 2009), and was further considered by Prof. of 95 percent. In addition, fuel savings of Lodoysamba at the National University of up to 50 percent have been observed and Mongolia. It has become apparent that most the stove may remain hot for much longer of the emissions come from the cold start-up periods (one model stays warm for over 10 phase of the stove and to a lesser extent from hours). In partnership with the World Bank the refueling. Measurements confirmed that and GIZ, the Ministry of Mineral Resources PM emission reductions of 60 to 80 percent and Energy’s (MMRE’s) Clean Air Project can be obtained when the fire is started supported by the Asian Development Bank differently and when the fire is not allowed to (ADB) is spearheading this activity, with the die down but continues to burn throughout establishment of the laboratory, the training the day and night. This scenario thus reflects of stove designers, and a pilot program to mainly behavioral changes and not necessarily test some more promising stove models. This large capital investments for the beneficiary. scenario refers to an effort to make cleaner There are two options to reduce the start-up stoves available to users on a large scale. emissions: Successful implementation would require a. The first is to light the fire in a different (a) an awareness campaign to convince manner. The fire is usually lit at the back households of the advantages of changing of the stove, toward the chimney. Instead their stove; (b) a financing mechanism with of using wood, another fuel such as LPG a possible subsidy component to enable is recommended to start the fire; simple households to purchase the stove as well LPG canisters can easily be used and are as to promote a wide variety of eligible available already in ger areas. stoves to address customer preferences and b. The second is to slightly modify the increase chances of rapid market penetration; stove and use the back-lighting method and (c) an eligibility program to select of starting the fire. The design of the the appropriate stoves for support and stove can be changed in by inserting dissemination, and to create a sustainable more firebricks into the stove. This production capacity of such stoves. The main will reduce the size of the combustion issue with adoption of this scenario will be chamber and result in a slower and the perceived benefits from the stove to the cleaner burn resulting in fuel savings of user. The likely questions include: “Does it some 15 percent. The starter fuel and the save fuel? How quickly does it give off heat, bricks are part of the cost for rolling out and is cooking within accustomed times and this scenario, and therefore a publicity/ methods possible? How often is refueling promotional program to convince needed? Does it smoke when opening households to start and continue using the door for refueling? How much does it a different firing technique would be cost?� The costs of realizing this scenario required. The main issue to address in will include the investment in new stoves, this scenario is the training and awareness the replacement of these stoves after their program.20 useful service life, removal of old stoves, a publicity/promotional program, a quality 20 Some measurements have now been conducted in UB during control mechanism to maintain a sustainable the ADB-supported Clean Air Policy Advisory Support and Technical Assistance Project; however, the measurements production capacity of these stoves, and a should be confirmed through additional testing. possible subsidy that might be required for 72 Air Pollution Abatement Options and Their Costs in Ulaanbaatar quick adoption. The new stoves will reduce to recurrent fuel consumption subsidies until fuel consumption and thus provide a benefit the economies of scale are achieved and/or to end-users. Another economic benefit for incomes rise to afford the more expensive this scenario is a reduction in implementation fuel. Therefore, the benefits need to be costs. confirmed conclusively before a program is ■ Scenario 3 (medium-term). Replace existing started. stoves and fuels. Semi-coked coal (SCC) ■ Scenario 4 (medium-term). Install electric receives much attention, and while SCC heating in existing ger homes. This approach can burn cleanly in an appropriate stove, is discussed in more detail in the Ger Heating there are two challenges associated with this Report (World Bank/ASTAE 2009). This scenario. SCC is difficult to light and its requires a large investment program to create production costs are higher than raw coal. the capacity to generate the power needed Raw coal burns very cleanly after the start-up to supply ger households with electricity for phase and actually transforms into coked coal heating (estimated at 1.7 GW by MMRE in with associated low emissions. The bulk of 2011). In addition, an equalization charge is emissions occur during the start-up phase. needed because the cost of electric heating is Since semi-coked coal is difficult to ignite due significantly higher than the cost of heating to the absence of volatiles, wood and other with coal. In this option, people do not start-up fuels are needed, which promote move into new homes but continue to live high emissions. Moreover, because the cost at their current residence and start using of producing semi-coked coal is high, the electric stoves for heating and cooking. The heating costs associated with the converted cost of electric stoves and heaters, the cost fuel are much higher than with raw coal, of electricity minus the savings of coal fuel, and equalization payments (subsidies) are and the cost of infrastructure for incremental necessary to avoid poor households paying generation and distribution capacity will more for heating. Since start-up emissions need to be incorporated into the cost constitute most of the total emissions, the analysis. The emission reduction can be large overall impact remains unclear if more wood (close to 100 percent; see also remarks on is needed to get the fire started compared to apartment buildings21), assuming that people raw coal. Tests so far have been inconclusive will actually refrain from using coal once as to whether SCC will reduce emissions. they obtained an electric heater. A limiting Furthermore, it is necessary to use new factor will be the infrastructure investments stoves to burn SCC cleanly. Because SCC is to supply the additional electricity and the more expensive than raw coal, the European willingness of households to pay for the Bank for Reconstruction and Development electricity, because although electric heating (EBRD)’s Clean Air Initiative has proposed is more convenient than coal heating it will a continuation of equalization payments also be more expensive (or, a subsidy may be until scale economies can be obtained and needed to equalize heating costs, but this has SCC could be sold without subsidies. This not been incorporated for now). The scenario scenario still requires additional basic research assumes that most people in ger districts to develop the emission details, and therefore will actually switch to electricity once the the study team needed to assume certain government announces the availability of this benefits that could be verified later. The option. scenario therefore includes setting up the ■ Scenario 5 (long-term). Relocation of ger production capacity of SCC and SCC stoves, households into apartments. This is the an awareness campaign to convince people preferred long-term option, indicated in the to start using it in new stoves, and recurrent Smokeless UB program. New apartment annual subsidies to enable the use of SCC buildings are established in newly developed at equal costs to raw coal. The industrial production of semi-coked coal from raw coal 21 The incremental emissions at the CHP plant have been requires a commitment by the government ignored in a first-order estimation. 73 Air Quality Analysis of Ulaanbaatar areas, in existing ger areas, and in other ■ Cost of additional infrastructure needed cities. The impact is relatively simple, as (additional production capacity of clean coal consumption can be avoided almost fuels, electricity and/or heat, electricity/heat completely (from heating in coal stoves). distribution network) There will be an increased contribution ■ Cost of incremental heating energy (or from the district heating system, but this saving, whichever applies; in the case of is estimated to be small compared to the apartments, coal and firewood saved, but consumption of coal used for heating in ger electricity and district heating energy is used stoves. The costs of construction, as well as instead) the incremental capacity needed for district heating, will need to be incorporated in the The net present value of the identified cost analysis, and these costs are very high. costs and possible direct benefits (fuel savings) are then calculated for the period 2010–2023, Results—cost implications using a discount rate of 10 percent. This is an indication of the costs required to roll out a particular scenario. Two of the scenarios The absolute level of impact on emission result in appreciable savings (negative costs, or reductions cannot be measured yet in all cases. benefits) and lead to low NPVs, while three As a result, the reduction is expressed in percent have relatively high costs. Benefits may accrue of the baseline, i.e., compared to as if no to the individual household through reduced intervention had taken place and trends would energy costs. The first four scenarios reflect have continued while business as usual takes estimated penetration rates; those expected to place. The emission reductions over time are then be realistic in the absence of estimated targets compared to the costs needed to realize these in the Smokeless UB program and which can reductions. be achieved in the time period indicated. The 5th scenario uses data from the Smokeless UB The costs for the intervention include the program. following types of investments and expenditures (see table 7-1). Results—emission reductions ■ Cost of the program to promote the intervention (publicity campaign, training, The results in terms of average emission including any subsidy needed, cost of testing reductions22 over the period 2010–23 are and certification) presented in table 7-3. When these data are ■ Cost of additional equipment needed (such discounted at the same rate and combined as stoves, electric heaters, apartment building, etc.) 22 Straight average over the entire period Table 7-1: Cost elements that play a role in the different scenarios Source: AMHIB study. 74 Air Pollution Abatement Options and Their Costs in Ulaanbaatar Table 7-2: Net present value of the costs and direct benefits Source: AMHIB study. Table 7-3: Average and maximum emission reductions Source: AMHIB study. with the costs for the scenario, an indicator is reductions over the 15-year period along with the obtained for the effectiveness of the emission maximum attainable. Relocation into apartments reduction scenario. Three of the scenarios have would have a higher contribution if it were not a very significant impact: modifying the stove for the lead times needed for construction and and lighting technique (1b), introducing cleaner relocation. stoves (2) and electric heating (4) in ger areas. These numbers do not reflect the viability of the Three of the scenarios are expected to scenarios, but the potential impact on reducing have a long-lasting impact on the emission emissions. The table simply reflects the “what if � levels as they would implement irreversible situation—what would be the emission reduction solutions. Reducing the start-up emissions will if x percent of the households moved into immediately produce pollution benefits (short- apartment buildings; if y percent adopted electric term); electric heaters will only have medium- heaters while continuing to live in their homes; term benefits, because first the generation and or if z percent replaced its old coal stove for a new distribution infrastructure need to be increased one, etc. to cater for the increased electricity demand; and moving into apartments will have long-term Figure 7-1 shows the emission reduction benefits only, because time is needed to raise the that would be obtained if the scenario were fully financing and construct the required number of implemented, and table 7-3 illustrates average buildings. 75 Air Quality Analysis of Ulaanbaatar Figure 7-1: Estimated emission reduction from ger area interventions Source: Authors’ illustration. Discussion the investments needed to convince people to replace their stoves or change their habits for Two of the scenarios yield more direct benefits lighting the fire are negative. This means that the (i.e., savings) than the costs needed to roll out the benefits exceed the costs. scenario. Reduced start-up emissions and certified stoves are options from both an economic There are significant differences between the standpoint23 as well as a beneficiary’s perspective. scenarios, both in terms of investment and in The other three scenarios imply more costs than terms of benefits, and a more detailed economic benefits and actually require significant levels of and technical analysis should be conducted to investment. determine the right mix of short- and long-term solutions. While this report explores a number of When assessing the costs in combination options, many more remain. with the potential emissions reduction, a useful measurement is the NPV of the investments When establishing priorities, this cost-benefit needed to reduce emissions by one percent analysis should not be considered in isolation compared to a baseline scenario. This allows of other important factors, such as technical for a direct comparison of the different options feasibility, security of supply (e.g. if a new fuel is (see table 7-4). For example, an investment of proposed), ease of implementing the proposed $1.6 million would be needed to obtain a one measure (e.g. changing behaviors is challenging), percent reduction in emissions for the SCC affordability, social acceptance, and strength of scenario, $22 million to obtain a one percent governance. With this caveat, the results of these reduction from enabling electric heating in ger estimated cost benefit analyses reveal that some areas, and $133 million for the same reduction measures generate significant financial benefits from constructing new apartment buildings and to consumers (fuel savings) which compensate associated infrastructure. However, the NPV of for the investment costs of the measures. Other measures require a much higher level of 23 Not counting health benefits or other societal benefits investment and may take longer to implement 76 Air Pollution Abatement Options and Their Costs in Ulaanbaatar Table 7-4: Relative cost-effectiveness of the different options Source: AMHIB study. with fewer financial benefits for consumers, but There are about 160 heat-only boilers in 89 reduce pollution by a significant amount. These boiler houses in the six central UB ger districts investments are also different, ranging from (Bayangol, Bayanzürkh, Chingeltei, Khan-Uul, one-off capital subsidies for stoves to recurrent Songinokhairkhan, and Sukhbaatar). A higher subsidies for fuel or electricity consumption over number of boilers in the three outlying UB several years to large infrastructure investments. districts are not connected to the district heating The optimal strategy will differentiate between grid (Baganuur, Bagakhangai, and Nalaikh). these differences and impacts to ensure that short- HOBs in the six central districts contribute to air term measures are implemented, while larger pollution in UB but not the others, as they are investments have the right amount of time to be physically too distant from UB (Baganuur is over adequately prepared over the medium-term. 150 km away). See Table F.1(Impact and costs of the Ownership of an HOB is partly public, different scenarios for PM emission reduction) in partly private. Most privately owned HOBs annex F. are already technically efficient as many have been replaced previously. Some of the public HOB emissions reduction option and costs HOBs are managed by private companies, and (Scenario 6) some public HOBs have already been replaced. In total, 40 percent of the HOBs in the 89 Scenario 6 (medium-term). This option does boilers houses do not need to be replaced, as not refer to households, but to public schools, they are efficient already. It must be noted that hospitals, and other buildings in ger areas that the actual efficiency of the HOBs depends on use a heat-only boiler (HOB). HOBs are not the operation and maintenance of the boilers. connected to the UB district heating system. Recent technical assistance work supported by While the capacity of household stoves typically JICA reveals a high variability in technical skills remains below 10 kW, an HOB can exceed and know-how among boiler operators. Thus, 1 MW. Their coal consumption and emission poor operation, even of newer boilers, could be a levels are much higher than household stoves contributing factor to higher than expected source which typically use 20–24 kg of coal per day in contributions from HOBs. January, compared to an HOB, which uses on average 4 t per day. However, although emissions In the following scenario, all remaining from ger area households are one of the two inefficient HOBs are replaced—the medium completely dominating contributors to total PM efficient HOBs over a 3-year period and the emissions in the city (the other one being the most inefficient HOBs over a period of 8 years. CHPs), HOBs make only a small contribution to Data from a 2009 market study of HOBs and total PM emissions. coal-fired water heaters, 2009, CBDICFP (JICA 77 Air Quality Analysis of Ulaanbaatar Table 7-5: HOB baseline data 2009) have been used for unit consumption data sidewalks, maintaining these roads and sidewalks, and efficiency data. The Smokeless UB program and sweeping.24 Based on the above assumptions, gave cost data for the replacement of HOBs at PM10 reduction from roads are estimated from $214/kW. This figure has been used, with some (a) newly paved roads and the sidewalk along the miscellaneous costs added, to give a total cost of newly paved roads; (b) sweeping of these newly replacement of $150,000 for an inefficient HOB paved roads; and (c) sweeping of existing paved and $125,000 for a medium-efficient HOB. All and unpaved roads. Only hard surfaced unpaved relative fuel savings and emission reductions have roads are supposed to be swept. been calculated using the data in table 7-5. Washing or wet methods of cleaning the See table A-2 (Results HOB Scenario) in the roads may further enhance the PM10 reduction, appendix to chapter 7. but are not included in the analysis due to the scarcity of water, dryness of the climate, and the Soil suspension reduction options and freezing winters. It is assumed that the dustiest their costs roads will be paved and swept first. Unpaved roads in the city center and areas nearby are Scenarios 7 and 8 include two additional scenarios assumed to have less traffic than paved roads but of fugitive dust reduction to address PM10 are higher traffic than the average unpaved roads. included: (1) paving of unpaved roads with the It is also assumed that the feasibility study, construction of sidewalks and sweeping the newly environmental impact assessments and other paved roads and existing paved and unpaved safeguard issues, financial mobilization, and roads; and (2) increasing vegetation in the city. construction will be conducted during 2010–15. Scenario 7: Road dust reduction. In ger Control techniques for fugitive dust areas of UB (including the city center, mid-tier, sources generally involve watering, chemical and fringe ger areas), the combined length of stabilization, or reduction of surface wind speed earthen and paved roads is 80,929 km with the with windbreaks or source enclosures. Watering, majority, 72,313 km or 89.4 percent, being the most common and, generally, least expensive earthen roads (Kamata et al 2010). It is anticipated method except in water-scarce and dry areas that 500k m of these earthen roads will be paved like Mongolia, provides only temporary dust and sidewalks constructed from 2016 to 2023. control. The use of chemicals to treat exposed Also, it is assumed that these newly paved roads surfaces provides longer dust suppression, but will be swept. In addition, it is also expected may be costly, have adverse effects on plant and that existing paved and unpaved roads will be animal life, or contaminate the treated material. swept, with an annual increase of 1,000 km for Windbreaks and source enclosures are often each category of road. Data for capital costs for impractical because of the size of fugitive dust road pavement and for sidewalks were obtained sources. The reduction of source extent and from Kamata et al. (2010); emission factors for paved and unpaved roads were obtained from 24 An estimate of operation and maintenance cost is obtained from Bank transport staff. An estimate of sweeping cost is Guttikunda (2007). The costs for this scenario are derived from a Bangkok Air Quality Management project mainly the costs of paving the road, constructing report (World Bank 2007). 78 Air Pollution Abatement Options and Their Costs in Ulaanbaatar the incorporation of process modifications or Maintaining equipment cabs in good adjusted work practices, both of which reduce operating condition also reduces operator exposure the amount of dust generation, are preventive to respirable dust. A study conducted on dozers techniques for the control of fugitive dust and drills demonstrated that properly maintained emissions. These techniques could include, for cabs can attain dust reductions of 90 percent for example, the elimination of mud/dirt carryout drills and between 44 percent and 100 percent on paved roads at construction sites. Conversely, for dozers. The variations in the dust reductions mitigative measures entail the periodic removal of for dozers were attributed to re-entrainment of dust-producing material. Examples of mitigative internal cab dust. An additional study completed measures include clean-up of spillage on paved or on haul trucks, which involved the retrofitting of unpaved travel surfaces and clean-up of material a cab with a filtration/pressure air conditioning spillage at conveyor transfer points. system to produce positive pressure in the cab, showed that properly maintained cabs are able For haul trucks, for example, there are several to produce a potential 52 percent reduction of dust reduction practices. These include reducing respirable dust (Reed and Organiscak 2007). haul truck speed and maintaining safe following distances, watering haul roads, treating haul These measures are not included in the cost- roads, and maintaining equipment cabs. Reducing benefit analysis of this work. the haul truck speed is the simplest method. Implementing a policy to ensure that trucks do Scenario 8: Greening the city with not follow within 20 seconds of another truck can vegetation. Suspension of dust from the soil result in a 41 to 52 percent reduction in airborne surfaces will be reduced by a mix of simple respirable dust exposure to the following truck. vegetation, grass, bush and trees because it binds and covers the open soil and thus prevents The use of water on haul roads is the most suspension of dust to a considerable degree if the common dust reduction method. Watering the vegetation cover is successfully established. In haul road on the test section in a study allowed addition, some of the dust already suspended, as instantaneous dust concentrations to remain well as a fraction of all particles in the air, will also below 2 mg/m3 for over three hours. Past research be captured by the vegetation, although this is has shown that watering haul roads with a water a minor effect compared to the reduction of the truck once an hour has a control efficiency of suspension. The number of areas to be vegetated 40 percent for total suspended particulates (TSP). and the costs of vegetation are derived from the If watering is increased to once every half hour, smokefree Ulaanbaatar national program for the control efficiency for TSP increases to 55 2010–16. In the program, 0.7 percent (about percent. The control efficiency was defined as a 952 hectares) of Ulaanbaatar’s land is annually comparison of the controlled (watered) emission vegetated during the period and the program is rate to the uncontrolled emission rate. The EPA expected to continue from 2011 to 2023. This reported several test results of watering haul roads. vegetation includes different of types of vegetation The results ranged from a control efficiency of (e.g., a mix of simple vegetation, grass, bushes, 74 percent for TSP for the three to four hours and trees), boundary marking, and protecting following the application of water at a rate of fences such as reforestation, planting broad- 2.08 L/m2 (0.46 gallons/yd2) to a control leafed trees upstream of the Tuul and Selbe rivers, efficiency of 95 percent for TSP for 0.5 hours after greenbelt establishment, land reclamation, and the application of 0.59 L/m2 (0.13 gallons/yd2). so forth. The program mentions other initiatives including a ger area garden, a garden near power Treating haul roads generally involves the plant number four, national park, and dust application of chemicals, and requires a significant reduction through re-vegetating city public space amount of road maintenance. In one study, and roads—although it does not specify how control efficiencies were 95 percent for magnesium many hectares will be vegetated. These actions chloride and 70 percent for a petroleum derivative are not included in the assumption of annual for controlling haul-truck-generated dust. vegetation of 952 hectare. Due to UB’s harsh 79 Air Quality Analysis of Ulaanbaatar environment, it is expected that half of the annual vegetation is not available, the estimate of PM increase in vegetation cover will die each year. reduction by the establishment of vegetation is considered a very tentative analysis. It is, however, This scenario incurs the cost of planting a potentially very important measure, since much and maintaining the vegetation. The Smoke Free of the PM10 problem in UB originates from Ulaanbaatar national program 2010–16 aims soil suspension. Studies of dust reduction by to plant vegetation across the city, not merely vegetation in UB should be conducted. alongside the roads. Thus, the estimated PM reduction due to suspension prevention operates See table F.3 (Paving Roads and Greening the on the total soil dust suspension emissions of the City) in annex F. whole city. The estimated reduction in PM10 due to the capture of already airborne particles by Summary the vegetation operates mainly on particles from all low-level PM10 emissions in UB (see table Table 7-6 gives a summary of the different 3-3). The PM reduction due to the vegetation’s scenarios and their specific costs, benefits, and prevention of suspension, when successfully impacts. It shows that the impact of the scenarios established, is estimated to be 80 percent. Local has a wide range, from about 5,000 tons to almost studies on this are not available, and results from 190,000 tons of PM reduced over a period of 15 studies carried out by Grantz et al. (1998) have years. The costs also vary widely— some scenarios been used (see annex F). PM10 deposition in air have direct benefits (i.e., fuel savings are larger to different types of vegetation is very specific than the direct costs of rolling out the scenario) to each situation (e.g., wind speed, surrounding while some costs are reduced by several hundred building, climate, temperature, latitude, type of thousand dollars per t. vegetation, and so on). Local data on the PM10 deposition rate to trees or any types of vegetation This analysis is intended to compare the is not available for UB. Thus, the estimate is various options, mainly based on data included in very conservative and will change depending on the Smoke Free UB plan. A more rigorous analysis the values used (Hewitt 2010). Since local data is required before any of the scenarios is put into on suspension reduction by and deposition to action. Table 7-6: Summary performance of the various options Source: AMHIB study. 80 8. Health Benefits of the Air Pollution Management Scenarios T his chapter estimates the monetized scenario, the PWE (for PM10) contribution for health benefits of the abatement the relevant source category in table 3.3 above options outlined in chapter 7. It (for the base year) is multiplied by the percentage estimates the year-to-year benefit reduction in emissions as calculated in chapter achievable in the period 2010–23, as well as the 7. By this method, it is possible to estimate how present value in 2010 of the benefit accruing much the total PWE in the city is reduced by over the full period. The largest benefit derives the abatement measures. As any value of PWE from ger electric heating ($1,803 million), corresponds to a certain health cost (which is the from introducing certified stoves ($1,605 sum of mortality and morbidity costs, see chapter million), and from reducing start-up emissions 6), the team is able to calculate the health benefit using the lighting technique ($1,599 million). (i.e. the avoided health costs) associated with each Other technical and nontechnical ger area scenario and each year of the scenarios. Only interventions—such as relocating ger residents the base case method for calculating mortality into apartments—also produce substantial health impacts is applied in this chapter. To account for benefits. Reducing emissions from HOBs, road population growth, the resulting figures for health dust reduction, and greening scenarios give some benefits with an annual growth rate of 5 percent health benefits, although they are small when are being adjusted. The annual average population compared to the other scenarios. growth rate in Ulaanbaatar in the period 1993– 2008 was 4.3 percent, while the growth rate in The previous chapter explored management the period 2000–08 was 4.7 percent. 5 percent is scenarios for reducing emissions from ger stoves suggested as a reasonable estimate in the base case. and HOBs in terms of their costs and potential Guttikunda (2007) assumes 5 percent, 8 percent, emission reductions, as well as options for and 10 percent annual population growth toward reducing road dust and windblown dust. The 2020; the 10 percent figure is described as following chapter estimates the health benefits unlikely. associated with the abatement scenarios year by year and calculates the present value of the benefit In addition to population growth and of each scenario. Finally, the health benefits income growth, other welfare developments are accruing from the scenarios are compared with likely to occur. This implies that estimates of the costs of implementing the measures needed to future potential benefits of interventions should realize the scenario. be adjusted. Approximately 85 percent of the benefit is related to either mortality impacts A simplified approach based on the and chronic bronchitis, both valued using the percentage reduction in emissions given in chapter value of statistical life (VSL). Therefore, the 7 is used for each scenario, which also applies income elasticity rate of VSL to adjust the benefit to the sources’ contribution to total population- estimates is used. Studies conducted largely in the weighted exposure (PWE) in Ulaanbaatar. United States suggest that the income elasticity Therefore, for each year (2010–23) of a given of WTP for mortality risk reductions is less 81 Air Quality Analysis of Ulaanbaatar than one, with elasticities averaging between 0.4 economic growth rate of 10 percent, this means and 0.6 (USEPA 1999). However, elasticities the benefit estimates are inflated with an annual greater than 1.0 are suggested by research on the growth factor of 15 percent. The discounted relationship between long-term economic growth annual health benefit from Scenarios 1–8 are and the VSL, by cross-country comparisons, shown in figure 8-1.26 Figures 8-2 and 8-3 show and by new research that estimates the VSL by how the benefit estimates change when higher income quartile. Moreover, studies suggest that or lower annual population growth rates and the elasticity varies by income level; that is, lower income elasticities of VSL are being assumed. The income levels are associated with higher income health benefit to some extent mimics the emission elasticities of VSL (Hammitt and Robinson reductions shown in figure 7-1 previously 2011).25 In Mongolia income levels are low, (especially in the low case), since the combined which suggests using an elasticity higher than effect of assumed population growth and increased 1.0. In the following calculations, an elasticity willingness to pay for risk reduction is partly of 1.5 is used in the base case and 1.0 and 2.0 canceled by the discount rate of 10 percent applied. in the low and high case, based on the study by Hammitt and Robinson (2011)25. Assuming an The value of the present health benefits for the various scenarios is presented in table 8-1. 25 See Hammitt and Robinson (2011). In Ulaanbaatar we do not have to transfer VSL estimates from other, higher income countries, and thus are primarily interested in the time-series development of VSL versus income in the city. Hammitt and 26 The remaining 15 percent of total health benefits that Robinson (2011) report elasticities in the range 1.5–3.0 in are not valued using the VSL estimate are related to the longitudinal studies. Since differences in longitudinal versus direct cost of hospitalization. It is possible to argue that we cross-section studies are yet not resolved, we choose to use an should have used the economic growth rate of 10 percent elasticity estimate based on the broader basis of studies covered to adjust hospitalization costs. This would however, have a by Hammitt and Robinson (2011) instead of the more limited minor impact on our results, and for simplicity we omit this evidence from longitudinal studies. adjustment. Figure 8-1: Annual health benefit from abatement scenarios (mill USD) (discounted values), base case Source: Authors’ illustration. 82 Health Benefits of the Air Pollution Management Scenarios Figure 8-2: Annual health benefit from abatement scenarios (mill USD) (discounted values), high case assuming an annual population growth rate of 8% and an income elasticity of VSL of 2 Source: Authors’ illustration. Figure 8-3: Annual health benefit from abatement scenarios (mill USD) (discounted values), low case assuming an annual population growth rate of 5% and an income elasticity of VSL of 1 Source: Authors’ illustration. 83 Air Quality Analysis of Ulaanbaatar Table 8-1: Present value of health benefits ($ million), accumulated for the period 2010–23 Source: AMHIB study. The largest benefit in terms of reduced coked coal (Scenario 3) and relocation into health costs derive from ger electric heating apartments (Scenario 5) entail a health benefit of, (Scenario 4), estimated at $1,803 million; respectively, $1,029 million and $597 million. from introducing certified stoves (Scenario 2), The HOB, road dust reduction and greening estimated at $1,605 million; and from reducing scenarios also give health benefits, although the start-up emissions using the lighting technique benefits are small when compared to the other (Scenario 1b), estimated at $1,599 million. scenarios. Reducing start-up emissions using propellant (Scenario 1a) entail somewhat lower health In chapter 9, the estimated health benefits benefits, about $866 million. Replacing existing are compared with the costs of implementing the stoves and fuels with cleaner stoves and semi- scenarios. 84 Health Benefits of the Air Pollution Management Scenarios As shown above, the different abatement avoided costs of some scenarios when compared scenarios being considered have quite different to other scenarios. present values of health benefits. An alternative comparison between the scenarios is to compare For example, a comparison of short-term the avoided costs between short-term measures to (immediately available) interventions like medium- and long-term measures. Medium– and application of certified stoves (scenario 2), long-term solutions require time to be prepared application of new lighting techniques (scenario properly, with substantive investments that have 1b) or propellant techniques (scenario 1a) to to be mobilized over several years, but people reduce startup emissions in the stoves compared continue to breathe unhealthy air, which results with the long-term intervention of relocating in an increase in both morbidity and mortality. inhabitants into new apartments, shows A relevant estimation is therefore to assess the significant avoided costs. Figure 8-4 illustrates Figure 8-4: Difference in health benefits ($ million) between five short to medium term measures and a long-term scenario (relocation into apartments). Source: Authors’ illustration. 85 Air Quality Analysis of Ulaanbaatar the extent of avoided cost between the short- and be between the same short-term interventions long-term interventions. and long-term interventions exemplified by relocation into apartments. This does not mean Table 8-2 shows how much the avoided costs that there should be entirely focus on short-term estimated in absolute mortality and morbidity interventions, but rather that short, medium (measured in cases of chronic bronchitis, and cases and long term interventions should be sought in for respiratory and cardiovascular diseases) may parallel. Table 8-2: The difference in health benefits ($ million and absolute numbers) between five short to medium term measures and a long-term scenario (relocation into apartments) * Applying WB 2007 dose response function. (World Bank 2007). Source: AMHIB study. 86 9. Cost-Benefit Considerations T his chapter compares the annual health The largest net benefit derives from benefits to the annual implementation scenario 1b (start-up emission reduction using costs for each scenario as well as improved lighting technique), while scenarios over the full implementation period. 1a, 2, 3, and 4 (start-up emission reduction/ Benefits are calculated in terms of their present backlighting, certified stoves, improved stoves/ values (PV) and costs are calculated as net present semi-coked coal, and electric heating) also values (NPV), since, as described above, there are yield substantial net benefits. Improved HOBs some cases where energy savings reduce the costs. produce a net benefit, but relatively smaller The PV and NPV of future annual costs and than that of the ger area interventions. The benefits represent the value in 2010 of costs and relocation of ger residents into apartments has benefits accruing over the period 2010–23 and the largest net cost—estimated to be more than enable a cost-benefit evaluation of the different $3 billion. While the health benefit of relocation scenarios. The largest net benefit is calculated for is estimated to be substantial, it is the lowest reducing start-up emissions using the lighting among the ger area interventions. Scenario 4 technique ($1,635 million) and from introducing (electric heating) is the option producing the certified stoves ($1,605 million), while relocating largest health benefit, but the costs are also ger area residents to apartments carries by far substantial, yielding the lowest net benefit largest net cost (about $3,497 million). among the ger area interventions. The health benefit and the abatement costs The dust reduction scenarios (7 and 8) are presented in the two left-most columns in yield relatively low health benefits. Scenario table 9-1. Both benefits and implementation 8 (greening) also has a low cost, resulting in a costs vary considerably between the scenarios. clearly positive net benefit. Reducing road dust Cost-benefit considerations of benefits and (scenario 7) appears to be a rather expensive abatement costs can be calculated in several option, yielding a net cost around zero and having ways. One way is to calculate the “net benefit� the highest cost/benefit ratio among the scenarios for each scenario. In table 9-1, the “net benefit� with a positive implementation cost. estimates are calculated by subtracting the net present value (NPV) of abatement costs from Cost-benefit considerations for each of the the present value (PV) benefit value for each scenarios scenario. Sensitivity calculations assuming higher or lower annual population growth rates The scenarios are discussed in the order of and income elasticities of VSL are included in intervention type, starting with the largest net the table (see details regarding the choice of benefit. Please refer to chapter 8 for a detailed parameters in chapter 8). description of the scenarios. 87 Air Quality Analysis of Ulaanbaatar Table 9-1: Comparison of present value (PV) of health benefits (base case) with net present value (NVP) of implementing costs , and net benefit (PV minus NPV) for the eight abatement scenarios, 2010 (mill USD) Note: Base case: Annual population growth: 5% and income elasticity of VSL: 1.5 Source: AMHIB study. Scenario 1b: Reduce startup emissions be above $1,599 million. It has a (small) negative (lighting technique). Net benefit: cost, since fuel is saved by the lighting method $1,635 million (less use of wood). Measurements confirm that PM emission Scenario 1a: Startup emissions reductions in excess of 80 percent can be obtained (backlighting). Net benefit: $919 million when the fire is started differently and when the fire is not allowed to die down but continues to This is similar to scenario 1b, except that the fire burn throughout the day and night. This scenario starting method is different. It uses another fuel— uses a back-lighting method of starting the fire, such as LPG—than wood to start the fire. Simple in combination with a slight change in the design LPG canisters can easily be used and are already of the stove that can be realized by inserting more available in ger areas. firebricks into the stove. This will reduce the size of the combustion chamber and result in a For this scenario the health benefit is very large, slower and cleaner burn. However, it should be estimated at close to $866 million, and also there noted that changing behavior, such as customary are (small) negative costs, because fuel is saved. methods of lighting a traditional stove, may be challenging. Scenario 2: Certified stoves. Net benefit: $1,605 million This represents a major health benefit from this scenario since it reduces the emissions from This scenario envisions replacing existing stoves the ger stoves very significantly and is estimated to with better coal stoves (“improved stoves�), 88 Cost-Benefit Considerations without changing the fuel. Measurements the savings in heating fuel, and the cost of taken elsewhere than in UB are conclusive and infrastructure for incremental generation and show that the correct type of stove can obtain distribution capacity is incorporated into the cost emission reductions of 50–80 percent. Similar analysis. The scenario assumes that most people measurements are currently being taken in UB. in ger areas will actually switch to electricity once This scenario attempts to make cleaner stoves the government announces that this option is now available to users on a large scale. This requires an available. awareness campaign to convince households of the usefulness of changing their stove, a financing The health benefit value is the highest of all mechanism with a possible subsidy component scenarios, almost $1,803 million, since emissions to enable households to purchase the stove and from ger homes are considered to be reduced to to promote a wide variety of eligible stoves to almost zero. There is, however, a large cost to address customer preferences and increase chances implement this scenario, over $1,400 million. of rapid market penetration. The scenario would This includes the infrastructure development also require an eligibility program to select the (power generation, lines, etc.). Thus the net appropriate stoves for support and dissemination benefit value is reduced to about $524 million, and create a sustainable production capacity of but it is still positive, meaning the health benefit such stoves. is considerably higher than the cost. Estimated health benefits are about $1,605 Scenario 5: Relocation of ger population million. The implementation cost is estimated to into apartments. Net benefit: be near zero (costs and savings equal each other). $3,497 million Scenario 3: Semi-coked coal (SCC) + new In this scenario, new apartment buildings are set stoves. Net benefit: $992 million up in newly developed areas, in existing ger areas, and in other cities. The impact is relatively simple, This scenario looks to replace existing stoves and as coal consumption can be avoided almost fuels. Semi-coked coal (SCC) receives significant completely (from heating in coal stoves) although attention, and while SCC can burn cleanly in there will be an increased contribution from the an appropriate stove, there are two problems district heating system. The very high costs of associated with this fuel. It remains difficult to construction, as well as the incremental capacity light and the production costs are higher than needed for district heating, is incorporated in the raw coal. The scenario therefore includes setting cost analysis. up the production capacity of SCC, an awareness campaign to convince people to start using it in Even if the emissions from the Gers are new stoves, and subsidies to enable the use of almost eliminated in this scenario, the health SCC at equal costs as raw coal. benefit, of almost $597 million is much less than for the electric stove scenario above, where The estimated health benefit is about ger stove emissions are also eliminated. This $1,029 million for this scenario. There is a is because of the slow implementation of the positive implementation cost of about $37 million, relocation scenario, as the apartment buildings giving a net benefit value of about $992 million. need to be built. In the meantime, ger residents are still exposed to emissions for many years, Scenario 4: Electric heating in Gers. Net reducing the health benefit that can be obtained benefit: $393 million in the given period of time. In this scenario people do not relocate but start The costs are very high for this scenario, using electric stoves for heating and cooking including apartment building construction, of in their current homes. The cost of electric approximately $4.1 billion. This gives a negative stoves and heaters, the cost of electricity minus net benefit of about $3,497 million, meaning 89 Air Quality Analysis of Ulaanbaatar that the costs are much higher than the estimated Summary value of the health benefits. Scenarios 1 through 5, which reduce ger area Scenario 6: Improvement of heat-only emissions substantially by various methods, boilers (HOBs). Net benefit: $14 million yield substantial health benefits, and are all strong candidates for consideration, although In this scenario, all remaining inefficient HOBs some of these scenarios are substantially more are replaced, the medium-efficient HOBs over costly. a 3-year period, and the most inefficient HOBs over a period of 8 years. Reducing the start-up emissions (Scenarios 1b and also 1a) and electric heating in the ger Both the health benefit and the costs of this homes (Scenario 4) appear to produce by far scenario are relatively small. HOB emissions are the largest health benefits. Start-up emission not significant and have a slight impact on the reduction has a much lower net cost than population-weighted exposure (PWE), while installing electric heating, so scenario 1 provides improvement costs are also relatively small. by far the largest net benefit value. There is a net benefit ratio of $14 million. As noted earlier, poor operation, even of newer Improved stoves and improved fuel (semi- boilers, could be a contributing factor to coked coal) (Scenarios 3 and 4) also produce higher than expected source contributions from substantial health benefits at fairly low cost. HOBs. The HOB (Scenario 6) and the dust Scenario 7: Reduction of road dust reduction (Scenarios 7 and 8) produce relatively suspension. Net benefit: $0 small health benefits, at fairly low cost, according to the methodology used for the assessment. The scenario envisions the paving of roads Dust from the extensive dry, uncovered to reduce fugitive dust. The costs include surfaces in UB is difficult to suppress under the maintenance of the roads. prevailing climatic conditions. The knowledge base regarding soil suspension and effects of abatement such as vegetation cover is weak. It The estimated health benefit is $67 million, seems worthwhile to look more closely at methods the same as the estimated costs. The net benefit to establish vegetation cover on uncovered thus is zero. surfaces in UB and study its effect to reduce dust suspension, as this is such an important PM10 Scenario 8: Greening of urban areas to source. prevent dust suspension. Net benefit: $0.7 million As noted earlier, factors other than cost- benefit considerations are important in the This scenario focuses on the greening of the process of selecting abatement scenarios, city by planting vegetation (e.g., a mix of including technical feasibility, affordability, simple vegetation, grass, bushes, and trees), implementation capacity, and social acceptance boundary marking, and protecting fences. It and strength of governance. includes reforestation, planting broad-leafed trees upstream of the Tuul and Selbe rivers, Conclusions greenbelt establishment, land reclamation, and so forth. This report presents the results of the implementation of a systematic approach by the The estimated health benefit is $58 million. AMHIB study to assess particulate air pollution in Given the relatively low cost of $2.5 million, the Ulaanbaatar and its effects. It evaluated costs and net benefit is $56 million. benefits of alternative abatement scenarios. 90 Cost-Benefit Considerations Socially acceptable and technically feasible governance associated with the abatement emission and pollution exposure reduction measure. targets are needed to guide the development of action plans. These targets are determined by the Due to the spatial distribution of available technical options and the ability and the population and UB’s extremely high willingness to pay for pollution reduction in the concentrations of pollution, short-term strategies society. The costs of air pollution are paid through to reduce air pollution as soon as possible could higher incidences of pollution-related illnesses, achieve large improvements in health for the and mitigation costs are covered from people’s population in a significant part of the city, even pocketbooks and the budget of the communities. though all parts of the city might not meet air What and how to pay for air pollution and its quality standards evenly. reduction is a choice to be made by civil society and its representatives. Due to the complex nature Additionally, the high peaks observed in daily of air pollution, an open discussion of options air pollution coincide with observed emission and their estimated impacts based on an analytical peaks from the ignition and re-loading phases of framework using best available data and analysis the burn cycle in ger stove heating. Because these methodologies is recommended. Cost-benefit peaks comprise a significant share of PM average analysis and estimating avoided health costs of concentrations in wintertime, it may be a sound each policy option, together with other factors strategy to focus on the ignition and reloading held important by Ulaanbaatar’s citizens, should phase of the burn cycle in abatement design. be considered in choosing clean air strategies. Laboratory testing of emissions during typical Setting targets that have been openly discussed burn cycles has confirmed this indication. helps build widespread support for pollution abatement activities that involve asking people The present assessment of the air quality to change environmentally damaging behaviors. situation in Ulaanbaatar and of the effects of Many in civil society, especially the poorest in some selected abatement measures has been based UB, will be asked to change their behavior in upon a wide range of existing data, reports and some way to improve air quality. They should information, as described in this report. The become active allies in the reduction of air assessment followed the basic concept for carrying pollution in UB. out air quality management work. This includes studying monitoring data for air pollutants and When faced with choices between proposed meteorology; carrying out emissions inventories abatement measures, policy makers need to and dispersion modelling; and calculating consider which criteria should be used to pollutant concentrations and their distribution prioritize measures. At the core of the local air spatially and temporally, and the contributions pollution abatement program is its ability to from the various main source categories. Such reduce pollution and the harmful effects it has calculations were carried out for the current on the population. The selection criteria could be situation—the AMHIB project used the period (a) the degree to which the abatement measure, June 2008–May 2009, the latest year with an or a package of measures, moves toward meeting extensive data base—as well as for the situation Mongolian or international air quality standards assuming that some selected abatement scenarios across all of Ulaanbaatar; (b) the net benefits of are implemented. Calculations also were made for abatement measures, or a package of measures, the reduction in population-weighted exposure taking into account of the value of health benefits to PM for a number of abatement scenarios, and versus abatement costs; (c) technical feasibility the corresponding benefits in terms of reduced of the measure; (d) affordability of the measure health costs. There are uncertainties related to this to the population and the government; (e) ease comprehensive chain of analysis. Uncertainties of implementation (how quickly can it have are related both to the data used as input to the an impact? Is it complicated to prepare and/ analysis, as well as to the analytical methods used or implement?); (f ) social acceptance (will it which correspond to state-of-the-art methods be accepted by society); and (g) strength of used by the scientific community worldwide. 91 Air Quality Analysis of Ulaanbaatar The use of two different methods to assess the particulate matter in UB must assume that the contribution to the PM exposure from the various health impacts of each unit change in pollution main source sectors, and the relatively good concentration is similar for each source. agreement of models with the PM measurements, lends credibility to the analytical results in The response of the population of terms of the population’s exposure to PM, its Ulaanbaatar to questions about their health and, spatial distribution, and the distribution of the ultimately, their expressed willingness to pay contributions from the various sources. Compared to lower these risks are not inconsistent with to other similar assessments, an important what members of the study team have seen in strength of the current study is the health benefit other countries. In all countries that have taken assessment, where local studies of the link surveys of the type administered in Ulaanbaatar, between air pollution and hospitalization rates including China and Japan, the public has and of the local population’s willingness to pay for evidenced a strong aversion to bearing mortality reducing mortality risks were carried out during risks and is willing to pay a significant fraction the course of the study and the results applied in of their income to reducing such risks. The the analysis. Uncertainties have to some extent estimates of WTP as a fraction of household been quantified. Data quality has been improved income are high in Ulaanbaatar compared to since 2009 when the preliminary analysis for China and most developed countries, but income Ulaanbaatar was published as a discussion paper underreporting may be a more serious problem in (World Bank 2009). Ulaanbaatar, which would inflate this percentage. The Ulaanbaatar population does appear to have While the health and environmental less of an understanding of risk than populations authorities are to be commended for undertaking in other countries. But whether people who a significant effort in recording air pollution understand risk are included or excluded from the data, there is a great need for additional data analysis does not have a significant effect on WTP. to be collected. Analysis of additional years of pollution data, especially for PM2.5 mass Based upon available data and the results and its constituents, would help confirm the of the comprehensive analysis presented in the epidemiological findings presented in chapter report, the following conclusions can be drawn: 4, which are based on only one year of data. In other cities around the world, previous studies ■ Ulaanbaatar is definitely one of the most examining the effects of daily exposure to air polluted cities, and it might be the most pollution on major outcomes such as mortality polluted city in the world in terms of annual and hospital admissions typically involve several particulate matter concentrations. Arguably, years of pollution and health data. Additional its severity is driven by extreme wintertime data in UB would ensure that the current findings PM concentrations (see chapter 3). were not due to chance alone and would provide ■ The population of UB is exposed to high additional statistical power to detect an effect concentrations of PM from different (i.e., find statistically significant impacts). Thus, sources and with different chemistry and additional data would reduce the uncertainty size fractions. The findings of the health in the epidemiology studies of health effects of effects study conducted as part of AMHIB air pollution in UB. In addition, measurement support conclusions that can be drawn of specific constituents (i.e., elemental carbon, from the health effects literature—that organic carbon, nitrate, metals, etc.) would there is a significant public health burden aid in conducting health studies on specific related to exposure to air pollution in UB sources of particulate matter. Specifically, the (see chapter 4). empirical association of each source with a given ■ The people in Ulaanbaatar are, similar to health outcome could be discerned. With this people in other cities and countries, willing, information, the efficacy of reducing specific when asked, to consider the value of reducing sources could be examined explicitly. Currently, the risks associated with high air pollution. the calculation of benefits of controlling A WTP of 159,000 tugrug for a 5-in-10,000 92 Cost-Benefit Considerations annual contemporaneous mortality risk at much higher costs. The SCC and electric reduction is concluded from the study. This heating solutions will require extended translates into a value of statistical life (VSL) subsidies to equalize heating costs (see of 319 million tugrug, or $221,000, based on chapter 7). the official exchange rate, or $493,000 based ■ Saved health costs are substantially larger on a purchasing power parity (PPP) exchange than the costs of the abatement for several rate (see chapter 5). of the abatement scenarios analyzed. Of ■ The needed effort to reduce air pollution the eight scenarios for which cost-benefit is considerable. A reduction of more than calculations have been carried out, the five 80 percent of the emissions across all four scenarios in which ger area emissions are main sources of air pollution (ger stoves, soil substantially reduced produce very large suspension, road traffic, heat-only boilers) health benefits in terms of avoided deaths, would come close to reaching the Mongolian with the present value of health benefits air quality standards in most of the UB city reaching $597–$1,803 million aggregated area, which are equivalent to the middle over the implementation period (2010–23). interim target set for developing countries by The short-term solutions produce substantial the WHO. To achieve WHO global guideline and immediate health benefits that aggregate values, the emissions reductions would need into very large benefit values over the to be about 95 percent (see chapter 6). A 2010–23 period (see chapter 8). combination of measures is thus needed to be ■ Five of these scenarios have a net benefit able to reach Mongolian air quality standards. (saved health costs minus abatement costs) ■ The annual cost of health damage in the range of $393–$1,635 million. This attributable to current levels of air pollution means that air pollution management in terms of particulate matter is estimated to in Ulaanbaatar can be carried out with a reach 18.8 percent of Ulaanbaatar’s GDP in substantial economic gain, when health costs 2008, or $463 million. Using an alternative are taken into account (see chapter 9). approach to health damage estimation, this amount could increase to 27.9 percent of It is recommended that policy makers Ulaanbaatar’s GDP or $687 million. The set targets such as the following and open a maximum achievable benefit from the discussion with civil society on the costs and described interventions (80 percent reduction benefits: in three sectors: ger area emissions, soil suspension, HOB emissions) is estimated to ■ Set targets that would reach Mongolian air equal 7 percent of GDP in Ulaanbaatar in quality standards for PM10 as soon as possible 2008, or $174 million (see chapter 6). and PM2.5 targets by 2020. ■ Different options for reducing emissions from ■ To achieve these targets, all main PM the ger heating systems have been proposed, sources in Ulaanbaatar need to be addressed. and their costs estimated. It is possible Reductions are needed primarily in the to reduce air pollution in the short term emissions from ger heating systems. Dust through a number of interventions, given suspension from uncovered soil surfaces, here according to increasing costs. These are: roads, and near-road surfaces are the second improved stove lighting methods, modifying most important PM10 source in UB. Its traditional stoves, or replacing traditional control is difficult, but should nevertheless stoves with cleaner stoves. The investment be pursued. Emissions reductions from costs are limited and will be more than the sources with less contribution to the compensated through a reduction of the fuel population’s exposure will also contribute to costs. Medium and long-term solutions— improved air quality. These sources include such as semi-coked coal in combination with HOBs, exhaust PM from road traffic, and SSC stoves, electric heating in the ger areas, CHP emissions. For CHPs, the obvious or moving into apartment buildings—will intervention is to improve the operation of also reduce air pollution by a large factor, but PM cleaning systems. The actual composition 93 Air Quality Analysis of Ulaanbaatar of the abatement scenario mix should be Air Quality Assessment guided by the cost-benefit considerations for the various options, as well as technical, ■ Air pollution monitoring: Monitoring with implementation and financial feasibility. the use of state-of-the-art monitors within ger ■ Recognizing socioeconomic constraints in areas should include at least PM2.5 and PM10, Ulaanbaatar, it is further recommended that SO2, NOX, and NO2. In this development, interim targets and reduction time lines be the suggestions presented by the AMHIB set for PM pollution reduction. If the team should be considered (see annex A). population exposure to PM could be reduced ■ Inventory of emissions: There remains a by 50 percent in a relatively short term, need to improve the existing emissions health costs of $86 million would be saved inventory including emission factors (EF) annually. To achieve this, it is recommended for the various sources, especially ger stoves; to target the ger areas immediately, where fueling practices (e.g. number and timing of pollution reduction benefits are greatest. fire starts); total amount of fuel burned per ■ Consider further the issue of soil suspension source category; traffic data on main roads; problem by better assessing the effect of and study of the soil suspension source. establishing vegetation cover, maintaining Population: the spatial distribution of the road/sidewalk/gutter surfaces, cleaning the population as used for this report should be roads and preventing dust from entering road confirmed, and the population growth and its surfaces. spatial distribution should be followed. ■ Install continuous emission monitoring systems in power plants to ensure better Environmental Damage Assessment operation of the flue gas cleaning systems. ■ Begin an open and candid discussion of ■ Studies of the health status of the population actual costs and benefits of abatement should continue. measures by (a) ensuring the abatement ■ Indications of the improved health status measures are technically feasible and their following abatement implementation should emissions reduction benefits are justified be studied. with sufficient evidence; and (b) appraising ■ The population exposure assessment should the full costs of abatement measures and be extended to take into consideration indoor their contribution to overall improvements air quality in ger areas. in ambient PM concentration reductions, so they can be compared to the health cost reductions. Abatement Options Assessment ■ Strengthen air quality monitoring and emissions inventories by providing sufficient ■ Several technical assessments have been operating budgets to key Mongolian air carried out recently. Studies of the quality institutions. performance of various ger stove designs, ■ Establish state-of-the-art pollution for various fuels, should be completed. monitoring stations within the ger areas as Large-scale demonstration projects need soon as possible. to be rolled out immediately to test proposed concepts, especially in ger areas The following is a summary of next steps where the success of abatement measures following the completion of this AMHIB study depends not only on technical effectiveness based on the basic steps in the Air Quality but also on socioeconomic and strong Management (AQM) process introduced in this cultural considerations. It is important to report. These recommendations are intended to systematically progress from demonstration improve the knowledge and analysis of the air projects to scale up as quickly as possible, but pollution situation in Ulaanbaatar, and the effects only when concepts show promising impacts. on improved health that should result from the ■ As outlined in chapters 7–9 of this report, abatement actions: some measures can be implemented over 1–2 94 Cost-Benefit Considerations years while others will take several years to ■ Strike an optimal balance between short, implement. Others may require feasibility medium, and long-term abatement study analyses reflecting the size of the measures. proposed investment and technical complexity. Optimal Control Strategy Cost-Benefit Analysis or Cost Effectiveness Analysis ■ Continue to set timetables and secure financing. ■ Cost-benefit analyses linking measures with ■ Establish a monitoring and evaluation results should continue to be updated. system that continually reports on air quality improvements and assesses impacts of Abatement Measures Selection abatement measures. ■ Indications of improved health status ■ Continued work should help to provide following abatement implementation should needed information to the government to be studied. develop and select abatement measures. 95 References Alberini, A., M. Cropper, A. Krupnick, and N. Brook, R.D., S. Rajagopalan, C.A. Pope, III, J.R. Simon. 2004. “Does the Value of Statistical Brook, A. Bhatnagar, et al. 2010. “Particulate Life Vary with Age and Health Status? Matter Air Pollution and Cardiovascular Evidence from the U.S. and Canada.� Journal Disease.� An Update to the Scienti�c of Environmental Economics and Management Statement From the American Heart 48: 769–792. Association. Circulation 121: 2331–2378. Alberini, A., M. Cropper, A. Krupnick, and Brown, Thomas C., Patricia A. Champ, Richard N. Simon. 2006. “Willingness to Pay for C. Bishop, and Daniel W. McCollum. 1996. Mortality Risk Reductions: Does Latency “Which Response Format Reveals the Truth Matter?� Journal of Risk and Uncertainty 32: about Donations to a Public Good?� Land 231–245. Economics 72(2): 152–66. Alberini, Anna, Alistair Hunt, and Anil Burnett. R.T.,J.R. Brook, T. Dann, C. Delocla, Markandya. 2006. “Willingness to Pay to O. Philips, S. Calmak, R. Vincent, Reduce Mortality Risks: Evidence from a M.S. Goldberg, and D. Krewski. 2000. Three-country Contingent Valuation Study.� “Associations between particulate- and gas- Environmental and Resource Economics 33(2): phase components of urban air pollution 251–264. and daily mortality in eight Canadian cities.� AirQUIS, 2008. Slørdal, L.H., H. Mc Innes, and Inhalation Toxicology 12 (Suppl. 4):15−39. Terje Krognes. The Air Quality Information Castillejos, M., V.H. Borja-Aburto, D.W. System AirQUIS. Norwegian Institute for Dockery, et al. 2000. “Airborne coarse Air Research (NILU), Kjeller, Norway. In particles and mortality.� Inhalation Information Technologies in Environmental Toxicolology 12:61–72. Engineering/ Volume 1, January 2008. Chestnut, L., R. Rowe, and W. Breffle. 2004. Aunan, K. and Pan, X.-C. 2004. Exposure- Economic Valuation of Mortality Risk response functions for health effects Reduction: Stated Preference Approach in of ambient air pollution applicable for Canada. Report for Health Canada prepared China—a meta-analysis. Science of the Total by Stratus Consulting. Environment 329 (1-3), 3-16. Cohen, A.J., et al. 2004. “Urban air pollution.� Banister, Judith, and Kenneth Hill. 2004. In M. Ezzati, A.D. Rodgers, A.D. Lopez, Mortality in China 1967-2000. Population and C.J.L. Murray, eds. Comparative Studies 58(1): 55–75. quanti�cation of health risks: Global and Boyle, K.J., and R.C. Bishop. 1988. “Welfare regional burden of disease due to selected major Measurements Using Contingent Valuation: risk factors, Vol. 2. Geneva: World Health A Comparison of Techniques.� American Organization. Journal of Agricultural Economics 70(1): Cropper, Maureen, and F. G. Sussman. 1990. 20–28. “Valuing Future Risks to Life.� Journal of 97 Air Quality Analysis of Ulaanbaatar Environmental Economics and Management Valuation Study.� Environmental and Resource 19: 160–174. Economics 33: 399–423. Desvousges, William H., V. Kerry Smith and Hammitt. J.K., and L.A. Robinson. 2011. “The Matthew P. McGivney. 1983. “A Comparison Income Elasticity of the Value per Statistical of Alternative Approaches for Estimating Life: Transferring Estimates Between High Recreation and Related Bene�ts of Water and Low Income Populations.� Journal of Quality Improvements.� EPA-230-05-83- Bene�t-Cost Analysis (forthcoming). 001. Washington, DC: Of�ce of Policy Hamermesh, D. 1985. “Expectations, Life Analysis, U.S. Environmental Protection Expectancy, and Economic Behavior.� Agency. Washington, DC. Quarterly Journal of Economics 100: Dockery, D.W., C.A.Pope III, X. Xu, J.D. 389–408. Spengler, J.H. Ware, M.E. Fay, B.G. Ferris, Hewitt, C.N. 2010. “Trees and urban air quality, and F.E. Speizer. 1993. “An association Birmingham as a case study.� Accessible at: between air pollution and mortality in six . 329:1753−1759. Holmes, T.P., and R.A. Kramer. 1995. “An Dominici, F., A. McDermott, S.L. Zeger, and S. Independent Sample Test of Yea-Saying and M.Samet. 2002. “On the use of generalized Starting Point Bias in Dichotomous-Choice additive models in time-series studies of air CV.� Journal of Environmental Economics and pollution and health.� American Journal of Management 29: 121–23. Epidemiology 156:193−203. Horioka, Charles Yuji, and Junmin Wan. 2006. Eftim, S.E., J.M. Samet, H. Janes, A. McDermott, “The determinants of household saving in and F. Dominici. 2008. “Fine particulate China: A dynamic panel analysis of provincial matter and mortality: a comparison of the six data.� National Bureau of Economic Research cities and American Cancer Society cohorts Working Paper Series 12723. Accessible at: with a Medicare cohort.� Epidemiology < http://www.nber.org/papers/w12723>. 19:209–216. Hurd, M., and K. McGarry. 1996. “The Franklin, M., A. Zeka, and J. Schwartz. 2007. Predictive Validity of the Subjective “Association between PM2.5 and all-cause Probabilities of Survival in the Health and and speci�c-cause mortality in 27 US Retirement Survey.� Stony Brook, NT: communities.� Journal of Exposure Science and Economics Department, SUNY. Environmental Epidemiology 2007 (17):279– Itaoka, K., A.J. Krupnick, M. Akai, A. Alberini, 287. M. Cropper, and N.B. Simon. 2005. “Age, Grantz, D.A. et al. 1998. “DustBusters reduce Health, and the Willingness to Pay for pollution, wind erosion…Though dif�cult to Mortality Risk Reductions: A Contingent achieve, revegetation is best way to stabilize Valuation Survey in Japan.� Discussion paper soil.� California Agriculture 52(4). 05-34. Washington DC: Resources for the Guttikunda, S. 2007. “Urban Air Pollution Future. Analysis for Ulaanbaatar.� Paper prepared for JICA (2009). Capacity Development Project for the World Bank. Washington, DC: World Air Pollution Control in Ulaanbaatar City, Bank. Mongolia Summary Report of 2nd JICA Haefele, M., R. Kramer, and T. Holmes. 1992. Detailed Planning Survey (March 15– “Estimating the total value of forest quality May 13, 2009) in high-elevation spruce-�r forests.� In Kamata, T., et al. 2010. Mongolia Enhancing The Economic Value of Wilderness. General Policies and Practices for Ger Area Development Technical Report SE-78. Research Triangle in Ulaanbaatar. Washington, DC: Sustainable Park, NC: Southern Forest Experiment Development Department, East Asia and Station. Paci�c Region, The World Bank. Hammitt, J.K. and Y. Zhou. 2006. “The Kan, H., Chen, B., Chen, C., Fu, Q. & Chen, M. Economic Value of Air-Pollution-Related (2004). An evaluation of public health impact Health Risks in China: A Contingent of ambient air pollution under various energy 98 Cost-Benefit Considerations scenarios in Shanghai, China, Atmospheric J.D. Kaufman. 2007. “Long-term exposure to Environment, 38 (1): 95–102. air pollution and incidence of cardiovascular Katsouyanni, K., G. Touloumi, E. Samoli, A. events in women.� New England Journal of Gryparis, A. Le Tertre, Y. Monopolis, G. Medicine 356:447–458. Rossi, D. Zmirou, F. Ballester, A. Boumghar, Ministry of Health (Mongolia), Public Health H.R. Anderson, B. Wojtyniak, A. Paldy, R. Institute (Mongolia), WHO. 2008. Braunstein, J. Pekkanen, C. Schindler, and Indoor Air Quality Survey (Survey Report). J. Schwartz. 2001. “Confounding and effect Ulaanbaatar. modi�cation in the short-term effects of Mitchell, R.C., and R.T. Carson. 1981. An ambient particles on total mortality: results Experiment in Determining Willingness from 29 European cities within the APHEA2 to Pay for National Water Quality project.� Epidemiology 12:521−531. Improvements. Draft report. Washington, Krupnick, A.J., A. Alberini, M. Cropper, N. DC: U.S. Environmental Protection Agency. Simon, B. O’Brien, R. Goeree, and M. Mitchell, R.C., and R.T. Carson. 1984. A Heintzelman. 2002. “Age, Health and Contingent Valuation Estimate of National the Willingness to Pay for Mortality Risk Freshwater Bene�ts: Technical Report to Reductions: A Contingent Valuation Study the U.S. Environmental Protection Agency. of Ontario Residents.� Journal of Risk and Washington, Dc: Resources for the Future Uncertainty 24: 161–186. Press. Krupnick, A., and A. Alberini. 2000. “Cost of National Oceanic and Atmospheric Illness and WTP Estimates of the Bene�ts Administration. 1993. “Natural Resource of Improved Air Quality in Taiwan.� Land Damage Assessments under the Oil Pollution Economics 76 (1): 37–53. Act of 1990.� Federal Register 58(10): 4601– Li, Y., M. Bai, W. Zhang, K. Yang, and X. Wang. 4614. 2002. “Analysis on the influence factors of OECD. 2000. “Ancillary Bene�ts and Costs of residents’ willingness to pay for improving Greenhouse Gas Mitigation.� Proceedings air quality in Beijing.� China Population, of an IPCC co-sponsored workshop, Resources and Environment 12(6): 123–126. Washington, DC, USA, March 27-29, 2000. Lopez, A. D., Shibuya, K., Rao, C., Mathers, Pope, C.A., Burnett, R.T., Thun, M.J., Calle, C. D., Hansell, A. L., Held, L. S., E.E., Krewski, D., Ito, K. & Thurston, Schmid, V. and Buist, S. (2006). Chronic G.D. 2002, “Lung cancer, cardiopulmonary obstructive pulmonary disease: current mortality, and long-term exposure to �ne burden and future projections. European particulate air pollution,� Journal of the Respiratory Journal 27: 397–412 DOI: American Medical Association, vol.287, no.9. 10.1183/09031936.06.00025805. Pope, C.A., M. Ezzati, D.W. Dockery, “Fine- Malig, B., and B. Ostro. 2009. “Coarse particles Particulate Air Pollution and Life Expectancy and mortality: evidence from a multi-city in the United States,� The New England study in California.� Occupational and Journal of Medicine, January 22, 2009. Environmental Medicine 66:832–839. Ready, R., J. Buzby, and D. Hu. 1996. Mar, T.F., G.A. Norris, J.Q. Koenig, et al. “Differences between Continuous and 2000. “Associations between air pollution Discrete Contingent Value Estimates.� Land and mortality in Phoenix, 1995–1997. Economics 72: 397–411. Environmental Health Perspectives 108: Ready, R.C., S. Navrud, and W.R. Dubourg. 347–53. 2001. How Do Respondents with Uncertain Mestl, H E S, Aunan, K., Fang, J, Seip, H M, Willingness to Pay Answer Contingent Skjelvik, J M, and Vennemo, H .2004. Valuation Questions? Land Economics 77(3): Cleaner production as climate investment— 315–26. Integrated assessment in Taiyuan City, China. Reaves, D.W., R.A. Kramer, and T.P. Holmes. Journal of Cleaner Production. 1999. Does Question Format Matter? Valuing Miller, K.A., D.S. Siscovick, L. Sheppard, K. an Endangered Species. Environmental and Shepherd, J.H. Sullivan, G.L. Anderson,and Resource Economics 14: 365–83. 99 Air Quality Analysis of Ulaanbaatar Reed, W.R. and J.A., Organiscak, 2007. tradeoffs for chronic bronchitis.� Journal of Haul Road Dust Control Fugitive dust Environmental Economics and Management characteristics from surface mine haul roads 21:32–51. and methods of control. Wang, H., J. Mullahy, D. Chen, L. Wang, and R. Rowe, R., W. Schulze, and W. Breffle. 1996. A Peng. 2001. “Willingness to pay for reducing Test for Payment Card Biases. Journal of the risk of death by improving air quality: a Environmental Economics and Management contingent valuation study in Chongqing, 31: 178–185. China.� Paper presented at the Third Slørdal et al. 2003, L.H., Walker, S.E., Solberg, International Health Economic Association S. The urban air dispersion model EPISODE Conference in York, UK. applied in AirQUIS2003. Technical Wang H., and D. Whittington. 2005. “Measuring description. Kjeller, Norwegian Institute for Individuals’ Valuation Distributions Using Air Research (NILU TR 12/2003). a Stochastic Payment Card Approach.� Smith, V.K., and W.H. Desvousges. 1987. An Ecological Economics 55: 143–54. Empirical Analysis of the Economic Value of Welsh, M., and G. Poe. 1998. “Elicitation Effects Risk Changes. Journal of Political Economy in Contingent Valuation: Comparisons 95: 89–113. to a Multiple Bounded Discrete Choice State Statistical Bureau of the People’s Republic Approach.� Journal of Environmental of China. China Statistical Yearbook Economics and Management 36: 170–185 . 2006. China Statistical Information and WHO (2005): Air Quality Guidelines for Consultancy Service Center. Beijing: 2006. particulate matter, ozone, nitrogen oxides U.S. EPA. 1995. AP 42. Fifth Edition, Volume I. and sulphur dioxide. Global update 2005. Chapter 13: Miscellaneous Sources. 13.2 Summary of risk assessment. http://eur- Introduction to Fugitive Dust Sources. lex.europa.eu/LexUriServ/LexUriServ. Accessible at: . http://www.epa.gov/air/criteria.html U.S. EPA, 1999. The Bene�ts and Costs World Bank. 2007. Cost of Pollution in China, of the Clean Air Act, 1990 to 2010, Economic Estiamtes of Physical Damages. USEPA,Washington, DC. Washington, DC: World Bank. U.S. EPA, 2006 & 2009. Regulatory Impact World Bank. 2009. “The Willingness to Pay Analysis for the Review of the Particulate for Mortality Risk Reductions in China.� Matter National Ambient Air Quality Discussion Paper Series draft prepared by Standards. Fed. Reg. 71(200) pp. 60853-54. Resources for the Future under a contract Oct. 17, 2006. (http://www.epa.gov/ttn/ecas/ with the World Bank, Washington DC. ria.html for full text of the RIA, accessed World Bank, ASTAE. 2009. “Heating in poor, Jan. 27, 2009). peri-urban Ger areas of Ulaanbaatar.� U.S. EPA 2008. Integrated Science Assessment for Zhang, X. 2002. Valuing mortality risk reductions Oxides of Nitrogen – Health Criteria (Final using the contingent valuation methods: Report). EPA/600/R-08/071. evidence from a survey of Beijing Residents Viscusi, W.K., W. Magat, and J. Huber. 1991. in 1999. Beijing, China: Centre for “Pricing environmental health risks: A Environment and Development, Chinese survey assessment of risk-risk and risk-dollar Academy of Social Sciences. 100 Introduction to Attached CD-ROM Annex A: Baseline PM Concentrations/Baseline as possible results from the different stations, in Ulaanbaatar June 2008 – May 2009 instrument inter-comparison campaigns were carried out during 4 different weekly periods The purpose of this annex is to present the during the year. The inter-comparison results measurements of particulate matter (PM10 and could be used to develop data correction PM2.5) that was carried out under the AMHIB procedures to arrive at a common basis for the study during the period June 2008–May 2009. data from most of the stations. There were 8 monitoring stations in operation under the AMHIB umbrella during the entire Still, these instrument problems cause period, although some shorter periods were uncertainty in the PM monitoring results. Due not covered at some stations due to instrument to the resulting uncertainty, it is judged to be the non-availability or operational problems. In correct procedure to present first in this Annex the addition, 4 German donor stations (called GIZ data as they were measured, without corrections. stations) were in operation starting in November/ Then, after applying the correction procedure, December 2008. the corrected data are presented and the PM concentration levels in UB under the AMHIB The results of the measurements provide a period is assessed. The uncertainty caused by the basis for assessing the ranges of concentrations instrument issue has been estimated. It is judged of PM10 and PM2.5 that were experienced in that the quality of the assessment of the PM UB during the AMHIB measurement period, concentrations in UB is acceptable, and gives an separately for Ger and non-Ger areas. adequate basis for assessing the health effects of the PM concentrations as well as to support the The instrumentation for PM measurements analysis of the effects of abatement measures. at the AMHIB stations was based on what was available in UB in 2008, owned and operated by The corrected results provide a basis for various institutions with mandate for and research comparison between measured and modeled interest in air pollution monitoring. As a result of concentrations using air pollution dispersion this, the stations were equipped with instruments models. The air pollution modeling efforts of different types and measurement principles. presented in Annex C provide a more complete Financial and time constraints prevented the picture of the spatial variations of the PM complete set up of a state-of-the-art monitoring concentrations in the UB area, and a more network with uniform instrumentation. complete basis for the assessment of health effects Preparations for the monitoring program revealed and effects of abatement measures. that the various instruments gave differing results, which was due to sampling and measurement The annex also contains recommendations artifacts of the various instruments. In order to regarding continuing air quality monitoring provide a basis for comparable and as correct in UB. 101 Air Quality Analysis of Ulaanbaatar Annex B: Identification and Apportionment of in the Discussion Paper (World Bank, 2009). New PM Pollution Sources by Receptor Modeling data has become available, and as a result of that the emissions inventory (EI) had to be updated. The purpose of Annex B is to present the results The Annex gives details on the updating of the of the source apportionment analysis that has been EI. Through the measurements carried out by the carried out based upon PM data from the three AMHIB study (see Annex A), there is now a better AMHIB stations 2 (NRC), 3 (Zuun ail) and 6 (3 basis for evaluating the model calculations, and Khoroolol). The Nuclear Research Centre (NRC) this is presented in the Annex. The Annex also of the National University of Mongolia (NUM) compares the results regarding source contributions has been participating in an international project to PM that come from the two methodologies: during many years, where particulate matter the statistical source apportionment based upon samples have been taken at the NRC station the elemental analysis of PM samples (described in and been subjected to multielemental analysis at Annex B) and the dispersion modeling described in the Institute of Geological and Nuclear Sciences this Annex, and develops conclusions on this issue. (GNS) in New Zealand. Receptor modeling has also been carried out on the data, lately using The dispersion model calculations provide the PMF methodology (see below). Under the the spatial distribution of the PM concentrations AMHIB project, this activity was enhanced such across the 30x30 km2 grid used, on an hourly that PM samples suitable for receptor modeling basis. When validated by the measurements, this were taken also at the AMHIB stations 3 and 6. spatial distribution provides the basis for the calculation of the exposure of the population to The annex presents the methodology and the PM pollution, e.g. in terms of the population detailed results of the source apportionment (SA) for averaged exposure. The methods and results of each of the PM size fractions sampled at the stations this calculation is presented. This population (fine PM fraction - PM2.5, coarse PM fraction - exposure figure is the input to the estimation of PM2.5-10, and the sum of fine and coarse - PM10). the health effects of the PM pollution in UB. The results provide a separate basis for Annex D: Estimating the Effects of Air Pollution assessing source contributions to PM in on Mortality and Hospitalization in Ulaanbaatar Ulaanbaatar. They can also be studied together with results of source contributions that can In this annex, the results of the analysis of the be extracted from the air pollution dispersion association between air pollution and both modeling work that has been carried out for mortality and morbidity in UB are presented. Ulaanbaatar as a part of the AMHIB project, Similar epidemiological studies, conducted over based upon an inventory of PM emissions the last decade on five different continents, have from the sources in Ulaanbaatar and upon demonstrated associations between short-term meteorological data. Annex C describes the (i.e., daily) changes in air pollution and premature dispersion modeling work and the comparison death. Many other adverse health outcomes from with the SA results from the receptor modeling daily, multi-day, or long-term changes in ambient described here in this Annex. air pollutants, including PM, have been reported. Annex C: Air Pollution Dispersion Modeling PM is a mixture of liquid and solid particles for Ulaanbaatar and Assessment of Source of different chemical constituents and sizes. PM Contributions is typically designated as either PM10 or PM2.5 (particles less than 10 or 2.5 microns in diameter, The purpose of Annex C is to present the basis for respectively) or as the difference between PM10 and the dispersion model calculations, the models, and PM2.5 (known as �coarse particles� or CP). PM2.5 the results. The dispersion models were used also (known as �fine particles�) is generated from many in the preliminary assessment of concentrations sources, including fuel combustion by mobile and contributions from sources that was presented sources (cars, trucks and buses), stationary sources 102 Introduction to Attached CD ROM (power plants and industrial boilers), and residential PM2.5, with a 0.8% (95% confidence interval (CI) sources (home heating and cooking). CP can also be = 0.5, 1.1) increase in daily total mortality per every generated by mechanical grinding during industrial 10 µg/m3 change in PM10. processing, by construction debris, and by natural sources such as sea salt and blowing dust. Fine Since this effort, several other multi-city particles are often thought to be more toxic on a studies have been published for both PM10 and weight-adjusted basis than coarse particles, since PM2.5. For example, in a study of 10 USA cities, they are more likely to penetrate deeply into the Schwartz (2000) examined the daily effects of lung. However, the evidence is somewhat mixed PM10 and reported that a 10 µg/m3 change in and may depend on the concentrations and the PM10 (measured as a two-day average of lag 0 and patterns of exposure and population characteristics. lag 1) was associated with a 0.7% increase in all- The various particulate matter metrics – including cause, daily mortality. In another multi-city study, PM10, PM2.5, black smoke, and sulfates – appear Burnett et al. (2000) analyzed total mortality data to show fairly consistent associations with both for 1986–1996 from the eight largest Canadian premature mortality and morbidity. The latter cities and found that both PM10 and PM2.5 were includes outcomes such as hospital admissions, associated with daily mortality. For PM10, a 10 emergency room visits, heart attacks, asthma µg/m3 increase was associated with a 0.7% (CI exacerbation, respiratory symptoms, work and = 0.2, 1.2) increase in daily mortality. Another school loss, and reduction in lung function. study involving 29 European cities reported an association between daily mortality and PM10, Similarly, associations in epidemiologic with an overall effect estimated at 0.6% per 10 studies have been observed between NO2 and µg/m3 (Katsouyanni et al., 2001). Dominici et al., both mortality and morbidity. The primary (2002) analyzed the 88 largest cities in the USA sources for NO2 are fuel combustion by mobile (NMMAPS) and found an association of about sources, and combustion of fossil fuels by 0.27% per 10 µg/m3 of PM10. Meta-analyses power plants, factories, and residences. The of earlier mortality studies suggest that, after epidemiologic studies indicating effects of PM converting the alternative measures of particulate and NO2 are also supported by findings from matter used in the original studies to an equivalent toxicological and clinical studies. PM10 concentration, the effects on mortality are fairly consistent. A recent meta-analysis of Because of data limitations, most studies to European studies suggested a mean increase of the date have examined the effects of relatively short- risk of 0.6% per 10 µg/m3 PM10 (WHO, 2004). term exposure. Specifically, time-series or case In addition, a meta-analysis of Asian studies crossover studies examine the correlation of daily indicated a mean increase of the risk of 0.4% to changes in air pollution, typically over several years, 0.5% per 10 µg/m3 PM10 (Wong et al. 2008). with daily changes in mortality. These studies control for other potential confounding factors that More recently, data on PM2.5 have become vary over time and may be associated with mortality, available to support analyses of effects on health. so that an independent effect of pollution can be For example, Ostro et al. (2006) analyzed nine quantified. With increasing statistical sophistication, large counties in California and reported an effect these studies have shown that either one-day or of 0.6% increase in mortality (CI = 0.2, 1.0) per multi-day PM average concentrations are associated 10 µg/m3 PM2.5. Fine particles were also associated with both total mortality and cardiopulmonary with cardiovascular and respiratory mortality, mortality. Among the first of the multi-city studies as well with all-cause deaths for those above age on mortality, Schwartz et al. (1996) examined 65. In a study of 25 U.S. cities, Franklin et al. data from the Harvard Six Cities study. This (2007) found an effect of 1.2% (CI =0.3, 2.1) database included monitors sited specifically to for a similar change in PM2.5. Finally, in a study support ongoing epidemiological studies and to of 112 (for PM2.5) and 47 (for PM10) U.S. cities, be representative of local population exposures. Zanobetti et al. (2009) reported effects of 1% (CI Consistent associations were reported between daily = 0.8, 1.2) and 0.5% (CI = 0.2, 0.7) for 10 µg/m3 mortality and daily exposures to both PM10 and changes in fine and coarse particles, respectively. 103 Air Quality Analysis of Ulaanbaatar It is important to note that much larger in Ulaanbaatar, Mongolia. The survey includes effects have been detected from the few studies both contemporaneous and latent risk that have examined long-term exposures to PM reductions of a magnitude typically achievable on cohort survival. In this type of study, a sample through clean air policy. The study looks at of individuals are selected and followed over time. mortality risk reductions of the magnitude For example, Dockery et al. (1993) followed and types typically resulting from air pollution approximately 8,000 individuals in six cities control policy. While the prime objective with in the eastern USA over a 15-year period (the the study is to estimate the willingness to pay Harvard Six Cities study); and Pope et al. (1995) for mortality risk reductions in Ulaanbaatar followed mortality rates over a 7-year period in in order to support the work in the AMHIB approximately 550,000 individuals in 151 cities in project, an additional intention is to build a the USA. These studies used individual-level data more solid bridge for benefits transfer between so that other factors that affect mortality could developed and developing countries. The survey be characterized and adjusted for in the analysis. was conducted in winter 2010. Estimates of Once the effects of individual-level factors were willingness to pay passed external and internal determined, the models examined whether longer- scope tests. Study results imply a value of term citywide averages in PM (measured as PM10, statistical life of $221,000 (319 million tugrug) PM2.5 or sulfates) were associated with different for contemporaneous 5 in 10,000 annual risk risks of mortality and life expectancies. The reduction using the official exchange rate to estimated mortality effects of long-term exposure to convert tugrug to U.S. dollars. fine particles (approximately 7 to 13% per 10 µg/ m3 of PM2.5) are much larger than those associated Annex F: Air Pollution Abatement Options and with daily exposure. Importantly, these study Their Costs in Ulaanbaatar results imply large differences in life expectancy. Specifically, 24 µg/m3 difference in PM2.5 between As several abatement options applied in chapters the cleanest and dirtiest cities is associated with an 7 to 9 involve various forms of stove improvement almost 1.5-year difference in life expectancy for the and replacement alternatives, this annex first city populations (Pope, 2000). The difference for presents some information about the emissions people who actually died from diseases associated from stoves. Then the underlying data that has with air pollution was estimated to be about 10 been applied to estimate the performance for years. This is because air pollution-related deaths each option as outlined in chapter 7 (startup make up only a small fraction of the total deaths emissions (backlighting, clean stoves, Semi-cocked in a city. Since these earlier studies were published, coal plus clean stoves, electric heating, relocation several additional and supportive studies, involving into apartments, HOBs, road dust and greening) other cohorts, have been completed (Eftim et and references to the applied data are presented. al., 2008; Puett et al., 2008; Miller et al., 2007; Finally, the costs, savings and PM reduction for Ostro et al., 2010). A comprehensive review of the the different scenarios are presented for each year existing studies and a discussion of the underlying in the 15 years projected period. biological mechanisms that underlie these effects are provided by Brook et al. (2010). Annex G: Data Annex Annex E: The Willingness to Pay for Mortality Section 1 of Annex G shows results of inter- Risk Reductions in Ulaanbaatar, Mongolia comparison campaigns between instruments participating in the AMHIB measurement This annex reports results from a stated network. Section 2 presents the data preference survey designed to estimate the correction procedures resulting from the willingness to pay for mortality risk reductions inter-comparisons. 104 Environment and Social Development East Asia and Paci�c Region THE WORLD BANK 1818 H. Street N.W. Washington, D.C. 20433 USA Tel: (202) 473-1000 Fax: (202) 473-6391 Internet URL: www.worldbank.org, worldbank.org/eapenvironment THE WORLD BANK OFFICE ULAANBAATAR 5F, MCS Plaza Building Seoul Street 4 Ulaanbaatar 210644, Mongolia Tel: (976-11) 312.647; 312.654 Fax: (976-11) 312.645 Internet URL: www.worldbank.org.mn