52970 Sustainable Development--East Asia and Pacific Region 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 Pollution in Ulaanbaatar Initial Assessment of Current Situation and Effects of Abatement Measures December 2009 THE WORLD BANK Air Pollution in Ulaanbaatar Initial Assessment of Current Situation and Effects of Abatement Measures Discussion Paper December 2009 Sustainable Development Series: Discussion Paper Sustainable Development Department East Asia and Pacific Region THE WORLD BANK © 2009 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, NW Washington, DC 20433 USA December 2009 All rights reserved. This study was prepared by the Sustainable Development Department (EASSD) of the East Asia and Pacific Region, and was mainly funded by the Bank-Korea Environmental Partnership (BKEP) and the Netherland-Mongolia Trust Fund for Environmental Reform (NEMO). Sustainable development issues are an integral part of the development challenge in the East Asia and Pacific (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 findings, 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 Office 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 ACKNOWLEDGEMENTS ..................................................................................................................... xi EXECUTIVE SUMMARY ................................................................................................................... xiii INTRODUCTION: NATURE OF THE AIR POLLUTION PROBLEM IN ULAANBAATAR ........................................1 Geology, topography and population................................................................................................. 1 Air pollution sources: Types, importance, spatial distribution ............................................................ 3 Meteorological conditions and impact on the pollution situation ...................................................... 4 The air pollution problem in Ulaanbaatar: A descriptive summary .................................................... 4 1. THE ANALYTICAL FRAMEWORK FOR AIR QUALITY MANAGEMENT IN UB .........................................7 Air quality management concept ....................................................................................................... 7 Air pollution and population exposure assessment, and impact on public health ............................... 9 Assessing the effects of abatement measures ..................................................................................... 10 2. THE MAIN PROBLEMS IN AIR POLLUTION ASSESSMENT IN UB .....................................................11 Monitoring system weaknesses ........................................................................................................ 11 Incomplete emissions inventory ...................................................................................................... 12 3. PRELIMINARY RESULTS OF ASSESSMENT WORK IN 2008 ..............................................................13 Extent of the assessment work for this Discussion Paper .................................................................. 13 Summary of present air quality situation ......................................................................................... 13 Current air quality situation in UB, assessed from pre-AMHIB data ............................................... 14 SO2 and NO2 data from the CLEM monitoring network in UB, 2001­2006 ......................... 14 Particulate matter (PM) concentrations measured by NUM, NAMHEM and NILU .............. 15 Source contributions to PM assessed from measurements ........................................................ 21 Sulphur dioxide (SO2) concentrations ..................................................................................... 21 The AMHIB baseline monitoring: Results and quality assessment ................................................... 23 The AMHIB monitoring stations ............................................................................................ 23 Results from the AMHIB PM measurements June­December 2008 ....................................... 23 AMHIB data quality assessment .............................................................................................. 27 Air pollution health effects in UB: Current situation and ongoing study ......................................... 27 International comparison of UB air pollution.................................................................................. 29 Emission inventory used for the preliminary air pollution assessment in this paper ......................... 29 The main air pollution sources in UB ..................................................................................... 29 Introduction to the development of an emissions inventory for assessment and modelling purposes........................................................................................................... 34 Summary of the emissions inventory used for the present assessment work ............................. 35 iii Air Pollution in Ulaanbaatar Modelled current spatial pollution distributions, and model evaluation ........................................... 40 Methodologies and modelling tools ......................................................................................... 40 Modelled current air pollution levels and distribution in UB ................................................... 42 Evaluation of the air pollution dispersion model ..................................................................... 50 4. ABATEMENT SCENARIOS AND THEIR BENEFITS IN TERMS OF REDUCED HEALTH COSTS ...................57 Abatement scenarios and reductions in PM pollution levels compared to Mongolian and International Air Quality Standards (AQS) ............................................................................... 57 Abatement scenarios, spatial impacts and reductions in health costs ................................................ 59 A 75%Ger/83%HOB/50% dust scenario--illustration of spatial impacts of abatement measures .................................................................................................................................. 59 The 30%/50%/80% scenarios--illustration of health impacts of abatement measures ............ 60 CONCLUSIONS ..............................................................................................................................73 REFERENCES .................................................................................................................................77 APPENDIX A: AIR QUALITY STANDARDS AND GUIDELINES FOR PM AND SO2 ........................................81 APPENDIX B: CRITERIA AND SUGGESTIONS FOR AN IMPROVED MONITORING NETWORK FOR AIR POLLUTION IN UB ............................................................................................................83 APPENDIX C: AMHIB DATA QUALITY ASSESSMENT .........................................................................87 APPENDIX D: EXAMPLES OF PM CONCENTRATIONS IN CITIES WORLDWIDE .........................................91 APPENDIX E: PRELIMINARY EMISSIONS INVENTORY FOR ULAANBAATAR .................................................95 APPENDIX F: EMISSION FACTORS FOR COAL AND WOOD COMBUSTION IN SMALL STOVES AND BOILERS ......111 APPENDIX G: ULAANBAATAR TRAFFIC DATA....................................................................................113 APPENDIX H: AIR POLLUTION MODELLING IN UB IN AMHIB: METHODS, TOOLS AND MODEL EVALUATION ............................................................................................................117 FIGURES Figure 1: The administrative districts of Ulaanbaatar ................................................................... 1 Figure 2: Location and topography of Ulaanbaatar ...................................................................... 2 Figure 3: Typical winter air pollution situation in Ulaanbaatar (Guttikunda, 2007) .................... 4 Figure 4: The height of the mixing layer (daily average) in Ulaanbaatar (meters) as a function of time of the year......................................................................................................... 5 Figure 5: Wind speed and direction distribution (wind rose) for Ulaanbaatar, 2007, UB meteorological station (Ref: NAMHEM)...................................................................... 5 Figure 6: Concept for development of cost-effective Air Quality Management Strategies ............. 8 Figure 7: SO2 and NO2 measurements in Ulaanbaatar, December 2001­October 2006, at stations UB1­4. Daily average concentrations (g/m3). (Guttikunda, draft 2007) ...... 15 Figure 8: Summary of PM measurements at the NUM station, 2006­2007 (monthly averages) (Lodoysamba et.al., 2008) .......................................................................................... 16 Figure 9: Individual PM fine and coarse measurements at the NUM station, 2007 ................... 17 Figure 10: Individual PM2.5 and PM10 measurements at NUM station in 2007 and 2008............ 18 Figure 11: Results of PM10 monitoring using a beta absorption method instrument (Japanese type `KOSA'), for the period February 2007­January 2008......................................... 19 Figure 12: PM monitoring in Ulaanbaatar 17­21 November, 2008, using a GRIMM 107 PM monitor (g/m3). Meteorological data are from NAMHEM ......................... 20 Figure 13: Estimated contributions to the fine (PM2.5) and coarse fraction (PM10­2.5) of PM at the NUM station, based upon the 2006­2007 measurement series .............. 22 Figure 14: SO2 data for the CLEM stations UB1­4, daily averages, for 2007. ............................. 22 Figure 15: Locations of the PM monitoring stations of the AMHIB network .............................. 24 Figure 16: AMHIB PM2.5 measurements during June­December, 2008. Daily and monthly average concentrations ................................................................................................ 25 iv Table of Contents Figure 17: AMHIB PM10 measurements during June­December, 2008. Daily and monthly average concentrations ................................................................................................ 26 Figure 18: Invasion of various particle size fractions in the human lung ....................................... 28 Figure 19: Spatial distribution of ger household PM10 emissions, in km2 grids (tons/year) ........... 38 Figure 20: Spatial distribution of HOB PM10 emissions, in km2 grid (tons/year) ......................... 39 Figure 21: Distribution of the emissions of PM10 from the vehicular traffic in UB (tons/year/km2) .................................................................................................... 40 Figure 22: Spatial distribution of suspended PM10 from road traffic in UB. Left: paved roads. Right: unpaved roads. Note the different scales in the two figures ............................... 41 Figure 23: Modelled spatial distribution of SO2 in Ulaanbaatar, 2007 ......................................... 42 Figure 24: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 43 Figure 25: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ...................................... 44 Figure 26: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 44 Figure 27: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 45 Figure 28: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 45 Figure 29: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 46 Figure 30: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 ....................................... 46 Figure 31: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ....................................... 47 Figure 31-a: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ....................................... 47 Figure 31-b: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ....................................... 48 Figure 31-c: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ...................................... 48 Figure 31-d: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ...................................... 49 Figure 31-e: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ...................................... 49 Figure 31-f: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 ...................................... 50 Figure 32: SO2 concentrations at stations UB 1­4 ....................................................................... 53 Figure 33: Measured and modelled PM10 (daily average) at the NUM measurement site ............. 53 Figure 34: The highest annual average PM10 (top figure) and PM2.5 (bottom figure) concentration in any grid cell in UB, 2007 and for various abatement scenarios (g/m3) ..................... 58 Figure 35: Calculated reduction in the PM10 grid resulting from the interventions (g/m3) ......... 61 Figure 36: Calculated remaining PM10 concentrations in the grid after implementing all three interventions ...................................................................................................... 63 Figure 37: Calculated reduction in the PM2.5 grid resulting from the interventions (g/m3)......... 64 Figure 38: Calculated remaining PM2.5 concentrations in the grid after implementing all three interventions ...................................................................................................... 65 Figure 39: Population weighted PM10 (top figure) and PM2.5 (bottom figure) reductions resulting from the 30%/50%/80% scenarios ............................................................... 67 Figure B1: Locations of currently operating air quality monitoring stations in UB, approximate locations ................................................................................................. 84 Figure C1: Results from PM sampler and monitor comparisons in Ulaanbaatar, for PM10 and PM2.5, 2008 ........................................................................................... 88 Figure D1: PM10 in selected cities, 2000­2004 ............................................................................. 93 Figure D2: PM10 in selected cities, 2005 ....................................................................................... 93 Figure E1: Spatial distribution of ger household emissions in km2 grid cells, PM10, 2007 (tons/year) ......................................................................................................... 98 Figure E2: Time variation (from day to day) of ger emissions, based upon temperature variations....... 99 Figure E3: Time variation across the day of ger emissions, based upon heating and cooking practices ...99 Figure E4: Location of a number of HOBs in Ulaanbaatar (Guttikunda, 2007)......................... 101 Figure E5: Spatial distribution of HOB emissions, 2007 (tons/year) .......................................... 101 Figure E6: Locations of the three CHP plants in Ulaanbaatar (Guttikunda, 2007) .................... 102 Figure E7: Traffic flow on the main road network in Ulaanbaatar, classified inbroad ADT classes: <20,000 (yellow), 20,000­40,000 (green), 40,000­60,000 (red), >60,000 (purple) ...................................................................................................... 104 v Air Pollution in Ulaanbaatar Figure E8: Vehicle and road data for Ulaanbaatar (Guttikunda, 2007) ....................................... 106 Figure E9: Spatial distribution of the PM emissions from exhaust particles from the road traffic in UB, 2007 (tons/year) .................................................................................. 107 Figure E10: Spatial distribution of suspended PM10 from road traffic in UB. Left: paved roads. Right: unpaved roads. ............................................................................................... 109 Figure H1: The domain used for air pollution modelling in this work (1 1 km2 grid) ............... 119 Figure H2: Meteorological conditions during the modelling period ............................................ 121 Figure H3: SO2 concentrations at stations UB 1­4. Measured and modelled daily average concentrations, 2007 (g/m3) ................................................................................... 122 Figure H4: Measured and modelled PM10 (daily average) at the NUM measurement site ........... 123 TABLES Table 1: UB population by district, 2007 (total population, 2008) ............................................ 2 Table 2: PM concentrations in UB, measured by NUM at its monitoring station, monthly averages, 2004­2006 .................................................................................... 16 Table 3: Summary of PM measurement results carried out by NUM ....................................... 16 Table 4: PM concentrations in UB during the week of 17­22 November, measured with the GRIMM monitor. Daily and max hourly averages, g/m3 .......................................... 19 Table 5: Monitoring sites and monitoring units, their characteristics ........................................ 23 Table 6: Summary of the emissions inventory for Ulaanbaatar, 2007 (tons/year) For details, see also Figure 7 ........................................................................................ 36 Table 7: Summary of PM10 emissions inventory for coal and wood combustion sources, winter season 2006/7. Including estimate of uncertainty () ...................................... 37 Table 8: Maximum PM10 concentrations by source and spatial distribution, 2007 ................... 51 Table 9: Maximum PM2.5 concentrations by source and spatial distribution, 2007 ................... 51 Table 10: Measured and modelled annual average SO2 (g/m3) for stations UB 1­4, 2007 ........ 52 Table 11: Measured PM10 and modelled source contributions at the NUM station, 2007 (g/m3) .............................................................................................................. 55 Table 12: Population weighted average PM concentrations (PWE) in Ulaanbaatar, and reductions from the 75%Ger/83%HOB/50% dust intervention scenario (g/m3)...... 60 Table 13: Population weighted average PM concentrations in Ulaanbaatar, and reductions from abatement scenarios (g/m3) ....................................................................................... 66 Table 14: Exposure-response coefficients (% change in incidence of health effect per g/m3 PM10), baseline incidence rates, willingness-to-pay (WTP) for avoiding premature death (long-term effect) and chronic bronchitis, and Cost of Illness (COI) of hospital admissions ...................................................................................... 69 Table 15: 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 million USD) ........................................................ 71 Table A1: Various guidelines, standards and limit values for PM2.5 and PM10 .............................. 81 Table A2: Basis for WHO Air Quality Guidelines (AQG) and Interim Targets ........................... 82 Table A3: Various SO2 guidelines, standards and limit values, SO2 guidelines, standards, limit values (all numbers in g/m3)...................................................................................... 82 Table C1: Characteristics of the instruments in the AMHIB network ......................................... 89 Table D1: Cities with highest air pollution (average PM10 concentrations 2004­05 and 2006­07) in China .............................................................................................. 92 Table D2: PM10 in 340 cities in the US, 2004 ............................................................................. 94 Table F1: Emission factors for coal and wood, used by Guttikunda (2007) .............................. 111 Table G1: Results from traffic countings at selected street sections in Ulaanbaatar..................... 114 Table G2: Traffic volume in streets in Ulaanbaatar .................................................................... 115 Table H1: Measured and modelled annual average SO2 (g/m3) for stations UB 1-4, 2007 ....... 122 Table H2: Measured PM10 and modelled source contributions at the NUM station, 2007 (g/m3) ............................................................................................................ 124 vi Foreword F or people living in Ulaanbaatar (UB) it bring the understanding of the severity of the has for quite some time been experienced situation forward. that the city unfortunately has high air pollution concentrations, particularly We would particularly like to emphasize that during the winter months, which also have the study has been initiated and shaped by our had severe impact on human health. Following Mongolian counterparts, which have left us with the drastic expansion of Ger areas surrounding a deep impression about their sincere wish for the traditional city centre, air pollution levels solving UB's critical air pollution challenges as apparently have increased. well their professionalism with regard to both air quality and health impact subjects. With these In order to get a better understanding of capacities we are convinced that UB one day will and responses to such questions like how high become both a clean and green place to live. are the pollution concentrations particularly in expanding Ger areas, what is the exact impact Ede Jorge Ijjasz-Vasquez on human health and what would the most Sector Manager cost effective interventions be, the World Bank China & Mongolia Sustainable Development Unit initiated in 2008 an "Air Monitoring and Health The World Bank Impact Baseline" (AMHIB) study based upon recommendations from several Mongolian Arshad M. Sayed counterparts. The study, which is part of a larger Country Manager UB Clean Air Program, also intends to provide Mongolia a reference (baseline) upon which the effect of The World Bank future intervention could be measured within such a program. Magda Lovei Sector Manager This discussion paper provides the Social, Environment and Rural Unit information from the first half of the AMHIB East Asia and Pacific Region period from June 2008 to May 2009. A The World Bank final report will be presented following the completion of the work in early 2010. Nevertheless, the findings in this paper clearly vii Acronyms ADB Asian Development Bank AMHIB The Air Monitoring and Health Impact Baseline Study AQ Air Quality AQG Air Quality Guidelines AQLV Air Quality Limit Values AQM Air Quality Management AQS Air Quality Standards BKEP Bank-Korea Environmental Partnership CHP Combined heat and power stations CLEM Central Laboratory for Environmental Monitoring EBRD European Bank for Reconstruction and Development EF Emission Factor EI Emissions inventory ESP Electrostatic Precipitator EU The European Union GDP Gross Domestic Product GTZ Deutsche Gesellshaft fur Technische Zusammenarbeit, GmbH HOB Heat only boilers JICA Japanese International Cooperation Agency NAMHEM National Agency of Meteorology, Hydrology and Environmental Monitoring of Mongolia NEMO The Netherland-Mongolia Trust Fund for Environmental Reform NILU Norwegian Institute for Air Research NO2 Nitrogen dioxide NRC Nuclear Research Center (of the NUM) NUM National University of Mongolia PM Particulate matter. Suspended particles in air PWE Population Weighted Exposure RH Relative humidity in air SO2 Sulphur dioxide TSP Total Suspended Particulate UB Ulaanbaatar US EPA Environmental Protection Agency of USA WB World Bank for Reconstruction and Development WHO World Health Organistion WTP Willingness-to-pay ix Acknowledgements T he AMHIB has brought together Gailius Draugelis, Senior Energy Specialist and Mongolian air pollution scientists, coordinator of the UB Clean Air Program, and leading air pollution monitoring Jostein Nygard, Senior Environmental Specialist officials, and public health experts to and lead manager of the AMHIB activities. take a synergetic approach of linking public health The primary authors of the analysis and draft and air quality issues. This Discussion Paper is report are Steinar Larssen, Li Liu, Sereeter built up from the work on air pollution analysis Lodoysamba, A. Enkhjargal, Kristin Aunan, performed by World Bank Consultant Dr. Sarath and Gailius Draugelis. Updates of the report are Guttikunda. Parts of his 2007 report are included made by Jostein Nygard and Steinar Larssen. in this report with updates. Dr. Bruce Denby of NILU contributed to model testing. Ms. Shawna Fei Lee, World Bank Junior This Discussion Paper greatly benefited Professional Associate, also contributed to the from cooperation and input of a wide range of description of the overall project as well as the stakeholders including the AMHIB Steering cover photo. In addition, the Discussion Paper Committee led by the Ministry of Nature, was peer reviewed by Jan Bojo, Lead Economist, Environment and Tourism and including, East Asia Operations and Policy Unit, Sustainable NAMHEM, the UB city administration and Development Department, East Asia and Pacific Ministry of Health. In addition to the co-project Region; Paul Procee, Environmental Specialist, managers of this activity Gailius Draugelis, Environment Unit, Latin America and Caribbean Senior Energy Specialist and Jostein Nygard, Region, World Bank; Alan Krupnick, Senior Senior Environmental Specialist, the AMHIB Fellow and Director, Quality of the Environment, team comprises Tumentsogt Tsevegmid, Senior Resources for the Future; and Mr. Taizo Yamada, Infrastructure Officer, World Bank, and experts Senior Advisor, Environmental Management from the National University of Mongolia, the and Planning, Japan International Cooperation Public Health Institute of Mongolia, and the Agency (JICA). The report was designed and Norwegian Institute for Air Research (NILU) and typeset by Shepherd Inc. Steinar Larssen, consultant and formerly with NILU. The AMHIB team also wishes to thank This Discussion Paper is part of a World the JICA air quality monitoring experts for their Bank response to the Government of Mongolia's kind support and peer review. Comments have request to mobilize a wide range of resources to been received from several stakeholders and would develop and support abatement measures for air like to acknowledge contributions from Messrs. pollution in UB. The wide range of activities is Robert van der Plas, Household Energy Specialist, called the UB Clean Air Program, and includes and Crispin Pemberton Pigott, Household Stove support from the Korean Environmental Specialist--both World Bank consultants. Management Corporation especially for this Discussion Paper and its related project, the Air This Discussion Paper was prepared Monitoring and Health Impact Baseline study. The under the guidance of co-Task Managers World Bank also contributed resources toward xi Air Pollution in Ulaanbaatar this study. Additionally, the Government of the have all contributed to various activities under Netherlands (through trust funds managed by the the UB Clean Air Program. The Bank has also World Bank and through a separate Government- been asked by the Government to prepare an executed, Bank administered trust fund project investment project--the proposed UB Clean "NEMO"), Japan (through a Government- Air Project. In addition, the Bank assists the executed, Bank administered trust fund project Government in raising awareness among external "Capacity Building for the Development of and internal financiers of opportunities to support Carbon Financing Projects in Mongolia"), and abatement measures and is working closely with Korea (through a Bank executed trust fund ADB, EBRD, GTZ, JICA, and many other project "Korea-Bank Environmental Partnership") partners. xii Executive Summary T he Discussion Paper presents an air pollution assessment to evaluate, prioritize preliminary findings for and measure results. dissemination to initiate a discussion on the short-term and This Discussion Paper lays out the long-term emission reduction strategies for approach that the ongoing World Bank study reducing UB's air pollution, given its changing will follow through to its conclusion in early demographics--growing population and a 2010. The World Bank and its study partners growing urbanization. This Discussion Paper invite comments on this approach and analysis assesses the current air quality situation in UB, so that it can be as helpful as possible to the based on available data from 2006 to 2008, and scientific community in air pollution analysis estimates effects of pollution abatement options and to Government for action planning. The on ambient concentrations of particulate matter World Bank launched with its partners a technical (PM). Population exposure estimates are used assistance project that intends to introduce a to assess current health damage attributable to systematic way of evaluating abatement options. air pollution in the city and the health benefits The AMHIB Technical Assistance Project that can be achieved from implementing the will develop a baseline for Particulate Matter abatement options. This World Bank Study (PM), source contributions and health effects launched a collection of PM data in seven and prepare a dispersion model that estimates monitoring stations in UB in July 2008. The final effects of abatement options. This Discussion paper is due in early 2010 and will incorporate Paper introduces the approach and model using more recent pollution measurements. data collected by Mongolian air quality experts, including data collected during the first half of Local authorities and central Government AMHIB. need to develop a well-defined process of action planning to reduce air pollution in UB, Air quality data in UB is improving, but preferably based on systematic analysis and there are significant shortcomings in data built on existing institutional frameworks. quality that should be addressed. Some of the There has been much debate on the most cost- data used in this Discussion Paper is collected by effective way to improve the UB's air quality government agencies and research institutions to acceptable levels, which pollution sources to prior to the start of the World Bank AMHIB target and the need to develop and implement an study,1 while other data presents the first seven effective action plan. The inter-agency National Coordination Committee chaired by the Ministry of Mineral Resources and Energy and vice-chaired by UB city administration is an excellent platform 1 Collected in conjunction with the air pollution analyses carried out by World Bank consultant Dr. Sarath Guttikunda to develop a well coordinated process for action in the report titled "Urban Air Pollution Analysis for UB" planning, linking proposed policy options with (draft 2007). xiii Air Pollution in Ulaanbaatar months of data collected through the AMHIB relatively most severe air pollution problem study up to the end of December 2008.2 Air in UB. In terms of PM, UB is among the most quality monitoring devices vary in quality and polluted cities in the world. Measurements state of operation and the inventory of emissions carried out in UB show that PM is by far the from main sources needs more accurate emissions most serious air pollution component. It is factors through measurements. Due to data well documented that particles (primary PM10, quality shortcomings, the key findings presented PM2.5, and secondary PM due to SO2 and NOx in this Discussion Paper must be considered emissions) cause negative health effects when preliminary. The results from the modelling of inhaled by people. In UB the main sources of estimated effects of abatement options could ground level PM2.5 (fine particle) concentrations be subject to changes when a better emissions are primary carbonaceous particles from coal inventory is constructed. The AMHIB team, combustion for heating and cooking (Ger however, considers the model itself is well suited households) and industrial activities (heat-only- for UB. It is able to replicate the current air boilers and power plants). Suspension of dust pollution patterns observed in data collected from from streets and other surfaces contribute to larger UB's air quality monitoring stations. Better data yet also harmful coarse particles contributing on emissions and air quality monitoring that also up to 50% of total annual average PM10 extends to the ger areas, like the AMHIB network concentrations in one part of the city (NUM starting in June 2008 for the entire period to May station). It is noted that the largest emissions 2009, may result in a need for model adjustment sources may not be the largest contributors to the and improvement. Additionally, data collected ground level pollution people inhale. from air quality monitoring stations provides sufficient observations for the AMHIB team to Inhaling Particulate Matter (PM) can establish findings about air pollution levels and its severely affect the lungs and the heart. The severity. size of the particles determines the potential for serious negative health problems when inhaled. There is room for optimism as newly The figure below indicates that the smaller the refurbished monitoring stations have been particles are, the further down into the respiratory installed by CLEM as well as through a GTZ system and lung they are transported and program and additional state-of-art monitoring deposited. Particles in the PM2.5 fraction (having stations supported by the French Government diameters smaller that 2.5 µm) passes into the are expected soon. JICA is working on upgrading smaller bronchi and air sacks and a fraction of emissions factors and inventories for boilers and them are deposited there, while more coarse power plants. This should contribute significantly particles (in the size range between approximately to data quality and accuracy of analysis in the 10 µm and 2.5 µm) are deposited in the upper future and hopefully prior to the final AMHIB respiratory tract. Once deposited in large enough report, due in early 2010. amounts, the particles can cause health damage. Although pollutants such as SO2 also In the most polluted parts of the city, are higher than international standards, based on available data, annual average Particulate Matter (PM) is the largest and concentrations of PM10 are 2­10 times higher than Mongolian and International Air Quality Standards (AQS). The PM2.5 concentrations are less well documented by measurements but 2 Here PM monitoring data were collected from research available data indicate that the PM2.5 situation institutions and government agencies, including National is equally severe compared to AQS. Annual University of Mongolia (NUM), National Agency of average concentrations of PM10, measured at the Meteorology, Hydrology and Environmental Monitoring of Mongolia (NAMHEM), and Mongolian Central Laboratory National University of Mongolia (NUM) campus for Environmental Monitoring (CLEM). NUM also area to the east of central UB (the station with the provided a statistical assessment of source contributions to PM emissions based upon the chemical analysis of the dust longest series of PM measurements) were as high particles collected at its monitoring station in UB. as 141, 157 and 279 µg/m3 for 2006, 2007 and xiv Executive Summary Source: Guttikunda, draft 2007. 2008 respectively. Measurements at monitoring above 200 µg/m3 up to a few years ago, but PM stations in other parts of UB since June 2008 levels are coming down in Chinese cities. The under this AMHIB study give even higher annual average PM10 concentrations in European concentration levels. Based on this AMHIB study's and US cities are much lower, the highest levels modeling results, the concentrations are likely to are in the range 60­100 µg/m3 (although some be higher in the north-central areas of UB than desert cities in the US have higher levels), and in at the NUM station. These measured UB PM10 most cities the concentrations are below levels are 2­5 times higher than Mongolia's AQS 40 µg/m3. of 50 µg/m3, 5­10 times higher than the WHO Guideline Value of Wintertime air pollution drives annual 20 µg/m3, and 3­7 times higher than the average concentrations of PM to their very high European Limit value of 40 µg/m3. WHO has set levels. Hourly average concentrations observed interim target values realizing that the Guideline thus far are at least as high as 2,300 µg/m3, and Values cannot be met in the short term in many daily averages above 1,000 µg/m3 in the most developing countries. The highest interim target polluted parts of the city (AMHIB monitoring value is 70 µg/m3. Thus, the present PM10 level data). These episodes occur regularly and often in UB is at least three times higher than this throughout the winter periods, caused by the target in the most polluted areas of the city. special climatic and meteorological situation of The spatial distribution of the pollution is wide UB, and bring the annual average concentration spread across UB city and its surroundings. The to its very high level. Some industrial cities in PM2.5 concentrations are less well documented Europe still have maximum daily concentrations by measurements but limited samples taken in in the range 400­700 µg/m3 approaching those in November 2008 indicate the severity of the PM2.5 UB, but they occur on very few days of the year, situation. The measuring equipment used until unlike in UB. Most cities in the US and Europe the end of 2008 is affected by sampling artifacts, have much lower daily maximums. resulting in too low PM2.5 levels. A measurement campaign during the last part of November 2008 The wintertime episodes of extremely high provided parallel PM2.5 and PM10 measurements hourly and daily concentrations are likely to indicating that 50­60% of PM10 was in the PM2.5 represent the highest urban scale short-term fraction on those days. The PM2.5 concentrations peak PM concentrations anywhere, ranging reached as high as over 400 µg/m3 as daily average from 4­14 times Mongolian and international and maximum hourly levels of up to 1300 µg/m3. standards. During winter, the limited data available show that daily average concentrations Some cities in northern China and south Asia of PM10 reach at least 7 times the Mongolian still have high annual average concentrations, i.e. AQS for 24-hour average concentrations, 4 times xv Air Pollution in Ulaanbaatar the highest interim 24-hour target values for loading phases for heating stoves combined with developing countries and EU standards, and the poor meteorological dispersion conditions at 14 times the WHO Global Guideline Value. those hours. There is an indication that, as a During the winter, AMHIB data show daily short-term measure, significant reductions in average concentrations reach at least as high as emissions can be achieved from changing the 1,000 µg/m3. These episodes occur regularly way raw coal is lighted for heating.3 throughout the winter months, and bring the annual average concentration to its very high Socially acceptable, technically feasible level. emission reduction targets should be set to give a clear direction for action plans. Targets A large share of the PM10 concentrations come from these wintertime peaks that may 3 See Lodoysamba & World Bank's Heating in Poor, Peri-Urban correspond to the cold start ignition and re- Ger Areas (June 2009). Comparison of UB PM Air Pollution and International Standards 2006­2008 PM2.5 PM10 PM10 PM2.5 PM10 24 hour 24 hour Annual daily Annual daily Hourly Hourly Max. average average average average average average hourly Measured PM concentrations in UB (pre-final AMHIB) (all numbers in g/m3) 2006 141 2007 157 600 1700 2008 279 November 17­22, 2008 max. 350 max. 600 2300 1300 2500 Guidelines, standards, limit values (all numbers in g/m3) Mongolia AQS 25 50 50 100 WHO Guidelines (2005) 10 25 20 50 WHO Interim Targets IT-1 35 75 70 150 IT-2 25 50 50 100 IT-3 15 37.5 30 75 USEPA Air Quality Standards (2006) 15 351 -- 150 EU limit values 253 -- 40 502 204 1 7 days above 35 per year is allowed. 2 35 days above 50 per year is allowed. 3 To be met by 2010. 4 To be met by 2020. xvi Executive Summary will be determined by technical options and openly discussed helps build widespread support the ability and willingness to pay for pollution for pollution abatement activities that involve reduction by civil society. The costs of air asking people to change behaviors. Many in civil pollution are paid from the pocketbook, the society, especially the poorest, will be asked to budget and future health costs through higher change their behavior in some way to improve air incidences of pollution related illnesses. What quality. They should become active allies in the and how to pay for air pollution is a choice to reduction of air pollution in UB. This approach be made by civil society and its representatives. provides policy makers with realistic options for Due to the complex nature of air pollution, an developing air quality management strategies that open discussion of options and their estimated are suited to the current socio-economic situation impacts based on an analytical framework in Mongolia. using best available data is recommended. Cost effectiveness or cost-benefit analysis can be used To achieve Mongolian AQS, 80% emissions for each policy option. These estimates together reductions are needed. This can only be with other factors that are considered important achieved realistically in the long term. An to civil society can be considered in choosing emissions reduction strategy that sets ambitious clean air strategies. Setting targets that have been but realistic short-term targets is recommended. xvii Air Pollution in Ulaanbaatar Different sources of air pollution show different collecting data from hospitals in UB located close impacts on air quality. To meet the Mongolian to the air pollution monitoring stations. These standard in UB for the most harmful particulates, include 8 family and village hospitals, 7 district PM2.5, the model predicts more than 80% of hospitals, 1 ambulatory facility, and 3 tertiary emissions reductions are needed. The results also hospitals. Data on daily admissions connected show that different source sectors have different to respiratory and cardiovascular diseases, based impacts on air quality. Based on available data, on diagnoses, will be collected to perform reducing emissions in the ger areas by half yields statistical analysis with variables connected to PM an improvement in PM concentrations by about concentrations. The analysis should be completed one-third--much more than similar emissions before the end of 2009. reductions in other source sectors. In order to make preliminary estimates of The AMHIB has also started a process to associated health costs, the Discussion Paper uses evaluate the health effects associated with air the population weighted average PM concentration, pollution in UB. Part of the AMHIB project is in short the population weighted exposure Estimated health damage due to PM pollution in Ulaanbaatar and number of cases avoided due to interventions (in annual number of cases) Hospital All-cause admissions Hospital mortality Chronic (respiratory admissions (chronic) bronchitis disease) (CVD) 2007 (current health damage) 614 379 735 448 80% in all 504 308 664 406 30% in all 129 107 245 151 50% in all 244 184 411 252 30% ger stoves / 80% HOB 146 119 271 167 80% ger stoves / 30% HOB 308 220 487 299 30% ger stoves 83 71 165 102 50% ger stoves 149 121 277 170 80% ger stoves 273 201 446 274 30% HOB 18 16 39 24 50% HOB 31 28 65 40 80% HOB 51 45 104 65 30% power plants 1 1 1 1 50% power plants 1 1 2 1 80% power plants 2 1 3 2 30% suspended dust 18 16 38 23 50% suspended dust 30 27 63 39 80% suspended dust 49 43 101 62 xviii Executive Summary (PWE). When compared to annual average multitude of data, ranging from data on measured concentrations, the PWEs more accurately reflect concentrations, inventory of population and exposure of UB's population to air pollution of the emissions and their spatial and temporal by adjusting for the spatial distributions of the distribution, meteorology and dispersion data, pollution and the population. The calculated and cost data on abatement measures. PWEs were used as basis for calculating health effects and their reduction as a result of the The monitoring data available for this work reductions in emissions. The base year of this was limited to 4 stations for SO2 and NO2, and assessment is 2007 and the threshold for excessive one station for PM2.5 and PM10, until June 2008, air pollution is taken as the WHO Guideline when the PM measurements at 7 additional Value. In 2007, it is estimated that the maximum stations started under the AMHIB project. health costs associated with the air pollution in During this work, all the available monitoring UB correspond to US$147 million equivalent, data from before June 2008 has been utilised, representing 8.0% of UB's GDP and 3.8% of and their quality assessed. The monitoring data national GDP. A sensitivity analysis using an was used partly to assess the present air pollution income elasticity of Willingness to Pay of 0.5 levels, partly to validate the dispersion modelling rather than 1.0 gives a health damage cost of carried out to map the pollution situation as 4.1% of UB's GDP and 1.9% of national GDP or a basis for assessing the contributions to the $74 million per year. pollution levels from the various main sources. A PM monitoring mission was carried out, in Note on Data Quality order to assess the quality of the PM measuring and Modelling Approach instruments presently used in UB. Data from the AMHIB monitoring network from the period This is an interim report intended to generate June­December 2008 became available after discussion on how best to measure and the modelling work had been carried out. The evaluate air pollution reduction programs in modelling work is futher updated in the final UB. Assessments of air pollution and the costs AMHIB report. and effects of specific abatement measures on selected source categories, carried out according The existing preliminary EI developed to air quality management practices, require a under a previous World Bank consultant report xix Air Pollution in Ulaanbaatar was improved somewhat based upon additional out more fuel-stove testing. JICA is preparing available data. Some road traffic data have been technical assistance to reduce emissions. established, based upon a small traffic counting effort. This improved, but still preliminary EI is Modelling Approach The effects of used as input to the dispersion modelling carried abatement measures on the annual average PM10 out under this work. and PM2.5 concentrations are simulated on the following basis. The Bank team intends to work with the Air Quality and Health Impact Baseline The links between the emissions from (AMHIB) Study Steering Committee the various source categories and the resulting (comprising Ministry of Nature, Environment contributions to ground level concentrations and Tourism (chair), NAMHEM, CLEM, UB were investigated using air pollution dispersion Air Quality Department), JICA, GTZ and models, calculating hour-by-hour concentration others to improve data quality to strengthen the distributions in the UB urban area over an entire analysis in the Final AMHIB paper. Preliminary year. The results from the modelling compared results shared in May 2009 from an ongoing well with the available pre-June 2008 monitoring JICA study of emissions factors and inventories results. Maps have been produced showing for HOBs, which also included measurements of the spatial distribution of the contribution to CHPs and a ger stove differ from the emissions PM concentrations (PM10 as well as PM2.5) for factors used in this Discussion Paper. The 2007 from the following source categories: Ger preliminary JICA results would modify the source household stoves, HOBs, CHPs, road traffic contributions to ground level PM concentrations. vehicle exhaust, and suspension of dry dust from The CHP contribution would be larger, and paved and unpaved roads. Industrial process heat-only-boilers (HOB) and ger contributions emissions affecting the populated area are limited relatively smaller. Time is needed to fully evaluate especially during the winter period and have not these most recent findings. The Bank endeavours been included in the calculations. to work closely with JICA on improving the emissions inventory by the final AMHIB paper The main sources of the high ground level due in early 2010. PM concentrations in UB are the approximate 130,000 Ger household (2007) heating systems The emissions inventory (EI) used for this (stoves, heating walls and coal water heaters) that work includes the main source categories but it use raw coal and wood for heating and cooking, requires improvement. Emission factors (EFs) the about 250 HOBs in the city using raw coal, used are uncertain and would benefit from more and the suspension of dry dust from paved and measurements, especially for the coal and wood unpaved roads and other surfaces. CHP plants combustion in ger heating systems and HOBs, as have very large emissions of SO2 and PM, well as the suspension of dry soil dust from roads although their tall stacks limit their contribution and surfaces. to the ground level concentrations. The spatial distribution of the sources, with ger households Several donor initiatives are underway distributed in the areas around the city central which will help improve data quality. The GTZ area, the HOBs distributed more close to the installed several air quality monitoring stations centre and the road traffic more concentrated in in January 2009 that have recently produced the central urban areas, provide the setting for air reliable air quality data. In cooperation with pollution modelling as a necessary method for World Bank consultants, EBRD has conducted assessing spatially distributed contributions. fuel-stove tests that should help improve emission factor estimates for ger heating systems. ADB has The quality of the AMHIB data is further proposed Technical Assistance that will purchase evaluated in Appendix C. equipment to set up a local laboratory to carry xx Introduction: Nature of the Air Pollution Problem in Ulaanbaatar Geology, topography and population UB is located in a valley within the northeastern mountainous area of Khentij, with Mongolia's continental position in east central the highest peaks up to about 2,800 meters Asia and its high altitude (average height above (Figure 2). UB is about 1,300 m.a.s.l., at the foot sea level is about 1,400 m) gives the country of the mountain Bogd Khan Uul to the south of a generally cold and dry climate. Its extreme the city. The valley floor with its river Tuul runs continental climate is characterized by long, cold due east-west. The north-south width of the valley winters lasting 7­8 months (mid-September to is some 4 km at the UB central area. mid-May) and short, temperate and relatively wet summers. The annual average temperature The current population of UB (2008) is is presently typically around 0°C, making it the 1,030,000 inhabitants. This covers the 9 districts world's coldest capital city. The temperature has shown in Figure 1. increased over the last 60 years of temperature recording, from around ­3­4°C 60 years ago. Table 1 gives the population by district. Monthly average temperatures are typically This work concentrates on the 6 westernmost ­20°C for winter months such as January and districts, see Figure 1, with a total population of February, and night time temperatures can go as 930,337 (2007), about 90% of the total 9 district low as ­40°C. July, the warmest month, has an population. Figure 2 shows a gridded area which average of 15­18°C. Precipitation is scarce, about is the area of air pollution concentration modeling 200 mm annually, falling mostly during the short in this work. summer period, so winters are extremely dry. Figure 1: The administrative districts of Ulaanbaatar 1 Air Pollution in Ulaanbaatar Table 1: UB population by district, 2007 (total population, 2008) District Population District Population District Population Bayangol 160,818 Khan Uul 90,925 Bagakhangai 3,827 Bayanzurkh 211,614 Songino Khairkhan 211,056 Baganuur 25,731 Chingeltei 132,883 Sukhbaatar 123,041 Nalaikh 27,297 Total population, 2008: 1,030,000 Figure 2: Location and topography of Ulaanbaatar Upper figure: Red lined area: Central UB (no gers). Yellow lined areas: Ger areas. Lower figure: Shows topography, as well as water (blue), roads (green) and urban districts. 2 Introduction: Nature of the Air Pollution Problem in Ulaanbaatar Air pollution sources: Types, importance, spatial streets, and traffic jams are frequent. Traffic distribution countings indicate that the most trafficked road sections have more than 60,000 vehicles per day. The main sources of air pollutant emissions in The main road network is mostly paved, while UB are related to the demand for heating and small roads, especially in the ger areas, are mostly cooking, as well as the road traffic and industrial unpaved. Suspension of road dust, contributing activities. The dry ground conditions also significantly to airborne particle pollution, takes represent a pollution source, a depot of particles place from all roads, and especially from the that become airborne with turbulent action from unpaved ones. the wind as well as from vehicles traveling the roads. Greater UB has a fairly broad industrial manufacturing base: machine tools, cement, About half of the population live in bricks, pharmaceuticals, carpets, textiles and apartments, about 80% of them supplied with food processing. These industrial activities need central heating and hot water from 3 combined energy which is mainly provided by small size heat CHPs located to the west of the city centre. The only boilers and by district heating provided by rest of the apartments are supplied from heating the three large power plants. In terms of process boilers; a few are also heated by individual stoves. emissions, the brick industry is the main emitter, The rest of the population live in traditional with PM and SO2 emissions from the burning of Mongolian tents (gers) or small individual houses. coal and other combustibles in the brick kilns. These are all heated by small, inefficient stoves fed mainly with lignite coal and also wood. The main source of fugitive dry dust other than from roads is the suspension of dust from The apartment houses are located in the open soil surfaces. Most of the land surfaces central areas of the city, while the ger areas are in UB have no vegetation, and the dry soil is scattered around the city, mainly in the east- available for suspension by wind action most of north-west sector. The ger/house heating systems, the time because of the dry climate. The top soil there are some 130,000 of them within the UB layer is very fine grained, and dust is easily picked populated areas, represent the dominating source up by wind action. The magnitude of this source of ground level pollution in UB because of their is difficult to assess, but the source apportionment distribution throughout the ger areas, obviously methods used and shown later in the report close to people's quarters, and their low level indicates that this dust source is significant of emissions (short stacks from the gers, some in terms of contribution to airborne particle 2­3 meters above ground). The power plants, concentrations. with large emissions, affect the average ground level urban air relatively less because of their tall Other fugitive dust sources include dust from stacks (100­250 meters), although they give large construction activities (from the construction concentrations occasionally when their plumes itself as well as construction related traffic on, to hit the ground during unstable meteorological and from the construction sites. This is a source conditions. of considerable strength which is difficult to assess accurately, but it is relatively small compared to Road traffic is a significant source of air the continuous dry PM source that the roads pollution in UB. There were in 2007 about represent, hour-by-hour and day-by-day. The 93,000 registered vehicles in UB, of which 75% power plant ash ponds are another noticeable are light duty, 15% trucks and 7% buses. Car fugitive dust source in UB. This source is ownership is thus low, less than 0.1 per person, intermittent, active in periods of gusty winds, and close to 10 times lower than in Europe and the although it is in such periods a dominating source US. However, they operate on a limited road in their neighborhoods, it is not a significant network, so traffic volume is high on the main source in terms of annual emissions. 3 Air Pollution in Ulaanbaatar Figure 3: Typical winter air pollution situation in Ulaanbaatar (Guttikunda, 2007) As in most cities in Asia, the uncontrolled directions are aligned with the valley axis, burning of garbage and waste contributes to the although a bit skewed towards the northwest and air pollution problem. Such emissions are visible southeast. Low wind speeds, below 2 m/s, occur also in UB. close to 40% of the total time, providing frequent potential for high pollution levels. Meteorological conditions and impact on the pollution situation The air pollution problem in Ulaanbaatar: A descriptive summary With its location and climate, the conditions in UB are well positioned for creating a winter The contributions from the main sources and air pollution problem (see Figure 3) while the their spatial distribution causes the air pollution summer period is unproblematic in that respect. problem to be widespread across UB city and its Its population of about 1 million people depends surroundings. The ger areas are more affected upon coal and to some extent wood for the than the central parts of the city. generation of energy for almost all purposes, most importantly for the space heating needs, The emissions are dominated by as well as for cooking. The cold climate means a contributions from the coal burning and from large coal consumption for heating, the lack of the road traffic. The road traffic contributes both precipitation means conditions for suspension of with suspension of road dust as well as exhaust dust from the ground and streets, and the valley/ emissions. mountain topography creates periods of calm conditions with temperature inversions that trap The resulting air pollution problem is the pollutant emissions within a relatively thin characterized by very high concentrations of layer near the ground, thus creating very high airborne particles, PM, and by less severe sulphur ground level concentrations of pollutants that dioxide and nitrogen oxides levels. Measurements the population is exposed to. It so happens that carried out in UB (see Chapter 3) shows that low wind conditions typically occur in the late PM is by far the most serious component of the air winter afternoon/evenings at the same time as the pollution problem. Annual average concentrations heating and cooking demands are at its highest, a of PM10 (particles that are inhaled into the lungs combination that results in periods of extremely and causes health damage), measured at the high air pollutant concentrations. National University of Mongolia campus area to the east of UB centre area, the only station with a The wind direction and speed distribution long series of measurements, were as high as 141, for UB (Figure 5) shows that the main wind 157 and 279 µg/m3 for 2006, 2007 and 2008 4 Introduction: Nature of the Air Pollution Problem in Ulaanbaatar Figure 4: The height of the mixing layer (daily average) in Ulaanbaatar (meters) as a function of time of the year* *Dotted line: Average mixing height during winter season (December­February) for all years included. Source: Guttikunda, 2007. Figure 5: Wind speed and direction distribution (wind rose) for Ulaanbaatar, 2007, UB meteorological station (Ref: NAMHEM) Source: NAMHEM respectively (see Chapter 3). Measurements at (which is 20 µg/m3), and the European Limit several new stations since June 2008 (under the value (which is 40 µg/m3). (Appendix A). WHO AMHIB study) give an even significantly higher has set interim target values, realizing that the concentration level. These levels massively exceed guideline cannot be met in the short term in the Mongolian air quality standard (which is many developing countries. The highest interim 50 µg/m3), as well as the WHO guideline value target value is 70 µg/m3, so the present PM10 level 5 Air Pollution in Ulaanbaatar in UB is several times higher than that target in The air pollution also affects the visibility in the most polluted areas of the city. the city to such an extent that airplanes at certain occasions are prevented from landing at the city The dynamics of the air pollution situation airport. The PM particles are hygroscopic due in UB results in extremely high short-term to their contents of sulphur. SO2 from the coal concentrations during the winter, with daily and burning is adsorbed to and converted to sulphate hourly average concentrations reaching above on the particles due to the sometimes very 1,000 µg/m3 and 2,000 µg/m3 respectively. These high relative humidity (RH) on evening/night/ episodes occur regularly throughout the winter morning hours, due to the low temperatures. At months. Taking into account that at least 50% RH above 67%, the hygroscopic particles grow of this PM10 mass is derived from emissions from much larger due to water absorption, and this coal combustion, there is no doubt that the PM reduces the visibility considerably, sometimes so pollution in UB affects the health of the UB much that dense fog forms. population in general. The measured PM levels are among the highest, probably the highest, measured in urban areas globally (see Chapter 3). 6 1. The Analytical Framework for Air Quality Management in UB Air quality management concept This concept, which provides the analytical framework for the air pollution and control In order to successfully implement effective analysis carried out in this work on air pollution quality management to combat urban air in UB, is visualized in Figure 6. pollution in UB, it is important to understand the characteristics of the various pollutants prevalent Air quality assessments can use two in the city, their sources and their effects. An different modelling approaches, separate and in integrated approach to air quality management is combination. The dispersion modelling approach needed. builds on an Emissions Inventory (EI) of main stationary sources of air pollution. This then The integrated approach to air quality feeds a dispersion model that together with management enables an effective assessment of metereological and other statistical data can the air pollution levels, the contributions from model the effects of sources on annual and the various sources to ground level concentrations daily concentrations of major pollutants such as well as to the exposure of the population to as particulate matter (PM), SO2, NOX ozone damaging air pollution, and to develop cost- (03) and others. The receptor modelling approach effective abatement strategies1. collects monitoring data from air pollution monitoring stations and uses sophisticated filter The basic concept of Air Quality analysis to determine sources of pollution. Both Management includes: are then compared to evaluate their completeness and accuracy. The AMHIB study used both Air Quality Assessment; approaches and despite significant shortcomings Environmental Damage Assessment; in the emissions inventories and monitoring data, Abatement Options Assessment; attempted to systematically put together an air Cost Benefit Analysis or Cost Effectiveness quality assessment for UB. This is meant to kick- Analysis; start an ongoing process to ensure new analysis Abatement Measures Selection (Action plan); can be as complete and accurate as the data can and offer. Optimum Control Strategy. In order to develop an emissions inventory, 1 The integrated air quality management concept is described in the basic methodology for estimating emissions detail in the Urban Air Quality Management Strategy in Asia from industrial or household fuel consumption, Guidebook: http://books.google.com/books?id=9G0cd7d_nQ cEC&printsec=frontcover&dq=Urban+Air+Quality+ as well as traffic activity, is a standard formula Management+Strategy+in+Asia&source=bl&ots=9pWuVliGhw& applied universally regardless of emission type: sig=XzXl8UYXbqqwUoVpiod5aQ3lJE&hl=en&ei=63JOS _fmN5XklAeokoSODQ&sa=X&oi=book_result&ct= result&resnum=3&ved=0CBEQ6AEwAg# Emissions Activity Emission Factor 7 Air Pollution in Ulaanbaatar Figure 6: Concept for development of cost-effective Air Quality Management Strategies For industrial or household sources with sources to urban air pollution in UB. In order emission reduction and control technology such to target these sectors for effective air pollution as scrubbers, electrostatic precipitators, and reduction measures, it is necessary to translate the desulfurization devices, the following equation emissions contributions into the contributions applies: they make to the ground level air pollution concentrations, using modelling methods. The Emissions Activity Emission Factor emission inventory also establishes baseline (1 Efficiency in %) emission levels to track changes in emissions and ground level air pollution in response to Emission factors are either developed new developments and policy measures for air empirically through source testing or cited from pollution abatement. Additionally, with a baseline publications by authorities such as the United emission inventory and modelling capabilities, States Environmental Protection Agency, the the Government of Mongolia can evaluate European Environment Agency, the International and compare policy options in terms of their Energy Association, and accredited academic respective effectiveness to curtail future emissions, sources. The selection of good emission factors is thus facilitating the selection of the optimal crucial to compiling an accurate and dependable control option(s) for implementation. emissions inventory; otherwise, an unreliable emission factor could translate into large Data from the emissions inventory are discrepancies in the total emission estimates, and maintained in a database which is then fed into an further in the concentration assessment. air pollution dispersion model. To provide a valid and unbiased basis for modeling air pollutant Emission estimates from all stationary concentrations in the city of UB, the project team (including point and area) and mobile sources introduced factors such as spatial distribution of provide a relatively complete emissions inventory emission sources, temporal variation of emissions, for policymakers to identify major contributing and other physical characteristics of the sources 8 The Analytical Framework for Air Quality Management in UB (e.g. height of the release point, temperature, etc.) cases'), RR is the relative risk of health effect to render an accurate depiction of UB's air quality. between two levels of pollution (here the The modeling tool employed to perform the current level and a lower level obtained from an spatial and temporal distributions of ground level intervention or the lower threshold level), fp is the air pollution concentration analyses, as well as current incidence rate of the health effect, and the air quality management software used for the POP is the exposed population considered. Except analysis carried out is described in Appendix H. for the mortality function, where we rely on WB (2007) (except we use 20 µg/m3 as a threshold Air pollution and population exposure level instead of 15), RR is given by: assessment, and impact on public health RR exp( * (C Ct)) In addition to the physical damages done to infrastructure and degradation of the where is the exposure-response coefficient, C environment, air pollutants such as SO2 and is the current pollution level and Ct is the target PM also affect the well being of the public. A pollution level obtained from an intervention or wide selection of literature and studies have from reaching the threshold value. We calculate documented the association between elevated the remaining number of cases attributable to levels of urban pollutants and increases in air pollution after each intervention, and derive mortality rates, and illness related rates such as the number of cases that can be avoided by respiratory infections, the number and severity subtracting these figures from the calculated of asthma attacks, and the number of hospital excess cases in the current situation (which is admissions.2 The concentrations of pollutants calculated by using the threshold levels described in the atmosphere and the size of the affected above). To determine costs of air pollution, this population determine the magnitude of the Paper uses a willingness-to-pay methodology (see, impacts on public health. Both of these two e.g., WB, 2007) to monetize health impacts and factors are high in UB, especially during the estimate the economic value of avoided health winter season, and thus, could lead to serious damage. public health implications. Due to the considerable attention these The analysis presented in this Discussion preliminary calculations may have when Paper is restricted to the three major so-called disseminated, more background information is "health end-points"--premature deaths, chronic provided in Chapter 4 bronchitis and hospital admissions. Based on exposure-response functions3 from the literature, In air pollution analysis worldwide, a lot health impacts are derived using the equation of attention is paid to population exposure, below. i.e. what is the actual concentration level that people are exposed to. It is of importance to E ((RR 1)/RR) * fp * POP assess the specific contributions of each key pollution source to population exposure, or where E is the number of cases of each health the `source sector specific contribution to endpoint attributed to air pollution (`excess population exposure'. This is a good indicator for comparing the importance of each source to the health effects of the population, as opposed 2 OECD: 2000, "Ancillary Benefits and Costs of Greenhouse Gas Mitigation", Proceedings of an IPCC co-sponsored to just comparing emissions amount per source, workshop, Washington, DC, USA or even the average ground level concentration 3 Exposure-response functions measure the relationship between exposure to pollution as a cause and specific outcomes as an contributed by each source. The population effect. They refer to damages/production losses incurred in a exposure should ideally be calculated based year, regardless of when the pollution occurs, per unit change upon data on each person's movements within in pollution levels. The function is defined as the percentage change in effects incurred per unit change in concentrations the various parts of the city day-be-day, or (µg/m3) per capita. even hour-by-hour. Obviously, such data are 9 Air Pollution in Ulaanbaatar not available in Mongolia, and this detail of 3. Air pollution modelling (see section from population exposure assessment has hardly been p. 40 in this paper). A Eulerian grid carried out anywhere in the world. model was used, which has embedded subgrid models for calculation of pollutant In this Discussion Paper we use a population concentrations resulting from different types exposure assessment typically used when the data of sources (area-, line- and point sources). described above is unavailable, and that can be The model solves the time dependent supported by data available in UB. This is the advection/diffusion equation in a 3 dimensional population weighted average exposure, PWE, which grid. The model grid used for the UB sums up the average pollution concentrations modelling is 30 30 km, and the grid cell size in 1 km2 cells on a distribution map (a grid of used is 1 1 km (Appendix H). 1 km2 cells in the six central districts of UB) Input to the model includes the emission multiplied by the total number of persons in each inventory and meteorological and population cell and divided by the total population. Because data. The emissions are preprosessed to pollution and people are unevenly distributed provide hourly emissions in each of the grid across UB, their exposure levels are different, cells to the model. depending on where they live. The result of The model gives as output hourly this calculation is the exposure summarised as concentrations throughout the modeled one number, PWE, as representing the whole period (here: one full year) in each of the grid population. This PWE exposure number is based cells, as well as in specific points, such as the on the outdoor concentration in the grid cell locations of the monitoring stations. where each individual lives, and does not, as 4. Calculating the annual average population mentioned above, take account of the difference weighted exposure, total and contribution in exposure that people get when moving outside from each of the source categories, by their area when going to work and school, etc. A combining the air pollution and population time-activity pattern would need to be established distributions in the grid. which is unavailable. In UB this may be less important, since the highest concentrations occur Assessing the effects of abatement measures during early morning and late afternoon/evening hours when people tend to be mostly around In this work, the effects of abatement measures on home. The PWEs are used to calculate health the reduction of their impact on people's health is effects. carried out as follows (Chapter 4 of this paper): The steps carried out in this work to assess 1. The emissions from each abated source the exposure and the effects of abatement of category is reduced by a given percentage sources on the exposure are as follows: amount, as a result from the abatement measure. 1. Analysis of all available monitoring data. This 2. The resulting reduction in average population gives an overview of the current air pollution exposure is calculated based upon the problem, and a basis for evaluating the air dispersion modelling runs. pollution dispersion model (sections within 3. This is the basis for evaluating the cost Chapter 3). effectiveness or cost-benefit ratio of the 2. Establishing an emissions inventory, which abatement measure, when combined with includes all main sources: ger heating systems, abatement cost data and health effects heat-only-boilers, power plants, road traffic reduction data and their monetary value. (section from p. 29 in this paper). 10 2. The Main Problems in Air Pollution Assessment in UB Assessing with certainty the air pollution levels regarding data on emissions from the main in an urban area and the effects of abatement sources. measures requires quality data and information of several types: Monitoring system weaknesses Measurements of pollutant concentrations Air pollution has been measured for several years An emissions inventory covering all the main at a few stations. The formal measurements sources carried out by the government has covered Meteorological data as input to dispersion SO2 and NO2 at four stations for many years, modelling while PM has been measured at one station, Population data and its distribution by a National University of Mongolia (NUM) Technical and cost data related to abatement group since 2006. Data from these stations has options provided a valuable basis for assessment, but The assessment should cover the spatial there are questions related to the data quality aspects of the pollution situation (mapping), and suitability of samplers used for routine and should also cover the temporal scale: monitoring purposes to assess compliance with air concentrations and contributions both on quality limit values. short term (hour, day) and longer term averages (annual) Thus, a proper baseline for air quality does In addition to describing the existing not yet exist for UB. The AMHIB project was situation, the assessment should also look launched to assist in setting this baseline. towards the future in terms of scenarios for development with and without abatement There is a need for upgrading of the actions monitoring system for air pollutants, both in When the assessment should also cover terms of number of stations, their location and health effects of the pollution level, data are type (urban, traffic, industrial, etc) and choice required on the present health situation of the of methods and equipment (from the present population, and dose-response relationships manual methods to continuous state-of-the-art between pollutants and their effects, monitors). Appendix B presents criteria that can applicable to the local area and population be used to improve the monitoring network, as in question well as suggestions for an improved network to Valuation of health costs is needed to complement ongoing efforts by NAMHEM/ establish net benefits CLEM supported by programs financed by Germany and France. In UB, data and information are scarce on all these topics, and the accuracy and quality of Upgrading activities started during the last the data has often not been assessed, especially half year of 2008. Germany and France had 11 Air Pollution in Ulaanbaatar offered assistance in improving the monitoring draft 2007). The EI used in this Discussion in UB, while NAMHEM/CLEM where Paper is a combination of the NAMHEM EI and implementing their own improvements, by improvements carried out by Guttikunda and establishing more stations and updating the AMHIB. However, this EI needs to be further methods and equipment. improved and completed to provide a proper basis for air pollution modelling. Incomplete emissions inventory A number of project activities are presently The existing preliminary emissions inventory (EI) (2009) underway that will provide improved for UB does not meet internationally acceptable data for EI, e.g. HOB inventory and emissions standards. In many areas it is incomplete or measurements from HOBs, CHPs and Ger needs updating. Based on available inventories stoves carried out under a JICA project, as and data at that time, a preliminary EI has been well as an EBRD project on Ger stove and fuel developed as part of the World Bank activities improvements, which also involve emissions in UB environment improvement (Guttikunda, measurements. 12 3. Preliminary Results of Assessment Work in 2008 Extent of the assessment work for this 2008 (first part of the AMIHB study). These Discussion Paper measurements are continuing. SO2 and NO2 are measured routinely in the CLEM monitoring In line with the concept of air quality network since several years. These measurements management (AQM) presented in Chapter 2, do not provide a sufficient basis for a definitive the objectives of the assessment work are: assessment of UB air quality, and this is the reason for launching AMHIB, but the data are to assess the present air pollution situation sufficient to give an indication of the severity of from available pollution concentration the situation. measurement data to map the spatial air pollution distribution UB's SO2 daily average values can be 125 µg/m3 based upon a preliminary emissions inventory or even higher which significantly exceed the and local meteorological data WHO Guideline of 20 µg/m3 (24-hour mean). to model the contributions to ground level According to data from these measurements, SO2 concentrations from the various main source levels are at times higher than AQ standards of the categories WHO Interim Target 1 (24-hour average, IT-1, to calculate the effects of some selected 125 µg/m3), the highest target for developing abatement measures on some sources, on the countries, but the excess is limited. air pollution levels and on the population exposure The Discussion Paper finds the PM problem to assess the health effects situation in UB in UB more severe than the SO2 problem. The to distribute the results of the assessment, PM concentrations in UB are extremely high. solicit comments and suggestions, to improve There is a very strong seasonal variation with the work leading to a construction of a very high winter concentrations and much lower baseline for 2008­2009 in a final report due summer concentrations. The annual average PM10 in early 2010. concentrations measured at the NUM monitoring station since 2006 where 141, 157 and 279 µg/m3 Summary of the present air quality situation for 2006, 2007 and 2008 respectively. The real concentration is expected to be higher, since the High concentrations of particles in air, PM, is sampler used underestimate the concentration currently the main air quality problem of UB. (section 4.3.2 and Appendix C). The measured This can be assessed from measurements carried concentration increase may indicate that the PM out since 2006 at the monitoring station at concentrations in UB have been increasing over the National University of Mongolia (NUM), the later years, although it is possible that the as well as measurements carried out at several seeming increase might be explained by factors stations as part of the AMHIB baseline project, such as meteorology. The measurements carried measurements carried out during June­December out under the AMHIB study at several new 13 Air Pollution in Ulaanbaatar stations since June 2008 give even considerably The PM2.5 concentrations are even less well higher concentration levels, confirming the documented by measurements, but limited severity of the PM concentration level in UB. samples taken in November 2008 indicate the severity of the PM2.5 situation. A measurement Annual concentrations in UB are very much campaign during the last part of November 2008 higher than international or Mongolian air quality provided parallel PM2.5 and PM10 measurements standards. The measured levels can be compared indicating that 50­60% of PM10 was in the PM2.5 with the Mongolian standard of 50 µg/m3, the fraction on those days. The PM2.5 concentrations WHO Guideline of 20 µg/m3, and the EU Limit reached as high as over 400 µg/m3 as daily average Value of 40 µg/m3 (Appendix A). WHO has set and maximum hourly levels of up to 1300 µg/m3. Interim Targets (IT), since it is not realistic that some developing countries will be able to meet Current air quality situation in UB, assessed the Guideline within reasonable time. The WHO from pre-AMHIB data IT-1 is at 70 µg/m3. The PM levels place UB among the most PM SO2 and NO2 data from the CLEM polluted cities in the world, and it is probably monitoring network in UB, 2001­2006 the most PM polluted. Compared to such high concentrations, some cities in northern China The Mongolian Central Laboratory for and south Asia also had concentrations above Environmental Monitoring (CLEM) under 200 µg/m3 up to a few years ago, but the PM NAMHEM is operating an air pollution levels in Chinese cities are being reduced over the monitoring network in UB. Until 2007 there latest years (see Appendix D). The highest annual were 4 stations in the network which had been average PM10 concentrations in European and US operated over several years. cities are much lower, in the interval 60­100 µg/m3, and in most cities the Figure 7 shows results from the daily concentrations are below 30 µg/m3. measurements taken of SO2 and NO2 at stations UB1­4, located as shown, between December Daily concentrations in UB are also 2001 and October 2006. Since 2007, the network much higher than Mongolian or international has been developed and has presently 6 stations in standards. The extremely episodic nature of operation. UB PM pollution, which is caused by the combination of ger heating practices and the The very strong seasonal variation in the meteorological situation, causes extremely high compounds are related to emissions from coal short-term PM concentrations. The extremely burning, such as SO2. NO2 is a compound more high hourly and daily concentrations may representative of emissions from road vehicles. represent the highest urban scale PM levels Here, the seasonal variation is less pronounced, anywhere, with hourly PM10 concentrations since the source is present at all times, but the approaching 2,500 µg/m3 or higher and daily generally poorer dispersion conditions in the averages above 1,000 µg/m3 in the most polluted winter still give higher concentrations. parts of the cities, i.e. the ger areas. The figure indicates that there are large These levels can be compared with the spatial differences in the pollution level, especially Mongolian standard of 100 µg/m3, and the during winter, with the highest SO2 levels at UB4 WHO Guideline of 50 µg/m3 (Appendix A). The in the east and lowest levels at UB1 in the south. EU Limit Value is 50 µg/m3 as 90th percentile The main wind direction in winter is from west (allowing 36 days per year to exceed that level). and northwest, making the UB4 station the most The US allows 150 µg/m3, which is the same as exposed one to the urban and ger emissions, as the WHO Interim Target 1 (IT-1) for developing they are brought by the wind across the city and countries. accumulated from west towards east. UB1 is far 14 Preliminary Results of Assessment Work in 2008 Figure 7: SO2 and NO2 measurements in Ulaanbaatar, December 2001­October 2006, at stations UB1­4. Daily average concentrations (g/m3). (Guttikunda, draft 2007) enough from the main emission areas to be less Figure 8 shows monthly average PM exposed. concentrations for 2006 and 2007. The results show that the coarse fraction5 (PM10­PM2.5) Particulate matter (PM) concentrations dominated over the fine fraction (PM2.5), and measured by NUM, NAMHEM and NILU both fractions have a very strong seasonal variation, high during winter and low during summer months. Monthly average PM10 Particulate matter (PM) data from concentrations reached close to 400 µg/m3 in National University of Mongolia December 2007. The main contributor to the fine fraction particles is the coal and wood burning in The Nuclear Research Centre of National the city, while the suspension of dry dust, partly University of Mongolia (NUM) has provided data from road surfaces and partly from open soil from measurements of PM2.5 and PM10 at the surfaces, contributes mainly to the coarse fraction NUM monitoring station to the east of UB centre in all seasons, and also gives a contribution to the area, for the period 2004­2008 (Tables 2 and fine fraction. 3). The monitoring is carried out using a GENT Sampler, Schulberger Model 2504, separating airborne particles into two size fractions: PM2.5 (fine fraction) and PM10­2.5 (coarse fraction). PM10 is the sum of the two fractions. Samples are taken 5 The coarse fraction is the share of particulates with size routinely on 2 different days each week. (diameter) between 2.5 and 10 µm. The fine fraction is the share of particulates with a size equal to or below 2.5 µm. This is important because the fine particulates are universally in literature shown to have particularly harmful effects on human 4 http://cat.inist.fr/?aModele=afficheN&cpsidt=2097174 health. 15 Air Pollution in Ulaanbaatar Table 2: PM concentrations in UB, measured by NUM at its monitoring station, monthly averages, 2004­2006 Table 3: Summary of PM measurement results carried out by NUM Year Period Fine PM Coarse PM PM10 2004 Oct­Dec 64 243 307 2005 Jan­April and Oct­Dec 77 159 236 2006 Jan­Dec 32 109 141 2007 Jan­Dec 33 124 157 2008 Jan­Dec 63 216 279 Figure 8: Summary of PM measurements at the NUM station, 2006­2007 (monthly averages) (Lodoysamba et.al., 2008) 16 Preliminary Results of Assessment Work in 2008 The data for the last 3 years cover the was 157 µg/m3. Parallel measurements with the whole year. The data indicate an increase in KOSA and NUM (Ghent) instruments (as well PM10 concentration at the NUM station, from as other instruments) co-located showed that the 141 µg/m3 in 2006, to 157 µg/m3 in 2007 and KOSA gives a much lower concentration than the a substantial increase to 279 µg/m3 in 2008. Ghent sampler. (In Appendix C, discrepancies Figure 10a shows the large increase to 2008 is between the results from various instruments in associated with a period with very large coarse UB are described in more detail). The strength of fraction concentrations during January­March the NAMHEM-Kosa results is that they confirm 2008, while PM2.5 is also elevated in general and the strong daily and seasonal variation in the PM especially during the fall period. These variations concentration. This variation is mostly due to the could result from meteorological variations, and variations in amount of coal combusted at various this warrants closer study. times determined by the temperature variations and daily domestic rhythm, while meteorological Figure 9 shows individual daily PM variations also play a role. The morning peak is concentrations for 2007. There is large variability generally shorter than the evening peak. This between individual days, and daily PM10 is so partly because the wind usually picks up concentrations were as high as 600­700 µg/m3 during the late morning hours to improve the during the winter period. dispersion of the emissions, while in the late afternoon the wind slows down and the inversion Measurements carried out by NAMHEM layer establishes itself again and builds up during provide another source of PM data in UB. the evening, while the heating of the gers sustains NAMHEM uses an instrument of type KOSA, a until the night hours. Japanese beta absorption method PM monitoring instrument. Results for 2007 are NILU carried out PM measurements during shown in Figure 11 for the NAMHEM station, the week 17­22 November 2008, using a GRIMM located on the 5th floor of the NAMHEM 107 PM monitor6. The instrument recorded building in UB centre. The average PM10 the concentrations of PM10, PM2.5 and PM1 concentration from these measurements was about 50­60 µg/m3, considerably lower than measured with the NUM method at the NUM station, 6 http://www.grimm-aerosol.com/Environmental-Dust- where the annual average PM10 concentration Monitors/107-spectrometer.html Figure 9: Individual PM fine and coarse measurements at the NUM station, 2007 17 Air Pollution in Ulaanbaatar Figure 10: Individual PM2.5 and PM10 measurements at NUM station in 2007 and 2008 continuously at 15-minute averages throughout the table, the actual concentrations in the air the week. This provides the possibility to see the during the first two days are about 60% of those variations in PM concentrations as a function of recorded by the instrument, due to the high time of the day as well as of the meteorological relative humidity (above 70%) which results conditions. Figure 12 shows the results, and in hygroscopic particle growth affecting the they are summarised in Table 4. As noted in response of the instrument. 18 Preliminary Results of Assessment Work in 2008 Figure 11: Results of PM10 monitoring using a beta absorption method instrument (Japanese type `KOSA'), for the period February 2007­January 2008 Source: NAMHEM. Table 4: PM concentrations in UB during the week of 17­22 November, measured with the GRIMM monitor. Daily and max hourly averages, g/m3 PM10 PM2.5 Date Daily average Max hour Daily average Max hour 17­18.11 1 620 1200 480 1080 18­19.11 1 770 1440 560 1080 19­20.11 487 1700 272 1000 20­21.11 619 2300 363 1300 21­22.11 703 1600 418 900 1Approximate concentrations. Concentrations recorded by the instrument during these days are affected by the high relative humidity, resulting in growth of hygroscopic particles. The concentration level is much higher in morning and evening inversion) combine these results with the GRIMM monitor than to give the strong daily variations in the PM the KOSA instrument operated by NAMHEM, concentrations. for reasons indicated above (see Appendix C). However, both instruments show similar Maximum daily concentration measured with variations across the day, demonstrating how the GRIMM instrument during this week was the heating practices and the meteorological about 700 µg/m3 for PM10 and about 400 µg/m3 conditions (low wind speed and ground level for PM2.5, adjusting for the effect of high relative 19 Air Pollution in Ulaanbaatar Figure 12: PM monitoring in Ulaanbaatar 17­21 November, 2008, using a GRIMM 107 PM monitor (g/m3). Meteorological data are from NAMHEM humidity on the instrument. The highest hourly daily average PM concentration is highest on days averages were extremely high, about 2300 µg/m3 with low temperature and wind speed during the and 1300 µg/m3 for PM10 and PM2.5 respectively. afternoon and evening/early night hours. The afternoon/evening PM peak is usually Another observation of interest is the very the longest and highest, and the one that most low PM2.5 concentrations between the morning determines the level of the daily average. The and the afternoon peaks, although the PM10 20 Preliminary Results of Assessment Work in 2008 concentration does not go that low (Figure 12). kind of statistical analysis that should be noted. An interpretation is that ger stove use is very low But it is clear that the `sulphur' source in this during mid-day, while the coarse PM fraction UB case is related to coal combustion, while the is contributed by dust suspension, probably `zinc' source is often interpreted as associated mainly from road dust, a source which operates with road traffic. However, it should be expected throughout the day. to find a `sulphur' source predominantly in the fine fraction, and the most recent version of their Source contributions to PM assessed from statistical analysis also shows a sulphur source in measurements the fine fraction. The PM sample filters taken by NUM are Still, the conclusion from the updated weighed (before and after exposure), and then analysis is that the main sources of PM2.5 are coal subjected to state-of-the-art analysis of elemental combustion, dry dust and motor vehicles (40%, composition.7 Such data provide the basis for a 38% and 18­22% respectively in Figure 13) statistical analysis which gives an estimate of the and to coarse fraction PM it is dry dust, coal main source categories to the PM pollution, as combustion and motor vehicles (63%, 22% and well as their relative contribution to the PM mass. 3% respectively). The approximately 100 filter samples taken over the period covered in Table 2 were subjected to There is a large contribution from the dust such analysis, providing the following estimate of source especially to the coarse fraction both in contributions to the PM10 concentrations at the summer and winter. This corresponds well with NUM station: coal burning: 35%, windblown the general knowledge that suspended dust from (suspended) dust: 50%, motor vehicles (exhaust the ground is predominantly in the coarse size particles): 12%, wood burning: 3%. fraction (and most of it even on particles much larger than PM10). It is also known from literature The Nuclear Research Centre (NRC) at that suspended dust has also a fine fraction, NUM has carried out updated statistical analysis especially when the suspension is caused by of the elemental composition of PM, for the vigorous turbulence such as created by vehicles fine and coarse fractions separately. The updated driving on roads. The results in Figure 13 on statistical analysis, based upon the 2006­2007 the fine fraction are in agreement with this. measurement series (see Table 4) modified The dust source is estimated to contribute with the picture regarding the source contributions 38% of the fine PM mass in summer, while it is somewhat. When studying the pie charts (see somewhat reduced to 35% in the winter, because figures on previous pages), note that the coarse then the coal and wood combustion sources fraction concentration is some 3­4 times larger are much stronger. Vehicle exhaust is estimated than the fine fraction. The updated analysis to contribute with 6­18% of the PM mass, indicated two separate sources of dry dust, dependent upon the size fraction and season. called `soil' and `construction'. It is believed that important sources to the `soil' factor are suspension Sulphur dioxide (SO2 ) concentrations of dust on roads due to traffic as well as the suspension of soil dust from all ground surfaces. A Measurement data for SO2 for 2007 was also `sulphur' source appeared in the coarse fraction, as provided. Figure 14 shows SO2 measured at well as a `zinc' source in the fine fraction. stations UB1-4 (locations: see Figure 14) by CLEM in 2007. The interpretation of sources, as they appear from the statistical analysis, is not always SO2 has an even more clear seasonal variation straightforward. There are uncertainties in this than PM has. The dominating source of ground level SO2 concentrations is the coal burning in 7 http://www.cse.polyu.edu.hk/~activi/BAQ2002/BAQ2002_ the gers and HOBs. Occasional peaks can be files/Proceedings/PosterSession/16.pdf expected also from the power plant stacks. The 21 Air Pollution in Ulaanbaatar Figure 13: Estimated contributions to the fine (PM2.5) and coarse fraction (PM10­2.5) of PM at the NUM station, based upon the 2006­2007 measurement series Lodoysamba et al., 2008. Figure 14: SO2 data for the CLEM stations UB1­4, daily averages, for 2007 ger contribution has a strong seasonal variation International guidelines and limit values which also shows up in the measurements. for SO2 are mainly given for the daily (24-hour) average value. According to the CLEM During the measurement campaign carried measurements, daily SO2 concentrations reached out by NILU during 17­22 November 2008, up to and above 80 µg/m3, highest at station 4 SO2 was measured by NILU using the passive to the east of the city centre. Allowing for the sampling method. Passive samplers were installed possibility that the method used gives too low at four of the UB monitoring stations. The values, the SO2 levels can at times be higher result was that the NILU passive samplers gave that the EU Limit Value of 125 µg/m3 (US AQ approximately twice as high SO2 concentrations as standards allow 365 µg/m3 as daily average). measured by CLEM, using their standard manual Levels are very much higher than the recent method. It is necessary to check better the quality WHO Guideline of 20 µg/m3, while the WHO of the CLEM SO2 measurements. Interim Target 1 (IT-1) for developing countries 22 Preliminary Results of Assessment Work in 2008 is 125 µg/m3, the same as the EU Limit Value, their locations are given in Figure 15. There are and the Interim Target 2 (IT-2) is 50 µg/m3 many different types of instruments used in the (Appendix A). network. The need for comparing the different instruments in terms of their PM results has been Thus, although SO2 is also a problem recognized. The instruments have been compared compared to limit values, its importance as an air with each other during three different periods, quality problem in UB is much less than the PM when the instruments were all brought together at problem. the same location. The AMHIB baseline monitoring: Results Results from the AMHIB PM measurements and quality assessment June­December 2008 The AMHIB monitoring stations PM concentrations A PM monitoring network has been established The results from the measurements carried out in UB as part of this AMHIB study. The network at the AMHIB stations during June­December consists of the stations described in Table 5, and 2008 have been reported by the local UB AMHIB Table 5: Monitoring sites and monitoring units, their characteristics Sampling site Mobile/ number/ PM10, Institute/Location Stationary Characteristics name Site position PM2.5 Mobile unit GENT Sampler, Schulberger Model 2 / NRC 106°58,311 PM10 and 250, Measures PM10 and PM2.5. PM2.5 47°54,811 National University of Polycarbonate, nuclear pore filters Mongolia(NUM) Mobile unit GENT Sampler, Schulberger Model 3 / 100 ail 106°55,343 PM10 and 250, Measures PM10 and PM2.5. PM2.5 47°55,975 Polycarbonate, nuclear pore filters Stationary Kosa Monitor, Measures PM10. 1 / NAMHEM 106°54,704 PM10 and National Agency unit And PM2.5, measurement on-line, PM2.5 47°55,220 for Meteorology principle Light Scattering, no filter. Hydrology and Can sample for each hour. Environmental Agency Mobile unit Partisol FRM-Model 2000, Measure 6 / 3-r 106°52,167 PM10 (NAMHEM) PM10 16.7 l/min, filter khoroolol 47°55,582 Central Laboratory Mobile unit Rotary Bebicon, Type 35RC-28SD5, 5 / CLEM 106°52,967 PM10 for Environmental Measure PM10 15 l/min, filter 47°53,64 Monitoring (CLEM) Mobile Dust Trak-8520, measure PM2.5 or 4 / buudal 106°54,159 PM2.5 PM10, based Laser 47°54,719 NAMHEM Mobile Dust Trak-8520, measure PM2.5 or 8 / Niseh 106°45,749 PM2.5. PM10, based Laser 47°51,865 Mobile Dust Trak-8520, measure PM2.5 or 7 / S.K.H.D PM2.5. PM10, based Laser 23 Air Pollution in Ulaanbaatar Figure 15: Locations of the PM monitoring stations of the AMHIB network team (Lodoysamba et. al., 2009). A summary Generally large differences between stations description of the results is included below. on any given day. These measurements from 2008 show higher Measurements were carried out two days average concentration level than measured per week (24-hour samples) during the entire in earlier years, at stations in ger areas where period, and in addition every day during one measurements have not been done before, as week in October, November and December. This presented in the previous sections above. resulted in 8­9 days of sampling per month in At the NUM station (station 2), the average June­September, and 11­15 samples per month PM10 for the months June­December has in November­December. Sampling is missing increased from 136 to 209 to 224 µg/m3 for on some stations some days because of sampler 2006, 2007 and 2008 respectively. problems. For PM2.5 there is no data from station However, maximum monthly and daily 7 before 3 September, from stations 4 and 8 averages measured at the NUM station in during July­August and Dustrac sampling at 2008 are about the same as in previous years. station 1 (NAMHEM) started in October. For PM10 there is no data from station 1 (NAMHEM, The extremely high concentrations measured KOSA instrument) from early August to at the ger area stations (stations 3, 4, 6 and 7) 25 September. in December especially, are much higher than The results are shown as time series in what has been measured at the NUM station as Figure 16 (PM2.5) and Figure 17 (PM10). Main described in the section on p. 21. There is one observations are: important point to consider regarding the quality of the measurements, and that is the effect of high As expected, the PM concentrations increase relative humidity (RH) on the response of the steadily and strongly into the winter months instruments measuring PM2.5 at the ger stations (after September). 4, 7 and 8 (see section from p. 15). The RH was Extremely high concentrations at some high enough in December to affect the response to stations some days, mostly during winter the extent that the measured PM2.5 concentrations months, up to 4000 µg/m3. should be reduced by about 40%. Even so, the remaining concentrations are still very high. 24 Preliminary Results of Assessment Work in 2008 Figure 16: AMHIB PM2.5 measurements during June­December, 2008. Daily and monthly average concentrations 25 Air Pollution in Ulaanbaatar Figure 17: AMHIB PM10 measurements during June­December, 2008. Daily and monthly average concentrations 26 Preliminary Results of Assessment Work in 2008 The AMHIB measurement period will run Details of the results from the monitor and until June 2009, and will, together with sampler comparisons are given in Appendix C. the on-going automatic measurements at It can be concluded that the Ghent (NUM and NRC) samplers, the Partisol and the Dustrac a number of stations starting early 2009, instruments give data of reasonable quality, given provide an improved baseline of PM the shortcomings described in Appendix C. These pollution in UB. samplers are used at stations 2, 3, 4, 6, 7 and 8. On the other hand, the Kosa (station 1) and the The quality of the AMHIB data is evaluated in C-20 instruments (station 5) somehow give too the next section and Appendix C. low PM concentrations. Source contributions Air pollution health effects in UB: Current situation and ongoing study The NUM group has carried out statistical source apportionment analysis on the June­December Health effects and air quality guidelines 2008 filters as well, based upon multiple for PM elemental analysis, similar to that described in section from p. 21. Particle pollution, also called particulate matter or PM, is a complex mixture of extremely small This 2008 analysis gives the same sources the particles and liquid droplets in the air. When responsibility for the PM concentrations. The size breathed in, these particles can reach the deepest of the various contributions differ somewhat, e.g. regions of the lungs. Figure 18 visualises where with more from `soil' and less from combustion particles of different sizes are deposited in the in the fine fraction than in the 2006­2007 data. lung when inhaled. Exposure to particle pollution The main conclusion is the same, however: The is linked to a variety of significant health largest contributions to PM at the NUM station problems, ranging from aggravated asthma to comes from resuspended soil particles and from premature death in people with heart and lung coal combustion, with motor vehicles as a third disease. Particle pollution also is the main cause of and less important contributor. visibility impairment. AMHIB data quality assessment Both WHO, the US Environment Protection Agency (USEPA) and the European Union (EU) The PM measurement equipment of the AMHIB have set Air Quality Guidelines (AQG), Standards network is provided by the various institutions (AQS) or Limit Values (AQLV) for PM, as a involved, and differs between the various stations. measure to protect the public from adverse health The instruments utilize different measurement effects (see Appendix A). principles. It is of interest to compare the instruments in terms of the PM concentration These guidelines and standards are based data they provide. To investigate this, co-located upon a large body of epidemiological and other comparison sampling has been carried out during types of studies worldwide. The majority of the three campaigns in 2008, each of 4­5 days epidemiological studies have used PM10 as the duration: 4­5 and 17­20 April, 1­6 July and exposure indicator, since PM10 has by far been 18­22 November. The two first campaigns were the most measured PM quantity. PM10 represents carried out at the NAMHEM monitoring station the particle mass that enters the respiratory tract, at the roof of the NAMHEM building, while and includes both the coarse fraction (particle the last one in November was carried out at the size between 10 and 2.5 µm) and fine particles meteorology station UB3 located in a Ger area (measuring below 2.5 µm, PM2.5), that are to the west of UB centre. NILU participated in considered to contribute to the health effects the last campaign providing a GRIMM 107 PM observed in urban environments. PM2.5 is now monitor. being measured to a larger and increasing extent. 27 Air Pollution in Ulaanbaatar Figure 18: Invasion of various particle size fractions in the human lung From Guttikunda, 2007. The present WHO AQGs are based upon concentrations, and setting a strict standard for studies that have used PM2.5 as the exposure PM2.5.9 indicator.8 WHO considers, however, that AQGs for PM2.5 alone will not provide protection against Current air pollution health effects situation in UB, harmful effects of coarse particles. WHO has and ongoing study consequently kept AQGs also for PM10 , and now set its value by applying a ratio between PM10 Many studies internationally have shown a and PM2.5 of 2, based upon evidence of this factor good relationship between air pollution and as measured in urban atmospheres in developed cardiovascular disease (Jamal et al, 2004; WHO (a factor of 2) and developing (factor range of 2002; Spengler 2005; Sunyer et al, 1996; Keil 1.25­2) countries. However, WHO does not 2005; Stansfield & Shepard 1993). There was consider that the evidence is sufficient to derive much research conducted and still running on air a separate AQG for the coarse particle fraction. quality effects on health. However, most of the WHO considers that the large body of literature air pollution health impact studies in Mongolia on effects of short-term exposures to PM10 is a are more concerned with the respiratory diseases good enough basis for developing AQGs and (Burmaa & Enkhtsetseg 1996; Spickett et al, Targets for 24-hour PM10 concentrations. 2002; Saijaa 2004; Tseregmaa, 2003). There were not many relevant data about relationship The EU has followed the WHO between cardiovascular disease and air pollution recommendation, and EU Limit Values are set for in Mongolia. A. Enkhjargal (2006) conducted both PM2.5 and PM10. The USEPA has, however, a survey on health impacts assessment of air decided differently. The lack of clear evidence pollution and cardiovascular and respiratory system of a link between health problems of long-term diseases in two cities. This study demonstrated exposure and coarse particle pollution has led significant correlations between respiratory and the EPA to revoke its annual PM10 standard, cardiovascular morbidity with NO2, SO2 and while keeping its standard for 24-hour average some meteorological parameters. The correlation 8 World health organization: Air Quality Guidelines for particulate matter, ozone, nitrogen oxides and sulphur dioxide. Global update 2005. Summary of risk assessment. 9 US Environment Protection Agency: PM Standards http://whqlibdoc.who.int/hq/2006/WHO_SDE_PHE_ Revision­2006. http://www.epa.gov/oar/particlepollution/ OEH_06.02_eng.pdf naaqsrev2006.html 28 Preliminary Results of Assessment Work in 2008 of mainly respiratory case admissions with polluted. Some cities in northern China meteorological parameters is because the cold and south Asia also had concentrations above winter conditions in the two cities result in the 200 µg/m3 up to a few years ago, but PM accumulation of pollutants in the atmosphere. pollution is coming down in the most polluted Thus population exposure to air pollution increases Chinese cities (see Appendix D). Compared to significantly in the winter months. such high concentrations, the highest annual average PM10 concentrations in European and US The rates of cardiovascular diseases in cities are much lower, some cities are in the range Mongolia during the last 15 years have increased 60­100 µg/m3, except some cities in dry regions 2 times. Cardiovascular disorders now affect in the US (Arizona and California), where PM10 is 11.0 per cent of the urban population and dominated by dry surface dust particles, and very 13.8 per cent of the rural population (Ministry little associated with fuel combustion. In most of Health of Mongolia, MOH, 2004). cities in the US and Europe, the concentrations are below 40 µg/m3. The AMHIB study is investigating, to the extent possible, the relationship between air The episodes of extremely high hourly pollution and health effects in UB. It is hoped and daily concentrations that are caused that this would kick-off more in depth and robust by the special climate and meteorological studies that can complement these efforts. situation of UB probably represent the highest urban scale PM levels anywhere, with hourly Part of the AMHIB project is to study the concentrations approaching 2,500 µg/m3 and current health effects situation and its relation daily averages approaching 700­800 µg/m3 over to the air pollution situation. Connected to the areas covering much of the city. These episodes AMHIB monitoring activities, health effects data occur regularly and often throughout the will be collected from hospitals in UB located winter periods, and bring the annual average close to the monitoring stations, at three levels: concentration to its very high level. US and European cities have highest daily averages Primary level: Family and village hospitals mostly below 200 µg/m3, while some industrial (8 hospitals) cities in the Eastern part of Europe still have Secondary level: District hospitals and high maximums, a few cities in the range ambulatories (7 hospitals and 1 ambulatory) 400­700 µg/m3. These maximum daily averages Tertiary level hospitals (3). approach those experienced in UB, but they occur very seldom, a few days per year. Data on admissions connected to respiratory and cardiovascular diseases, according to a Emission inventory used for the preliminary air number of diagnosis, will be collected on a daily pollution assessment in this paper basis from these hospitals, and statistical analysis applied to the data, using the SPSS version The main air pollution sources in UB 11.5 tool. Independent variables collected are PM concentrations (PM2.5 and PM10) as well as It is important to note that the main source of meteorological data (temperature, wind speed and emissions in a city may not necessarily be the relative humidity). dominant source of air pollution people breathe. The inventory of the air pollution emissions Data collection started in June 2008 and will in UB was investigated under a previous WB continue till June 2009, whereafter data analysis activity (Guttikunda, draft 2007), based upon a will take place. significant effort although with limited resources. The main sources of the emissions were listed as International comparison of UB air pollution follows: UB is among the most polluted cities in the 1. Stoves in households in Ger areas world in terms of PM, and it probably the most 2. Stoves in kiosks and food shops 29 Air Pollution in Ulaanbaatar 3. Power plants boiler. Statistically, most households report to use 4. Heat only boilers coal with wood for ignition. Anecdotally, burning 5. Vehicle exhaust emissions of tires, garbage, tarred bricks are said to be used 6. Fugitive dust--transport and non-transport by particularly lower income households. 7. Construction industry--Bricks 8. Garbage burning Heating Stoves and Wood and Coal Bundles for Sale Summarised descriptions of each of those source categories and their emissions are as follows: 1. Stoves in households in the Ger areas In UB, the largest source of coal and wood combustion related air pollution at the ground level is from heating/cooking stoves in Ger areas. Many homes are Gers, the traditional Mongolian dwellings consisting of a wooden frame beneath several layers of wool felt. However, by 2007 the majority of Ger residents have built wooden homes within their hashaa, or homestead. Depending on the level of income, these wooden homes are heated with stoves with heating walls or coal fired individual household boilers. The households in the Ger areas in UB can be divided into 4 categories in terms of the types of stoves used (World Bank, 2008a): households living in a ger and using a heating stove with chimney; households living in a small detached house and using a heating stove with chimney to directly heat their home; Source: Guttikunda, 2007. households living in a medium size detached house and using a stove attached to a heating The average winter period consumption of wall; and raw coal and wood per household is 4.2 tons and households living in a larger detached house 4.6 m3 respectively. The coal consumption varies and using a small low pressure boiler (LPB) between 3.5 tons for ger stoves and 6.2 tons for attached to a system with radiators and households with household boilers, most of this circulating water. obviously consumed during the 8 winter months. The total annual coal consumption for this source The emissions from the household stoves category is about 550,000 tons of raw coal and are dominated completely by the small stoves of 415,000 tons of wood. the traditional type. There are a total of about 130,000 stoves in the UB Ger areas, of which there are 91% small stoves used in the three first types of households in the above list, while only 9% of the households have a individual household 30 Preliminary Results of Assessment Work in 2008 Unconventional Fuels Used in Ger Areas 2. Stoves in kiosks and shops A rather unreported category, though less important than household heating, are kiosks and shops. This is a less important source category, however included here because the same type of stove is used also in the kiosks and small shops. The number of such kiosks and shops was estimated to be 4,500 in 2005. In the same period, the number of food shops more than doubled. Total emissions from this category were some 5% of the Ger area household emissions. 3. Power plants There are three coal-fired CHP Plants in UB, located as shown in Figure E4 in Appendix E). They provide nearly all of the installed power capacity in the city and also are the main source of the district heating system providing space heating and domestic hot water to the apartment and office buildings in the central parts of UB, covering 80% of the energy need for about 60% of the households in central UB. The power plants consumed almost 3.4 million tons of coal in 2007. The pollution control technology is in poor condition. Power Plants #2 and 3 use wet scrubbers with cleaning efficiency estimated to be around 70% while Power Plant #4 operates an Electrostatic Precipitator (ESP) although it is unlikely it achieves close to the reported 95% cleaning efficiency. As a reference, ESPs usually operate at an efficiency of 99.95%. JICA recently discovered that one or two of the chimneys in the power plants do not have an adequate access to measure flue gases. In addition to stack emissions, another unaccounted for source in the emissions inventory is the fly ash from the ponds where fly ash is disposed. Removed fly ash is sent to settling tanks where the sedimented dust is collected and sent to the ash ponds. These ash ponds are continually subjected to wind erosion in the dry season as seen in the figure on the next page. The emission rate is likely to be a function of Source: Dr. Sarantuya Myagmarjav, MNE. 31 Air Pollution in Ulaanbaatar wind speed, particle size, and area exposed and owned) (World Bank, 2008b). The boilers therefore is a very intermittent source difficult to are of various types: mainly Russian, Chinese, calculate without specific measurement exercise. Mongolian (German design). Several new boiler Anecdotally, it is a common sight in the spring installations are not included in the inventory. and summer months along with dust storms from The efficiency of the boilers vary from about 40% deserts. Due to high moisture content and snow for the Russian boilers to better than 80% for the cover, it is unlikely to account for much of the Mongolian type of German design. The boiler pollution sources in winter. designs have an impact of the emission factor of PM, as well on the emission conditions, such Fly Ash from Power Plant Ash Ponds as height, temperature and speed at the outlet (stack). The HOBs are mainly located in central UB and along an east-west axis along the main road. 5. Vehicle exhaust emissions As mentioned, about 93,000 vehicles are in use in UB (2007). About 75% of these are passenger vehicles, 15% are trucks and 7% are buses. The toal number of motorized vehicles is growing rapidly (Guttikunda, 2007) due to increased incomes and availability of affordable vehicles. Use of individual automobiles has already caused a severed congestion problem for the municipality. The passenger fleet is a mixture of old and newer vehicles, while the trucks and buses are mostly of older types. The vehicle emissions are spatially distributed along the road network, which is concentrated to the central areas, and less concentrated in the ger areas. The annual daily traffic is large on many sections of the network, the most trafficked sections having an annual daily traffic of more than 60,000 vehicles. The vehicles emit large amounts of nitrogen oxides and carbon monoxide, while the 4. Heat only boilers (HOB) vehicle exhaust PM emissions in UB accounts for less than 10% of the ger household PM emissions. In UB, mainly public facilities and industry not supplied by the central District Heating 6. Suspension of dry surface dust System from the Combined Heat and Power Plants use small HOBs. These are used in small The dry climate of UB creates a large town centers where extending the District potential for dust suspension from surfaces, both Heating is not feasible. Use of HOBs could from roads and other near-road surfaces, and increase if the District Heating system is unable other dry non-vegetated surfaces, especially in the to expand to support a growing number of winter. The largest dust suspension problem is apartment and commercial buildings. The latest created by the traffic, as the turbulence from the inventory of HOBs gives a number of 145 boiler vehicles stirs up the dust deposited on the streets. houses with a total of 267 boilers, of various This is especially observable on unpaved roads, ownership (main city, the military and privately which are dominating in the ger areas. According 32 Preliminary Results of Assessment Work in 2008 to the results from the PM10 monitoring in UB, Other non-transport dry dust suspension the dry dust source in UB is of about the same sources include for example the ash ponds of magnitude as the coal combustion source, in the power plants (resuspension of deposited fly terms of its contribution to ground level PM10 ash), and the movement of construction vehicles concentrations. on temporary roads to and from construction sites. The ash ponds are localised sources where dust is suspended by wind action at times with The main source of fugitive dry dust other relatively high wind speeds, making that source than from roads is the suspension of dust from limited both in time and space. The construction- open soil surfaces. Most of the land surfaces associated traffic is rather widespread, and is a in UB have no vegetation, and the dry soil is more continuous source, more like the regular available for suspension by wind action most of road traffic. It is difficult to assess the magnitude the time, because of the dry climate. The top soil of this non-transport PM source. layer is very fine grained, and dust is easily picked up by wind action. The magnitude of this source Non-transport Related Suspended Dust is difficult to assess, but the source apportionment methods used and shown later in the report indicates that this dust source is significant in terms of contribution to airborne particle concentrations. Vehicular Suspended Dust Examples in UB Guttikunda, 2007. 7. Industry The main industrial activities in UB creating process emissions to air are the brick industry and the construction industry. There are 4 larger brick factories and 10 smaller brick kilns in UB. The annual coal consumption in these factories Guttikunda, 2007. and kilns has been estimated to about 50,000 33 Air Pollution in Ulaanbaatar tons (World Bank, 2008b). Bricks are produced Introduction to the development of an mainly in the period between mid-March to early emissions inventory for assessment and December. Thus, the contribution to UB air modelling purposes pollution in winter is limited to the November­ December period. The main objectives of the emissions inventory are: 8. Open burning of garbage To calculate the total emissions per source Backyard burning of trash, waste and garbage category and type, as a basis for a preliminary is a rather common practice in UB. Garbage is assessment of the importance of each of them burned in stoves, backyards and landfill sites. to the air pollution situation in the city Landfill site burning is anecdotally done due to To provide input to air pollution (dispersion) low tariffs (to reduce collection costs) and poor modelling of the air pollution concentrations collection. Based upon a study on amount of in the city, which determines the actual garbage produced in UB in different seasons, importance of each source. Guttikunda (2007) estimated the PM emissions from this source category to be close to 20% of Total emissions versus emission height the ger area household emissions. Much of the and location garbage is not collected, and burned or illegally dumped. The garbage burning is distributed The first objective relates to calculating total across all seasons. There is no independent emissions (e.g. tonnes per year), irrespective of emissions inventory for garbage burning but the locations and time variations of the emissions. recent surveys show that second to heating, To meet the second objective, it is necessary the collection of solid waste is among the top to specify the locations/spatial distribution of priorities for Ger residents.10 the emission sources, the time variation of the emissions (seasonal/monthly as well as hourly) as Hospital Waste Burning. In addition to well as the emission conditions of each source: household garbage, another unreported though height above ground, temperature, etc. smaller source is hospital medical waste burning. Hospitals are required to install incinerators that The source-wise total emission assessment burn trash and infectious medical waste. About points out the main sources in terms of emissions 35 hospitals practice bio-hazard waste burning, amount. The second step, assessing also their but do not use regulated incinerators. There is time and space distribution and emission height, very little information on this source, though it is usually modifies the impression of importance considered minor. of sources: location away from population Garbage Burning Example centres as well as tall stacks (often the situation for power plants and large industries) reduces significantly their importance for the urban air pollution levels, while smaller source categories in terms of total emissions may be much more important, when they are located throughout the population centres and emit at low heights. This is the situation for small scale domestic heating by combustion, as well as for road traffic. Pollutants In line with the assessment of PM and SO2 as the main pollution problems of UB, the emission Guttikunda, 2007. inventory in this paper will be limited to PM and 10 World Bank, 2009. SO2. 34 Preliminary Results of Assessment Work in 2008 Methods in this Discussion Paper is a combination of the NAMHEM EI and improvements carried The basic method for emissions inventorying is out by Guttikunda (draft 2007) and AMHIB. utilized in this paper: However, this EI needs to be further improved and completed to provide a proper basis for air Emissions are the product of an activity (e.g. pollution modelling. Main uncertainties are amount of fuel burnt, kms driven) and an associated with the emission factors for ger stoves emission factor (EF) (e.g. amount per fuel and road dust suspension, for PM10 and PM2.5, as used or km driven). well as the amount of suspension of soil dust Emission cleaning is either accounted for in from dry open surfaces which is not included the EF, or by reducing the emissions above by in the inventory. a factor (1-cleaning efficiency). The EF depends upon many factors that As mentioned in section 3.2, there are project need to be taken into account: e.g. type of activities presently (2009) underway that will combustion process (such as boiler/stove provide improved data for some of the sources in type), fuel specifications, process technology the EI. (such as engine and exhaust cleaning technology of a vehicle), etc. Summary of the emissions inventory used For each source category/type, the emissions for the present assessment work can be assessed by top-down or bottom-up methodology. Each of the main air pollution sources is treated separately. For each source the basis for the Example of top-down method: e.g. when the emissions inventory is described: total fuel consumption for small-scale combustion for space heating has been estimated (e.g. from fuel sales or consumption statistics or data), Description of the source and the total emissions by applying an EF is Calculation method distributed spatially and over an area where the Emission factor(s) fuel is burned, as a function of the distribution Total emissions of the population/density of households. A time Spatial distribution variation function can be overlaid, such as daily, Time variation based upon heating practices and seasonally based Uncertainties upon temperature statistics. Example of bottom-up method: e.g. for Appendices E, F and G describe the details vehicle exhaust emissions: when the road traffic of the inventory of the emissions per source volume (vehicles per day) and vehicle type category. The summary of the inventory is shown distribution is known/estimated for each road in Table 6. Below are given summaries of the link of the total urban road network: the EF is information, data and calculations per main applied to each type of vehicle in the traffic flow, source category. The industry (mainly brick) and and emissions calculated for each road link. The waste burning sectors have not been included in spatial distribution of the emissions is then known this assessment work. The brick plants are located from the locations of the road links, and thus away from main populated urban areas and they specified in the input to the model. The time are operated mainly during the summer season variation is also often known from traffic data, or when air pollution is low. The open waste burning it is estimated from the general activity patterns source is distributed across the year and urban/ger for the city. areas. It has been estimated to contribute possibly up to 10% of the PM emissions (Guttikunda, There are many uncertainties linked to this draft World Bank 2007), but its assessment is preliminary emissions inventory. The EI used uncertain. 35 Air Pollution in Ulaanbaatar Table 6: Summary of the emissions inventory for Ulaanbaatar, 2007 (tons/year)1 For details, see also Figure 7 Source/ Height of calculation emissions, method2,3 PM10 PM2.5 SO2 meters Spatial distribution Ger households2 16,363 13,262 7,0844 5­10 Throughout ger areas Figure 19 HOBs2 6,480 3,888 4,3605 20­30 Figure 20 CHPs3 6,290 2,516 33,6006 100­250 3 point sources to the west of UB centre Appendix E Figure E4 Vehicle3 exhaust 1,161 1,161 1,3547 <1 Figure 21 Dry dust from roads3 Paved 5,142 771 <1 Mainly throughout the central city Figure 22 Unpaved 4,812 722 <1 Mainly throughout the ger areas Figure 22 1 Last update: for all source categories, emissions were updated as part of this work, to various extent, see Appendix E. 2 Top-down calculation 3 Bottom-up calculation 4 0.65% S in coal (75% Nalaigh coal w/0.7%S/25% Baganuur coal w/0.5%S)11 5 0.5% S in coal 12 6 0.5% S in the coal (Baganuur coal) 7 From Guttikunda (2007). Data for 2006 Background material "Air Pollution Sources Inventory of UB City." Ministry of Environment, National The main background sources for the inventory Agency for Hydrology Meteorology and are: Environmental Monitoring, 2007. (Referred to as "NAMHEM, 2007") The emissions inventory recently developed by S. Guttikunda for the World Bank The ger households are the largest source of (Guttikunda, draft World Bank 2007) PM emissions. Their total emissions is estimated The draft report: Mongolia: Energy Efficient to be about 2.5 times larger than the HOB PM and Cleaner Heating in Poor, Peri-urban emissions. The ger emissions are split about 50/50 Areas of UB. Summary Report on Activities between coal and wood (see Table 7). The ger (World Bank, 2008a). (Referred to below as and the HOB sources emit at fairly low heights "The heating report, 2008") and are distributed throughout large parts of the Report "Small boiler improvement in UB". urban area, thus giving important contributions (World Bank, 2008b). (Referred to below as to the population's exposure to PM pollution. "HOB report, 2008") The CHPs, which have PM cleaning equipment, 11 Source: Japan Coal Energy Center, 16 June 2008 have about the same total emissions as the 12 Source: World Bank, 2008b HOBs. Their emissions take place at large heights 36 Preliminary Results of Assessment Work in 2008 Table 7: Summary of PM10 emissions inventory for coal and wood combustion sources, winter season 2006/7. Including estimate of uncertainty () Specific Total Emission Total Number of ger consumption consumption factor, PM10 emissions Source households/HOB (tons/winter) (tons) (kg/ton) (tons/year) Ger coal 130,000 4.19 545,000 16 8,715 1 10% 5% 10% 50% 51% Ger wood 130,000 3.18 413,400 18.5 7,648 1 10% 5% 10% 50% 51% HOB 267+ 400,000 16.2 6,480 5% 20% 15% 25% CHP 3,360,000 19.5 2 6,290 3 2% 25% 25% 1 The emissions during the summer season come in addition. For coal, this addition is small, for wood it is larger, since households use mostly wood for cooking during the summer. We add 5% to the consumption to account for the wood use in summer. 2 This is uncleaned EF. The weighted average cleaning efficiency of the 3 CHPs used is 90%. 3 Cleaned emissions. (through the CHP stacks of 100­250 meters), Ger area households which limits their contribution to population exposure. The households in the ger areas live in gers or in small houses, about 50/50 each. Vehicle exhaust particles are a less important source than gers, HOBs and CHPs in terms of Figure 2 shows the ger area locations around mass of emissions. The suspension of dry particles and close to UB. The number of households is from roads is, however, a more important PM10 about 130,000, and this includes the households source, with a total emission mass about 50% in the 6 districts closest to UB as well as gers close larger than HOBs. The suspension particles are to UB centre. The stove types used in the various mainly in the coarse fraction. Suspension is much types of dwellings has been inventorized (World less important for PM2.5. Bank, 2008a). The average consumption of coal and wood in the ger household stoves, for the The suspension of soil particles through wind winter season 2006/7, was there estimated to be action from open surfaces, apart from the roads, 4.19 tons/year and 3.18 tons/year. has not been included in the inventory. Source apportionment indicates this as an important The emission factor for PM emissions from PM10 source, its strength being difficult to assess ger stoves is very uncertain, there is no good basis with existing methods. in the literature for determining the EF. From what is available, the PM10 emission factor (EF) The CHPs are the dominating sources of for ger coal is estimated to 16 kg/ton, and the SO2, but their tall stacks limit their contribution ratio between PM2.5 and PM10 is set at 0.6. For ger to ground level concentrations relative to sources wood, the PM10 EF is estimated to 18.5 kg/ton, with much lower heights of emissions. with a PM2.5 / PM10 ratio of 0.9 (Appendix F). 37 Air Pollution in Ulaanbaatar Figure 19: Spatial distribution of ger household PM10 emissions, in km2 grids (tons/year) The resulting spatial distribution of ger estimated to 27 kg/ton. The ratios between PM10 household emissions of PM10 is shown in and TSP, and between PM2.5 and PM10 are both Figure 19. The ger areas far to the north in UB set to 0.6 (Appendix F). are located behind the hills and do not affect UB air quality directly, the emissions there The resulting spatial distribution of HOB are thus not included. There is a considerable, emissions of PM10 is shown in Figure 20. HOB unknown, uncertainty associated with the spatial emissions take place at a higher level than from distribution, which cannot be reduced until ger household, some 10­15 meters, and are improved data of population density distribution somewhat more removed from direct exposure of is available. the population. Even so, this is also an important source to the population's exposure. Uncertainties in the various data and factors are as estimated in Table 7. The uncertainty Combined heat and power (CHP) plants estimates are of course themselves uncertain. The fuel consumption data are considered to have The total consumption of coal in the three CHP the least uncertainty, due to the extensive work plants is 3.36 million tons for 2007 (statistical reported in the `heating report'. The emission data from NAMHEM). This is distributed factors are very uncertain. between the plants with 5%, 25% and 70% for the plants CHP1, CHP2 and CHP4 respectively. Heat only boilers For emission factors, those used by The emission inventory of the HOBs is based Guttikunda (2007) is used also in this report: mainly upon the data in the World Bank 2008b 19.5 kg/ton for PM10 and 7.8 kg/ton for PM2.5. reference. The number of HOBs is somewhat The efficiency of the flue gas cleaning for particles higher than 267, while the total annual is set to 80% for CHP 2 and 3, and 95% for consumption of coal is estimated to be 500,000 CHP 4. tons. The basis for the emission factor is actual measurements carried out on a few HOBs in UB. The CHP emissions take place at a high level Based upon this, the average HOB EF for TSP is (100­250 meters). 38 Preliminary Results of Assessment Work in 2008 Figure 20: Spatial distribution of HOB PM10 emissions, in km2 grid (tons/year) Vehicle exhaust traffic volume, the length of the road sections and the EFs. Figure 21 shows the spatial distribution Among the referred 92,706 registered vehicles of the emissions from the registered part of the in UB (2007), about 64% of the vehicles run on road network (the 100 links). In addition to this, gasoline, about 34% on diesel and the remaining the emissions from the traffic on the smaller roads about 2% on gas. (estimated to make up 30% of the total main road traffic activity) is distributed over the ger areas in The existing vehicle exhaust regulations in the same way as the ger combustion emissions are UB does not limit the types of vehicles allowed distributed, reflecting the population distribution. on the roads. No specific information is available on the exhaust emission levels of the vehicles on Suspension of dust from roads UB roads, old or new. As about 50% of the cars were older than 11 years in 2006, it is fair to Dry dust on road surfaces is whipped up, assume that they are of medium and low technical suspended in air, from vehicle turbulence as they standard. Mongolian gasoline still contains lead, travel the road. The extent of suspension increases so to the extent that new vehicles bought and with the speed of the traffic by about the square used include catalysts, they do not function after of the speed. Naturally, suspension takes place short time driving on local gasoline. Most of the only when the surface is dry, which is most often trucks are also old (mostly of Russian types), the case in UB, although conditions with ice on while the bus fleet has a wide age spread, some the road may limit suspension. Large vehicles, fairly new. EFs have been estimated based upon trucks and buses, suspend much larger amounts this information, see Appendix F. of dust than small vehicles. The dust suspension is very much larger from unpaved than from A limited program on traffic counting on the paved roads. The dust suspension problem is main road network has been carried out as part much larger per vehicle in the ger areas with the of this work, and as a result of that, the traffic unpaved roads than in UB central areas, although volume has been estimated on as many as about there is substantial suspension also from paved 100 road links (Figure E7 in Appendix E). The roads in the centre, since there is always a depot of exhaust gas emissions are calculated from the dust on the road surfaces. 39 Air Pollution in Ulaanbaatar Figure 21: Distribution of the emissions of PM10 from the vehicular traffic in UB (tons/year/km2) Most of the mass of the suspended dust is road emission distribution has the population on particles larger than 10 micrometers, thus distribution as a basis, and thus depends on the larger than those affecting humans by breathing. correctness of that distribution. However, a substantial amount is also below 10 micrometers (PM10), and a fraction also below Stoves in kiosks and shops 2.5 micrometers PM2.5). There are some 4,500 kiosks and shops in The suspension emissions are calculated UB which are heated by the same type of using the same road and traffic data as for the stoves as used by the ger households. This is vehicle exhaust. The source strength is calculated a less important source, and we do not have using suitable formulas (Appendix E, section its spatial distribution across UB. We use a 2.5). The formulas take account of the dryness of previous estimate (Guttikunda, draft 2007) and roads and paved/unpaved surface, the speed and distribution of this source. We allocate 5% of the fraction of large vehicles (trucks and buses). the ger household emissions to this source, and Figure 22 shows the resulting spatial distribution distribute it as the household emissions. of suspended PM10 from roads. Modelled current spatial pollution This estimate of suspension of PM from distributions, and model evaluation roads was used in this work in the first attempt to account for road dust suspension in UB. Methodologies and modelling tools The representativeness of the formulas have not been tested for UB conditions, although the The spatial distribution of ground level air resulting modelled PM concentrations, with pollution concentrations in UB is assessed using all sources included, correspond well with the a dispersion model developed for urban scale measured concentrations at the NUM site, the applications. The model, EPISODE, is a Eulerian only site with data available at the time when the grid model with embedded subgrid models for modelling work was carried out. The unpaved calculation of pollutant concentrations resulting 40 Preliminary Results of Assessment Work in 2008 Figure 22: Spatial distribution of suspended PM10 from road traffic in UB. Left: paved roads. Right: unpaved roads. Note the different scales in the two figures from different types of sources (area-, line- and 2008; Slørdal et al, 2008). The AirQUIS system point sources) (Slørdal, et al, 2003). EPISODE is an integrated air quality management system solves the time dependent advection/diffusion on a software platform, described in Appendix H. equation on a 3 dimensional grid. The size The integrated system contains different modules, of the grid elements in UB is 1 1 km2, and including emission inventory module, GIS related the size of the grid is 30 30 km. The model geographical information module, measurement operates with 10 vertical layers, the lowest one module, and the models module. having thickness 20 meters. The model calculates hourly concentrations over the entire modelling The meteorological model TAPM (`The Air period, in this case one full year, based upon Pollution Model', see Appendix H), was used to meteorological and emissions data that are also supplement the locally measured meteorological measured, calculated and input to the model on data for 2007, where data was missing for periods an hourly basis. The hourly concentrations can be extending over several weeks. aggregated to daily, monthly and annual averages. The typical results from the model calculations is iso-lines of ground level concentrations over the The input data to the modelling includes 30 30 km area. geographical data (topography), meteorological data and emissions data, as well as population The EPISODE model is embedded in the air distribution data and measured pollution quality management system AirQUIS (AirQUIS, concentration data. 41 Air Pollution in Ulaanbaatar Figure 23: Modelled spatial distribution of SO2 in Ulaanbaatar, 2007 Isolines of concentration (g/m). Appendix H describes details of the models described in the evaluation section from p. 50. and tools used, as well as the input data. The correspondence between measured and modelled concentrations at the measurement sites Modelling results are given in the next is fairly good. From this, it can not be established sections, partly in the form of maps. On a how the modelled concentrations, especially for background which shows the main road system, PM, correspond with reality in other parts of UB. rivers, railroad and district borders, the maps Preliminary data from the AMHIB monitoring show modelled isolines of concentrations. network for PM may indicate that the model gives too low PM concentrations in the ger areas. This Modelled current air pollution levels could mainly be the result of the uncertainties of and distribution in UB the emissions inventory. The EI, providing spatial and temporal SO2 distributions of emissions to the model grid, is input to the dispersion model system. The annual average SO2 concentrations for 2007 The sections below present modelled spatial are shown in Figure 23 as isolines, based upon distributions for SO2, PM10 and PM2.5. the km2 gridded concentrations. Sources included are ger households, HOBs and CHPs. Including The uncertainties of the EI propagate the traffic SO2 contribution would increase the through the model and give corresponding concentrations in the central UB area somewhat. uncertainties in the modelled concentration distributions. The model is evaluated by means PM10 of comparing the modelled concentrations with those measured, at 4 sites for SO2 and at only Isolines for calculated annual average PM10 one site for PM, the NUM site, located a few concentrations for 2007 are shown in Figure 24. km east of UB centre, the only site at which The spatial contributions from each main source PM measurements were available at the time the are shown in Figures 25­30. modelling work was carried out. The evaluation is 42 Preliminary Results of Assessment Work in 2008 Figure 24: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Isolines of concentration (g/m3). How to read isolines? Modelling results are presented in the form of maps with what look like contour lines on topographical maps. These are called isolines. Like contour lines on topographical maps which connect points with the same elevation, isolines go through the points that have the same pollution concentration. The isolines envelop the areas of consecutively higher concentrations. The maps show that the various sources expose areas in UB differently to various concentrations, dependent upon the location of the sources and the height of the emissions, as well as the wind distribution (direction and speed). One way to read these maps is to take the table of international and Mongolian standards and follow the isolines to see in which specific areas of UB the standards are exceeded. This mapping shows that i) the contribution from pollution sources to average concentrations of air pollution are significantly different and expose different areas to high concentrations; and ii) the effects of air pollution on the population are not distributed evenly and depend on the spatial distribution of ambient air pollution and the spatial distribution of the population. PM2.5 suspension contribution is less for PM2.5 (only 15% of PM10) The northern maximum shows Isolines for calculated annual average PM2.5 up also for PM2.5, as a result of a combination of concentrations for 2007 are shown in Figure 31. ger household and HOB emissions as well as the Contributions from ger households, HOBs, traffic contribution. CHPs and road traffic (exhaust particles and suspension of dry dust from paved and unpaved The PM2.5 distribution is quite similar to roads) have been included. The combined the SO2 distribution, although there are some PM2.5 spatial distribution differs from that of differences because of the traffic contribution to PM10. The PM2.5 exposure in the city centre is PM. Maps showing source-wise contributions are less than for PM10 mainly because the dry dust shown in Figures 31­a to 31­f. 43 Air Pollution in Ulaanbaatar Figure 25: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ ger households. Figure 26: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ heat only boilers. 44 Preliminary Results of Assessment Work in 2008 Figure 27: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ dust suspension from paved roads. Figure 28: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ dust suspension from unpaved roads. 45 Air Pollution in Ulaanbaatar Figure 29: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ vehicle exhaust particles. Figure 30: Modelled spatial distribution of PM10 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ CHPs. 46 Preliminary Results of Assessment Work in 2008 Figure 31: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Isolines of concentration, g/m3. Figure 31-a: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ ger households. 47 Air Pollution in Ulaanbaatar Figure 31-b: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ HOBs. Figure 31-c: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ dust suspension, paved roads. 48 Preliminary Results of Assessment Work in 2008 Figure 31-d: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ dust suspension, paved roads. Figure 31-e: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ vehicle exhaust particles. 49 Air Pollution in Ulaanbaatar Figure 31-f: Modelled spatial distribution of PM2.5 in Ulaanbaatar, 2007 Contributions from each main source category (isolines of concentration (g/m3) ­ combined heat and power plants. The maps show significantly different Evaluation of the air pollution dispersion patterns of PM10 and PM2.5 concentrations for model each main source across the city. PM contributed by ger households are distributed from east The dispersion model used, the EPISODE to west, in the ger areas, to the north, with model, is evaluated against measurements maximum annual average contributions in parts carried out during 2007, the same period as for of these areas of 120 µg/m3. The modelled PM10 the modelling. The model, its input data and contributions of unpaved roads are smaller than evaluation is described in Appendix H. that of paved roads because traffic density on unpaved roads is much smaller than that of It is important to note that any comparison unpaved roads. Should traffic patterns change in of an air pollution model with observations entails the ger areas due to improved transport services or not just an evaluation of the model itself but also increased use of personal cars, without mitigation of the input data used to drive the model, i.e. measures such as paving of roads, the contribution emissions and meteorology. Unfortunately, errors of dust suspension from unpaved roads could rise in input data are difficult to separate from errors dramatically. in the model. When there are sizable uncertainties in the input data then these will propogate Based upon the spatial information in the through the model to the resulting concentration Figures 23­30, the maximum impact areas for fields. each of the main sources are listed in Tables 8 and 9 below: 50 Preliminary Results of Assessment Work in 2008 Table 8: Maximum PM10 concentrations by source and spatial distribution, 2007* Source Maximum Impact Area (see Figures 24­30 above) All sources 190 -- Ger households 120 Extended area from east to west in ger areas in north Paved roads 70 UB centre area, smaller in the ger areas Unpaved roads 30 Across the ger areas HOBs 30 West and north of city centre Vehicle exhaust 10 UB centre area, smaller in the ger areas CHPs 8 Areas south of city centre *The figures are maximum concentration contributions in different areas. Thus, the maximum concentration of the map where all the source contributions have been added, is lower than the sum of all the separate maximums, because they do not occur in the same point (grid cell). Table 9: Maximum PM2.5 concentrations by source and spatial distribution, 2007* Source Maximum Impact Area (see Figure 31, sub a­f above) All sources 90 Ger households 70 Extended area from east to west in ger areas north Paved roads 10 UB centre area, smaller in the ger areas Unpaved roads 4 Across the ger areas HOBs 18 West and north of city centre Vehicle exhaust 10 UB centre area, smaller in the ger areas CHPs 4 Areas south of city centre *The figures are maximum concentration contributions in different areas. Thus, the maximum concentration of the map where all the source contributions have been added, is lower than the sum of all the separate maximums, because they do not occur in the same point (grid cell). Sulfur Dioxide--SO2 sulphur in the coal combusted is the dominant sulphur source. Most of the sulphur in the coal is Emissions of sulphur dioxide, SO2 from fossil fuel emitted as SO2, only a few percent is converted combustion can be relatively accurately assessed, to sulphate (SO4) before being emitted to the when the amount of fuel burnt and its sulphur atmosphere, or kept in the emitted coal particles content is known, as well as the efficiency of any or in the bottom ash. There is some sulphur emissions cleaning equipment used. In UB, the also in diesel and gasoline combusted by road 51 Air Pollution in Ulaanbaatar Table 10: Measured and modelled annual average SO2 (g/m3) for stations UB 1­4, 2007 Mon. Station Measured Modelled Comment UB-1 14.2 19.2 1 Jan­10 Nov UB-2 28.4 35.9 Entire year UB-3 23.0 19.6 Entire year UB-4 31.1 30.6 Entire year vehicles, however, this source is much smaller. its spatial distribution and time variation, are (Guttikunda, 2007) estimated the vehicles source according to the section on emissions inventory of SO2 to be about 6% of the sulphur from the from p. 29. The modelled concentrations follow coal combusted. the seasonal variation of the measurements well, and also reflect the different levels at the four Since SO2 is also rather stable in the urban stations, see Figure 32. There are some deviations atmosphere in smaller cities like UB where the from the measurements in some periods, such as residence time of the gas is short, typically a for the UB-2 station, located a bit to the west of couple of hours, SO2 is a suitable compound for the UB centre area: overestimation during early validating atmospheric models. January and in November­December, and in early March there is overestimation at 3 stations, Table 10 shows the comparison of modelled UB-1,2 and 3, but not at UB-4. Apart from with measured SO2 concentrations at the four this, the correspondence between measured and CLEM monitoring stations (Stations UB 1­4, modelled concentrations is quite good. Table 10 located as shown in Figure 7 on p. 15). The shows measured and modelled annual averages. measured concentrations were adjusted according to the preliminary results of the comparative These results indicate that the model and results described in section on SO2 concentrations emission data represent the observed seasonal on p. 21 by multiplying with a factor 2. This trends and annual means of SO2 to within the obviously has to be confirmed through further expected uncertainty. However, given the current quality checking. As a result, the uncertainty of the uncertainty in the monitoring data, as well as measured SO2 concentrations are considered high the emissions, it is not possible to give a more until further quality control is carried out through conclusive assessment concerning the validity of a comparison study. The measured SO2 levels vary the emissions. considerably between the sites, with high winter concentration levels from as low as 50 µg/m3 at PM10 the UB-1 station to about 100 µg/m3 at the UB-2 station. The seasonal variation is similar at all PM was in 2007 measured at only one station, stations (Figure 32), with very low concentrations the NUM station at the National University of in the summer, reflecting the low consumption Mongolia. As described above (section from of coal for ger heating and for HOBs then. The p. 14), the NUM sampler gives reasonably CHPs are also operating in the summer, and the good data for PM10, while it underestimates the low summer concentrations reflect the limited concentration of PM2.5. Thus comparison of contribution from the CHPs to ground level modelled concentrations with measured ones can concentrations, because of the tall stacks. only be done for PM10. The emissions that are input to the model Figure 33 shows measured data and modelled calculations, total emissions per source and contributions to PM10 from a number of sources. 52 Preliminary Results of Assessment Work in 2008 Figure 32: SO2 concentrations at stations UB 1­4 Measured and modelled daily average concentrations, 2007 (g/m3). Figure 33: Measured and modelled PM10 (daily average) at the NUM measurement site Contributions from various sources. 53 Air Pollution in Ulaanbaatar The measurements are taken generally over coefficient of determination (R2), that represents 24 hours on two days per week. On days with how much of the observed variability is explained very high PM2.5 concentrations, the sampling by the model, is quite low at 14%. This reflects period is shorter, down towards 6­8 hours from the models inability to capture the day-to-day the starting time, due to sampler clogging. The variations in the emissions. This is particularly brown line in Figure 33 shows the total modelled important for resuspension, which is strongly PM10 concentration, with contributions from Ger dependent on surface conditions, though these and kiosks coal and wood, HOBs, CHPs, vehicle are not described in the model. The fact that exhaust particles and suspended dust from roads. a much larger amount of the SO2 variability is explained by the model, 37­55 %, indicates that Note that Figure 33 represents the NUM the major part of the unknown variability does site only. As we saw in the modelling section not come from the coal combustion sources but from p. 40, the contributions from the various is, as indicated, the result of uncertainties in the PM sources to ground level concentrations varies emission of PM10 through resuspension. substantially throughout the city: e.g. in ger areas the ger household emissions will dominate more, while in the city centre the road dust suspension Despite the strong variability, the model and exhaust emissions will be more dominant. predicts an annual average PM10 of 163 µg/m3, all source contributions added. The model overestimates the measurements The measurements give 157.7 µg/m3 (see early in the year (January­February) and Table 11 below). Each of the modelled source underestimates at the end of the year (November­ contributions are based upon separate scientific December) 2007. Inspection of the measurements considerations and upon the input in the as shown in section from p. 15 shows that the emissions inventory section from p. 29. The amount of suspension of dry dust (the coarse estimated annual average concentration from fraction of PM) was small in January­February the measurements is made up of data from only and very large in November­December in that 2 days per week and thus has a considerable year. Suspension of dust from surfaces is naturally uncertainty, while the modelled annual average a very non-steady mechanism. Dust is building is based upon hourly data throughout the up on the surfaces during humid and low wind year. The uncertainty related to the 2-day- periods, and then released when exposed to per-week sampling instead of all days in the turbulence when it is very dry. Data on dryness/ year is estimated to about +/­ 20%, assuming freezing/wetness conditions on the roads have the variability of the about 73 samples is not been collected so far. Our model is presently representative of the other 292 days when no run for constant dry conditions where the dust samples were taken. This comes in addition is always available for suspension, i.e. a steady to the the uncertainty in each of the sampled suspension from the roads, hour-by-hour only values, which can be estimated to be about dependent upon traffic amount. Thus, as the +/­ 15­20%, giving a composite uncertainty of model is run presently, it does not necessarily about 26%. Another factor, which is discussed reproduce the winter-time suspension dynamics in Appendix C, is that the sampler used for these day-by-day. It is possible to refine these model measurements operates, when the concentrations runs, if relevant information on road conditions are high, only during part of the day, from day-by-day can be found. During the summer 10 AM and then 6­8 hours. Due to the daily period, suspension is the completely dominating cycle of the concentrations (see Figures 11 and PM10 source. 12) this will introduce an extra uncertainty when comparing the daily mean values that Figure 33 shows that the model estimates the are predicted in the model. On this basis, the average summer-time PM10 reasonably well. agreement between the measured and modelled annual average is considered to be within the Correlation between the measured and the uncertainties of the measurements and the modelled PM10 concentrations is quite low, with a emissions. 54 Preliminary Results of Assessment Work in 2008 required for a thorough assessment of this and Table 11: Measured PM10 and modelled source will be incorporated into the Final Report due in contributions at the NUM station, 2007 (g/m3) early 2010. Measured Modelled The dispersion modelled PM10 Concentration 157.7 163 concentrations give the following contributions from the main source categories at the NUM Source contributions station, as summarised in Table 11: Coal and wood combustion 36% 44% Coal and wood 71 µg/m3 or 44% Suspended dust 58% 50% combustion: Vehicle exhaust 6% 6% Suspended dust: 82 µg/m3 or 50% Vehicle exhaust: 10 µg/m3 or 6% The relative contributions of the different The statistical source apportionment from modelled sources to PM10 can also be compared the NUM measurement data give a basis for to source apportionment studies made at the estimating the following contributions to PM10: NUM station (Table 11). The modelled source 36% for combustion particles, 58% for the soil contributions, on average, are very similar to those and construction particles and 6% for motor estimated from the monitoring data and well vehicle particles. Our model thus gives about within the uncertainties related to the data. This 10% higher combustion particle contribution at further supports the model results for the total the expense of the soil/suspension contribution, PM10 concentration, indicating that the model while the vehicle exhaust particle contribution is producing realistic long-term average source is the same for the two methods. The dispersion contributions to PM10, within the uncertainties modelling does not include construction particles. related to both the monitoring and the emissions. This is important for the application of the model As mentioned, the preliminary data from to assess the source contributions to population the AMHIB monitoring network for PM10 and exposure and to provide useful information for PM2.5, indicate that the PM concentrations in the abatement analysis. ger areas could be considerably higher than has been modelled here. Further evaluation of the The available data is very limited for making model can be made after improvements of the EI any solid conclusions concerning the performance has been carried out and when quality assured of the model for PM10. More information will be AMHIB data are available. 55 4. Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs W hen faced with choices between Improvements in the emissions inventory may proposed abatement measures, result in an even increased importance of the ger policymakers should use a basis stove and road/soil dust suspension sources for the for selection. The core of a local average population exposure to PM. air pollution abatement program is its ability to reduce pollution and the harmful effects it has Global Concerns. Considering global climate change on the population. The selection criteria could be concerns and that the nature of the problem in UB is the a) the degree to which the abatement measure, burning of carbon, reducing carbon emissions could also be or package of measures, moves toward meeting considered as a selection criteria. Mongolian or International AQS across all of UB and/or b) the degree to which an abatement measure, or a package of measures, can reduce Abatement scenarios and reductions in PM health costs, which may mean that, depending pollution levels compared to Mongolian and upon the health cost reduction target, not all International Air Quality Standards (AQS) of UB would meet air quality standards due to the spatial distribution of the population and For the purpose of estimating the emission pollution. reductions needed to reach Mongolian and international standards for the entire UB The types of assessments shown below population, Figure 34 shows the annual average provide policy makers with options regarding how concentration in the km2 grid cell with the to meet their targets, whether they are based on highest concentration, for the 2007 situation as reaching concentration targets or on health effects the basis, and for the various reduction scenarios, reductions. which were: The indication from the first preliminary Impacts of reducing emissions by 30%, data from the AMHIB PM monitoring network 50%, and 80%--irrespective of abatement is that the PM concentrations in the ger areas are method--from each of the main sources, underestimated. The main combustion source to respectively: ger heating systems, HOBs, PM in the ger areas is the ger stoves themselves, CHPs and dust suspension from paved roads, while suspension of dust from roads and surfaces and is the main source to the coarse fraction. Both of Impacts of reducing emissions in all four these sources could be underestimated in the ger main sources by 30%, 50%, and 80%-- areas. The scenario analysis below shows that even irrespective of abatement method. if the ger stoves source strength is underestimated, it is still the main combustion source to the UB The highest modelled annual average grid population exposure to PM, and the combustion cell concentration is 211 µg/m3 for PM10 and source which is most important to control. 106 µg/m3 for PM2.5. The figure compares these 57 Air Pollution in Ulaanbaatar Figure 34: The highest annual average PM10 (top figure) and PM2.5 (bottom figure) concentration in any grid cell in UB, 2007 and for various abatement scenarios (g/m3) annual average concentrations with various estimate based upon assumed functioning PM standards. On the basis of the available data cleaning equipment. The higher emission estimate used in the model, the ger stove abatement gives based upon the JICA testing will increase the the largest effect to reduce the concentrations. calculated reductions in concentrations resulting The WHO IT-1 target can be met only with from percentage reductions in CHP emissions, an 80% reduction in the emissions from all the by about a factor of 3. This should be taken into four sources. To meet the Mongolian standard account in the consideration of the results below. in the entire UB area, even larger reductions are The CHP contribution is important, although needed. the mentioned correction will not change substantially the overall conclusions, since the Regarding the contributions from the CHPs, CHP contribution to ground level concentrations the analysis here is based upon the low emission is still limited. 58 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Abatement scenarios, spatial impacts and A 75%Ger/83%HOB/50% dust scenario-- reductions in health costs illustration of spatial impacts of abatement measures Population weighted average exposures, PWE, (see Chapter 1) were calculated with the The Discussion Paper simulates the impacts of dispersion model to assess the health effects as a reducing emissions by 75% from ger households, result of simulated emissions reductions. 83% from heat only boilers and 50% from dust suspension on paved roads. These are emission The purpose of the simulations is to reductions based on technical options currently demonstrate the spatial impacts of interventions discussed in UB, including combined fuel and and health impacts. stove switching programs, replacing boilers, and cleaning roads. The simulations are the same as in the section above and an additional simulation: The reduction resulting from each of the interventions is substantial, especially from (To demonstrate spatial impacts) Impacts interventions in the gers and HOBs. With all of reducing emissions by 75% from ger three interventions, the population weighted PM households, 83% from heat only boilers and are reduced by 60­65%. The remaining PWE is 50% from suspended dust on roads 30.3 µg/m3 for PM10 and 12.4 µg/m3 for PM2.5 Impacts of reducing emissions by 30%, (Table 12). This can be compared with the WHO 50%, and 80%--irrespective of abatement Guidelines, which are 20 µg/m3 for PM10 and method--from each of the main sources, 10 µg/m3 for PM2.5. The WHO Guidelines are respectively: ger heating systems, HOBs, set at the level where only minimal health effects CHPs and dust suspension from paved roads, should occur. The remaining PM pollution after and these three interventions is still associated with Impacts of reducing emissions in all four some health effects on the UB population. main sources by 30%, 50% and 80%-- irrespective of abatement method. However, since PWE is a weighted average based on the distribution of the population, Based on available data used in the model, a significant part of UB will remain exposed the PWE in 2007 for PM10 is 76.7 µg/m3 to concentrations above the average with the and for PM2.5 is 37.6 µg/m3, which are much highest modelled PM10 concentrations ranging lower numbers than the maximum grid average between 70­80 µg/m3 and average PM10 concentrations of 211 µg/m3 and 106 µg/m3, concentrations in central UB ranging from respectively, in section 4.1. The figures need to be 40­60 µg/m3. treated separately because they are very different concepts. This is discussed below. Spatial distribution of PM10 under the 75%/83%/50% Scenario The maximum annual average concentration indicator is the highest cell in the grid and Figure 35 shows the reductions in the PM10 this is what the Mongolian standard should be concentrations in the grid resulting from the compared with. The population weighted average interventions, for each of the three interventions is calculated by multiplying each grid PM value separately, as well as for all three combined. The with the fraction of the population living in that ger stove/fuel improvements give the largest grid. This weighted average is a good estimate reduction, up to 90 µg/m3 in the most affected for the exposure that the population as a whole area, while the intervention to reduce dry experiences in UB. This indicator is used for dust suspension from roads gives the smallest estimating the resulting population-wide health reduction, concentrated in the central UB area effects of the reduced PM pollution (see below). where most of the traffic is. 59 Air Pollution in Ulaanbaatar Table 12: Population weighted average PM concentrations (PWE) in Ulaanbaatar, and reductions from the 75%Ger/83%HOB/50% dust intervention scenario (g/m3) PM10 PM2.5 Population weighted average, PWE, 2007 76.7 37.6 Reductions in PWE from interventions: Ger stoves 32.7 19.6 HOB 8.7 5.2 Dust suspension 5.0 0.4 Total reduction 46.4 25.2 Remaining PWE 30.3 12.4 The total effect of the three interventions is The total effect of the three interventions is to reduce the PM10 concentration by as much as to reduce the PM2.5 concentration by as much as 120 µg/m3 in the most affected area (north of UB 60 µg/m3 in the most affected area north of UB centre). This corresponds to about 65% of the centre. As for PM10, this corresponds to about present PM10 concentrations. 65% of the present PM2.5 concentrations (see Chapter 3.8.2). Figure 36 shows the remaining PM10 concentrations after the three interventions. The Figure 38 shows the remaining PM2.5 highest concentrations then would be 70­80 concentrations after the three interventions. The µg/m3, while the average concentration in the highest concentrations then would be 30 µg/m3, UB central area would be about 40­60 µg/m3. while the average concentration in the UB central area would be about 20­30 µg/m3. These concentrations are still high compared These concentrations are still high compared to the WHO and other guidelines. However, they to the WHO and other guidelines. However, as is approach the WHO Interim Target IT-2 of the case also for PM10, they approach the WHO 50 µg/m3, and the EU Limit Value of Interim Target IT-2 of 25 µg/m3, and the EU 40 µg/m3. The Mongolian AQ standard of Limit Value of 20 µg/m3. The Mongolian AQ 50 µg/m3 would still be exceeded moderately in standard is exceeded only moderately. much of the central- northern of UB city area. The 30%/50%/80% scenarios--illustration Spatial distribution of PM2.5 under of health impacts of abatement measures the 75%/83%/50% Scenario In order to make preliminary estimates of Figure 37 shows the reductions in the PM2.5 associated health costs, the Discussion Paper uses concentrations in the grid resulting from the the population weighted average PM concentration, interventions, for each of the three interventions in short: PWE. Compared to annual average separately, as well as for all three combined. As concentrations, the PWEs more accurately reflect for PM10, the ger stove/fuel improvements gives exposure of UB's population to air pollution the largest reduction, up to 50 µg/m3 at the most by adjusting for the spatial distributions of the affected area. The intervention to reduce dry dust pollution and the population. suspension from roads gives only a very small reduction since only 15% of the suspended dust is The PWE data in Table 13 are used to in the PM2.5 fraction. estimate health benefits of the reductions. 60 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Figure 35: Calculated reduction in the PM10 grid resulting from the interventions (g/m3) Ger stoves emissions: --75% HOB emissions: --83% Continued 61 Air Pollution in Ulaanbaatar Figure 35: Continued Dust suspension from roads: --50% PM10 reduction for all 3 interventions 62 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Figure 36: Calculated remaining PM10 concentrations in the grid after implementing all three interventions The reduction in the PWE shows that while reduction scenarios. These reductions drive Ger interventions give the largest reduction, the health cost calculations given in the next interventions in HOBs and dust suspension are section. However, Figure 39 could be used to also more important than their relative shares give an indication of the relative health impacts in average concentration contributions would of different interventions based on the spatial initially suggest, especially for PM10. This is distribution of the population and the spatial because the central areas where the HOB and distribution of PM and its sources. For example, a suspension sources have a larger PM contribution, 30% reduction in ger emissions would result in a have a larger population density than the ger reduction in the population weighted exposure of areas. PM10 by 17%, compared to a maximum average concentration reduction of 19% (Figure 34). Figure 39 should be interpreted carefully, This, however, does not correspond directly with especially when comparing it to average annual a reduction in health costs because these also concentrations shown in Figure 34 above. It depend on a variety of factors. The 30% example is not correct to compare directly Figure 35 reduction yields a 13% reduction in calculated with Mongolian AQ standards or international health costs as shown in the next section. standards. Portions of the population still remain exposed above the standards even when the Avoided premature deaths, cases of chronic weighted average and the standard values are bronchitis, and hospital admissions equal. A large number of studies were conducted and are Figure 39 shows the population weighted being conducted around the world to document average concentrations under different emission a consistent association between elevated 63 Air Pollution in Ulaanbaatar Figure 37: Calculated reduction in the PM2.5 grid resulting from the interventions (g/m3) Ger stove emissions: --75% HOB emissions: --85% Dust suspension from roads: --50% PM2.5 reduction for all 3 interventions ambient PM10 and PM2.5 levels to an increase in UB, both are true, especially in the winter season, mortality rates, respiratory infections, number with high ambient PM concentrations and people and severity of asthma attacks and the number of in high density areas of Gers being constantly hospital admissions.13 Actual health impacts of exposed to them. air pollution are determined by two factors, i.e., by sufficiently high concentrations of pollutants in the atmosphere and the presence of people in This analysis applies the exposure-response the region affected by these pollution levels. In functions applied in the assessment of costs of air pollution in China by the World Bank (2007) to calculate the number of premature deaths 13 OECD. 2000, `Ancillary Benefits and Costs of Greenhouse Gas Mitigation', Proceedings of an IPCC co-sponsored (i.e. deaths brought forward due to air pollution workshop, Washington, DC, USA exposure), new cases of chronic bronchitis, 64 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Figure 38: Calculated remaining PM2.5 concentrations in the grid after implementing all three interventions and hospital admissions for respiratory and are estimated to establish a baseline value and cardiovascular diseases, that can be avoided by then to determine what are considered "excessive" implementing the interventions described above. ambient concentrations. The excessive ambient Premature deaths and enhanced rates of chronic concentrations are in some ways subjective obstructive lung diseases (of which chronic because they determine the level of pollution bronchitis typically is the most prevalent) are concentrations that are deemed by society to two major health impacts associated with long- be unacceptable. Proxies for this excessive level term exposure to PM pollution. In addition a could be WHO or other international guideline range of other impacts, e.g., acute respiratory values or local standards. WHO guideline infections and more frequent asthma attacks, are values are typically determined by the level of found to be linked to air pollution exposure on pollution concentrations that are identified in a shorter time-scale. The different short-term epidemiological studies as `threshold' levels for and long-term symptoms and diseases can lead observable effects. Thus, they are a metric for to school and work absenteeism, emergency the actual physical impacts rather than what room visits and hospitalization, which will have may be defined as the target or acceptable level an economic impact on society. Using data for as defined in specific settings. This study uses number of hospital admissions for cardiovascular 20 µg/m3 PM10 and 10 µg/m3 PM2.5 as threshold and respiratory diseases in district level secondary levels for the effect (the WHO guideline hospitals, we include these end-points in the values). Second, the excessive concentrations calculation below. are combined with the population at risk of exposure and dose responses--the estimated In this Discussion Paper, health impacts health impacts of exposure to excessive levels of are estimated in the following way: First, the ambient concentrations of PM. The AMHIB corresponding changes in ambient concentrations is currently reviewing public health records 65 Air Pollution in Ulaanbaatar Table 13: Population weighted average PM concentrations in Ulaanbaatar, and reductions from abatement scenarios (g/m3) PM10 PM2.5 Present situation (2007) 76.7 37.6 Reductions in population weighted average from interventions: 30% reduction of Ger stoves 13.1 7.9 50% reduction of Ger stoves 21.8 13.1 80% reduction of Ger stoves 34.9 20.9 30% reduction of HOBs 3.1 1.9 50% reduction of HOBs 5.2 3.1 80% reduction of HOBs 8.3 5.0 30% reduction of power plants14 0.1 0.06 50% reduction of power plants 0.17 0.1 80% reduction of power plants 0.27 0.16 30% reduction of suspended dust 3.0 0.2 50% reduction of suspended dust 5.0 0.4 80% reduction of suspended dust 8.0 0.6 30% reduction of all 4 sectors 19.3 10.0 50% reduction of all 4 sectors 32.2 16.7 80% reduction of all 4 sectors 51.4 26.7 30% reduction of Ger stoves and 80% reduction of HOBS 21.4 12.8 80% reduction of Ger stoves and 30% reduction of HOBS 38.0 22.8 to evaluate correlations between incidences of should help to understand better the orders of hospital admittances for various respiratory magnitude in the trade-offs between various costs illnesses and relatively high concentrations of of abatement options and their corresponding PM--to provide more context to this analysis benefits. since epidemiological studies establishing local exposure-responses take years to complete. Third, The analysis presented in this Discussion a quantitative value is calculated on the health Paper is restricted to the three major so-called impacts using mathematical tools and generally "health end points"--premature deaths, chronic accepted methodologies explained in more bronchitis and hospital admissions. Based on detail below. While not perfect, these exercises exposure-response functions15 from the literature, 14 Regarding the contributions from the CHPs, the analysis substantially the overall conclusions, since the CHP here is based upon the low emission estimate based upon contribution is still limited. well functioning PM cleaning equipment. The higher 15 Exposure-response functions measure the relationship between emission estimate based upon the JICA testing will increase exposure to pollution as a cause and specific outcomes as an the calculated reductions in PWE resulting from percentage effect. They refer to damages/production losses incurred in a reductions in CHP emissions, by about a factor of 3. This year, regardless of when the pollution occurs, per unit change should be taken into account in the consideration of the in pollution levels. In this table, the function is defined as results below. The CHP contribution is of importance, the percentage change in effects incurred per unit change in although the mentioned correction will not change concentrations (µg/m3) per capita. 66 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Figure 39: Population weighted PM10 (top figure) and PM2.5 (bottom figure) reductions resulting from the 30%/50%/80% scenarios health impacts are derived using the equation RR is the relative risk of health effect between two below. levels of pollution (here the current level and a lower level obtained from an intervention or the E ((RR ­ 1)/ RR)* fp * POP lower threshold level), fp is the current incidence rate of the health effect, and POP is the exposed where E is the number of cases of each health end population considered (for hospital admissions point attributed to air pollution (`excess cases'), we replace fp*POP with the annual number of 67 Air Pollution in Ulaanbaatar hospital admissions (see below)). Except for the annual number. Note that the hospital admissions mortality function, where we rely on WB (2007) estimates most likely represent only a fraction (except we use 20 µg/m3 as a threshold level of the entire effect related to this end point instead of 15), RR is given by: since only district level secondary hospitals are included. RR exp( *(C ­ Ct)) Since no long-term epidemiological cohort where is the exposure-response coefficient (see studies on mortality rates and air pollution have Table 12 where betas are given as percentage been carried out in Asia, the well-known, large values), C is the current pollution level and Ct study in the USA by Pope et al. (2002) is used by is the target pollution level obtained from an the WB (2007) to establish an exposure-response intervention or from reaching the threshold function. A direct application of the exposure- value. We calculate the remaining number of response function in Pope et al. may lead to cases attributable to air pollution after each implausibly high damage estimates in polluted intervention, and derive the number of cases that regions in Asia, and the US results were therefore can be avoided by subtracting these figures from calibrated towards the few cross-sectional studies the calculated excess cases in the current situation on mortality rates that were available for high (which is calculated by using the threshold pollution cities in China (see WB (2007) for levels described above). To determine costs of air details). The result is an exposure-response pollution, this Paper uses a willingness-to-pay function that flattens towards higher PM10 levels. methodology (see, e.g., WB, 2007) to monetize However, there are particularly large uncertainties health impacts and estimate the economic value related to this adjustment. New findings from of avoided health damage. short-term studies in Asia find that the pattern of the exposure-response functions appears linear Due to the considerable attention these over a fairly large range of ambient concentrations preliminary calculations may have when up to and sometimes exceeding 100 µg/m3. disseminated, it is necessary to provide some In addition to the estimated premature deaths background in academic literature to disclose key resulting from the adjusted exposure-response assumptions so that others could use this work function in WB (2007), this Discussion Paper and improve it. provides an estimate of the health effect using the original (linear) exposure-response function from An important reason for limiting the number Pope et al. (2002). This indicates to some extent of health end points is the lack of background the sensitivity to the final results of the choice of data in UB when it comes to prevalence rates function for the mortality impact. We use a cut- for different diseases, hospitalization rates, etc. off of 20 µg/m3 PM10 when estimating mortality Nevertheless, assumptions are made in this impacts (10 µg/m3 for PM2.5 in the sensitivity test Discussion Paper about the prevalence of chronic based on the approximate PM2.5/PM10 ratio) (see bronchitis in UB. According to Lopez et al. above). (2006) the prevalence rate in China and Mongolia is around 3% in adults above 30 years of age, with The exposure-response functions for large uncertainties in the estimate. In WB (2007) chronic bronchitis, hospital admissions for a prevalence rate of 3.4% and a corresponding respiratory diseases, and hospital admissions for annual incidence rate (new cases per year) cardiovascular diseases (CVD) applied in WB of 0.15% were used. This incidence rate was (2007) are based on a meta-analysis of several assumed in this Discussion Paper. As mentioned Chinese studies (Aunan and Pan, 2004). above, we used data for the number of hospital admissions for cardiovascular and respiratory As done in previous applied studies (e.g. diseases in district level secondary hospitals in Mestl et al, 2004; Kan et al, 2004), the Discussion UB in the calculation below. As the data were for Paper uses as mentioned the PWE estimates 6 months (1 June 2008 till 15 December 2008), for the whole region (i.e. the concentration we multiplied with 2 to obtain an estimate of the times population in each grid averaged for the 68 Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs Table 14: Exposure-response coefficients (% change in incidence of health effect per g/m3 PM10), baseline incidence rates, willingness-to-pay (WTP) for avoiding premature death (long-term effect) and chronic bronchitis, and Cost of Illness (COI) of hospital admissions Exp-resp Health end point coefficient (PM metric) Baseline incidence rate (USD per case) Premature death (WB method) See WB (2007) 0.0067 200,133 (WTP) Premature death (Pope et al. (2002) 0.6 (PM2.5) 0.0067 200,133 (WTP) Chronic bronchitis 0.48 (PM10) 0.0148 64,042 (WTP) Respiratory hospital admissions 0.12 (PM10) See text 800 (COI) CVD hospital admissions 0.07 (PM10) See text 1300 (COI) total population in all grids) as input to the in 2007 was 1507 USD (WB database). This health benefit analysis. Given that the exposure- yields an estimated VSL of about 200,100 USD response function for mortality (taken from WB, for Mongolia. As in WB (2007) we multiply 2007) method) is non-linear, this Discussion this value with 0.3218 to obtain an estimate of Paper's results probably deviate slightly from the WTP for avoiding a new case of chronic the results that would have been obtained using bronchitis (see Table 14). To obtain an estimate of geographically disaggregated PWE values. This the Cost of Illness related to hospitalization, we uncertainty, however, is regarded by the team to adjust the estimates from China (WB, 2007) in be minor compared to other uncertainties in the the same way as was done for VSL. analysis. Table 15 shows the estimated number of Monetized health benefits cases attributable to PM pollution in the current situation and the number of cases that can be This Discussion Paper relies also on WB (2007) avoided from implementing the interventions to derive the unit costs of a premature death, new described above. The current health damage cases of chronic bronchitis, and hospitalization. corresponds to 8.0% of the GDP in UB19 and Based on willingness-to-pay studies in China 3.8% of GDP in Mongolia. In the sensitivity (WB, 2007) which derives a Value of Statistical calculation using the linear exposure-response Life (VSL) of 1.4 million Yuan, i.e. the value function from Pope et al (2002) directly, current placed on avoiding premature death. We use damage corresponds to 12.2% of GDP in UB the ratio between this value and the GDP/ and 5.7% of GDP in Mongolia. The maximum cap in China combined with the GDP/cap of achievable benefit (80% reduction in all 4 sectors) Mongolia in 2007 to calculate the corresponding corresponds to 6.6% of GDP in UB in 2007 VSL for Mongolia.16 In addition to this direct (3.1% of GDP in Mongolia). In the sensitivity conversion, we in a sensitivity estimate assume an income elasticity of WTP of 0.5.17 The VSL/ GDP per capita ratio is 133 for China (2003 18 This factor is derived from a study indicating that people's choices imply that the utility of living with chronic bronchitis figures), while the nominal GDP/cap in Mongolia 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 risk-risk and risk-dollar tradeoffs for chronic bronchitis." Journal of Environmental 16 I.e. VSL (Mongolia) = [VSL(China)/GDP per cap Economics and Management 21:32­51.) (China)]*GDP per cap (Mongolia). 19 GDP in UB and Mongolia was 2.16 billion and 3.9 billion, 17 A range of studies indicate the income elasticity of WTO respectively, in current USD (WB database, for 2007, is below unit (see e.g. Pearce et al (eds), 2002. Valuing the available: http://econ.worldbank.org/WBSITE/EXTERNAL/ environment in developing countries. Case studies. Edward EXTDEC/0,,menuPK:476823~pagePK:64165236~piPK: Elgar Publishing. 567 pp.) 64165141~theSitePK:469372,00.html). 69 Air Pollution in Ulaanbaatar calculation this figure gets 11.6% of GDP in UB Mongolia. The maximum achievable benefit (5.4% of GDP in Mongolia). In the sensitivity (80% reduction in all 4 sectors) corresponds to calculation where we assume an income elasticity 3.3% of GDP in UB in 2007 (1.6% of GDP in of WTP of 0.5, current damage corresponds Mongolia). to 4.1% of GDP in UB and 1.9 % of GDP in 70 Table 15: 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 million USD) Share of GDP in Ulaanbaatar Annual number of cases Monetized (million USD) (2007) Hospital Hospital All-cause admissions Hospital All-cause admissions Hospital mortality Chronic (respiratory admissions mortality Chronic (respiratory admissions (chronic) bronchitis disease) (CVD) (chronic) bronchitis disease) (CVD) SUM 2007 (current 614 379 735 448 123 24 0.59 0.58 148 8.0 % health damage) 30% reduction of 83 71 165 102 17 5 0.13 0.13 21 1.2 % Ger stoves 50% reduction of 149 121 277 170 30 8 0.22 0.22 38 2.1 % Ger stoves 80% reduction of 273 201 446 274 55 13 0.36 0.35 68 3.7 % Ger stoves 30% reduction of 18 16 39 24 4 1 0.03 0.03 5 0.3 % HOBs 50% reduction of 31 28 65 40 6 2 0.05 0.05 8 0.4 % HOBs 80% reduction of 51 45 104 65 10 3 0.08 0.08 13 0.7 % HOBs 30% reduction of 1 1 1 1 0 0 0.00 0.00 0 0.0 % power plants 50% reduction of 1 1 2 1 0 0 0.00 0.00 0 0.0 % power plants Abatement Scenarios and Their Benefits in Terms of Reduced Health Costs 71 Continued Table 15: Continued 72 Share of GDP in Ulaanbaatar Annual number of cases Monetized (million USD) (2007) Hospital Hospital All-cause admissions Hospital All-cause admissions Hospital mortality Chronic (respiratory admissions mortality Chronic (respiratory admissions (chronic) bronchitis disease) (CVD) (chronic) bronchitis disease) (CVD) SUM Air Pollution in Ulaanbaatar 80% reduction of 2 1 3 2 0 0 0.00 0.00 0 0.0 % power plants 30% reduction of 18 16 38 23 4 1 0.03 0.03 5 0.3 % suspended dust 50% reduction of 30 27 63 39 6 2 0.05 0.05 8 0.4 % suspended dust 80% reduction of 49 43 101 62 10 3 0.08 0.08 13 0.7 % suspended dust 30% reduction of 129 107 245 151 26 7 0.20 0.19 33 1.8 % all 4 sectors 50% reduction of 244 184 411 252 49 12 0.33 0.33 61 3.3 % all 4 sectors 80% reduction of 504 308 664 406 101 20 0.53 0.52 122 6.6 % all 4 sectors 30% reduction of 146 119 271 167 29 8 0.22 0.22 37 2.0 % Ger stoves, and 80% reduction of HOBS 80% reduction 308 220 487 299 62 14 0.39 0.38 76 4.1 % of Ger stoves, and 30% reduction of HOBS Conclusions This Discussion Paper lays out a systematic When faced with choices between proposed approach that the AMHIB project will follow abatement measures, policymakers should use through to its conclusion in early Fall 2010. a basis for selection. At the core of a local air The World Bank and its AMHIB partners pollution abatement program is its ability to invite comments on this approach and analysis reduce pollution and the harmful effects it has so that it can be as helpful as possible to action on the population. The selection criteria could be planning. a) the degree to which the abatement measure, or package of measures, moves toward meeting There is a need to set socially acceptable, Mongolian or International AQS across all of technically feasible emission reduction targets UB; b) the cost of abatement measures per unit to give a clear direction for action plans. Targets of emission reduced; and/or c) the net benefits of will be determined by technical options and abatement measures, or package of measures. the ability and willingness to pay for pollution reduction by civil society. The costs of air Due to the spatial distribution of the pollution are paid from the pocketbook, the population and UB's pollution, short term budget and future health costs through higher strategies could achieve improvements in a incidences of pollution related illnesses. What significant share of the city even though all parts and how to pay for air pollution is a choice to of the city might not meet air quality standards be made by civil society and its representatives. evenly. Due to the complex nature of air pollution, an open discussion of options and their estimated Additionally, the high peaks in daily air impacts based on an analytical framework pollution observed coincide with observed using best available data is recommended. Cost emission peaks from the ignition and reloading effectiveness analysis (cost per unit emission phases of the burn cycle in heating. Because reduced) and estimating avoided health costs these peaks comprise a significant share of PM of each policy option can, together with other average concentrations in wintertime, focusing factors considered important to UB's citizens, be on the ignition and reloading phase of the burn considered in choosing clean air strategies. Setting cycle in abatement design may be a good strategy. targets that have been openly discussed helps Additional testing is needed to confirm this build widespread support for pollution abatement indication. activities that involve asking people to change environmentally damaging behaviors. Many in Data quality needs improvement and this civil society, especially the poorest in UB, will be is reflected in the uncertainty in the modelling asked to change their behavior in some way to of this Discussion Paper, which will be improved improve air quality. They should become active in the Final Report, due in early 2010. This allies in the reduction of air pollution in UB. assessment of the air quality situation in UB 73 Air Pollution in Ulaanbaatar and of the effects of some selected abatement The cost of health damage attributable to measures have been based upon a wide range current levels of air pollution in terms of of existing data, reports and information, particulate matter is estimated to be 8.0% as described in this Discussion Paper. The of UB's GDP in 2007, or US$ 148 million assessment followed the basic concept for doing (preliminary estimate). This is 3.8% of air quality management work. This includes Mongolia's GDP. Using sensitivity analysis, looking at monitoring data for air pollutants this estimate could fall to 4.1% of UB's GDP and meteorology, emissions inventorying, or US$ 76 million, or about 1.9% of national dispersion modelling and calculating pollutant GDP. concentrations and their distribution spatially The maximum achievable benefit from the and temporally, and the contributions from the described interventions (80% reduction in all various main source categories. Such calculations 4 sectors) is estimated to be 6.6% of GDP in are done for the current situation (for which UB in 2007, or 122 million US$ (preliminary the year 2007 was used, the latest year with an estimate). This is 3.1 % of Mongolia's GDP. extensive data base) as well as for the situation assuming some selected abatement measures are It is recommended that policy makers implemented. Calculations were also made of the set targets such as the following and open a reduction in population weighted exposure to discussion with civil society on the costs and PM for a number of abatement scenarios, and the benefits. These targets can be adjusted as better corresponding benefits in terms of reduced health data are obtained: costs. There are shortcomings in the needed input data for this kind of assessment. The main set targets that would reach Mongolian Air shortcomings are related, critically, to air pollution Quality Standards as soon as possible and monitoring data and emissions data. Therefore, PM2.5 targets by 2020. while this assessment can be considered complete, to achieve these targets, reductions of the this Discussion Paper provides what should be emissions from the ger heating systems considered only preliminary estimates. and the HOBs as well as dust suspension reduction from roads and near-road surfaces Given what is known, and based upon are all needed. available data, the following preliminary recognizing socio-economic constraints in conclusions can be drawn: UB, it is further recommended that interim targets are set of a reduction in emissions UB is definitely one of the most polluted of 50% across all four sectors resulting in cities, and it might be THE most polluted PM2.5 below 60 µg/m3 and PM10 120 µg/m3 city, in the world in terms of annual which would visibly improve smoke in UB. Particulate Matter concentrations and Currently calculated health costs (2007) of its severity is driven by arguably extreme US$ 49 million equivalent would be saved. wintertime PM concentrations. target the ger areas immediately, where The needed effort to reduce air pollution is pollution reduction benefits are greatest. considerable. An 80% emissions reduction target boilers also as they have a relatively across all four main sources of air pollution larger impact on the population due to their could come close to reaching the Mongolian distribution among population centers. Air Quality Standards in most of the UB install continuous emission monitoring city area, which are equivalent to the middle systems in the power plants to ensure better interim targets set for developing countries by operation of the flue gas cleaning systems. the WHO. To achieve WHO global guideline begin an open and candid discussion of actual values, the emissions reductions would need costs and benefits of abatement measures to be over 80%. by (i) ensuring the abatement measures are There is no magic bullet, no one solution that technically feasible and their emissions reduction can reach Mongolian Air Quality Standards. benefits are justified with sufficient evidence A combination of measures is recommended. (ii) appraising the full costs of abatement 74 Conclusions measures so they can be compared to the Studies of the health status of the health cost reductions and their contribution population should continue. The final to overall improvements in ambient PM AMHIB report shall share results of concentration reductions. its preliminary survey of air pollution Strengthen air quality monitoring and related health issues. emissions inventories by providing sufficient Indications of the improved health status operating budgets to key Mongolian air following abatement implementation quality institutions. should be studied. Abatement Options Assessment The following is a summary of next steps Several technical assessments have been in the AMHIB study based on the basic steps done over 2 years but remain largely in the Air Quality Management (AQM) process at a proof of concept stage. Large scale introduced in this paper. The following list points demonstration projects need to be to important points where an improved data base rolled out immediately to test proposed is needed. The AMHIB will attempt to address concepts especially in ger areas where these either directly through its study or through success of abatement measures depends recommendations that others can help address. not only on technical effectiveness but also on socio-economic and strong Air Quality Assessment cultural considerations. It is important to Air pollution monitoring: monitoring systematically move from demonstration with the use of state-of-the-art monitors projects to scale up as quickly as possible at more stations in UB is needed. This but only when concepts show promising monitoring should include at least PM2.5 impacts. and PM10, SO2, NOX and NO2. This Some measures can be implemented over activity is already underway in UB, but 1­2 years while others may take several data from this increased monitoring years to implement. activity is only partially available. The Cost Benefit Analysis or Cost Effectiveness AMHIB will set a baseline value for PM Analysis in 2008­2009 by the end of the study, Cost-benefit analyses linking measures but it is very important to establish with results should continue. The a long-term program for training of AMHIB will continue to work with monitoring network operators and a data Mongolian counterparts to build quality control program to continually capacity in this field. improve measurements and monitoring Abatement Measures Selection results. The AMHIB and complementary Inventory of emissions: the existing activities by the World Bank and emissions inventory needs to be others should help to provide needed improved. Main points needing information to the government to improvement include: emission factors develop and select abatement measures. (EF) for the various sources; amount of Optimum Control Strategy fuel burned per source category; traffic Set timetables and secure financing. data on main roads, study of the soil Establish a monitoring and evaluation suspension source. There are programs system that continually reports on air and studies ongoing that are expected quality improvements and assesses to improve EFs that will be used, if impacts of abatement measures. available, in AMHIB's final report. Indications of improved health status Population: improved data for the spatial following abatement implementation distribution of the population is needed. should be studied. Environmental Damage Assessment Improvements in air quality and health data will drive improved assessments 75 References AirQUIS 2008: http://www.nilu.no/airquis/ fulfilment for the degree of Masters of Anfossi, D., D. Oettl, G. A. Degrazia, and science, Kuala Lumpur Malaysia. A. Goulart (2005). An analysis of sonic Enkhtsteteg, Sh. (2000). Hygienic assessment of anemometer observations in low wind speed influence of atmosphere air to indicators of conditions. Bound.-Layer Meteor., 114, immune factors of children who are living in 179­203. living ecological condition of in Mongolia, Aunan, K. and Pan, X.-C. (2004). Exposure- PhD thesis, Irkitsk, Russia. response functions for health effects Guttikunda, S. (2007). Urban air pollution of ambient air pollution applicable for analysis for Ulaanbaatar. Consultant draft China--a meta-analysis. Science of the Total Report to World Bank. Environment 329 (1­3): 3­16. Hurley, P. J., W. L. Physick, and A. K. Luhar Burmaa, B., Enkhtsetseg, Sh. (1996). Results of (2005). TAPM: A practical approach to the study on correlation between morbidity prognostic meteorological and air pollution among population and the children of modelling. Environ. Modelling and Software, Ulaanbaatar and air pollution, Healthy 20, 737­752. city-wonderful life, Ulaanbaatar, Mongolia Haakonsen, G. and Kvingedal, E. (2001). MOH. 2004. Health sector 2003, Emissions to air from fuelwood use in Ulaanbaatar, Mongolia. Norway. Oslo, Statistics Norway (Reports Burmaa, B. (2001). Child health with correlation 2001/36) (In Norwegian). of ecological factors of Mongolia, diagnostics Jamal, H. H., Pillay M. S., Zaliha H., et al. and prevention, PhD thesis, Irkitsk, Russia. (2004). A study of Health Impact and Risk Chow, J., Watson, J. G., Egami, R. T., Frazier, Assessment of Urban Air Pollution in the Klang C. A. and Lu, Z. (1990). Evaluation of Valley, Malaysia, A Research Project Report, regenerative-air vacuum street sweeping on UKM. geological contributions to PM10. J. Air Jaszczur, T., Lewandowski, M., Szewczyk, W., Waste Manage. Assoc., 40, 1134­1142. Zaczkowski, A., Butcher, T. (1994). Cowlin, S. C., Kaufmann, R. B., Edwards, R., Performance of Polish home stoves. U.S. Smith, K. R. (2005). Impact of Improved Department of Energy Report (BNL­60367). Stoves on Indoor Air Quality in Ulaanbaatar, Kan, H., Chen, B., Chen, C., Fu, Q., Chen, M. Mongolia. ESMAP report 313/05. www An evaluation of public health impact of .esmap.org (accessed May 20, 2009). ambient air pollution under various energy Enkhjargal, G., et al. (2004­2005). Indoor air scenarios in Shanghai, China. Atmospheric quality assessment, Ulaanbaatar, Mongolia. Environment 38:95­102. Enkhjargal, A. (2006). "Health impact assessment Keil, C. (2005). Air Quality (online), http://www of ambient air pollution in two cities in .cdc.gov/nceh/ehs/Docs/NALBOH%20Web/ Mongolia." Medical Faculty University NALBOH-2.pdf (15 March 2005). Kebangsaan Malaysia, Research thesis in 77 Air Pollution in Ulaanbaatar Kukkonen, J., M. Pohjla, R. S. Sokhi, L. Luhana, applied in AirQUIS2003. Technical N. Kitwiroon, and et al. (2005). Analysis and description. Norwegian Institute for Air evaluation of selected local-scale PM10 air Research, Kjeller (NILU TR 12/03). pollution episodes in four European cities: Slørdal, L. H., McInnes, H., Krognes, T. (2008). Helsinki, London, Milan and Oslo., Atmos. The Air Quality Information System Environ., 39, 2759­2773. AirQUIS. Information Technologies in Lodoysamba, S., Shagjjamba, D., Gerelmaa, G. Environmental Engineering 1, 40­47. (2008). Characterization of PM2.5-0 and Smith, K. R., Mehta, S. and Maeusezahl-Feuz, M. PM10-2.5 air pollution of Ulaanbaatar city. (2005). Indoor Air Pollution from Household Presented at the CAI-Asia Better Air Quality Use of Solid Fuels. Environmental and Conference, Bangkok, 12­14 November, occupational risk factors, chapter 18, World 2008. Health Organization: 1435­1493. http:// Lopez, A. D., Shibuya, K., Rao, C., Mathers, www.who.int/publications/cra/chapters/ C. D., Hansell, A. L., Held, L. S., Schmid, V. volume2/part2/en/index.html. and Buist, S. (2006). Chronic obstructive Spengler, J. (2005). Air pollution sources and pulmonary disease: current burden human health effects, lecture, (online) http:// and future projections. European courses.dce.harvard.edu/~environment/ Respiratory Journal 27: 397­412 DOI: AP_Sources_Health_Effects_12_01_2004/ 10.1183/09031936.06.00025805. sld001.htm (7 April 2005). Mestl, H. E. S., Aunan, K., Fang, J., Seip, H. M., Spickett, J., Baker, A., Burma, B., Enkhtsetseg, Skjelvik, J. M., Vennemo, H. (2004). Sh., et al. (2002). "Preliminary Health Risk Cleaner production as climate investment-- Assessment of Environmental Pollution in Integrated assessment in Taiyuan City, China. Ulaanbaatar City, Mongolia." World Health Journal of Cleaner Production 13, 57­70. Organization, Ministry of Health, and Public Mestl, H. E. S., Aunan, K., and Seip, H. M. Health Institute, Ulaanbaatar. (2007). Health benefits from reducing indoor Stansfield, S., Shepard, D. (1993). ACUTE air pollution from household solid fuel RESPIRATORY INFECTION, Jameson, D., use in China--three abatement scenarios. Mosley, W., Measham, A., Bobadilla, J. Environment International, 33, 831­840. (eds.) Disease control priorities in developing Ministry of Health, MOH, Health Sector countries, pp 67­90, Oxford. Strategic Master Plan Initiative. 2004. Sunyer, J., Castellsague, J., Saez, M., et al. (1996). Synthesis papers. Second revised reprint. Air pollution and mortality in Barcelona. Ulaanbaatar, Mongolia. J Epidemiol Community Health. 50 Suppl Pope, C. A. III, Burnett, R. T., Thun, M. J., 1:s76­80. Calle, E. E., Krewski, D., Ito, K., Thurston, Tseregmaa, Ts. (2003). Study of respiratory system G. D. (2002). Lung cancer, cardiopulmonary morbidity of elders living in Ulaanbaatar city mortality, and long-term exposure to fine in connection with some influencing factors. particulate air pollution. J Am Med Assoc Ulaanbaatar, Mongolia, National University 2002;287(9):1132­1141. of Medical Science. PREGA (2006). Energy efficiency study of U.S. EPA (1995). Compilation of air pollutant thermal power plant #4 Ulaanbataar, emission factors. AP-42, Fifth edition. Mongolia. A technical study report. Draft Research Triangle Park, U.S. Environmental final report. Promotion of Renewable Energy, Protection Agency. Energy Efficiency and Greenhouse Gas U.S. EPA (1998). Residential wood combustion Abatement (PREGA) Report to the Asian technology review. Volume 2. Appendices. Development Bank. Research Triangle Park, U.S. Environmental Saijaa, N., et al. (2004). Adverse health effect of Protection Agency (EPA-600/R-98-174b). air pollution, Ulaanbaatar. World Bank (2007). Cost of Pollution in Slørdal, L. H., Walker, S. E., Solberg, S. (2003). China. World Bank: Washington, DC. The urban air dispersion model EPISODE http://siteresources.worldbank.org/ 78 References INTEAPREGTOPENVIRONMENT/ World Bank (2009). Mongolia--Enhancing Resources/China_Cost_of_Pollution.pdf Policies and Practices for Ger Area (conference version).(accessed May 20, Development in Ulaanbaatar. Urban 2009). Development Sector Unit, East Asia and World Bank (2008a). Mongolia energy efficient Pacific Region. and cleaner heating in poor, peri-urban areas WHO (2002). Health impact assessment of air of Ulaanbaatar. Summary report on activities. pollution in the eight major Italian cities. Asia Sustainable and Alternative Energy World Health Organization Regional Office Program (ASTAE). for Europe, Martuzzi, M., Galassi, C. 2002. World Bank (2008b). Small boiler improvement (online) http://www.euro.who.int/document/ in Ulaan Bataar consultant report. Part of E75492.pdf. (12 March 2005). the UB Clean Air Program mission Clean air action plan for Ulaan Bataar Mongolia. World Bank Consultant Mission Report. 79 Appendix A: Air Quality Standards and Guidelines for PM and SO2 Air quality standards, limit values values (LV) for PM10 and PM2.5 and international and guidelines for PM and SO2 standards for SO220. WHO Guidelines are the lowest. They Particulate Matter--PM 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), The tables below summarize Mongolian air realising that in many developing countries, the quality standards (AQS) as well as WHO WHO guideline cannot be met in the short term. guidelines, USEPA standards and EU limit Table A1: Various guidelines, standards and limit values for PM2.5 and PM10 Guidelines, standards, limit values (all numbers in g/m3) PM2.5 PM10 Annual 24 hour Annual 24 hour average average (daily) average average (daily) Mongolian Standards, 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 USEPA AQS, 2006 15 35 1 -- 150 EU LV 253 -- 40 502 204 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 20 World health organization: Air Quality Guidelines for particulate matter, ozone, nitrogen oxides and sulphur dioxide. Global update 2005. Summary of risk assessment. http://www.epa.gov/air/criteria.html http://eur-lex.europa.eu/LexUriServ/LexUriServ .do?uri=COM:2005:0447:FIN:EN:PDF 81 Air Pollution in Ulaanbaatar Table A2: Basis for WHO Air Quality Guidelines (AQG) and Interim Targets Basis for WHO Air Quality Guidelines (AQG) and Interim Targets Basis for selected level Interim Target ­ 1 (IT-1) These levels are associated with about 15% higher long-term mortality risk relative to the AQG level IT-2 In addition to other health benefits, these levels lower the risk of premature mortality by approximately 6% (2­11%) relative to the IT-1 level IT-3 In addition to other health benefits, these levels reduce the mortality risk by approximately 6% (2­11%) relative to the IT-2 level. Air Quality Guideline These are the lowest levels at which total, cardiopulmonary, and lung cancer mortality have been shown to increase with more than 95% confidence in response to long-term exposure to PM2.5. Source: WHO, Krzyanowski, Update of WHO Air Quality Guidelines, February 22, 2008. Table A3: Various SO2 guidelines, standards and limit values, SO2 guidelines, standards, limit values (all numbers in g/m3) Annual average 24 hour average (daily) 10 minute average WHO Guidelines, 2005 -- 20 500 WHO Interim Targets (IT) IT-1 125 -- IT-2 50 -- USEPA AQS, 2006 78 365 -- EU LV -- 125 -- USEPA standards and EU limit values differ. They represent to some extent what is The EU LV is stricter than the US AQS for PM10, politically and technically feasible to meet while it is more lax for PM2.5. presently. 82 Appendix B: Criteria and Suggestions for an Improved Monitoring Network for Air Pollution in UB C riteria and suggestions for an 5. 1­2 stations should be established within the improved monitoring network for air most polluted area in UB. pollution in UB 6. 2­3 stations should be located in the vicinity of streets in residential/city centre areas. The following criteria can be listed as a basis 7. 1­2 unpaved roads should also be covered, for designing an improved air quality monitoring within the ger areas. network for UB: 8. Pollutants: As the main source of air pollution in UB is 1. The long term monitoring network in coal combustion, SO2 and PM are the most UB should be based upon the automatic important pollutants to cover extensively. monitors, as presently installed in some PM2.5 and PM10 should be covered at all or NAMHEM/CLEM stations as well as in new most stations. stations provided by donors. The car exhaust source should be covered by 2. In addition, the two Gent filter sampler measuring NOX and NO2 at traffic related equipments of NUM should be part of the stations and at 1­2 urban stations. CO could network, for PM10 measurements, not for be measured at the traffic stations. Benzene PM2.5. should be measured at 1­2 traffic stations. 3. The locations of the AMHIB project Ozone is most probably not an important (Locations 1­8) will be operated until mid pollutant in UB, being a relatively small city 2009, and those locations are considered for away from other urbanised areas. It could fixed. be measured at one urban station. 4. In order to establish a link between the 9. Topography considerations: baseline monitoring done at the AMHIB UB is located in a wide valley, with hills network until mid 2009 and the continuing, to the North and South of the valley, and long-term monitoring with the new heights up to several hundred meters. Winds NAMHEM/CLEM and donor equipment, are predominantly along the valley axis such new equipment should be co-located during autumn and winter, when the air at some of the stations used for the AMHIB pollution is highest, while during spring and study (the "AMHIB stations"). summer it is turned more from the northerly It is suggested to locate this new automatic direction. Due to this topography/wind equipment at at least 2­3 of the AMHIB patterns, the wide UB valley is to a large stations no. 1, 3, 4, 6 and 7. (AMHIB station extent separated from the parallel valleys no. 5, which is same as CLEM station UB-1, behind the hills during the high pollution will already be equipped with automatic periods, and the emissions flow mostly along monitors). the valley axis. 83 Air Pollution in Ulaanbaatar The ger areas which are growing behind in this note. In such work, it would be the hills are thus to a large extent separated important to monitor indoor and outdoor pollution-wise from the main valley. Those pollution simultaneously. ger areas thus have to some extent their own air pollution domain. With more monitoring equipment available, it is suggested to check on the air pollution Suggested further improvements of the concentrations also in those areas. monitoring network 10. Indoor air pollution considerations The air pollution inside the gers have been Figure B1 shows the locations of the stations studied to some extent in UB. One report has now operating in UB. They include the CLEM found extremely high concentrations, using stations UB 1­4 and two new CLEM stations, passive samplers, while the report from the the AMHIB stations (marked 1­8) and Public Health Institute found moderate PM newly installed or soon to be installed donor- concentrations, using active samplers (the supported stations (marked D). The locations are NUM samplers). approximate. All stations except UB-3, UB-4 and More work is needed to establish the extent the new UB station measure PM concentrations of indoor air pollution. (PM2.5, PM10 or both). The areas with the highest Such work should be carried out in parallel PM concentrations, according to the modelling, with the monitoring network discussed are marked on the map. Figure B1: Locations of currently operating air quality monitoring stations in UB, approximate locations Yellow: CLEM stations. White: AMHIB stations. Blue: Donor-supported stations. 84 Appendix B: Criteria and Suggestions for an Improved Monitoring Network The monitoring stations are of two types: 1. One more automatic station should be established in the most polluted residential UB: urban background station: location area. A station at or near AMHIB station 4 is not dominated by any particular nearby is suggested. New equipment should be run source. in parallel with the AMHIB equipment, T: traffic station, located close to street/road to establish the connection with AMHIB or crossing. baseline. 2. One more station located near a heavily Except for station UB-2 and the nearby trafficked street is recommended. It could be D station, which are traffic (T) stations, the in the area around AMHIB station 3. stations are all urban background (UB) stations. 3. The pollution situation near unpaved roads should be covered, by placing a station in a Based on the above-listed criteria, stations at ger area near an unpaved road relatively large the following locations are suggested: traffic. 85 Appendix C: AMHIB Data Quality Assessment AMHIB data quality assessment None of the instruments can be considered reference instruments. However, under the The PM measurement equipment of the prevailing instrument operating conditions of the AMHIB network is provided by the various institutions in UB, it is fair to consider the Ghent institutions, and differs between the various (NUM and NRC) sampler and the GRIMM stations. The instruments utilize different monitor (under low RH conditions) as the better measurement principles and are in different instruments against which the other ones can be operating conditions, which affects data quality. compared, on the condition (for PM2.5) that the The following is a comparison of the PM Ghent sampler operates throughout the sampling concentration data provided by these instruments. period without getting clogged. The comparison has been made through Figure C1 shows the results of the co-located comparison sampling that has been comparisons. There are at times discrepancies, carried out three times in 2008, each of 4­5 days some large, occur between many of the samplers. duration: 4­5 and 17­20 April, 1­6 July and Some of the discrepancies are partly the result of 18­22 November 2008. The two first campaigns the characteristics of the instruments described were carried out at the NAMHEM monitoring in Table C1, e.g. that the Ghent (NUM) sampler station at the roof of the NAMHEM building, does not operate throughout the day (has to be while the last one in November was carried out shut off due to clogging that reduces the air flow at the meteorology station UB3 located in a Ger significantly), and that the Dustrac and GRIMM area to the west of UB centre. During the last monitors are affected by hygroscopic particle21 campaign, NILU provided a GRIMM 107 PM growth at high humidity (such as on the 17th, monitor. 18th, and 21st November). However, many discrepancies are not easily explained, e.g. the low The AMHIB team carried out calibration of concentrations measured by the C-20 (CLEM) the air flow through the instruments used, and sampler and by the KOSA instrument, although of the filter weighing procedure at NUM, where the KOSA has increased response during the all the filters are weighed. The uncertainty of November comparison. the air flow and particle weight determinations were generally on the order of 10%. These Table C1 provides observed characteristics of uncertainties are acceptable, and cannot explain the instruments in the AMHIB network. the larger instrument discrepancies. However, larger weighing errors have been detected 21 Hygroscopic particles are particles which readily take in and occasionally; therefore, there can be larger retain moisture under certain conditions of humidity and discrepancies for individual data. temperature. 87 Air Pollution in Ulaanbaatar Figure C1: Results from PM sampler and monitor comparisons in Ulaanbaatar, for PM10 and PM2.5 , 2008 Comment to Figure: ­ GRIMM1 to be compared with NRC and NUM. ­ GRIMM2 to be compared with Dusttrac 1 and 2. ­ due to different operation times of the instruments. 88 Appendix C: AMHIB Data Quality Assessment Table C1: Characteristics of the instruments in the AMHIB network Instrument/ Measurement Institution principle Characteristics Ghent sampler w/10 m Filtering. The sampler separates the particles in 2 size fractions: Fine (PM2.5) and coarse PM inlet. Gravimetric analysis (PM10­2.5). NUM. The first (coarse fraction) filter clogs easily. This has 2 effects: Stations 2 and 3 1) The coarse fraction filter will then collect also some of the fine fraction particles, (Sampler is denoted NUM resulting in too large coarse fraction and too small fine fraction. and NRC at stations 2 and 3 respectively) 2) During days with high pollution, the sampler cannot operate for longer than a few hours before clogging. C-20 sampler w/ special Filtering. The inlet to the sampler is designed to cut particles larger than 10 m. There is a inlet. Gravimetric analysis question whether the inlet cuts away smaller particles. The performance of the CLEM. inlet is not tested. Station 1 Partisol 2000. Filtering. State-of-the-art PM sampler, USA. PM10 NAMHEM. Gravimetric analysis Station 6 KOSA Beta absorption Japanese instrument. Measures PM2.5 and PM10. NAMHEM. Station 1 Dustrac 8520 Light scattering State-of-the-art monitor, USA. Measures PM2.5 or PM10. Stations 1,4,7,8 The response of the sampler, in UB conditions, increases significantly during high relative humidity (RH) conditions, when hygroscopic particles (containing sulphate) grow in size. GRIMM 107 Light scattering State-of-the-art monitor, Germany. The same problem as above with increased response during high RH. Main observations: The KOSA instrument gives very low readings compared to the Ghent, Dustrac and The Ghent (NUM) sampler agrees fairly well GRIMM instruments. with the GRIMM sampler when the relative The C-20 sampler generally gives much lower humidity is below about 70%. concentrations than the Ghent, Dustrac and The Partisol 2000 sampler varies widely GRIMM instruments. compared to the Ghent sampler. Most often it gives lower PM, but sometimes much In conclusion, the Ghent (NUM) samplers, higher. This is surprising because the Partisol the Partisol and the Dustrac instruments give is a state-of-the-art instrument. data of reasonable quality, given the shortcomings The Dustrac and GRIMM instruments listed in Table C1, while the Kosa (station 1) and agree rather well. They utilise the same the C-20 instruments (station 5) somehow give measurement principle. They both give PM concentrations that are too low. readings that are too high at high relative humidity conditions, especially during the winter period. 89 Appendix D: Examples of PM Concentrations in Cities Worldwide Examples of PM concentrations in cities These concentrations, with annual average worldwide PM10 up to 279 µg/m3, and maximum daily average up to 700­800 µg/m3 can be compared PM10 concentrations in Ulaanbaatar with concentrations in other cities in the world, see below. The overview below is rather complete for cities in the US and in Europe, while for the The PM concentrations in UB are very high. other regions data availability is incomplete. There is a very strong seasonal variation with very high winter concentrations and much lower summer concentrations. The annual average PM10 Some cities in China have annual average concentrations measured at the NUM monitoring PM10 approaching and exceeding somewhat 200 station since 2006 where 141, 157 and 279 µg/ µg/m3, and some other cities, such as Karachi and m3 for 2006, 2007 and 2008 respectively. The Cairo have similar levels. The levels are much lower real concentration is somewhat higher, since the in the US and in Europe, where most cities have samplers used underestimate the concentration PM10 below 40­50 µg/m3, with a few cities above (Chapter 3 and Appendix C). The increase in 100 µg/m3 and up to 180 µg/m3 in dry regions in measured concentration may indicate that the the US (Arizona and California). Here, the PM PM concentrations in UB have been increasing is dominated by dry surface dust particles. As for over the later years, although it is possible that maximum daily PM10 averages, US and European the apparent increase might be explained by cities have mostly below 200 µg/m3, while some factors such as meteorology. The measurements industrial cities in the Eastern part of Europe still carried out under the AMHIB study at several have high maximums, a few cities in the range new stations since June 2008 give similarly high 400­700 µg/m3. These maximum daily averages concentration levels, confirming the very high approach those experienced in UB. PM10 concentration level in UB. PM10 in recent years in Chinese cities The extremely episodic nature of UB PM pollution, which is caused by the combination The table below lists cities with highest PM10 of ger heating practices and the meteorological concentrations in 2004­05 and 2006­07 situation, causes extremely high short-term PM respectively. In 2004­05, 9 of the cities had concentrations. The extremely high hourly and annual average PM10 over 150 µg/m3 (i.e. higher daily concentrations may represent the highest than class 3 classification in China that illustrates urban scale PM levels anywhere, with hourly the worst/lowest air quality classification for PM10 concentrations approaching 2,500 µg/m3 PM10). In 2006­07, these were reduced to 2 cities and daily averages approaching 700­800 µg/m3 (Lanzhou and Beijing). The current target set by over areas covering much of the city. the government in China, is that all Chinese cities 91 Air Pollution in Ulaanbaatar Table D1: Cities with highest air pollution (average PM10 concentrations 2004­05 and 2006­07) in China PM10 concentrations PM10 concentrations Northern cities 2004­05 2006­07 Southern cities 2004­05 2006­07 Linfen (Shanxi) 0.202 0.141 Panzhihua (Yunnan) 0.256 0.112 Datong (Shanxi) 0.171 0.133 Xiangtan (Hunan) 0.138 0.128 Weinan (Shaanxi) 0.171 0.131 Yueyang (Hunan) 0.138 0.125 Baotou (Neimeng) 0.165 0.141 Zhuzhou (Hunan) 0.136 0.105 Lanzhou (Gansu) 0.165 0.161 Changsha (Hunan) 0.131 0.108 Kaifeng (Henan) 0.163 0.110 Chongqing 0.131 0.110 Pingdingshan (Henan) 0.162 0.129 Zigong (Sichuan) 0.125 0.88 Taiyuan (Shanxi) 0.157 0.133 Wuhan (Hubei) 0.125 0.122 Changzhi (Shanxi) 0.148 0.117 Chengdu (Sichuan) 0.120 0.117 Luoyang (Henan) 0.147 0.116 Luzhou (Sichuan) ... 0.123 Fushun (Liaoning) 0.145 0.113 Beijing 0.145 0.155 Yangquan (Shanxi) 0.144 0.098 Tongchuan (Shaanxi) 0.139 0.112 Jinan (Shandong) 0.139 0.116 Urumqi (Xinjiang) 0.114 0.144 Note: Concentrations in the table are in mg/m3. Multiply by 1000 to convert to g/m3 (0.1 mg/m3 corresponds to 100 g/m3). Source: World Bank staff estimates based upon China Environmental Yearbooks 2005­08. should reach at least a class 2 level (i.e. It is clear that the annual average PM10 < 100 µg/m3 within the ongoing 11th Five Year concentration in China cities is above 100 µg/m3 Plan period (i.e. 2006­2010). However, the target in a large number of cities, (i.e. above China's may be too challenging to achieve and a more Class 2 standards). The example from Shanxi realistic target may be through the 12th Five Year province indicates that control measures during Plan period (i.e. 2011­2015). the later years are being effective in reducing the PM10 level in many cities. PM10 data for 11 cities in Shanxi province for 2004­2007 were as follows (µg/m3, approximate data): Comparison of PM10 levels in cities worldwide Average Highest Collection of data from highly polluted cities 2004 163 220 worldwide have been collected from various 2005 140 200 assessments and given in the figures below. The 2006 147 200 two figures below are examples. The first one 2007 108 120 is from the preparatory work for the WHO AP 92 Appendix D: Examples of PM Concentrations in Cities Worldwide Figure D1: PM10 in selected cities, 2000­2004 Figure D2: PM10 in selected cities, 2005 guideline update 2005. The data are from the PM10 in USA period 2000­2004, different for the different cities. The second one is from a similar collection The following summarizes PM10 collected at 1115 of 2005 data. stations in the US, 200722 For Europe and USA, see more recent data overview below. 22 http://www.epa.gov/oar/data/ 93 Air Pollution in Ulaanbaatar Highest daily (24-hour average) was 172 Table D2: PM10 in 340 cities in the US, 2004 µg/m3, while most stations had less than PM10 concentrations (g/m3) 100 µg/m3. Average Range PM levels in Europe 340 cities 340 cities In 2007 PM10 was monitored continuously at Annual average 25.8 14­63 more than 2000 stations in some 700 cities plus 24 h average 33.8 24­268 rural areas in Europe. PM2.5 was monitored at 308 stations in about 210 cities. Data are reported to the central data base Airbase, and data can be searched there.24 Highest annual average: PM10 4 stations has annual average PM10 over 100 µg/m3 In most of the cities, the annual average PM10 the 3 highest annual average level is below, and mostly well below, 50 µg/m3. concentrations: 181, 168, 137 µg/m3 However, levels are high in a few industrial cities these high concentrations occur in in some countries in eastern Europe. Annual Arizona and California, in very dry average PM10 concentration in the 60­100 µg/m3 locations exposed to dusty conditions range were found in some 10 cities or industrial Highest daily (24-hour average) areas in Macedonia, Bulgaria and Poland, concentrations: including a few areas in Italy and Spain. The highest measured daily concentrations: highest daily averages measured were 864 and 10,000, and some stations around 674 µg/m3 in Tetovo city, Macedonia, and in 2,500 µg/m3 Pernik city, Bulgaria respectively. About 15 cities these occur in the same dry locations as and areas have maximum daily concentrations above, and are probably associated with dust above 400 µg/m3. storms 13 stations have max daily concentrations PM2.5 above 500 µg/m3, and 37 stations have max daily concentrations above 300 µg/m3 In most European cities, the annual average PM2.5 level is less than 30 µg/m3. A few cities, including PM10 levels in most population centers in the also in Macedonia, Bulgaria and Romania, have US are much lower than these maximum levels, concentrations in the range 30­50 µg/m3. The with annual averages mostly below 40 µg/m3 and highest daily averages measured in those areas max daily values mostly below 200 µg/m3. was 366 µg/m3, while 16 cities in all Europe had highest daily average above 150 µg/m3. The following summarizes measurements of PM2.5 at 1136 stations in the US, 2007 23 23 US Environment Protection Agency: PM Standards Revision­2006. http://www.epa.gov/oar/particlepollution/ naaqsrev2006.html Highest annual average was 22.5 µg/m3, and 24 http://air-climate.eionet.europa.eu/databases/airbase/ most stations had less than 15 µg/m3. index_html 94 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Introduction to development of the emissions large industries) is less significant than what may inventory be smaller total gross emitters that are located throughout the population centres and emit at The main objectives of the emissions inventory low heights. The latter is the situation for small are: scale domestic heating by coal combustion, as well as for road traffic. To calculate the total emissions per source category and type, as a basis for a preliminary Pollutants assessment of the importance of each of them to the air pollution situation in the city In line with the assessment of PM and SO2 as the To provide input to air pollution (dispersion) main pollution problems of UB, the emission modeling of the air pollution concentrations inventory in this Discussion Paper is limited to in the city, which determines the actual PM and SO2. importance of each source. Methods Total emissions versus emission height and location The basic method for emissions inventorying is utilized in this paper: The first objective listed above relates to calculating total emissions (e.g. tonnes per year), irrespective of Emissions are the product of an activity (e.g. the locations and time variations of the emissions. amount of fuel burnt, kms driven) and an To meet the second objective, it is necessary emission factor (EF) (e.g. amount per fuel to specify the locations/spatial distribution of used or km driven). the emission sources, the time variation of the Emission cleaning is either accounted for in emissions (seasonal/monthly as well as hourly) as the EF, or by reducing the emissions above by well as the emission conditions of each source: a factor (1-cleaning efficiency). height above ground, temperature, etc. The EF depends upon many factors that needs to be taken into account: e.g. type The first step, the per-source total emission of process (such as boiler/stove type), fuel assessment, identifies the main sources based on specifications, process technology (such as amount of emissions. The second step, assesses the engine and exhaust cleaning technology of a relative importance of each source by introducing vehicle) etc. spatial distribution, timing, and emission heights. For example, the importance of pollution sources For each source category/type, the emissions at locations away from population centers or tall can be assessed by top-down or bottom-up stacks (as is often the case with power plants and methodologies. 95 Air Pollution in Ulaanbaatar Example of top-down method: First,the Report "Small boiler improvement in UB". total emissions can be calculated after the total (World Bank 2008b) (Referred to below as fuel consumption for small scale combustion "HOB report, 2008"). for space heating has been estimated (e.g. from fuel sales statistics), and the total emissions has Inventory been estimated by applying an emissions factor (EF) established through testing for stove-fuel Below each of the main air pollution sources is combinations. The resultant emissions are treated separately. For each source the basis for the distributed spatially over the area where the emissions inventory is described: fuel is burned as a function of the distribution of the population/density of households. A Description of the source time variation function can be overlaid--daily Calculation method variation, based upon daily heating practices Emission factor(s) and seasonal variation, based upon temperature Total emissions statistics. Spatial distribution Time variation Example of bottom-up method: First, for the Uncertainties total vehicle exhaust emissions can be calculated for each road link when the road traffic amount (vehicles per day) and vehicle type distribution is Emissions per source known/estimated for each road link of the total urban road network and the EF is applied to each Ger area households type of vehicle in the traffic flow. The spatial distribution of the emissions in then known from References: World bank, 2008a;. NAMHEM, the locations of the road links, and thus specified 2007. in the input to the model. The time variation is also often known from traffic counting, or it is estimated from the general activity patterns for Calculation method the city. Top-down: This type of source category, a large number of small and rather similar stoves Background material of emissions inventory without specific knowledge of their location and in this Discussion Paper characteristics, is treated as an `area source'. Total emissions are calculated as the product of fuel The main background sources for the inventory consumption and emission factors, then spatially are: distributed according to population distribution. Air Pollution Sources Inventory of UB Description of source and its fuel City. Ministry of Environment, National consumption Agency for Hydrology Meteorology and Environmental Monitoring, 2007. (Referred Number of households to as "NAMHEM, 2007"). Urban air pollution analysis report. The Heating report gives 100,941 households Draft Consultant Report to World Bank (2007) in the 6 ger districts nearest to UB. (Guttikunda, 2007). The NAMHEM, 2007 statistics gives 119,210 Mongolia: Energy Efficient and Cleaner households (2007). In addition to the 6 khoroos Heating in Poor, Peri-urban Areas of UB. included in the Heating report, there are a few Summary Report on Activities (World Bank, districts in town that also have ger household 2008a). (Referred to below as "The heating in addition to apartments. When these areas report, 2008"). are included, the number of households was estimated to about 130,000. 96 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Our model area covers much of the 6 districts Total of the Heating report, but not all of its districts. Pollutant Emission factor emissions However, the number of ger households in the Coal Wood areas that are outside our model grid is limited, although there are some populated ger areas PM10 16 kg/ton 18.5 kg/ton 16,363 tons north of our grid area (stated based upon careful PM2.5 9.6 kg/ton 16.7 kg/ton 12,133 tons inspection of Google earth map), but still within SO2 6.5 kg/ton .... 3,542 tons the districts. The number of households inside our grid area is thus somewhat less than the about The basis in measurements for setting EFs for ger 130,000 of the Heating report. Considering the stoves is very weak. uncertainty of this number, we still use 130,000 ger households as a best estimate for the number Spatial distribution within our model area. The emissions are distributed over the ger areas Fuel consumption per type of household/stove the same way as done by Guttikunda (2007), based upon information about the distribution The Heating report gives the average of households across the areas. The distribution consumption of coal and wood per ger area is shown in Figure E1 . This spatial household household as 4.19 tons and 4.68 m3 (3.18 tons), distribution has not been quality assured, and for the winter season 2006/7. This is based upon the uncertainty is not known. E.g., from field statistics on types of households: gers, houses observations about household densities, it seems with and without heating wall and houses with like the estimated density in the central-western low pressure boilers, LPB). Gers have the lowest part of the ger areas is too low. average consumption (3.97 tons) and LPBs the highest (6.17 tons), while for wood the differences between the households types are less. Time variation The emissions are distributed with time as Total fuel consumption follows: The Heating report gives its estimate of total ger Seasonal: 95% of the consumption is used in area household consumption as 546,000 tons of the 8 winter months October­April. Within coal and 415,000 tons of wood for the 2006/7 this period, the consumption is distributed heating season. No estimate is given for the between the days dependent upon the daily summer consumption, while it is clear it is much average actual temperature for each day. less than the winter consumption. The rest 5% is evenly distributed over the summer season. The "NAMHEM, 2007" statistics for 2007 The resulting time variation of the ger gives the consumption as 403,459 tons of coal household emissions across the year is shown and 237,196 tons of wood, quite a bit lower than in Figure E2. the heating report gives. We put emphasis on the Daily: the variation follows the typical data from the heating report. heating and cooking schedule of the households, as shown in Figure E3. Emission factors and total emissions from ger area households Uncertainties Appendix F describes the available sources of The main uncertainties in the total emissions information on emission factors for small stoves. from this source is associated with the average Based upon this, the following EFs are used in emission factors, as discussed in Appendix F. the calculations, resulting in the following total There is not enough information available to emissions for the winter season 2006/7: assess the uncertainty statistically. 97 Air Pollution in Ulaanbaatar Figure E1: Spatial distribution of ger household emissions in km2 grid cells, PM10, 2007 (tons/year) Heat only boilers Description of source and its fuel consumption Reference: World Bank, 2008a. Number and types of boilers Calculation method According to the `HOB report', the UB Top-down: The number of HOBs is limited, municipality has identified 145 boiler houses in about 267, but not enough information is UB with 267 boilers (list from March 2007). This available on the location, size and characteristics list does not include industries and commercial of each HOB. The emissions are calculated from owners. Most of the `old' industries are supplied an estimate of the total fuel consumed in the from the CHP plants, while some new industries HOBs and multiplied by an average emission (the Coca Cola plant is mentioned as an example) factor. The emissions are then distributed spatially are equipped with new efficient boilers. evenly over the areas where most of them are located. Boiler types are mainly: Russian BZUI 100, improved Russian type named DTH, boilers 98 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Figure E2: Time variation (from day to day) of ger emissions, based upon temperature variations Figure E3: Time variation across the day of ger emissions, based upon heating and cooking practices 99 Air Pollution in Ulaanbaatar of Chinese type, and locally available improved Emission factors and total emissions boilers. The efficiency of the boiler types is from HOBs listed as 34%, 71%, 63% and 85% for the old Russian, DTH, Chinese and improved boiler Based upon limited testing, the EF for TSP respectively, based upon limited testing. There (total suspended particles) is given as 29.1, 26.8, also exist a number of other types, and the East 28.4 and 4.4 kg/ton for the old Russian, DTH, European type CARBOROBOT is mentioned Chinese and improved boilers respectively. The specifically, apparently with good efficiency and improved boiler thus appears to have substantially low emissions. less emissions, while the 3 other types are remarkably similar, although their designs and A clear inventory of numbers of each type is efficiencies differ. It seems fair to use 28 kg/ton not available presently. It appears that most are of for the 3 types and 4.4 for the improved boiler. the old Russian type, with a certain number of the others, e.g. it is mentioned that about 10% of the The average EF to use and combine with the boilers are of the Chinese type. total fuel consumption depends on the fraction of improved boilers among the boilers presently (see Total fuel consumption Appendix C). A 1%/5%/10%/20% fraction gives an average EF of 27.8/26.8/25.6/23.3 kg/ton The total fuel consumed by the HOBs is not well respectively, and the variation from the average known. The following is cited from the `HOB of these numbers is about +­9%, i.e. a limited write-up': "The total annual coal consumption uncertainty. It seems like the fraction of improved is assessed at 150.000 tonnes and a conservative boilers is still low, probably less than 5%. If so, an multiplication factor of 2 (when assessing the effects average HOB EF is about 27 kg/ton. on air pollution) gives an annual consumption of 300.000 tonnes of lignite. (...) Allowing for other Total annual HOB TSP emissions are then industrial and commercial consumers another 400,000 tons 27 kg/ton, giving 10,800 tons. 200.000 tonnes may be added to the HOB Using the fractions of PM10 and PM2.5 of TSP consumption". from Appendix F, 0.6 and 0.36 respectively, the resulting emissions of PM10 and PM2.5 are 6,480 This adds up to 500,000 tons annually tons and 3,888 tons respectively. consumed by HOBs in UB. The uncertainty of this estimate is apparent. Spatial distribution Using the fuel consumption data from the We use the information collected by Guttikunda HOB testing in the HOB-2008 report (300 kg/hr (2007) as a basis for distributing the HOB for the Russian type and 220­286 for the other emissions spatially. Figure E4 shows locations types), 5000 operating hours per year and 267 of HOBs according to his data. The resulting boilers, we arrive at an annual consumption of spatial distribution of HOB emissions is as about 380,000 tons per year. This assumes that shown in Figure E5, where the total emissions the hourly fuel consumption used in the tests have been distributed rather evenly in the grids is representative for all hours during the whole corresponding to the green dots in Figure E4. winter. Then, there are a few more than 267 There is no specific information yet on the actual boilers. The fairly new Coca Cola boiler, using distribution of the emissions. It is possible that 20,000 tons per year, is mentioned as an example. the distribution used overestimates the HOB emissions in the central urban areas. In conclusion, we use 400,000 tons as annual coal consumption, stating that there is a fairly large uncertainty to this number. 100 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Figure E4: Location of a number of HOBs in Ulaanbaatar (Guttikunda, 2007) Figure E5: Spatial distribution of HOB emissions, 2007 (tons/year) 101 Air Pollution in Ulaanbaatar Time variation Combined heat and power (CHP) plants The working hours of HOBs in UB is assessed to Calculation method be 5,000 annually (HOB write-up), distributed across the winter season. We distribute the Bottom-up: The locations are known, and the emissions evenly over all hours from early specific fuel consumption per plant is known. October to late April. Emissions are calculated per plant, and inserted in the dispersion model at their respective location. Uncertainties Emissions are calculated as a product of fuel consumption and EF, modified by efficiency of The main uncertainty in the estimate of HOB the cleaning equipment. emissions lies with the total fuel consumption. The uncertainty may well be of the order of 30%. Description of the source and its fuel A better estimate is needed here. The uncertainty consumption of the EF is not known, but based upon the limited testing done (see above), it seems to be There are three CHP plants located as shown in not very large. The spatial distribution and time Figure E6. variation estimates do not appear to introduce major uncertainties in the calculation of the contribution from HOBs to PM concentrations. Characteristics of the CHP plants: CHP2 CHP3 CHP4 Reference Capacity, MWe 21.5 148 540 PREGA study, 2006 No. of boilers 5 12 8 PREGA study, 2006 Power produced, 2004, GWh 106 565 2150 PREGA study, 2006 Coal consumption, 2006, million tons 0.182 0.888 2.42 Guttikunda, 2007 Height of stacks, m 100 120&150 250 Figure E6: Locations of the three CHP plants in Ulaanbaatar (Guttikunda, 2007) 102 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Total coal consumption in 2006 was 3.49 Description of the source million tons (Guttikunda, 2007) and in 2007 it was 3.36 million tons. It seems power production Vehicle data and coal consumption for the CHP plants vary little between years. The number of vehicles has grown steadily and sharply over the later years, see Figure E8, Emission factors and total emissions from while the length of improved roads (i.e. paved roads) has increased little since 1992. This has the CHP plants (see Appendix F) lead to increased traffic loads on the roads and frequent and widespread traffic jams. As of 2005, PM10: 19.5 kg/ton passenger cars made up about 75% of the vehicles PM2.5: 7.8 kg/ton population. Old cars (>11 years) dominate, and made up about 50% of the passenger cars in On top of this: Flue gas cleaning efficiency: 2006. Although public transport by micro-buses 80% for CHP 2 and 3, and 95% for CHP 4. and taxis has increased substantially since about 1998, private passenger vehicles dominate the Time variation transport and traffic sector in UB. The total load of the three CHPs is rather Vehicle data for 2007 (NAMHEM, 2007): constant during the period from 8 AM to 10 PM Total number of registered vehicles: 92,706 every day, while the lowest load during night time is about 25% lower, lowest at 3­5 AM. Seasonally Of which: the day-load varies from about 410 MW in Passenger cars: 69,502 74.9% December­January and down to 250 MW in July, Trucks 14,205 15.3% that +­ 30% from a yearly average load (PREGA, Public transport (buses) 6,440 7.0% 2006). Special vehicles 2,559 2.8% Number of vehicles using: In our model calculations, we enter the CHP Gasoline ca 64% emissions as constant over all hours of the year. Diesel ca 34% This introduces an overestimation for the summer Gas ca 2% months and underestimation for the winter months. This introduces only a small error in the annual average concentration contributions from Traffic data CHPs. Data for the traffic flow in streets are needed in order to model and assess the contribution from vehicular traffic to the spatially distributed Vehicle exhaust concentrations of air pollutants. Since no traffic flow data were available previous to this work, an Calculation method effort was made to provide such data. As part of the AMHIB study team, NILU provided a note Bottom-up: The locations of the road nodes for how to count traffic in sections in a simplified between road sections are known for a large manner as a basis for assessing the flow of part of the main road network. Traffic amounts traffic, its variation across the day and its vehicle and distribution between vehicle types has been composition. As part of the AMHIB study team, estimated for about 100 road sections (links), the NUM team was asked to carry out a limited and emission factors have been estimated for each counting effort, which they did successfully vehicle type. The emissions from each road link using students. Traffic was counted and vehicles are added in the km2 grid, and treated as area classified at 9 selected street sections. The results source. of this counting, and of subsequent further flow 103 Air Pollution in Ulaanbaatar estimation at a number of other sections, are Figure E7 shows the part of the main street given in Appendix G. network in UB with the traffic flow indicated by color coding. On each section the composition of A summary of the results: the traffic in terms of vehicle class contribution is also estimated, based upon the vehicle classifying The traffic flow at the 9 sections varied countings. between about 14,000 and 57,000 vehicles per day (ADT). Light duty vehicles This traffic flow counting and estimating dominated the traffic flow at all sections. activity provides input data for including The diesel fuelled share of the light duty emissions from the vehicular traffic to air vehicles varied within 10­15%. Diesel pollution modeling for UB. fuelled vehicles have much larger exhaust PM emissions than the gasoline fuelled ones. The heavy duty share of the traffic flow Emission factors varied between about 3% and 12% at the 9 sections. The vehicle exhaust emission factors depend Buses dominated the heavy duty share upon type of vehicle, their technology (engine at some sections, while light duty trucks and exhaust cleaning) and fuel used, their age and dominated at others. technical condition, as well as the driving speed While most heavy duty vehicles in Europe and road inclination. are diesel fuelled, with relatively large exhaust PM emissions, the situation in UB For UB, classes of vehicles and fuel used is different: most heavy vehicles are gasoline were defined as in the Table below. Their fuelled, with relatively small exhaust particle correspondence with the vehicle types classified in emissions. the counting exercise (Appendix G) is also given Rough estimates of the traffic flow, in 4 broad there. classes from <20,000 to >60,000 ADT was provided for 48 additional street sections The existing vehicle exhaust regulations in within and around UB centre area. UB does not limit the types of vehicles allowed Figure E7: Traffic flow on the main road network in Ulaanbaatar, classified inbroad ADT classes: <20,000 (yellow), 20,000­40,000 (green), 40,000­60,000 (red), >60,000 (purple) 104 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar on the roads. As Mongolian gasoline still fair to assume that they are of medium-to-low contains some lead, the catalysts on new or technical standard. Most of the trucks are also secondhand vehicles will not function after a old (mostly Russian), while the bus fleet has a short time when driving on local gasoline. No wide spread in ages, some fairly new. Based upon specific information is available on the exhaust this limited information, EFs for the UB vehicle emission levels of the vehicles on UB roads, old fleet cannot be set with high accuracy. The basic or new. As about 50% of the cars were older EFs in the Table below are used in this work, for than 11 years in 2006 (still older now), it is the vehicle classes defined above. Correspondence to vehicle types classified in the Emission factor Vehicle class, fuel countings, Table G1 g/km Light duty vehicles, gasoline Private car, taxi, micro bus, 0.1 40% of jeeps Light duty, diesel 60% of jeeps 2 Buses, diesel `Big passenger bus' 2 Light heavy duty vehicles, diesel Trucks up to 2.5 tons 2 Medium heavy duty vehicles, diesel 70% of trucks above 2.5 tons 2 Medium heavy duty vehicles, gasoline 30% of trucks above 2.5 tons 0.4 Regarding traffic speed, we use 30 km/h proportional to the population density in central area, 50 km/h on ring roads, etc, and distribution. 70 km/h on main roads out from UB. The resulting spatial distribution of the Spatial distribution exhaust particle emissions is shown in Figure E9. The road traffic in UB is considered in two parts: Total emissions a. The traffic on the main road network, as The total vehicle exhaust emissions are calculated shown in Figure E7. by multiplying the traffic amounts on each road b. It is clear that the road network defined in section with its length and the EFs for each of Figure E7 is not complete. There is traffic the vehicle classes, and summing up over all road on more large roads, as well as on numerous sections. The 30% additional small road traffic small roads, mainly unpaved, linking between is then added. For this traffic, the average vehicle the main roads, especially in the ger areas. composition of the main road traffic is used. As a gross estimate (e.g. based upon similar Time variation experiences in Oslo), we add small road traffic corresponding to 30% of total vehicle-km of Based upon the counting described in Appendix G, traffic under a) above. These vehicle-kms are the traffic varies during the day for the different distributed across the km2 grid cells in the ger roads counted. The variation in the daytime hours areas in the same manner as the household is generally not large. In this work, the hourly heating emissions are distributed, i.e. largely traffic is taken as constant from 7 AM till 7 PM, and 105 Air Pollution in Ulaanbaatar Figure E8: Vehicle and road data for Ulaanbaatar (Guttikunda, 2007) Centre figure: Age distribution of passenger cars. close to zero from 7 PM to 7 AM. This is an obvious Suspension of dust from roads simplification, but does not introduce significant error for the present analysis. This variation is used Calculation method for all days of the year, although it is known that the variation is different on weekends and holidays. Bottom-up: Calculated similarly as vehicle It is considered that the error introduced by this exhaust emissions, based upon the same data for simplification is quite small. traffic and roads. 106 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Figure E9: Spatial distribution of the PM emissions from exhaust particles from the road traffic in UB, 2007 (tons/year) Description of the source the centre, since there is always a depot of dust on the road surfaces. Dry dust on road surfaces is whipped up, suspended to the air, from vehicle turbulence as Most of the mass of the suspended dust they travel the road. The extent of suspension is on particles larger than 10 micrometers, increases with the speed of the traffic by about thus larger than those affecting humans by the square of the speed. Suspension takes place breathing. However, a substantial amount is also obviously only when the surface is dry, which is below 10 micrometers (PM10), as well as below most often the case in UB. Large vehicles, trucks 2.5 micrometers PM2.5). and buses, suspend much larger amounts of dust than small vehicles. The dust suspension is very The source emits parallel with the exhaust much larger from unpaved than from paved emissions from the traffic. However, it has a roads. The dust suspension problem is more different spatial distribution and time variation, substantial in the ger areas with the unpaved mainly because suspension only takes place when roads than in UB central areas, although there is it is dry, and because of the importance of surface substantial suspension also from paved roads in conditions (paved/unpaved). 107 Air Pollution in Ulaanbaatar Emission calculation Scandinavia, and we use the formula as if 10% of the vehicles had studded tires, i.e. ST 0.1. On The emissions are calculated based upon the same unpaved roads, the suspension is very much higher, traffic and road data as used for vehicle exhaust, and there we used, as a first attempt, ST=1.0, i.e. as with the addition of some parameters: if all vehicles had studded tires. state of dryness on the roads Traffic induced suspension also gives paved/unpaved road surface. a contribution to fine particles in air. The suspension of fine fraction particles, QPMF, was set The dust suspension is translated into to 15% of the coarse fraction, QPMC, based upon emissions of PM into the air by using a method experience from Oslo. that takes into account the variation of suspension as a function of vehicle speed and vehicle types The suspension emissions were calculated in (heavy duty vehicles suspend much more particles a similar way as for exhaust particles, using QPMF than light duty vehicles because of the more and QPMC as average emission factors per vehicle. intense turbulence around the larger vehicles). For the paved roads, the spatial distribution The following theoretical/empirical algorithm is given by the traffic distribution, using the is used, which was developed based upon PM same distribution as when calculating exhaust measurements in Oslo and Stockholm as a particle emissions. For the unpaved roads, the function of traffic parameters: spatial distribution follows the population distribution, the same way as the distribution of QPMC C * (A * TT + B) * (VD / VDref )2 exhaust particles is calculated for the small road * (0.98 * ST 0.02) traffic (see spatial distribution under the vehicle exhaust section on p. 105). The resulting spatial in which QPMC is the average PM coarse particle distribution of the emissions from paved and suspension emission per vehicle in the traffic flow; unpaved roads is shown in Figure E10. C is emission of exhaust particles at the reference site; TT is the percentage of heavy vehicles; A This estimate of suspension of PM from and B are the constants in the function relating roads was used in this work in the first attempt suspension to the share of heavy duty vehicles; VD to account for road dust suspension in UB. The is the driving speed and VDref is the driving speed representativeness of the formulas have not been at the reference site; ST is the share of vehicles tested for UB conditions, although the resulting using studded tires. The parameters A, B and modelled PM concentrations, with all sources C are empirical constants, these parameters are included, correspond well with the measured dependent on the local road and traffic conditions. concentrations at the NUM site, the only site with In this work, the values used are 0.62, 3.32, and data available at the time when the modelling work 0.54 respectively, values taken from previous was carried out. The spatial distribution of the experiments both in Oslo and Stockholm. The unpaved road emissions is determined by the ger algorithm is developed for sites where a share of area population distribution, and thus depends on the vehicles use studded tires, which is normal in the correctness of that distribution. Scandinavian cities for increased friction on icy roads. The wear of the road surface due to the studs Stoves in kiosks and shops in the tires create a depot of dust particles that are suspended due to the turbulence created by the There are some 4500 kiosks and shops in UB vehicles in traffic. The depot of dust on the road which are heated by the same type of stoves surfaces in UB come from other sources, mainly as used by the ger households. This is a less dry particles from open surfaces nearby. Thus, important source, and we do not have its spatial while the mechanisms behind the creation of the distribution across UB. We use Guttikunda's dust depot are different from Scandinavia, the (2007) estimate and distribution of this source: suspension mechanism is largely the same. It was we allocate 5% of the ger household emissions estimated, as a first attempt, that the dust depot on to this source, and distribute it as the household the UB paved road surfaces is similar to the roads emissions. with a rather moderate share of studded tires in 108 Appendix E: Preliminary Emissions Inventory for Ulaanbaatar Figure E10: Spatial distribution of suspended PM10 from road traffic in UB. Left: paved roads. Right: unpaved roads. Note the different scales. 109 Appendix F: Emission Factors for Coal and Wood Combustion in Small Stoves and Boilers Coal combustion Fraction of Ash households PM emission factors for coal combustion can Coal mine contents using the coal roughly be based upon ash contents in the coal, Nalaikh 16.5 % about 75 % fraction of the ash emitted as TSP, and then ratios Baganuur 13.1 % about 25 % between TSP and PM10 or PM2.5. Significant improvements are necessary through field and This gives an average ash contents in Ger laboratory testing of stove-fuel combinations to stoves of 15.65%, and thus a PM10 EF of about accurately adjust these emissions factors. 16 kg/ton. The table below gives the actual fractions We have searched the open literature for used by Guttikunda (2007) for UB. data PM emissions from small stoves like those in gers. We have found data from Polish ceramic Updating the emission factors stoves, they are quite a bit larger than ger stoves, and smaller than the HOBs we are looking at in Ger stoves UB. Their measurements gave TSP emissions of 14­17 kg/ton (Jaszczur, 1994). A Chinese study For the ash contents and consumption of coals gives, for bituminous coal in improved stove types used in UB, we have the following data: (`high efficiency household coal stoves') a PM Table F1: Emission factors for coal and wood, used by Guttikunda (2007) Ash TSP/Ash PM10/TSP PM2.5/PM10 EF PM10 EF PM2.5 Source % ratio ratio ratio kg/ton kg/ton CHP, coal 15 0.2 0.65 0.4 19.5 7.8 Ger stove, coal 25 0.2 0.5 0.6 25 15 Ger stove, wood 3.8 2.3 HOB, coal 15 0.2 0.6 0.6 18 10.8 111 Air Pollution in Ulaanbaatar (TSP) emission factor of 14.8 kg/ton (Zhi et al, agrees favourably with the HOB EF in Table F1 2008). There are no data in AP 42 for small scale above, when it is converted to PM10 emissions by coal stoves. With a ratio between PM10 and TSP multiplying with 0.6, which gives 16.2 kg/ton. of 0.5, the Polish and Chinese stoves have a PM10 EF of about 7­8 kg/ton. Wood in gers The old ger stoves in UB should have larger The emission factor used by Guttikunda is emissions than the larger Polish stoves and the 3.8 kg/ton. more efficient Chinese stoves above. The PM emissions from small scale wood It is apparent that the basis in measurements stoves have been studied in Norway, where the for setting EFs for ger stoves is weak, and that the use of such stoves is wide spread. Haakonsen EFs for both PM10 and PM2.5 are very uncertain, and Kvingedal (2001) have summarised EFs for including the PM2.5/PM10 ratio of 0.6. such stoves. The EF for PM from stoves varies significantly with the feed rate. At the typical feed HOB rate used Norway, 1­1.5 kg/hour, the emission factor for PM10 is as large as 40 kg/ton. US The WB Consultant's `HOB report' also gives EPA AP 42 gives a factor of 15.3 kg/ton (US data from measurements of TSP emissions EPA, 1995) , and 18.5 kg/ton in a more recent from three `old' and one new boiler type. TSP publication (US EPA, 1998). Thus, there is a wide emissions from the `old' ones are 27­29 kg/ton. range of EFs for wood burning in small stoves. The EF measured for the newer improved HOB type was much lower, 4.4 kg/ton. Estimating In the lack of EF data for UB stoves, we the number of improved boilers at 5% of the suggest to use the most recent US EPA factor, total HOBs in UB, the average TSP EF from 18.5 kg/ton. the UB measurements is about 27 kg/ton. This 112 Appendix G: Ulaanbaatar Traffic Data Ulaanbaatar traffic counting method provides for a rough estimate of the ADT at each of the streets, with a limited accuracy Results from traffic countings roughly estimated to be within +/­20%. Traffic was counted at 9 street sections in UB The estimated ADT varied between about during the period, at the request of the NILU 33,000 and 57,000 at 8 of the 9 street sections, part of the AMHIB team. The counting activity while the 9th section was relatively low traffic, was organised by the AMHIB team, executed by with an ADT of about 14,000. NUM, based upon a note developed by NILU. The traffic was counted for 3 15-minute periods The traffic flow was dominated by light during one day at each of the 9 streets, separately duty vehicles at all counted sections. The heavy in each traffic direction. The 3 15-minute duty fraction varied between about 3% (at the periods were in morning rush hour (period no. 1 section: Bayangol Hotel and the no. 4 between 9 and 10 AM), midday traffic (period section: Library Gorki) to about 12% (at the between 1 and 2 PM) and during evening rush no. 8 section: Traffic Police). Buses dominated the hour (between 6 and 7 PM). The vehicles were heavy duty fraction at some sections, and light classified and counted in 7 different vehicle duty trucks at other sections. classes. See details in Table G1. The diesel fuelled share of the light duty Based upon information provided by NUM vehicles varied within 10­15%. about the gasoline/diesel fuel mix within each of the counted classes, the 7 classes were distributed While the heavy duty share of vehicles in across 6 vehicle/fuel classes suitable for input Europe is generally dominated by diesel fuelled to emission factor assessment and input to air vehicles with large emissions of exhaust PM, the pollution modelling, as described in section situation in UB is quite different, where almost below. all buses and light and medium heavy trucks are gasoline fuelled, with comparatively low exhaust PM emissions. In the right-most sections in Table G1, the average daily traffic has been estimated, based upon the counting and the 6-class classification. Ulaanbaatar vehicle classes and emission The average count in the 3 15-minute periods was factors for exhaust PM considered to represent the average traffic count over the entire day (ADT). It was considered The following vehicle/fuel classes are defined for that this traffic lasted for 13 hours, between 7 AM UB, based upon information provided by NUM and 8 PM, and that the total evening-nighttime about the fuel mix within each of the vehicle traffic was about 10% of this daytime traffic. This classes counted, see section above: 113 Air Pollution in Ulaanbaatar Table G1: Results from traffic countings at selected street sections in Ulaanbaatar 114 Appendix G: Ulaanbaatar Traffic Data i. Classes / fuel: Further traffic volume estimations for UB 1. Light duty vehicles, gasoline: `Motor car', street sections taxi, micro bus, 40% of jeeps 2. Light duty, diesel: 60% of jeeps Based upon the counting described above, the 3. Buses, diesel: `big passenger bus' NUM team was asked to classify the traffic flow 4. Light heavy duty vehicles, diesel: `Truck in a number of other street sections in UB, the up to 2.5 tons' (might be added to objective being to provide rough estimates that class 1). could be used as a basis for traffic and emissions 5. Medium heavy duty vehicles, diesel: 70% input to the air pollution modelling. of `Truck above 2.5 tons' 6. Medium heavy duty vehicles, gasoline: The team was asked to classify the traffic flow 30% of `Truck above 2.5 tons' as follows: ii. Emission factors, PM, basic factors (at 60 km/h): Code Estimated traffic flow, ADT --Vehicle class: 1. 0.10 g/km 1 < 20,000 2. 2 g/km 2 20,000 ­ 40,000 3. 2 g/km 3 40,000 ­ 60,000 4. 2 g/km 4 > 60,000 5. 2 g/km 6. 0.4 g/km The results of this classification are shown --Traffic speed in Table G2. Figure E7 in Appendix E shows the Use 30 km/h in central area, 50 km/h on ring results of the traffic flow classification. roads, etc, and 70 km/h on main roads out from UB. Table G2: Traffic volume in streets in Ulaanbaatar No Name of points Number of points Traffic volume code 1 100 ailiin toiruu 10 1 2 Baraan zakh 11 2 3 Denjiin myanga 12 3 4 III surguul 13 3 5 Tasganii Ovoo 14 3 6 Ekh nyalkhasiin emneleg 15 3 7 Geseriin zuun tald 16 2 8 Narnii titem 17 3 9 Bombogor 18 3­4 10 Bombogoroos 4 zam khurtel 19 3 11 ETN Ordon 20 3 12 Durslekh urlagiin muzei 21 2­3 13 Y surguuli 22 2 14 Germany Elchin 23 1 15 EZD Surguuli 24 2 (continued on next page) 115 Air Pollution in Ulaanbaatar Table G2: Continued No Name of points Number of points Traffic volume code 16 Khogjimiin colloge 25 3 17 SHUTI Surguuli 26 3 18 Bagshiin ikh surguuli 27 3­4 19 Lenin club 28 3­4 20 Ikh tengeriin zam 29 2 21 Shine pioneriin ordon 30 3 22 Parkiin urd 31 3­4 23 Emiin zavod 32 3­4 24 Teeveriin tovchoo 33 4 25 YII guanz 34 2­3 26 Bars zakh 35 4 27 25 aptek 36 4 28 Od kinoteatr 37 3­4 29 Moskva restoran 38 4 30 28-r surguuli 39 2 31 Emkh taivnii guuriin urd 40 3 32 Zaisan 41 <1 33 18-r surguuli 42 3 34 Tunnel 43 3 35 Bokhiin orgoo 44 3­4 36 Khavdar sudlal 45 4 37 13-iin dund zam 46 1 38 Narantuul (baruun zam) 47 4 39 Narantuul (uulzvar) 48 4 40 15 khoroolol 49 2 Street sections outside UB centre area No Name of points Number of points Traffic volume code 1 Bayankhoshuu (araar) 1 2 2 Bayankhoshuu (urduur) 2 2 3 Tolgoit 3 1 4 5 Shar 4 3­4 5 Gurvaljin guur 5 3 6 IY stantsiin yrd zam 6 <1000 7 Yaarmag 7 3 8 Bayanzurkh duureg 8 2­3 116 Appendix H: Air Pollution Modelling in UB in AMHIB: Methods, Tools and Model Evaluation Modelling of air pollution in UB using the NILU Modelling tools applied in AMHIB and this AirQUIS system Discussion Paper The AirQUIS system To carry out the air quality modelling for UB city, the air quality model EPISODE and the The AirQUIS system (AirQUIS, 2008; meteorological model The Air Pollution Model Slørdal et al, 2008) is an integrated air quality (TAPM. See below) were used. TAPM has been management system that contains different used for the purpose to prepare the meteorological modules, such as emission inventory module, data for the modelling period, since there were GIS related geographical information module, significant gaps in the local meteorological measurement module, models module, etc. measurement data. The EPISODE model is an urban air quality model integrated in the The models used in the present AirQUIS AirQUIS system. These models were run for system include models for calculating emissions, UB for evaluating the present air pollution state, dispersion and exposure on urban scales: human exposure and the effects of planned interventions to improve the air quality. Emission model Wind field model--The diagnostic wind The EPISODE model field model (Mathew) Pollution dispersion model--The urban The dispersion model EPISODE (Slørdal et al, dispersion model (EPISODE) 2003) is a Eulerian grid model with embedded Exposure model--For stationary population subgrid models for calculation of pollutant exposure assessment concentrations resulting from different types of sources (area-, line- and point sources). EPISODE The combined functionalities of emission solves the time dependent advection/-diffusion inventory, numerical modelling, on-line equation on a 3 dimensional grid. The EPISODE monitoring data collection and statistical model has been applied for the calculation of assessment methods, within an operable and pollution compounds such as SO2, CO, O3, NO2, functional GIS platform, makes AirQUIS NOx, PM10 and PM2.5. an effective tool for air quality management, assessing present air quality and projecting future In addition to the Eulerian grid model, air quality and evaluating available abatement EPISODE also contains different sub-grid models options and strategies. for refined calculations in areas close to important sources. 117 Air Pollution in Ulaanbaatar The sub-grid line source model within the Meteorological data EPISODE is based on a standard integrated Gaussian model, HIWAY-2. This model calculates Meteorological data is prepared for input to the concentration levels of non-reactive pollutants EPISODE model. The meteorological data can from road traffic at distances from a few to be acquired from local meteorological stations, hundreds of meters downwind of the road. or simulated by meteorological models. The Each lane of traffic defined in the road system is required meteorological parameters are: wind treated as a straight, continuous, finite length, (speed and direction), temperature, atmospheric line source with a uniform emission rate. In the stability, horizontal and vertical turbulence and UB modelling, the traffic source has been treated mixing height. Cloud cover, relative humidity and as an area source, thus this line source model was precipitation are asked as optional. not activated there. Background concentrations Two different types of point-source sub- grid models can be applied optionally in the The contributions of the different species to the EPISODE. One is based on a segmented plume/ pollution of the air transported into the urban trajectory model, while the other is the puff/ area (the regional air pollution component trajectory model INPUFF. In both models the from sources outside the city and transported emissions from individual sources are treated as from larger distances) are specified at the open a temporal sequence of instantaneous releases boundaries of the model domain as a constant of a specified pollutant mass. The air pollution value. This value can be user specified or taken resulting from the power plant emissions in UB from a background measurement station. was calculated using the embedded INPUFF model. The TAPM model Input data required by EPISODE TAPM (`The Air Pollution Model') (Hurley et al, 2005) is a PC-based nestable prognostic Emissions inventory meteorological and air pollution model. TAPM was in this work as a meteorological model, to The emissions inventory module contains data close the data gaps in observational data for parts such as fuel consumption, emission factors, of the modelling period. physical description of stacks and processes, traffic load etc. Estimates of hourly emissions of the AirQUIS input and model set up different air quality components are calculated by application of the emission model. The emission Geographical data inventory includes three categories of sources: Topography Point source emission: Emissions from power plants, large industrial plants and significant The terrain information has been downloaded single sources (e.g. larger boilers). from http://www.esri.com/data/resources/ Line source emissions: Emissions from road geographic-data.html and prepared for modelling traffic. In the calculation, roads with annual domain on the 1 km resolution. daily traffic above a user defined limit value are included as line sources. The emissions Model domain and grid from the roads with lower annual daily traffic are treated as area sources. In this GIS based system, UTM coordinates Area sources emissions: Both stationary are applied for this work. In order to have all sources that are too small to be regarded as information consistent, all collected information point sources as well as road traffic emissions based on Lat/Lon coordinates have been from roads with low traffic loads. converted to UTM coordinates. The model 118 Appendix H: Air Pollution Modelling in UB in AMHIB: Methods, Tools and Model Evaluation Figure H1: The domain used for air pollution modelling in this work (1 1 km2 grid) domain lower left point coordinates are the same for supplying electricity for the city, therefore, as in Guttikunda's (2007) model domain. There no time variation factors applied for these power are exactly 30 30 grids on a 1 km resolution, plants. It is assumed that the SO2 are directly covering a 900 km2 area. The model domain emitted without any cleaning process, and for the covers the urban area of UB and its surroundings dust, the cleaning efficiencies are 80%, 80% and (Figure H1). 95% separately. Emission data and inputs Area sources The details of the emissions inventory used for In UB, based on the pollution released height, UB in this work are given in Appendix E. The emissions from households in gers, kiosks, heat UB sources are introduced into the model system only boilers (HOBs), brick plants, and waste as either point sources or area sources. In the burnings are considered as area sources. Those emissions inventory, the traffic source is treated emissions data are taken from Guttikunda (2007) as line sources (road sections). For the modelling, grids emission, as well as the distribution factors. the traffic source emissions are entered as area When importing the data to the AirQUIS system, sources. the emissions in the grids was projected from Lat/Lon coordinates to UTM coordinates by Point sources its locations, and then redistributed on the new grids. There are 3 power plants located in the southwest of UB city. The stacks in these power plants The coal consumption in gers and HOBs have been treated as point sources. Stacks is used for heating and cooking, therefore, physical information, such as exact coordinates, an obvious seasonal variation is applied. This stack height, width, flow rates and cleaning emission time variation is based on the estimated efficiencies have been input for the simulatio n. energy consumption during a year. The kiosks, The power plants are fully running over the year and waste burning are activities do not change 119 Air Pollution in Ulaanbaatar much with seasons, so no time variation factors parameters under most conditions, except an have been applied. The brick making plants are underestimation of low wind speed during only in operation about 7 months from April to night time. The difficulties of reproducing the October and closed in the rest of the year. low wind speeds is also discussed by previous studies (Anfossi et al, 2005). It is a common Line sources problem both for TAPM and MM5 and other similar meteorological models, when turbulent Traffic related emissions include not only direct motions may be of the same order as the wind exhaust emissions. The fugitive dust from the speed (Tang et al, 2009). For better simulation roads (dry dust on the roads suspended in air due of the distribution of air pollutants, the wind to the action of the turbulence created by passing speed from the TAPM model was adjusted vehicles) is a very import source of PM2.5 and during night time for winter months, where PM10 in UB. observation data were missing. The adjustment is based on the observation that the low wind Meteorological data speed is dominating in winter nights, and a low wind speed 0.7 m/s was fixed during the nights One year's meteorological observation data for in January and December, between 16:00­03:00 UB (year 2007) were provided by NAMHEM. hours. The meteorological data is available for most of the year, but data in January, and the first half Air quality measurements of February, April, and December are missing. The meteorological parameters available At the start of the AMHIB study, there were are wind speed, wind direction, surface limited PM observation data available from temperature, precipitation, radiation, humidity UB. There are 4 stations in UB with SO2 and pressure. measurements, the data from those 4 stations has been obtained. The measurements are used The missing local meteorological data were for model evaluation. SO2 measurements are then supplemented by data calculated by the available for most of the year 2007 as daily TAPM model. TAPM was run to calculate the average. The PM data is measured as coarse and vertical thermal structure over the model domain fine PM at University site about two times a as the input for EPISODE model. In this work, weeks. TAPM set up on a 300 km 300 km horizontal domain, and nested for UB region on finer scales Model evaluation by comparison with (27 km 27 km, 9 km 9 km, and 3 km measurements 3 km). Six hourly synoptic analyses data on a longitude/latitude grid at 0.75- or 1.0-degree grid Model set up for year 2007 spacing (approximately 75 km or 100 km) were used for driving TAPM, prepared from LAPS or GASP analysis data. The modelling year was set to 2007. Meteorological data were available, and the emissions inventory was updated for 2007. In air pollution modelling, it is important to capture the local-scale meteorological conditions. Especially the stable and low wind speed The air quality model was run for the full situations often lead to air pollution episodes year of 2007. The model has meteorological (Kukkonen et al, 2005). Meteorological data and emissions input for every hour, and the simulated for UB by TAPM has been evaluated model results are also provided on the hourly through comparing with the local meteorological base. Because all measurements from UB are measurements for periods when both where daily averages, the model results have to be available. The results shows that TAPM generally averaged and compared with measurements on does well at simulating the meteorological the daily basis. The model has a 1 km horizontal resolution, and vertically there are 10 modelling 120 Appendix H: Air Pollution Modelling in UB in AMHIB: Methods, Tools and Model Evaluation layers, from surface to 2750 m height, the first are high, creating episodes of very high pollution layer is 20 meter, and 3 layers close to surface are levels. in the lowest 100 meter. Model simulated SO2 concentrations and Meteorological conditions during this period comparison with measurements Local wind measurements have been acquired Figure H3 shows the comparison of modelled from NAMHEM, and wind rose fractions and with measured concentrations of SO2 at the stability frequency for this period are presented four CLEM monitoring stations UB 1­4, here. It shows that the dominating winds during located as shown in the report. The measured 2007 are westerly with wind speeds generally concentrations have been adjusted according below 4m/s. The second dominating wind sectors to comparison results described in section are from northwest and east/southeast. Because 4.3.4 below. The measured SO2 levels vary the large point sources are located to the west of considerably between the sites, with high winter the city and the often-occurred stable conditions, concentration levels from as low as 50 µg/m3 at their influence over the urban areas of UB is UB-1 to about 100 µg/m3 at UB-2. The seasonal expected to be small. The stability frequencies variation is similar at all stations, with very low show that the stable, light stable and neutral concentrations in the summer, reflecting the conditions occupy most of the evening, night low consumption of coal for ger heating and for time and early morning hours, when the coal HOBs then. The CHPs are also operating in the consumption in ger households also peaks. These summer, and the low summer concentrations kinds of meteorological conditions limit the reflect the small contribution from the CHPs to vertical dispersion of air pollutants, especially in ground level concentrations, because of the tall the winter days when air is stable and emissions stacks. Figure H2: Meteorological conditions during the modelling period a) Wind rose with wind classes b) Stability frequency by hour of the day 121 Air Pollution in Ulaanbaatar Figure H3: SO2 concentrations at stations UB 1­4. Measured and modelled daily average concentrations, 2007 (g/m3) Table H1: Measured and modelled annual average SO2 (g/m3) for stations UB 1-4, 2007 Mon. station Measured Modelled Comment UB-1 14.2 19.2 1 Jan­10 Nov UB-2 28.4 35.9 Entire year UB-3 23.0 19.6 Entire year UB-4 31.1 30.6 Entire year The emissions that are input to the model is overestimation at 3 stations, UB-1,2 and 3, calculations, total emissions per source and but not at UB-4. Apart from this, the agreement its spatial distribution and time variation, between measured and modelled concentrations are according to the emissions inventory in is quite good. Table H1 shows measured and Chapter 4. modelled annual averages. The modelled concentrations follow the Model simulated PM concentrations and seasonal variation of the measurements well, and comparison with measurements also reflect the different levels at the four stations. There are some deviations from the measurements For the purpose of the model simulation, PM is in some periods, such as for the UB-2 station, measured at only one station, the NUM station at located a bit to the west of UB centre area: National University of Mongolia. As described in overestimation during early January and in Appendix C, the NUM sampler gives reasonably November­December, and in early March, there good data for PM10, while it underestimates the 122 Appendix H: Air Pollution Modelling in UB in AMHIB: Methods, Tools and Model Evaluation Figure H4: Measured and modelled PM10 (daily average) at the NUM measurement site Contributions from various sources. concentration of PM2.5. Thus comparison of exposed to turbulence when it is very dry. We do modelled concentrations with measured ones can not have independent data for dryness/wetness only be done for PM10. to study such details. Our model gives a steady suspension from the roads, hour-by-hour only Figure H4 shows measured data and dependent upon traffic amount. Thus, the model modelled contributions to PM10 from a number cannot reproduce the winter-time suspension of sources. The measurements are taken generally dynamics day-by-day. During the summer period, on two days per week. The brown line shows suspension is the completely dominating PM10 the total modelled PM10 concentration, with source. Figure H4 shows that the model estimates contributions from ger and kiosks coal and wood, the summer-time PM10 reasonably well on the HOBs, CHPs, vehicle exhaust particles and average. suspended dust from roads. The model predicts an annual average PM10 Note that Figure H4 represents the NUM of 163 µg/m3, all source contributions added. The site only. The contributions from the various measurements give 157.7 µg/m3 (see Table H2 PM sources to ground level concentrations varies below). Each of the source contributions have substantially throughout the city: e.g. in ger areas been modelled based upon separate scientific the ger household emissions will dominate more, considerations and upon the input of emissions as while in the city centre the exhaust emissions and inventoried in Chapter 3. The estimated annual suspension will be more dominant. This is shown average concentration from the measurements is in detail in Chapter 3. made up of data from only 2 days per week. The annual average produced by this time coverage will deviate from an annual average if all days The model results overestimate the had been sampled, while the modelled annual measurements early in the year (January­ average is based upon hourly data throughout February) and underestimates at the end of the the year. The uncertainty related to the 2-day- year (November­December) 2007. Inspection of per-week sampling instead of all days in the year the measurements as shown in Chapter 3 shows is estimated to about ­/+ 4%. This comes on that the amount of suspension of dry dust (the top of the uncertainty in each of the sampled coarse fraction of PM) was small in January­ values, which can be estimated to be +/­ 20­25%. February and very large in November­December Another factor, which is discussed in Appendix C, in that year. Suspension of dust from surfaces is is that the sampler used for these measurements to a large extent a non-steady mechanism. Dust operates, when the concentrations are high, only is building up on the surfaces during humid during part of the day, from 10 AM and then and low wind periods, and then released when 6­10 hours. The effect of that is that the sampler 123 Air Pollution in Ulaanbaatar Table H2: Measured PM10 and modelled source contributions at the NUM station, 2007 (g/m3) a. Concentrations Source Measured Modelled Ger coal and wood + kiosks 45 HOBs 25 CHPs 1 Vehicle exhaust 10 Suspension Paved roads 70 Unpaved roads 12 Total 157.7 163 b. Source contributions Source Source apportionment Dispersion modelled Coal and wood combustion 35% 44% Suspended dust 58% 50% Vehicle exhaust 6% 6% generally misses parts of the afternoon-evening The statistical source apportionment from peak as well as much of the morning peak, and the NUM measurement data (Chapter 3) give thus generally gives too low PM values when 26­35% for combustion--sulphate particles, compared to 24-hour averages, which is what the 51­58% for the soil--construction particles modelled levels give. Thus, the measured annual and 6% for motor vehicle particles. Our model average tends to underestimate the PM level thus gives about 10% higher combustion systematically. particles contribution at the expense of the soil contribution, while the vehicle exhaust particle The modelled PM10 concentrations (see the contribution is about the same for the two table above) give the following contributions from methods. This deviation is within what can the main source categories at the NUM station: be expected, given the uncertainties of both methods. Coal and wood 71 ug/m3, or 44% combustion: Suspended dust: 82 ug/m3, or 50% Vehicle exhaust: 10 ug/m3, or 6% 124 Environment and Social Development East Asia and Pacific 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