BAYESIAN MAXIMUM ENTROPY SPACE/TIME ANALYSIS OF AMBIENT PARTICULATE MATTER AND MORTALITY IN THAILAND Sitthichok Puangthongthub A dissertation submitted to the faculty of the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Environmental Sciences and Engineering, School of Public Health. Chapel Hill 2006 Approved by Advisor: Marc Serre Reader: Douglas Crawford-Brown Reader: Richard Kamens Reader: David Leith Reader: Dana Loomis Reader: Karin Yeatts © 2006 Sitthichok Puangthongthub ii ABSTRACT SITTHICHOK PUANGTHONGTHUB: Bayesian Maximum Entropy Space/Time Analysis of Ambient Particulate Matter and Mortality in Thailand (Under the direction of Marc L. Serre, Ph.D.) Several epidemiological studies have confirmed the associations and estimated the risk of cardiovascular and pulmonary mortality due to short term exposure to PM10. However the errors introduced by techniques previously used to interpolate PM10 in characterizing the strength of the associations remain in question. Therefore we incorporate the Bayesian Maximum Entropy (BME) method of modern spatiotemporal Geostatistics in the exposure assessment to investigate the effect that daily PM10 has on cardiovascular and respiratory mortality in different regions of Thailand. The BME maps show that the PM10 field across Thailand exhibits considerable space/time variability. These maps suggest that the PM10 daily average concentration did not comply with the PM10 standard, which is supported by the maps of PM10 daily maximum concentration, and confirmed by BME maps of non-attainment areas. Furthermore, the BME analysis targeted districts in the Northeast region as sites where new monitoring stations should be added in order to improve the Thailand’s existing monitoring network. These maps provided the most accurate estimate of PM10 exposure obtained to date for each district location and day for which cardiovascular and respiratory mortality iii was reported in Thailand during 1998-2003. The strength of the associations was then investigated through an analysis using a case-crossover design. The observed associations were stronger for pulmonary mortality than for cardiovascular mortality. The high odds ratios observed in Bangkok and the central region were possibly due to industrialization, construction, and traffic emission sources while positive associations found in other regions could be related to sources from agricultural biomass burning, forest fire, wind- blown dust, and sea spray. Furthermore, a holistochastic human exposure analysis propagated mapping and epidemiologic uncertainty to obtain the lower-bound and upper-bound estimates of the number of deaths from cardiovascular and respiratory causes that resulted from acute health response to short term exposure to PM10 across Thailand. An Elasticity uncertainty analysis suggests that the uncertainty in the assessment of the number of deaths caused by PM10 can be improved by adding monitoring stations in the Northeastern region, and by improving the epidemiologic study in the Northern region. iv ACKNOWLEDGEMENTS This dissertation could not have been written without the significant contributions of a number of people. I would like to acknowledge my advisor Dr. Marc Serre for his excellence in teaching and advising as well as his unfailing motivation and encouragement. I feel extremely privileged to work with him and I have learned a great deal from him, above and beyond his contributions to my research. I am grateful to my dissertation committee members: Professor Douglas Crawford-Brown, Professor Richard Kamens, Professor David Leith, Professor Dana Loomis and Dr. Karin Yeatts. Each of them offered remarkable guidance, advice, criticism, and insight from his or her unique expertise. The consistent motivation, support, and encouragement from all of my committee members have meant so much to me. Without their support at the initial stage of forming this research prospectus, I would not be able to continue my PhD study. Special thanks go to Maryanne Boundy, a director of the Baity Air and Engineering Lab and Professor Michael Flynn for their support and critical advice when I was trying to continue my PhD study. As for SAS® programming assistance, my appreciation goes to Chris Weisen of the Odum Institute for Research in Social Science at UNC for his skillful editing, as well as for his useful ideas and thoughts on several issues that made my analysis more efficiently. Without his help, my analysis might have taken many more months to complete. As for traveling assistance, I would like to acknowledge the University Center v for International Studies at UNC for awarding me the Doctoral Research Travel Award which allowed me to meet and present my research proposal to Thai agencies and obtain data sets needed for this study. Several individual and agencies in Thailand played major roles in making my doctoral research possible. Special acknowledgment goes to Dr. Warangkana Polprasert of Sukhothai Thammathirat Open University for her assistance in obtaining GIS and demographic data. I would like to thank the Pollution Control Department of Ministry of Natural Resources and Environment, the Bureau of Policy and Strategy of The Ministry of Public Health, the National Statistical Office and the Meteorological Department of Ministry of Information and Communication Technology, and the Department of Provincial Administration of Ministry of Interior for providing data sets needed for this research. I also would like to recognize the Office of Civil Service Commission for awarding me a full M-PhD scholarship which gave me an opportunity to experience the world-class teaching and research at UNC for my professional development. I would like to thank all of my friends, both in Thailand and in the U.S., especially those in Chapel Hill, for their encouragement and support. I am gratified to my lab mates at the BME Lab for their friendship and assistance throughout this analysis. My deep appreciation goes to the Thai community and friends in North Carolina and the university with whom I have developed close friendship. They have made my stay in the U.S. very delightful. This special feeling has been invaluable and has made me feel at “home” while I have been away from my home in Thailand. Most importantly, I am forever deeply indebted to my parents, Chalong Sriwattananun and Somboon Puangthongthub, as well as my respected and beloved vi relatives, Somsak Chaijaroenwatana and Bin Poolprasert. Their unconditional love, inspiration, and motivation have been with me throughout these many years. With all of these wonderful components from them, I was able to maintain a solid foundation for my personal and professional development that has helped me to achieve this goal with joy, enthusiasm and excitement. vii TABLE OF CONTENTS Page LIST OF TABLES……………………………………………………………………….xi LIST OF FIGURES……………………………………………………………………...xii LIST OF ABBREVIATION…………………………………………………………….xv Chapter I. INTRODUCTION……………………………………………………………..1 Significances of PM10 and Cardiovascular and Respiratory Mortality…..1 Research Motivation……………………………………………………..3 Statement of Study Objectives…………………………………………...5 II. MANUSCRIPT 1: MODELING THE SPACE/TIME DISTRIBUTION OF PARTICULATE MATTER IN THAILAND AND OPTIMIZING ITS MONITORING NETWORK ……………………………………………………7 Abstract………………………………………………………………….7 Introduction………………………………………………………….…..9 Methods…………………………………………………………………10 Results…………………………………………………………………..15 Discussion………………………………………………………………23 Reference……………………………………………………………….27 III. MANUSCRIPT 2: REGIONAL ESTIMATES OF THE ASSOCIATION BETWEEN DAILY ATMOSPHERIC PARTICULATE MATTER AND CARDIOVASCULAR AND RESPIRATORY MORTALITY IN THAILAND……………………………………………………………………41 viii Abstract………………………………………………………………..41 Introduction……………………………………………………………42 Methods………………………………………………………………..46 Results…………………………………………………………………53 Discussion……………………………………………………………..58 Reference……………………………………………………………...66 IV. MANUSCRIPT 3: A COMPREHENSIVE SPACE/TIME ASSESSMENT OF CARDIOVASCULAR AND RESPIRATORY MORTALITY RISK ASSOCIATED WITH SHORT TERM EXPOSURE TO PM10 IN THAILAND……………………………………..82 Abstract………………………………………………………………82 Introduction…………………………………………………………..83 Methods………………………………………………………………84 Results and Discussion……………………………………………….89 Conclusion………………………………………………………..…104 Reference………………………………………………………..…..105 V. CONCLUSIONS Summery of Findings………………………………………….…....116 Strengths and Limitation…………………………………….……...117 Future Research Directions and Recommendations……….……….119 APPENDICES……………...………………………………………………………121 APPENDIX A Geographic locations of air quality monitoring network of Thailand………………………………………………....122 APPENDIX B Locations of calculated centroids for all 925 district polygons……………………………………………………124 ix APPENDIX C Graphs of the region-specific pooled estimates of odds ratios by cause of death (i.e. ratio of the odds of 3 death for a 1µg/m increase in daily PM10)……………..144 APPENDIX D Graphs of the region-specific and country-wide pooled estimates of the odds ratios showing results by region....147 x LIST OF TABLES Table Page Table 2.1 Covariance model coefficients, explaining the space/time variability of PM10 across Thailand from January 1998 to June 2004…………………………………………………………...........29 Table 2.2 Districts prioritized to be the site of a new monitoring Station……………………………………………………………………29 Table 3.1 The meta-analysis results of odds ratios and 95% CI for associations between PM10 and cardiovascular