AMBIENT PM
2.5EXPOSURE AND ASSOCIATED PREMATURE MORTALITY BURDEN FOR INDIA IN PRESENT AND FUTURE
CLIMATE
SOURANGSU CHOWDHURY
CENTRE FOR ATMOSPHERIC SCIENCES INDIAN INSTITUTE OF TECHNOLOGY DELHI
OCTOBER 2019
©Indian Institute of Technology Delhi (IITD), New Delhi, 2019
Ambient PM2.5 exposure and associated premature mortality burden for India in present and future climate
by
Sourangsu Chowdhury Centre for Atmospheric Sciences
Submitted
In fulfilment of the requirements of the degree of DOCTOR OF PHILOSOPHY
to the
Indian Institute of Technology Delhi October 2019
Dedicated to my father
Certificate
This is to certify that the thesis entitled “Ambient PM2.5 exposure and associated premature mortality burden for India in present and future climate” being submitted by Mr. Sourangsu Chowdhury for the award of the degree of Doctor of Philosophy, is a record of the original bonafide research work carried out by him. He has worked under my guidance and supervision and has fulfilled the requirements for the submission of this thesis. The results presented in this thesis have not been submitted in part or full to any University or Institution for award of any degree/diploma
Dr. Sagnik Dey Associate Professor,
Centre for Atmospheric Sciences, Indian Institute of Technology Delhi, Hauz Khas, New Delhi-110016, India
i
Acknowledgements
First of all, I would like to express my sincere gratitude towards my supervisor, Prof. Sagnik Dey for his valuable guidance, constructive criticism, affection and constant encouragement. He is the epitome of goodness and dedication for me. He helped me develop into a competent researcher and a confident man. His calmness inspired me through my toughest times. He is the person I credit with making my PhD tenure ‘happy’.
I would also like to thank Prof Kirk R. Smith for his valuable guidance, help and support during all phases of my PhD. He always initiated interesting discussions which helped to develop my critical and analytical thinking. I also thank him for mentoring me during my Fulbright tenure at the University of California Berkeley. I sincerely thank the United States-India Educational Foundation and the J. William Fulbright Foreign Scholarship Board for awarding me the Fulbright- Nehru Doctoral Scholarship to pursue research for 9 months at the University of California Berkeley.
I thank Prof S.N. Tripathi, Prof Kalpana Balakrishnan, Prof Somnath Baidya Roy, Dr. Santu Ghosh, Dr. Amit Mishra, Dr. Ajay Pillarisetti, Dr. Zoe Chafe, Dr Nick Lam, Prof Joshua Apte, Dr.
Charlotte Smith, Dr. Sarath Guttikunda, Dr. Sumit Sharma, Dr. Gufran Beig, Prof Partha Sarathi Dutta, Prof Ambuj Sagar and all the Centre of Atmospheric Sciences faculty members for helpful discussions during various stages of my PhD tenure. I especially thank Prof Pradip Swarnakar for being a philosopher, guide and friend to me. California would not have been good without him.
I thank my friends Dr Pratik Chakraborty, Bodhibrata Mukhopadhyay, Abhishek Upadhyay and Debarghya Dutta for always being the friends in need. I also thank my friends Ayan Pan, Nayan Ranjan Saha, Saptarshi Sarkar, Anupam Sengupta, Rajib Paul, Saptarshi Chatterjee,
ii
Anirban Banerjee and Siladitya Ghosh for being there for me throughout. I thank my seniors Dr Kamalika Sengupta, Mr Arjya Sarkar, Dr Parul Srivastava, Dr Pushp Raj Tiwari and Dr Sushant Das for helping and supporting me whenever required. I also thank Vijay for helping me with the logistics.
Finally, I thank my parents for taking all the efforts to nurture me. They were my first teachers.
They scolded and moulded me as a boy and provided me with freedom as a man. My father taught me to dream big, to understand the virtues of life, to be good, to help and to cook. I dedicate my thesis to him. Special thanks to my beautiful wife, Shamika for always being there for me, her constant encouragement and support.
My mother would have been a very happy person now.
Date: Sourangsu Chowdhury
iii Abstract
Exposure to ambient PM2.5 (particulate matter less than 2.5µm in aerodynamic diameter) is causally associated with morbidity and adult premature mortality from chronic obstructive pulmonary diseases, ischemic heart diseases, stroke and lung cancer. In this work, ambient PM2.5
exposure and associated premature mortality burden was estimated at the state level and further at district level over India. Health benefit assessments have been carried and plausible policies to mitigate ambient PM2.5 exposure have been proposed. Finally, ambient PM2.5 exposure and associated premature mortality burden for India was projected for future decades under climate change scenarios.
A robust ground-based pollution measurement system is expected to provide benchmark estimates of PM2.5 exposure. Considering dearth in quality-controlled data with adequate spatial coverage in India, a satellite-based PM2.5 retrieval model has been used which provides spatially and temporally varying estimates of PM2.5 was used over India. The long-term (2001-2015) population-weighted ambient PM2.5 exposure over India was estimated to be 58µg/m3 with exposure in north Indian districts exceeding more than 2 times the Indian National Ambient Air Quality Standard (NAAQS) of 40 µg/m3. The global human settlement data was utilized to link the change in PM2.5 exposure to urban expansion in 60 major Indian cities. All the cities expanded significantly over the past 15 years with an overall increasing trend in PM2.5 exposure with varying proportions. Delhi is the capital and one of the most polluted cities in the world, Spatio-temporal variation of PM2.5 was analyzed with a focus of transport of pollution from upwind regions affected by agricultural waste burning during October-November which climatologically results in the first major pollution episode during the week of intense fire burning. The second episode in PM2.5 was
iv
identified during early January which was attributed to adverse meteorology and secondary aerosol formation.
Premature mortality attributed to exposure to ambient PM2.5 was estimated at the state level and further down to district level over India. As most of the previous estimates used a single value of baseline disease rate for the entire country, a model was developed to provide spatially varying estimates of baseline disease rate. Using a set of exposure-response functions (ERFs) used for Global Burden of Diseases (GBD) studies 0.8-1.7 million premature mortality was estimated at an annual scale in India that can be attributed to ambient PM2.5 exposure. It was identified that irrespective of the use of different ERFs’, Uttar Pradesh and Bihar were unanimously the most vulnerable states, where maximum benefits would be obtained (in terms of averted premature mortality) if mitigation measures are undertaken. Henceforth, the health benefits of recent policy measures as undertaken by the Government of India were assessed. Household air pollution (HAP) has been consistently identified as the leading contributor towards ambient PM2.5 across India. A band of tenable scenarios to control sources of HAP and their plausible benefits towards ambient PM2.5 exposure reduction and associated health benefits were then assessed. It was also estimated that the recent odd-even traffic intervention policy in Delhi did not provide desired benefits, which are attributed to the small contribution of transport sector towards ambient PM2.5 within Delhi during that period.
It was estimated that PM2.5 exposure has increased significantly (by > 1.5µg/m3) over most parts of the country during the last two decades, most of which occurred from October to January.
Ambient PM2.5 exposure up to 2100 was then projected under (Representative Concentration Pathway) RCP4.5 and RCP8.5 scenarios using data from an ensemble of 13 CMIP5 models.
Premature mortality burden was estimated using projected demographic and socioeconomic data
v
from the 5 SSP (Shared Socioeconomic Pathways) scenarios. Due to projected improvement in socioeconomic conditions over India, the premature mortality burden was projected to decrease at the end of the century under all the combined RCP-SSP scenarios. The results presented as part of this thesis is expected to help the Indian policymakers in identifying the areas where mitigation policies should be directed and formulate more efficient air quality management plans so that maximum health benefit can be obtained.
vi
साराांश
परिवेशी PM2.5 (एयिोडायनामिक व्यास िें 2.5 µm से कि पामटिकुलेट िैटि) का क्रोमनक रूप से अनावृमि
क्रोमनक ऑब्सटरक्टिव पल्मोनिी मडजीज, इस्केमिक हाटि मडजीज, स्ट्रोक औि लंग कैंसि से िागी भाव औि
वयस्क सिय से पहले िृत्यु से जुडा हुआ है। इस काि िें, परिवेशी PM2.5 जोक्टिि औि संबद्ध सियपूवि िृत्यु
दि का अनुिान िाज्य स्ति पि औि आगे भाित िें मजला स्ति पि लगाया गया। स्वास्थ्य लाभ का आकलन
मकया गया है औि परिवेशी PM2.5 को कि किने के मलए प्रशंसनीय नीमतयां प्रस्तामवत की गई हैं। अंत िें, जलवायु परिवतिन परिदृश्ों के तहत भमवष्य के दशकों िें भाित के मलए परिवेशी PM2.5 जोक्टिि औि संबद्ध सिय से पहले िृत्यु दि का अनुिान लगाया गया था।
िजबूत धिातल्य प्रदुषणं प्रबन्धनं प्रणाली से PM2.5 एक्सपोज़ि के तल मिह्न अनुिान प्राप्त किने की उम्मीद है। भाित िें पयािप्त स्थामनक व्याक्टप्त के साथ गुणविा-मनयंमित डेटा िें किी को ध्यान िें ििते हुए, एक उपग्रह
आधारित PM2.5 पुनप्रािक्टप्त िॉडल का उपयोग मकया गया है जो भाित िें PM2.5 के स्थामनक औि सािमयक रूप से मभन्न अनुिान प्रदान किता है। भाित िें दीर्ािवमध (2001-2015) जनसंख्या के आधाि पि भारित
परिवेशी PM2.5 अनाविण 58 µg/m3 है, उिि भाितीय मजलों िें जोक्टिि भाितीय िाष्ट्रीय परिवेशी वायु गुणविा
िानक (NAAQS) 40 µg/m3 से 2 गुना अमधक पाया गया है। वैमिक िानव मनपटान डेटा का उपयोग 58 प्रिुि भाितीय शहिों िें शहिी मवस्ताि के साथ PM2.5 के परिवतिन से जोडने के मलए मकया गया था। मपछले
15 वषों िें, इन सभी शहिों िें अमभप्रायपूणि मवस्ताि हुआ है, औि PM2.5 के प्रवृमि िें कुल वृक्टद्ध अलग-अलग अनुपात िें हुआ है। मदल्ली िाजधानी है औि दुमनया मक सबसे प्रदूमषत शहिों िें से एक है, PM2.5 के अनुपात- लौमकक मभन्नता का मवश्लेषण, अिूबि-नवंबि के दौिान कृमष अपमशष्ट् जलने से प्रभामवत होने वाले क्षेिों से
प्रदूषण के परिवहन पि ध्यान केंमित कि मकया गया, मजसके परिणािस्वरुप पहले प्रिुि प्रदूषण प्रकिण तीव्र
vii
आग जलने के सप्ताह के दौिान देिा गया है। PM2.5 िें एक दूसिे एमपसोड की पहिान जनविी की शुरुआत
िें की गई, मजसे प्रमतकूल िौसि मवज्ञान औि िाध्यमिक एिोसोल के गठन के मलए मजम्मेदाि ठहिाया गया।
परिवेशी PM2.5 के संपकक से समयपूर्क मृत्यु दर को राज्य स्तर पर अनुमाननत नकया गया था और भारत में
निला स्तर अनुमाननत नकया भी पर गया था। िैसा नक नपछले अनुमानों में से अनिकांश ने पूरे देश के नलए बेसलाइन रोग दर के एकल मूल्य का उपयोग नकया था, बेसलाइन रोग दर के स्थाननक रूप से नभन्न अनुमान प्रदान करने के नलए एक मॉडल नर्कनसत नकया गया था। ग्लोबल बडकन ऑफ नडिीि (GBD) के अध्ययन के नलए उपयोग नकए िाने र्ाले एक्सपोज़र ररस्ांस फंक्शंस (ERF) के एक सेट का उपयोग करते हुए, भारत में र्ानषकक पैमाने पर 0.8-1.7 नमनलयन समयपूर्क मृत्यु दर का अनुमान लगाया गया था, निसे पररर्ेशी PM2.5
िोखिम के नलए निम्मेदार ठहराया िा सकता है। यह पहचान की गई नक नर्नभन्न ईआरएफ के उपयोग के
बार्िूद, उत्तर प्रदेश और नबहार सर्कसम्मनत से सबसे कमिोर राज्य थे, िहााँ शमन उपाय नकए िाने पर अनिकतम लाभ (औसत समयपूर्क मृत्यु के संदभक में) प्राप्त नकए िाएंगे। इसके बाद, भारत सरकार द्वारा नकए गए हानलया नीनतगत उपायों के स्वास्थ्य लाभों का आकलन नकया गया। घरेलू र्ायु प्रदूषण (एचएपी) को
लगातार पूरे भारत में पररर्ेशी PM2.5 की नदशा में अग्रणी योगदानकताक के रूप में पहचाना िाता है। HAP के
स्रोतों को ननयंनित करने के नलए दस पररदृश्ों का एक बैंड और पररर्ेशी PM2.5 िोखिम में कमी और संबंनित स्वास्थ्य लाभों के नलए उनके प्रशंसनीय लाभों का मूल्यांकन नकया गया। यह भी अनुमान लगाया गया था नक
नदल्ली में हाल ही में नर्षम-यातायात हस्तक्षेप नीनत ने र्ांनछत लाभ नहीं नदए थे, िो नक उस अर्नि के दौरान नदल्ली के भीतर पररर्ेश PM2.5 के नलए पररर्हन क्षेि के छोटे योगदान के नलए निम्मेदार हैं।
यह अनुमान लगाया गया था नक नपछले दो दशकों के दौरान देश के अनिकांश नहस्ों में PM2.5 का एक्सपोज़र काफी बढ़ गया है (> 1.5 µg/m3), िो अक्टूबर से िनर्री के दौरान हुआ। 2100 तक का पररर्ेशी PM2.5
एक्सपोज़र (प्रनतनननि एकाग्रता मागक) RCP4.5 और RCP8.5 पररदृश्ों के तहत अनुमाननत नकया गया था,
viii
निसमें 13 CMIP5 मॉडल के एक संयोिन से डेटा का उपयोग नकया गया था। 5 SSP (साझा
सोनशयोकोनानमक पाथर्े) पररदृश्ों से अनुमाननत िनसांखिकीय और सामानिक-आनथकक डेटा का उपयोग करके समय से पहले मृत्यु दर का अनुमान लगाया गया था। भारत पर सामानिक आनथकक खस्थनतयों में
अनुमाननत सुिार के कारण, सभी संयुक्त RCP-SSP पररदृश्ों के तहत समय से पहले मृत्यु दर का बोझ सदी
के अंत तक कम होने का अनुमान लगाया गया था। इस थीनसस के भाग के रूप में प्रस्तुत नकए गए पररणामों
से भारतीय नीनत ननमाकताओं को उन क्षेिों की पहचान करने में मदद करने की उम्मीद है िहां शमन नीनतयों
को ननदेनशत नकया िाना चानहए और अनिक कुशल र्ायु गुणर्त्ता प्रबंिन योिना तैयार करनी चानहए, तानक अनिकतम स्वास्थ्य लाभ प्राप्त नकया िा सके।
ix Contents
Acknowledgements ... i
Abstract ... iii
Hindi Abstract ... vi
Contents ... ix
List of Figures ... xiii
List of Tables ...xx
Chapter 1: Introduction ...1
1.1 Air pollution and aerosols ...2
1.2 Environmental health pathways ...4
1.3 Health impact of PM2.5: Global and national status ...7
1.4 Air quality guidelines and standards ...12
1.5 Status of air pollution in India ...13
1.6 Definition of the problem...19
1.7 Objectives ...22
1.8 Air quality guidelines and standards ...23
Chapter 2: Ambient PM2.5 exposure over India ...24
2.1 Rationale for satellite-based approach ...25
2.2 Ambient PM2.5 exposure ...27
2.2.1 Satellite retrieval ...27
2.2.1.1 MISR retrieved AOD ...27
2.2.1.2 Scaling factors ...28
2.2.1.3 Calibration and validation of MISR derived PM2.5 ... 31
2.2.2 Spatio-temporal variability ...32
2.2.3 PM2.5 exposure at district and state level ...38
x
2.3 Change in PM2.5 with urban expansion ...44
2.3.1 Approach ...45
2.3.1.1 Choice of cities ...45
2.3.1.2 The Global Human Settlement Layer ...45
2.3.1.3 MODIS MAIAC AOD ...48
2.3.2 Urban expansion of Indian cities ...52
2.3.3 Change in PM2.5 in the cities ...68
2.4 PM2.5 concentration over Delhi NCR at 1km resolution ...70
2.4.1 Approach ...72
2.4.1.2 Analysis of meteorology ...74
2.4.1.3 Analysis of fire count data as a proxy of open biomass burning ...75
2.4.2 Spatial and temporal variation of PM2.5 over Delhi NCR ...76
2.4.2.1 Pollution build up in dry season...77
2.4.2.2 Is Diwali effect detectable at weekly scale? ...86
2.4.2.3 Post 2009 enhancement in the first pollution episode ...88
2.4.2.4 The rural-urban divide in pollution build-up in Delhi NCR ...89
2.5 Summary ...94
Chapter 3: Estimation of premature mortality associated with chronic PM2.5 exposure in India 96 3.1 Introduction ...97
3.2 Approach ...98
3.2.1 Estimation of baseline mortality ...99
3.2.2 Estimation of relative risk from exposure-response functions ...102
3.2.2.1 The Integrated Exposure-Response functions ...102
3.2.2.2 Global Exposure Mortality Model (GEMM) ...104
3.2.3 Exposed adult population ...106
3.3 Estimation of premature mortality across Exposure Response Functions ...107
3.4 Sensitivity of premature mortality burden to baseline mortality ...116
xi
3.5 Interpretation for policy ...117
3.6 Summary ...122
Chapter 4: Expected health benefits from various mitigation policies ...123
4.1 Introduction ...124
4.2 Expected health benefits for achieving various air quality guidelines and standards125 4.3 Mitigating household air pollution ...129
4.3.1 Approach ...130
4.3.1.1 Development of mitigation scenarios ...130
4.3.1.2 Contribution of individual household sources towards ambient PM2.5 ...133
4.3.2 Mitigation of emission from various household sources ...136
4.3.3 Expected health benefits ...140
4.4 Assessment of the recent odd-even traffic restriction policy in Delhi ...141
4.4.1 Approach ...144
4.4.1.1 Analysis of satellite-based PM2.5 at 3km ...144
4.4.1.2 Meteorological analysis ...147
4.4.1.3 Analysis using chemical transport models ...148
4.4.2 Change in ambient PM2.5 during pre and post-intervention period ...150
4.4.2.1 By using satellite data ...150
4.4.2.2 Analysis of chemical transport model output ...155
4.5 Summary ...158
Chapter 5: Projection of ambient PM2.5 exposure and associated premature mortality burden in India for future climate...161
5.1 Introduction ...162
5.2 Projection of PM2.5 exposure ...163
5.2.1 Approach ...163
5.2.1.1 Satellite derived PM2.5 for the baseline period (2001-2005) ...163
5.2.1.2 Analysis of CMIP5 models for estimating PM2.5 exposure ...164
xii
5.2.2 Projection of ambient PM2.5 ...169
5.3 Projection of socioeconomic factors ...173
5.3.1 Socioeconomic drivers in the SSP scenarios ...176
5.3.2 Population distribution under SSP scenarios ...177
5.3.3 Projection of baseline mortality ...184
5.4 Projection of premature mortality ...187
5.5 Attribution of projected premature mortality burden to key factors ...197
5.6 Summary ...205
Chapter 6: Summary and future directions...207
6.1 Summary of results ...208
6.2 Future prospects ...211
References ...213
Abbreviations ...250
Appendix A: Geographical location of the Indian states ...255
Appendix B: Publications from thesis ...256
Appendix C: Other publications ...259
Brief Biodata...261
xiii List of figures
Figure Captions Page Number
Fig.1.1 The environmental health pathway spanning from sources of air pollution to health impacts.
6 Fig.2.1 Location of PM2.5 monitoring stations in India as of December 2018.
Colour of each of the points show the percentage of days in 2018 when PM2.5
was monitored.
26 Fig. 2.2 Spatial distribution of the conversion factor (ratio of simulated PM2.5
and simulated AOD, averaged through 2004-2011) across India obtained from GEOS-Chem chemical transport model simulation for January through
December (a-l).
30
Fig. 2.3 (a) Spatial distribution of PM2.5 (over India. (b) Rate of change of PM2.5/year from 2001 to 2015. The bold black lines depict regions where α<0.1.
34 Fig 2.4 Total PM2.5 emission in kg/m2/year (a) 2001, (b) 2015 and (c) the
change in PM2.5 emission from 2001 to 2015.
34 Fig. 2.5 Annual PM2.5 concentration (µg/m3) from 2001 through 2015. 35 Fig. 2.6 Monthly PM2.5 concentration (µg/m3), January through December (a-
l) averaged from 2001 through 2015.
39 Fig. 2.7 Rate of change of monthly PM2.5 concentration (µg/m3/year) January
through December (a-l) averaged from 2001 to 2015. Values with significance level, α<0.1 are marked by bold black lines.
40 Fig. 2.8 District level PM2.5 concentration (µg/m3) for a)2001, b)2015 and c)
average over a 15-year period from 2001 to 2015.
41 Fig. 2.9 a) 15 year (2001-2015) average population-weighted PM2.5
concentration µg/m3 at state-level. b) % of districts in a state having
population-weighted PM2.5 exposure above the NAAQS standard of 40 µg/m3 in 2001 and c) % of districts in a state having population-weighted PM2.5 exposure above the NAAQS standard of 40 µg/m3 in 2015.
43
Fig. 2.10 depicts the geographical location of the cities which were included in the study. The size of the dots indicates the area by which the city
expanded from 2001-2015 (km2).
47 Fig. 2.11 Sample domains which were considered around each city for the
analyses. C denotes the city centre.
48 Fig. 2.12 Validation of MODIS MAIAC PM2.5 with CPCB measured PM2.5.
Each point indicates annual PM2.5 concentration for a CPCB site.
50 Fig. 2.13 Column 1 indicates the settlement classification for Y1 and Column
2 indicates the settlement classification for Y2. ‘0’, ‘1’, ‘2’ and ‘3’ denotes
‘no-settlement’, ‘rural’, ‘low-density urban’ and ‘high density urban’
respectively. The last column indicates % change in PM2.5 from Y1 to Y2.
The cities are arranged by descending order of population within the city boundaries (Census 2011).
58
Fig. 2.14 The ratio of CU/UU for 3 groups of cities ‘large’, ‘medium’ and
‘small’ for the 3 considered domains ‘limited’, ‘intermediate’ and ‘big’.
68
xiv
Fig. 2.15 % change in PM2.5 in the 3 domains (150km ie. big, 100km ie.
intermediate and 50km ie. limited) 100 for the three settlement classes (CU, UU and RR) in the three groups of cities (a) large, (b)medium and (c) small.
70 Fig. 2.16 Scatterplot between bias-corrected MAIAC-PM2.5 (after calibration
with ~1400 data points) and coincident remaining in-situ PM2.5 measured at CPCB sites in Delhi. PM2.5 measurements from ground-based sites for the year 2016 are randomly selected for validation. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
73
Fig. 2.17 Total number of pixels in a week (colour scale) when AOD data from MODIS MAIAC algorithm is available within the boundaries of NCR that are used to derive the weekly statistics. Y-axis represents the weeks. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
74
Fig. 2.18 Study area of interest, the light blue shade indicates NCR (national capital region of Delhi), the dark blue shade indicated NCT (National Capital Territory). The magenta box indicates the extent where the fire events were estimated. The Figure is overlain by prevailing surface winds during October and November. represents 10 m/s. This figure has been published in
Chowdhury et al., 2019a (see Appendix B).
75
Fig. 2.19 (a) Mean ambient PM2.5 ground-level concentrations (µg/m3). The two blue boxes represent the upwind (A) and downwind (B) flanks of Delhi NCR and (b) trend (µg/m3 per year) over the NCR during the dry season for the period 2001-2016. The green contours mark the trend estimated with p- value < 0.1. Delhi NCT boundary is marked by a bold line. Surface wind is also marked in the left panel. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
78
Fig. 2.20 PM2.5 exposure (in µg/m3) over the NCR with the NCT marked with bold lines during (a) post-monsoon (October to November), (b) winter (December, January, February) and (c) summer (March, April, May, June).
represents 10 m/s. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
79
Fig. 2.21 Spatial pattern of PM2.5 (µg/m3) over the Delhi National Capital Region during the 15 dry periods considered in this study. National Capital Region is marked by a bold blue line. Mean wind during the dry period is overlain. Delhi NCT is marked by bold blue lines. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
80
Fig. 2.22 Comparison between in-situ CPCB data and MODIS-MAIAC retrieved PM2.5. Each point for satellite-based retrieval represents the average PM2.5 over Delhi obtained from MODIS-MAIAC, with the error bars
representing the variability within the National Capital Region of Delhi. Each point for in-situ data represents the average PM2.5 over Delhi obtained from the CPCB sites spread across Delhi, with the error bars representing the spread across the monitoring sites within the National Capital Region of Delhi. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
81
xv
Fig. 2.23 Average boundary layer depth across the dry seasons over the National Capital Region of Delhi. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
81 Fig. 2.24 (a) Weekly PM2.5 concentration over Delhi NCR and NCT. The
blue and orange lines depict mean (shaded areas represent ±1σ, σ is standard deviation) PM2.5 concentration; (b) Weekly PM2.5 concentration over the upwind and downwind boxes marked as A and B in Fig. 2a, and (c) Weekly mean boundary layer depth over Delhi NCR (in blue) and the number of fire events in the upwind region (demarcated by a magenta box in Fig. S2 in SI) in red with shades showing ±1σ. This paper is published in Chowdhury et al., 2019a (see Appendix B).
82
Fig. 2.25 Spatial weekly PM2.5 concentration over Delhi NCR (marked by blue boundaries) and NCT (marked by bold blue boundaries). Week 1 indicates 24-30 September.
83 Fig. 2.26 Weekly PM2.5 exposure derived from CPCB data obtained from ~12
sites within the NCT for the dry season of 2014-15 and 2015-16. The red line and the black line shows the fire count and boundary layer depth respectively for the above mentioned years. The shaded portion of the curve shows ±1 SD.
The two peak episodes of PM2.5 concentration identified from 15-years of satellite data (in Fig. 3) concur with the ground-based measurements (though in-situ data are for only two years). This figure has been published in
Chowdhury et al., 2019a (see Appendix A).
86
Fig. 2.27 Box plot (showing 5th to 95th percentile ranges) of ambient PM2.5
concentrations during “Diwali” week with and without “Diwali” days. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
87 Fig. 2.28 Inter-annual variability in weekly (a) ambient PM2.5 concentrations
(in µg/m3) in Delhi NCR and (b) fire count in the upwind regions. This paper has been published in Chowdhury et al., 2019a (see Appendix B).
90 Fig. 2.29 Weekly PM2.5 concentration (in blue) and fire events (in red) in the
pre-2009 (dashed lines) and post-2009 era (solid lines) over NCR. This paper has been published in Chowdhury et al., 2019a (see Appendix B).
90 Fig. 2.30 Weekly PM2.5 concentration (in blue) and fire events (in red) in the
pre-2009 (dashed lines) and since-2009 (solid lines) over NCT (a), Upwind box A (b) and downwind box B (c).This figure has been published in Chowdhury et al., 2019a (see Appendix B).
91
Fig. 2.31 Identification of 1km grids in NCR as urban (dark green), rural (light green) and no settlement (white). The two boxes A, B depict the upwind box A and the downwind box B. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
92
Fig. 2.32 Weekly PM2.5 concentration over the urban (orange), rural (blue) and minimal-settlement (black) settlement classes of Delhi NCR. The solid lines depict mean (shade represents ±1σ) PM2.5 concentration. This figure has been published in Chowdhury et al., 2019a (see Appendix B).
93
Fig. 3.1 Cumulative distribution plot of ambient PM2.5 exposure among Indian.
97
xvi
Fig. 3.2 Non-linear fit for per capita GDP against baseline mortality for (a) COPD (b) IHD and (c) stroke. The function for COPD and stroke is
GDP=a×yb and GDP=a×〖(1+y)〗b, where ‘a’ and ‘b’ are disease-specific constants. This figure has been published in Chowdhury et al., 2016. (see Appendix B).
101
Fig. 3.3 State-specific baseline mortality for (a) COPD, (b) IHD and (c) Stroke. The transition from green shade to red signifies crossing over the all India average baseline mortality. This figure has been published in
Chowdhury et al., 2016 (see Appendix B).
101
Fig. 3.4 Shape of the exposure-response curves for (a) COPD (b) IHD (c) Stroke and (d) Lung cancer.
105 Fig. 3.5 (a) Exposed population (Census, 2011) at the district level over India.
(b) The exposed population at the state level over India.
106 Fig. 3.6 Comparison of premature mortality burden and the relative shares of
each disease to the burden attributed to ambient PM2.5 exposure in India in 2015 estimated using three different ERFs’.
109 Fig. 3.7 Total Premature mortality at district level over India using (a)
IER2010, (b) IER2015 and (c) GEMM exposure-response function.
113 Fig. 3.8 Population weighted PM2.5 exposure at the state level over India. 114 Fig. 3.9 Premature mortality at the state level over India using (a) IER2010
(b) IER2015 and (c) GEMM exposure-response functions.
115 Fig. 3.10 Percentage change in estimated premature mortality adjusted for
state-specific baseline mortality relative to uniform baseline mortality. Shades of red (blue) tinge denote lower (higher) estimate in annual premature death using a single value of baseline mortality across India.
119
Fig. 3.11 Relative share (in %) of the burden for each state to the total premature mortality burden due to ambient PM2.5 exposure in India for (a) IER2010 (b) IER2015 and (c) GEMM exposure-response functions.
120 Fig. 3.12 Deviation (in %) of state-level premature mortality rates per
100,000 population with respect to all-India mean estimated using (a) IER2010, (b) IER2015 and (c) GEMM exposure response functions.
121 Fig. 4.1 (a) Location of districts having PM2.5 exposure within specified
standards and guidelines. (b) Mass of PM2.5 (µg/m3) that needs to be mitigated in each district to meet the Indian NAAQS standard of 40 µg/m3.
126 Fig. 4.2 Health benefit of meeting different air quality guidelines and
standards.
128 Fig. 4.3 District wise HAP (household air pollution exposure) in µg/m3,
downscaled from state-level estimates by Balakrishnan et al., 2012 with the help of Census 2011 household (HH) data.
130 Fig. 4.4 Relative contributions (%) of biomass use for (a) solid fuel cooking,
(b) space heating and (c) water heating and kerosene use for (d) lighting to annual ambient PM2.5 exposure at the district level at baseline, 2015. This figure is published in Chowdhury et al., 2019b (see Appendix B).
135
Fig. 4.5 The percentage of ambient PM2.5 exposure that can be attributed to household PM2.5 sources at baseline, 2015. This figure has been published in Chowdhury et al., 2019b (see Appendix B).
137
xvii
Fig. 4.6 (a) Percentage of districts where ambient PM2.5 exposure exceeds various guidelines before and after mitigation of household emissions.
Definition of the scenarios in the x-axis are provided in Table 1 and (b) annual PM2.5 exposure that needs to be further mitigated in each district after complete mitigation of household PM2.5 in order to achieve Indian standard in that district. This figure has been published in Chowdhury et al., 2019b (see Appendix B).
137
Fig. 4.7 Changes in population-weighted mean (±1σ shown by error bars) annual ambient PM2.5 exposure (upper panel) and percentage averted premature mortality based (bottom panel). The range of baseline PM2.5
estimate is shaded in the top panel; the mean value is indicated by the dotted line. The dashed horizontal line in the top panel represents the Indian
NAAQS. See text for details. This figure has been published in Chowdhury et al., 2019 (see Appendix B).
139
Fig. 4.8 (a, b) Bias correction of satellite-derived PM2.5 against in-situ PM2.5
concentration archived at CPCB stations. (c) validation of bias-corrected satellite data against PM2.5 concentration archived by SAFAR. This figure has been published in Chowdhury et al., 2017 (see Appendix B).
146
Fig. 4.9 (a) 13-year average wintertime PM2.5 (in µg/m3) over the National Capital Region (NCR) of Delhi, the arrows indicate the wind direction and the size of the arrows indicate the wind speed. The dotted lines indicate major highways. (b) Major point sources of emission (major industrial clusters, coal and gas-based power plants) in NCR. This figure has been published in Chowdhury et al., 2017 (see Appendix B).
151
Fig. 4.10 (a) 13-year Average PM2.5 exposure (in µg/m3) during the 3-time segments over NCR. (b) 13 year mean Stability parameter (Difference
between the temperature at 1000hPa and 850hPa) during the 3-time segments.
(c) 13 year mean wind speed and wind direction during the three-time segments. This paper has been published in Chowdhury et al., 2017 (see Appendix B).
152
Fig. 4.11 Stability parameter or the difference between the 1000hpa temperature and the 850hpa temperature during (a) in the pre-intervention period (16th to 31st December 2015), (b) during the traffic restriction period (1st January to 15th January 2016) and (c) in the post-intervention period (16th January to 31st January. This figure has been published in Chowdhury et al., 2017 (see Appendix B).
152
Fig. 4.12 Wind speed and wind direction (a) in the pre-intervention period (16th to 31st December 2015), (b) during the traffic restriction period (1st January to 15th January 2016) and (c) in the post-intervention period (16th January to 31st January). This figure has been published in Chowdhury et al., 2017 (see Appendix B).
153
Fig. 4.13 Anomaly in Stability parameter or the difference between the
1000hpa temperature and the 850hpa temperature w.r.t long term mean during (a) in the pre-intervention period (16th to 31st December 2015), (b) during the traffic restriction period (1st January to 15th January 2016) and (c) in the post-intervention period (16th January to 31st January. Wind speed and wind
153
xviii
direction (d) in the pre-intervention period (16th to 31st December 2015), (e) during the traffic restriction period (1st January to 15th January 2016) and (f) in the post-intervention period (16th January to 31st January. This figure has been published in Chowdhury et al., 2017 (see Appendix B).
Fig. 4.14 Anomaly of PM2.5 (Deviation of PM2.5 in the 3 time segments from 13 year average exposure) in the (a) pre-intervention (16th December 2016 – 31st December 2016), (b) intervention (1st January 2016 – 15th January 2016) and (c) post-intervention (16th January 2016 – 31st January 2016) periods with respect to 13 year mean in Delhi NCT. The same for larger NCR is shown as an inset. This figure has been published in Chowdhury et al., 2017 (see Appendix B).
154
Fig. 4.15 Percentage change in PM2.5 concentration (during 1-15 Jan 2016) due to reduction in emissions due to the odd-even rule. This figure has been published in Chowdhury et al., 2017.
156 Fig. 5.1 MISR derived ambient PM2.5 exposure over the Indian landmass for
the baseline period (2001-2005).
164 Fig. 5.2 Spatial distribution of bias for RCP4.5 scenario (a) and RCP8.5
scenario (b) linear regression statistics from the comparison of CMIP5 model- derived ambient PM2.5 exposure against MISR-retrieved PM2.5 in (c) RCP4.5 and (d) RCP8.5 scenario for the period 2011-2015. The Figures in the right panel compares model derived PM2.5 (y-axis) and satellite-derived PM2.5 (x- axis). Each red dot represents average PM2.5 exposure for each grid cell in the Figures on the left panel. The black box encompasses the desert region in Rajasthan where the CMIP5 models overestimate PM2.5 exposure relative to MISR-derived PM2.5. This figure has been published in Chowdhury et al., 2018 (see Appendix B).
170
Fig. 5.3 Projected PM2.5 exposure averaged over India for the future under both RCP4.5 and RCP8.5 scenarios. The shaded part represents the range (1- 99%) in all-India averaged PM2.5 across the 13 CMIP5 models. The black line represents the all-India averaged baseline (2001-2005) PM2.5 exposure level.
Solid and dashed lines indicate the projected exposure using approach 1 and approach 2 respectively. This figure has been published in Chowdhury et al., 2018 (see Appendix B).
171
Fig. 5.4 Spatial distribution of the projected changes in PM2.5 exposure (ΔPM2.5) from the baseline (2001-2005) exposure under (top panel) RCP4.5 and (bottom panel) RCP8.5 in (a and d) near future (2031-2040), (b and e) distant future (2061-2070) and (c and f) far future (2091-2100). Reddish (bluish) tinge signifies projected to increase (decrease) in PM2.5 in future.
This figure has been published in Chowdhury et al., 2018 (see Appendix B).
172
Fig. 5.5 Spatial patterns of ambient PM2.5 concentration (µg/m3) in three representative decades of (left) near future (2031-2040), (middle) distant future (2061-2070) and (right) far future (2091-2100) under (top panel) RCP8.5 and (bottom panel) RCP4.5 scenario. This figure has been published in Chowdhury et al., 2018 (see Appendix B).
174
Fig. 5.6 Depiction of the challenges for mitigation and adaptation (in relative scale) for the 5 SSP scenarios.
177
xix
Fig. 5.7 (a) Total population for the year 2010. (b) Projected total population for each of the five SSP scenario populations over the Indian landmass.
179 Fig. 5.8 Spatial distribution of the total projected population over India for the
near future (2031-40), distant future (2061-70) and far future (2091-2100) for all the 5 SSPs. The x-axes in each sub-plot are Latitude (N) and the y-axes represent Longitude (E).
181
Fig. 5.9 (a) Exposed population (above 25 years) for the year 2010. (b) Projected exposed population (above 25 years) for each of the five SSP scenario populations over the Indian landmass.
182 Fig. 5.10 Spatial distribution of the projected exposed population (>25 years)
over India for near future (2031-40), distant future (2061-70) and far future (2091-2100) for all the 5 SSP scenario. The x-axes in each subplot are Latitude (N) and the y-axes represent Longitude (E).
183
Fig. 5.11 Projected changes in demographic patterns of 2100 relative to 2010 for the 5 SSP scenario population. This figure has been published in
Chowdhury et al., 2018 (see Appendix B).
184 Fig. 5.12 Projected changes in baseline mortality (per 100,000 population) for
(a) COPD, (b) IHD and (c) stroke in five SSP scenarios using the present-day relationship between baseline mortality and GDP. The shaded region in the plots represents the uncertainty range in baseline mortality estimated using the ranges in the coefficients of the non-linear GDP-baseline mortality functions. This figure has been published in Chowdhury et al., 2018 (see Appendix B).
186
Fig. 5.13 Projected percentage changes in premature mortality burden from ambient PM2.5 exposure in India for the five SSP scenario population using projected baseline mortality under (a) RCP4.5 and (b) RCP8.5 scenarios. The range (1-99%) of premature mortality (as a function of standard error in baseline mortality model and the range of PM2.5 from CMIP5 models) is shown as shades around the mean values as bold lines. The combination of the RCP8.5 scenario and SSP1 is practically impossible. This figure has been published in Chowdhury et al., 2018 (see Appendix B).
188
Fig.5.14 Spatial distribution of premature mortality per 0.5º×0.5º grids over the Indian landmass for the 3 representative decades (2031-40), (b) 2061-70 and (c) 2091-100 under RCP4.5 scenario PM2.5 exposure. The SSPs’ are listed chronologically from left to right. The x-axes in each subplot are Latitude (N) and the y-axes represent Longitude (E).
195
Fig. 5.15 Spatial distribution of premature mortality per 0.5º×0.5º grids over the Indian landmass for the 3 representative decades (2031-40), (b) 2061-70 and (c) 2091-100 under RCP8.5 scenario PM2.5 exposure. The SSPs’ are listed chronologically from left to right. The x-axes in each subplot are Latitude (N) and the y-axes represent Longitude (E).
196
xx List of Tables
Table Captions Page
Number Table 2.1 Area under 3 settlement transitions (CU, UU and RR) for the cities
included in the analysis (in km2). The estimates in each column for the 3 domains 50km, 100km, 150km are separated by ‘,’. The cities are arranged in descending order of population within the city boundaries (Census 2011).
54
Table 2.2 The population-weighted PM2.5 for Y1 and Y2. The estimates in each column for the 3 domains 50km, 100km, 150km are separated by ‘,’. The cities are arranged in descending order of population within the city
boundaries (Census 2011).
56
Table 2.3 Mean (±1σ) ambient PM2.5 concentration (µg/m3) in various settlement classes in our study region.
92 Table 3.1 State level baseline mortality for the 3 diseases (5-95% CI) and
premature mortality (per year, 5-95% CI) estimated for all the states of India using the 3 exposure-response functions IER2010, IER2015 and GEMM. For the location of the Indian states please see Appendix A)
109
Table 3.2 State level baseline premature mortality (per year, 5-95% CI) estimated for all the states of India using the 3 exposure-response functions IER2010, IER2015 and GEMM. For the location of the Indian states, please see Appendix A.
111
Table 4.1 Definition of the target scenarios and premature mortality achieved if the respective targets are achieved.
128 Table 4.2 Definition of the scenarios formulated for the study. The numeric
values depict the percent mitigation of the particular household source. C = cooking, L = lighting, S = Space Heating, A= aspirational progress, M=
moderate progress, S= slow progress.
131
Table 5.1 List of CMIP5 models along with their spatial resolution (in degrees); the output of which is analyzed to project ambient PM2.5
concentration and the surface concentration of the PM2.5 species produced as output from the considered models.
166
Table 5.2 Projected mean estimates (±uncertainty) of premature mortality burden per year (in million) India due to ambient PM2.5 exposure till the end of the century for the ten combined RCP-SSP scenarios. Premature mortality during the baseline period was estimated as 0.81(±0.29) million
189
Table 5.3 Projected mean estimates (±uncertainty) of crude decadal premature mortality burden per 100,000 exposed population from ambient PM2.5
exposure for RCP4.5 and RCP8.5 scenarios. Total premature deaths for decades 2031-2040 (near future), 2061-2070 (distant future) and 2091-2100 (far future) are shown in Table 5.1. The crude mortality rate (per 100,000 population) for the baseline period was estimated to be 32.2(±12.1)
191
Table 5.4 Framework of the sensitivity study where the upward and downward arrows indicate an increase and decrease of the respective parameter relative to the baseline period in that particular sensitivity study;
197
xxi
while ‘no change’ is indicated by a horizontal arrow. The purpose of each sensitivity study and its interpretation are provided in the last column.
Table 5.5 Changes in projected premature mortality burden for the four
sensitivity studies (SA) relative to the burden in the baseline period. Values are in 1000s. The details of the four sensitivity studies are summarized in Table 5.4.
203
