monitoring of fine particulates in hong kong and pearl river delta region using remote sensing

Copyright Undertaking

This thesis is protected by copyright, with all rights reserved.

By reading and using the thesis, the reader understands and agrees to the following terms:

1. The reader will abide by the rules and legal ordinances governing copyright regarding the use of the thesis.

2. The reader will use the thesis for the purpose of research or private study only and not for distribution or further reproduction or any other purpose.

3. The reader agrees to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.

IMPORTANT

If you have reasons to believe that any materials in this thesis are deemed not suitable to be distributed in this form, or a copyright owner having difficulty with the material being included in our database, please contact [email protected] providing details. The Library will look into your claim and consider taking remedial action upon receipt of the written requests.

Pao Yue-kong Library, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

http://www.lib.polyu.edu.hk

MONITORING OF FINE PARTICULATES IN HONG KONG AND

PEARL RIVER DELTA REGION USING REMOTE SENSING

MUHAMMAD BILAL

Ph.D

The Hong Kong Polytechnic University

2014

lbsys

Text Box

This thesis in electronic version is provided to the Library by the author. In the case where its contents is different from the printed version, the printed version shall prevail.

The Hong Kong Polytechnic University

Department of Land Surveying and Geo-Informatics

Monitoring of Fine Particulates in Hong Kong and Pearl River

Delta Region Using Remote Sensing

Muhammad BILAL

A thesis submitted in partial fulfillment of the requirements for

the degree of Doctor of Philosophy

July 2013

Certificate of Originality

CERTIFICATION OF ORIGINALITY

I hereby declare that this is my own work and that, to the best of my

knowledge and belief, it reproduces no material previously published

or written nor material which has been accepted for the award of any

other degree or diploma, except where due acknowledgement has

been made in the text.

(Signed)

Muhammad BILAL (Name of Student)

Dedications

DEDICATION

I would like to dedicate this thesis to

my beloved family

Monitoring of PM2.5 in Hong Kong and PRD Region Using Remote Sensing Abstract

i

ABSTRACT

Abstract of thesis entitled:

Monitoring of Fine Particulates in Hong Kong and Pearl River

Delta Region Using Remote Sensing

Supervised by

Professor Janet E. Nichol

Submitted by

Muhammad BILAL

For the degree of Doctor of Philosophy at the Hong Kong Polytechnic

University in July 2013

The main objective of this study is to monitor and understand the

behavior of fine particulate matter (PM2.5) over the complex terrain of Hong

Kong and the Pearl River Delta (PRD) region. These experience some of the

worst air quality conditions in the world due to local emissions from vehicles as

well as regional aerosol particles from ocean–going vessels (OGV) and industries

throughout the PRD region. Aerosol particulate concentrations have been

investigated at regional to global scales using satellite aerosol products.

However, to obtain accurate PM2.5 estimates, better aerosol retrieval algorithm

which operate at high resolution and over the mixed surface types in the study


ii

region, is required. Most satellite–based aerosol optical depth (AOD) algorithms

operate at spatial resolutions of several to several tens of kilometers. Also, most

use a Radiative Transfer Model (RTM) to construct a look-up table (LUT) to act

as a map between measurements and physical quantities. In the current study, a

Simplified Aerosol Retrieval Algorithm (SARA) was developed from MODIS

images for use over Hong Kong at high (500 m) spatial resolution and without

using a LUT. Instead, RTM calculations were applied directly to the MODIS data,

with the aerosol properties derived from a local urban Aerosol Robotic Network

(AERONET) station at the Hong Kong Polytechnic University, and surface

reflectance from the MOD09GA level–2 daily surface reflectance product. The

500 m AOD retrieved from the SARA showed a high consistency with ground-

based AOD measurements, with average correlation coefficient (R) ~ 0.963, Root

Mean Square Error (RMSE) ~ 0.077, and Mean Absolute Error (MAE) ~ 0.065

than MOD04 C005 AOD (R ~ 0.883, RMSE ~ 0.140, and MAE ~ 0.123).

In order to use the satellite–derived AOD to develop a PM2.5 model for

accurate prediction of PM2.5, the SARA–retrieved AOD at 500 m resolution was

constrained using surface meteorological variables including temperature

(STEMP), relative humidity (SRH), and wind speed (SWS), and Planetary

Boundary Layer Height (PBLH) and surface pressure at 500 m resolution. The

SARA PM2.5 model of the AOD–PM2.5 relationship was developed at four

urban/suburban air quality stations based on bins of meteorological variables

for the years 2007 and 2008 (autumn and winter). An almost perfect correlation

(R ~ 0.99) between AOD and PM2.5 was found with the parameters 2.53–3.40

m/s WS, 47–79% RH, and ranges of PBLH from 300–449 m. The SARA PM2.5

model was evaluated for derivation of PM2.5 concentrations by comparison with

ground–level PM2.5 at urban/suburban (Central, Tsuen Wan, Tung Chung and

Yuen Long) and rural (Tap Mun) air quality stations in Hong Kong. The results

demonstrate a better agreement between SARA predicted PM2.5 and ground–

level observed PM2.5 than the existing AOD model, explaining around 80% to

82% of the variability in PM2.5 concentrations. Therefore, it can be concluded


iii

that the SARA PM2.5 model is superior to previous method of monitoring of PM2.5

over urban regions. It can be used for detailed monitoring of PM2.5 over regions

other than the present study area, by adopting two methodologies: (i) the SARA

method of retrieving high resolution (500 m) AOD from satellite images, and (ii)

refinement of regression coefficients based on bins of meteorological variables,

as proposed in this study.

Keywords: AERONET; MODIS; AOD; New Simplified Algorithm; 500 m

resolution; PM2.5; WRF; Hong Kong

Publications

iv

PUBLICATIONS ARISING FROM THE THESIS

JOURNAL

Bilal, M., Nichol, J.E., Bleiweiss, M.P., DuBois, D.W. (2013). A Simplified high

resolution MODIS Aerosol Retrieval Algorithm (SARA) for use over mixed

surfaces, Remote Sensing of Environment, 136, 135-145.

CONFERENCE

Bilal, M., Nichol, J.E., Bleiweiss, M.P., and Dubois, D. (2012). Retrieving MODIS

Aerosol Optical Depth in real time at 500 m resolution: urban–scale

evaluation over Hong Kong, European Aerosol Conference (EAC), Granada,

Spain, 2nd – 7th September.

Acknowledgements

v

ACKNOWLEDGEMENTS

All praises to Almighty Allah, the creator of the universe, who blessed me

with the knowledge and enabled me to complete this thesis. All respects to the

Holy Prophet Muhammad (May Allah grant peace and honor to him and his

family), who is the last messenger, whose life is a perfect model for the whole

humanity.

I would like to express deeply-felt thanks and highest respect to my Chief

Supervisor, Professor Janet E. Nichol, for being a wonderful, resourceful and

always-available supervisor who guided me during my research studies at the

Hong Kong Polytechnic University (PolyU), professionally as well as personally.

She is really a kindhearted, generous and one of the best supervisors in the

PolyU.

I am very grateful to the head of LSGI, chair of departmental research

committee and all the other faculty members of LSGI for their support, especially

Professor Bruce King to manage the workstation in Photogrammetry and

Remote Sensing lab for simulation of Weather Research and Forecasting (WRF)

model. I am sincerely thankful to the General Office (GO), the Research Office

(RO), the Finance Office (FO) and the Student Affair Office (SAO) for their

continuous support during the academic period. I am also thankful to Mr.

Kenneth Yau (technician) for his kind support to manage my PC throughout my

research period.

This study is not merely dependent upon the research work conducted at

PolyU, in fact, half of it is based on the work completed at New Mexico State

University (NMSU). Therefore, I am especially indebted to Mr. Max P. Bleiweiss

(Research Scientist at Center for Applied Remote Sensing in Agriculture,

Acknowledgements

vi

Meteorology, and Environment, Department of Entomology, Plant Pathology, and

Weed Science, NMSU, USA) and Dr. David W. Dubois (Department of Plant and

Environmental Sciences, NMSU, USA) for hosting me as a visiting research

scholar to work on a new aerosol retrieval algorithm. I would like to give a

heartfelt, special thanks to Mr. Max Bleiweiss for his continuous support,

enormous advices and kind help throughout my career (undergrad to PhD). He is

really a kind and gentleman, and always supports me whenever I need his help.

Without his endless support and valuable suggestions, this study would not have

been possible.

This study received very useful suggestions from Professor Tracey

Holloway (Nelson Institute Center for Sustainability and the Global Environment

(SAGE), University of Wisconsin–Madison, USA). She suggested me to include

WRF model in the research for monitoring of Fine Particulates over Hong Kong.

She helped me to contact with Dr. Scott Spak (Civil and Environmental

Engineering, the University of Iowa, USA) to visit the Center for Global and

Regional Environmental Research (CGRER) as a visiting research scholar to work

on WRF model. I am very obliged to her for fruitful pieces of advice and

suggestions; moreover, I am also thankful to Dr. Scott for providing me invitation

letter to visit his research lab. Dr. Scott is really a nice man and spent a lot of time

on me to understand the mechanism of WRF model. I acknowledge the help from

postgraduate students (Mr. Pablo Saide and Ms. Pallavi Marrapu) at CEGRER for

compiling and simulating the WRF model.

I appreciate the kind support of Dr. Waqas A. Qazi (Colorado University,

USA) during my training session of WRF at National Center for Atmospheric

Research (NCAR) in Boulder, USA. I am also very grateful to Dr. Muhammad

Saeed (the University of Iowa) for his hospitality and support during my visit to

CGRER, the University of Iowa.

I would like to extend my gratitude to my best friend Mr. Muhammad

Imran Ghafoor (National University of Agricultural Sciences, Pakistan), Mr. Majid

Acknowledgements

vii

Nazeer, Ms. Thom Vi, Mr. Nithiyanandam Yogeswaran, Dr. Wong Man Sing

(Remote Sensing Research Group, LSGI), and Qari Abdul Rasool Attari, Qari

Muhammad Anis Attari, Qari Ramzan Raza Attari and Mufti Muhammad Aqeel

Attari (Dawat-e-Islami, a global non-political movement for the propagation of

the Holy Quran and Sunnah, Hong Kong) for their prayers, support and

encouragement. I am very grateful to Mr. Hafeez-ur-Rahman (Hong Kong

resident) for his kind support during my PhD degree.

I would like to acknowledge NASA Goddard Space Flight Center for MODIS

sensor data and Dr. Brent Holben for helping me with the AERONET stations in

Hong Kong. I am very thankful and appreciate the help of Dr. P.W. Chan for

providing me Sun photometers (Hong Kong International Airport and the City

University of Hong Kong) data for validation purposes. I am very thankful to Dr.

Devin White (Oak Ridge National Laboratory) for providing me MODIS

Conversion Tool Kit (MCTK), and also for his kind support.

I am very grateful to the examiners (Professor Chengcai Li, Peking

University, Beijing, China, and Professor Sundar A. Christopher, University of

Alabama In Huntsville, USA) for evaluating my PhD thesis, and providing me

valuable suggestions and comments to improve the quality of this thesis.

I should express my indebtedness and my deepest sense of gratitude to all

of my family members for their endless love, moral support, encouragement, and

prayers for successful completion of this study.

Lastly, I am very thankful to Research Grants Council (RGC), Hong Kong,

for awarding me Hong Kong PhD Fellowship, which supported me during my

studies and gave me such wonderful opportunities for cultural and personal

engagement.

Table of Contents

viii

TABLE OF CONTENTS

ABSTRACT ............................................................................................................................................ i

PUBLICATIONS ARISING FROM THE THESIS ...................................................................... iv

ACKNOWLEDGEMENTS ................................................................................................................. v

TABLE OF CONTENTS ................................................................................................................. viii

LIST OF FIGURES ............................................................................................................................ xi

LIST OF TABLES ............................................................................................................................. xv

Chapter 1

Introduction ......................................................................................................................... 1

1.1 Background ............................................................................................................................. 1

1.1.1 Fine Particulate Matter (PM2.5) ............................................................................. 1

1.1.2 PM2.5 in Hong Kong .................................................................................................... 2

1.2 Measurement of PM2.5 ......................................................................................................... 5

1.3 MOD04 C004/C005 aerosol retrieval algorithm .................................................... 11

1.4 Research objectives ............................................................................................................ 13

1.5 Dissertation overview ....................................................................................................... 14

Chapter 2

Study Area and Data Used ............................................................................................. 15

2.1 Study Area .............................................................................................................................. 15

2.2 Data Used ............................................................................................................................... 17

2.2.1 Ground–based PM2.5 Concentration Data ....................................................... 17

2.2.2 Ground–based AOD Measurements .................................................................. 17

2.2.3 MODIS Satellite Data ............................................................................................... 19

2.2.4 Ground–based Meteorological Measurements ............................................. 20

2.2.5 WRF Meteorological Data ..................................................................................... 21

Table of Contents

ix

Chapter 3

Characteristics of Particulate Matter (PM2.5) and Influence of

Meteorological Variables .............................................................................................. 23

3.1 Spatio–temporal variations in PM2.5 ............................................................................ 23

3.2 Influence of Seasonal Meteorological Factors ......................................................... 31

Chapter 4

Methodology ...................................................................................................................... 39

4.1 Development of a Simplified high resolution MODIS Aerosol Retrieval

Algorithm (SARA) ................................................................................................................... 39

4.1.1 Statistical analysis .................................................................................................... 46

4.1.1.1 Correlation coefficient (R) ...................................................................... 46

4.1.1.2 Root mean square error (RMSE) .......................................................... 47

4.1.1.3 Mean absolute error (MAE) .................................................................... 47

4.1.1.4 Expected error (EE) ................................................................................... 47

4.1.1.5 Fraction of expected error (FOE) ......................................................... 48

4.2 Development of a SARA PM2.5 model ........................................................................... 49

Chapter 5

Results .................................................................................................................................. 53

5.1 Validation of SARA AOD with AERONET, Sky-radiometer and Microtops II

Sun photometer AOD ......................................................................................................... 53

5.2 Comparison between SARA and MOD04 C005 AOD ............................................. 57

5.3 Sensitivity analysis for SARA methodology .............................................................. 59

5.4 Spatio–temporal pattern of AOD ................................................................................... 61

5.5 Exploring the AOD–PM2.5 relationship with time of day ..................................... 64

5.6 Relationship between PM2.5 and satellite aerosol products ............................... 66

5.7 Development of PM2.5 model using SARA AOD (500 m), HKO and WRF (500

m) data .................................................................................................................................... 70

5.8 Refinement of SARA–retrieved PM2.5 at 500 m resolution using bins of

meteorological variables .................................................................................................. 73

5.9 Validation of SARA–Retrieved PM2.5 at urban/suburban air quality stations

using data of 2009 .............................................................................................................. 80

Table of Contents

x

5.10 Comparison of SARA with previously developed PM2.5 model for Hong Kong

85

5.11 Spatial Variation of SARA–retrieved PM2.5 (500 m) over Hong Kong and PRD

Region: an example for high pollution episode ....................................................... 89

Chapter 6

Summary and Conclusions ............................................................................................ 92

REFERENCES ................................................................................................................................... 95

List of Figures

xi

LIST OF FIGURES

Figure

1.1 Size of Fine Particulate Matter (PM2.5) [Image curtsey of US EPA Office of

Research & Development] ............................................................................................. 1

1.2 Common sources of PM2.5 [image curtsey of Fairbanks North Star Borough

Air Quality Division] ......................................................................................................... 3

1.3 The Hong Kong Air Pollution Index (API) based on all particulates as well

as other gases selected by the Environmental Protection Department

(EPD) ...................................................................................................................................... 4

1.4 Temporal trend of annual PM2.5 concentration in Hong Kong from 2005 to

June, 2012 (Retrieved from: http://www.hongkongcan.org/doclib/2012

%20Jan-Jun%20Review%20-%20ENG_Final.pdf) ............................................... 4

2.1 Study area and the locations of Ground–based Sun Photometers (circles),

air quality stations (triangles) and meteorological stations (round square)

in the complex terrain of Hong Kong ....................................................................... 16

2.2 WRF–ARW Modeling System Flowchart ................................................................ 21

3.1 Four year (2005–2008) daily means of PM2.5 concentrations at five air

quality stations (Central, Tsuen Wan, Tung Chung, Yuen Long and Tap

Mun) ..................................................................................................................................... 27

3.2 Four year (2005–2008) monthly means of PM2.5 at five air quality stations

(Central, Tsuen Wan, Tung Chung, Yuen Long and Tap Mun) ....................... 28

List of Figures

xii

3.3 Four year (2005–2008) weekdays and weekends means of PM2.5 at five

air quality stations (Central, Tsuen Wan, Tung Chung, Yuen Long and Tap

Mun) ..................................................................................................................................... 29

3.4 Four year (2005–2008) hourly means of PM2.5 at five air quality stations

(Central, Tsuen Wan, Tung Chung, Yuen Long and Tap Mun) ....................... 30

3.5 Temporal variations in daily mean PM2.5 concentrations and

meteorological variables (TEMP, RH, and WS) for years 2007–2008 at (a)

Central, (b) Tsuen Wan, (c) Tung Chung, (d) Yuen Long, and (e) Tap Mun

air quality stations. The Y-axis on left represents daily mean PM2.5

concentrations (g/m3, dark grey bars), and four Y-axes on right

represent TEMP (oC, black), RH (%, violet), and WS (m/s, green). .............. 37

3.6 Hourly PM2.5 concentrations from all air quality stations including (a)

Central, (b) Tap Mun, (c) Tsuen Wan, (d) Yuen Long and (e) Tung Chung

as a function of corresponding hourly surface wind direction in Hong

Kong (2007–2008) .......................................................................................................... 38

4.1a Methodology of the Simplified high resolution Aerosol Retrieval

Algorithm (SARA) ............................................................................................................ 40

4.1b Development of SARA PM2.5 model for prediction of PM2.5 ............................ 52

5.1 Validation of (a) SARA AOD, and (b) MOD04 C005 AOD against Hok Tsui

(rural) AERONET AOD from 2007 to 2009. The lines are as follow: EE

envelope = dotted orange, 1:1 line = dashed green and regression line =

solid red. .............................................................................................................................. 54

5.2 Fraction of expected error (FOE) in AOD retrievals of SARA and MOD04

C005 algorithms ............................................................................................................... 55

5.3 Validation of SARA AOD and MOD04 C005 AOD against (a & c) CityU Sky-

radiometer, and (b & d) HKIA Microtops II AOD measurements. The lines

List of Figures

xiii

are as follow: EE envelope = dotted orange, 1:1 line = dashed green and

regression line = solid red. ........................................................................................... 56

5.4 Fraction of Expected Error (FOE) in AOD retrievals of SARA and MOD04

C005 against CityU Sky-radiometer and HKIA Microtops AOD ..................... 57

5.5 Spatial pattern of SARA (a), and C005 (b) AOD for a high aerosol loading

event (30th January 2007) over the Pearl River Delta (PRD) region and

Hong Kong. Also shown, in panel (c), is the SARA AOD under-laid with

road data over Kowloon and Hong Kong Island. ................................................. 62

5.6 Variations in AOD between hilly and urban regions of Hong Kong on (a)

low (10th November 2008), and (b) high (26th January 2007) aerosol

loading conditions. .......................................................................................................... 62

5.7 Temporal trends of (a) MOD04 C005 AOD at 10 km resolution, and (b)

SARA AOD at 500 m resolution from 2007 to 2009 ........................................... 63

5.8 Relationship of 4–hr (09–12hr) averaged PM2.5 concentrations (g/m3)

with SARA (500 m) and MOD04 AOD (10 km) at (a) Central, (b) Tsuen

Wan, (c) Tung Chung, (d) Yuen Long, and (e) Tap Mun air quality stations

in Hong Kong for autumn and winter of 2007 and 2008 ................................. 68

5.9 Relationship of 4–hr (09–12hr) averaged PM2.5 concentrations (g/m3) at

all five air quality stations in Hong Kong in autumn and winter (2007–

2008) with (a) SARA AOD (500 m), and (b) MOD04 AOD (10 km) ............. 69

5.10 Relationship of SARA AOD at all five air quality stations in Hong Kong for

autumn and winter of 2007 and 2008 with (a) bins (interval of 5 g/m3)

of 4–hr (09–12 hr) averaged PM2.5, and (b) monthly means of 4–hr (09-12

hr) averaged PM2.5 .......................................................................................................... 69

List of Figures

xiv

5.11 Descriptive Statistics of surface–level PM2.5, SARA AOD, and

meteorological variables from both WRF–ARW and HKO at four air

quality stations for autumn and winter of 2007 and 2008 ............................. 72

5.12 Validation of SARA–retrieved PM2.5 (g/m3) at 500 m based on (a) Eqs.

5.2 (left) and 5.3 (right), (b) Eqs. 5.4 (left) and 5.5 (right), (c) Eqs. 5.6

(left) and 5.7 (right), (d) Eqs. 5.8 (left) and 5.9 (right), (e) Eq. 5.10 (left)

and 5.11 (right), (f) Eqs. 5.12 (left) and 5.13 (right) and (g) Eqs. 5.14 (left)

and 5.15 (right) using ground–based observed PM2.5 at four

urban/suburban air quality stations in Hong Kong for autumn and winter

of 2009 ................................................................................................................................. 83

5.13 Descriptive Statistics of surface–level observed PM2.5 (g/m3) and SARA–

retrieved PM2.5 (g/m3) using 14 PM2.5 models (Eq. 5.2 to 5.15) at four air

quality stations for autumn and winter of 2009 ................................................. 84

5.14 Evaluation of (a) SARA and (b) Wong et al. (2011) PM2.5 models at Tap

Mun in Hong Kong. .......................................................................................................... 86

5.15 Temporal variations of Observed, SARA, and Wong et al. PM2.5

concentrations at Tap Mun air quality station in Hong Kong for autumn

and winter of the years 2007 to 2009. Red boxes indicate the under and

overestimation of Wong et al’s PM2.5 model. ........................................................ 88

5.15 Spatial distribution of SARA–retrieved PM2.5 over (a) Pearl River Delta

(PRD) region and (b) Hong Kong during a high pollution episode (4th

December 2007) .............................................................................................................. 90

5.16 (a) HYSPLIT four days back-trajectory at three different altitudes 500 m

(red), 2500 m (blue) and 5000 m (green) AGL, and (b) WRF streamlines

showing 10m WD at 4.5 km resolution, during a high pollution episode

(4th December 2007) ...................................................................................................... 91

List of Tables

xv

LIST OF TABLES

Table

1.1 A summary of previously reported MODIS aerosol algorithms in Hong

Kong, as well as new Simplified high resolution MDOIS Aerosol Retrieval

Algorithm (SARA) developed in this study .............................................................. 9

1.2 Description of previously reported PM2.5 studies, as well as this new PM2.5

study in Hong Kong ......................................................................................................... 10

2.1 Latitudes/Longitudes of ground stations .............................................................. 16

2.2 Characteristics of ground–based Sun Photometers ........................................... 19

2.3 A Summary of MODIS Sensor Products .................................................................. 20

2.3 Physics Schemes for WRF–ARW model Simulations ......................................... 22

3.1 Total number of exceedances of 4–year (2005-2008) daily mean PM2.5

mass concentrations....................................................................................................... 24

3.2 Descriptive statistics of 4–year (2005–2008) daily means of PM2.5 ............ 26

3.3 Correlation between daily mean PM2.5 and meteorological variables for

years 2007 and 2008 ..................................................................................................... 34

5.4 Regression coefficients for the prediction of PM2.5 concentrations using

SARA AOD, HKO and WRF meteorological variables at four air quality

stations for autumn and winter of 2007 and 2008 ............................................ 71

List of Tables

xvi

5.5 Regression coefficients for the prediction of PM2.5 concentrations using

SARA AOD, WTEMP and WSH at four air quality stations for autumn and

winter of 2007 and 2008 .............................................................................................. 71

5.6 AOD–PM2.5 correlation based on bins of meteorological variables from

both HKO and WRF model at five air quality stations in Hong Kong for

autumn and winter of 2007 and 2008 .................................................................... 74

5.7 Descriptive statistics of predicted PM2.5 by SARA and Wong et al’s PM2.5

models at Tap Mun air quality station in Hong Kong for autumn and

winter of 2007 to 2009. ................................................................................................ 87

5.8 Wind direction from AWS at HKO, TMS, SLW, WLP and TM during a high

pollution episode (4th December 2007) ................................................................. 89

Introduction Chapter 1

1

Chapter 1

Introduction

1.1 Background

1.1.1 Fine Particulate Matter (PM2.5)

Fine Particulate Matter (PM2.5, Fig. 1.1) has recently been identified as a

severe health hazard (Bell et al., 2007; Pope and Dockery, 2006; Ward and Ayres,

2004). Studies have reported the association of PM2.5 with lung, (Kappos et al.,

2004), respiratory (Gotschi et al., 2008), mutagenic (Fang et al., 2002), and

cardiorespiratory disease (Englert, 2004) and mortality (Dominici et al., 2006;

Gent et al., 2009). For example, an increase of 10 g/m3 in PM2.5 is said to

increase by 4%, 6%, and 8% the rate of cardiopulmonary diseases, lung cancer

and mortality respectively (Pope et al., 2002).

Figure 1.1: Size of Fine Particulate Matter (PM2.5) [Image curtsey of US EPA

Office of Research & Development]


2

Fine Particulate Matter (PM2.5) can be emitted from natural and

anthropogenic sources (Dubovik et al., 2002; El-Fadel and Hashisho, 2001;

Wallace and Hobbs, 2006) or can be formed from gaseous pollutants. Common

sources of PM2.5 are (Fig. 1.2) wood–burning stoves, garbage, open burning,

forest and bush fires, vehicles, and point location sources such as power plants

and industrial sources. In urban areas PM2.5 is normally associated with local

emissions from automobile exhausts (Ho et al., 2006). These are not only the

most important source of PM2.5 (Fraser et al., 2003; Nolte et al., 2002) but also

the main source of secondary particles in the atmosphere through chemical

transformation (gas-to-particles) (Mysliwiec and Kleeman, 2002). Studies

indicate that tropical Asia contributes most to world air pollution due to the

significant increase in aerosol pollutants from anthropogenic and natural

sources (Chung et al., 2005; Lee et al., 2001). PM2.5 directly influences the Earth’s

energy budget, surface temperature and precipitation, and degrades atmospheric

visibility through light extinction, and it affects climate indirectly by changing the

optical properties of clouds. Thus it creates uncertainty in the prediction of

climate change (Cruz and Pandis, 1997; Kaufman et al., 2002; Sun and Ariya,

2006). PM2.5 is measured in units of particle mass per unit volume (g/m3),

which is the standard unit of concentration used by the World Health

Organization (WHO) and the United States Environmental Protection Agency

(EPA).

1.1.2 PM2.5 in Hong Kong

In 2009, the Hong Kong Environmental Protection Department (EPD)

announced the Air Quality Objective (AQO) for PM2.5 according to Air Quality

Guidelines (AQG) announced by the World Health Organization (WHO) on 5th

October 2006 for global air quality applications. The recommended WHO Air

quality Standard (AQS) is 25 g/m3 for 24-hr PM2.5, and for annual PM2.5 it is 10

g/m3. The HK AQO was defined by EPD using PM2.5 data for 2008 at ambient air


3

quality monitoring stations Central, Tsuen Wan, Tung, Chung, Yuen Long, and

Tap Mun, and these are 35 g/m3 and 75 g/m3 for annual and 24-hr PM2.5

respectively (EPD, 2009). The HK 24–hr AQO is three times less stringent than

the WHO AQS. The EPD defined Air Pollution Index (API) ranging from low to

severe, is based on all particulates as well as other gases in Hong Kong (Fig. 1.3).

Figure 1.2: Common sources of PM2.5 [image curtsey of Fairbanks North Star

Borough Air Quality Division]

Fine Particulates in Hong Kong come from both local and regional sources

(Tsang et al., 2008; He et al., 2008). They are spatially uniform, and the PM2.5

mass in Hong Kong comprises approximately 30% sulfate, 40% organic carbon

and 10% elemental carbon (Kwok et al., 2010). Annual PM2.5 concentrations

measured at general as well as roadside air quality stations (Fig. 1.4) show a


4

decreasing trend in Hong Kong from 2005 to June, 2012. In the recent years 2009

to 2012, annual PM2.5 concentrations measured at general air quality stations

have been lower than the HK annual AQ objectives but PM2.5 measured at

roadside stations is higher than the objectives. The PM2.5 at general stations is 3

to 4 times, and at roadside stations 3 to 5 times higher than WHO annual AQ

standard.

Figure 1.3: The Hong Kong Air Pollution Index (API) based on all particulates as

well as other gases selected by the Environmental Protection Department (EPD)

Figure 1.4: Temporal trend of annual PM2.5 concentration in Hong Kong from

2005 to June, 2012 (Retrieved from: http://www.hongkongcan.org/doclib/2012%20Jan-

Jun%20Review%20-%20ENG_Final.pdf)


5

1.2 Measurement of PM2.5

Because fine particulate pollution has become a global issue, ground–

based air quality stations have been established in most large cities worldwide

for regular measurement of PM2.5 mass with high temporal sampling frequency

(Al-Saadi et al., 2005; Gomišcek et al., 2004). Studies have been undertaken using

ground–based meteorological variables such as air temperature (STEMP),

relative humidity (SRH), wind speed (SWS), wind direction (SWD) and mixing

height (MH) to understand the formation and dispersion of PM2.5 in the

atmosphere (Dawson et al., 2007; DeGaetano and Doherty, 2004; Elminir, 2005;

Gupta et al., 2006a; Hien et al., 2002; Tran and Mölders, 2011; Wang et al., 2009;

Wise and Comrie, 2005; Zhao et al., 2009). Resulting from these studies, higher

values of 24-hr average PM2.5 appear to be associated with low temperatures

(Dawson et al., 2007; Elminir, 2005; Tran and Mölders, 2011). The

concentrations of PM2.5 also appear to be influenced by mixing height,

precipitation, cloud cover and wind speed and direction (Dawson et al., 2007;

Elminir, 2005; Wise and Comrie, 2005), but the association between PM2.5 mass

concentrations and meteorological variables varies by region. Gupta et al.

(2006a) reported correlation coefficients of -0.27 to -0.06, and -0.85 to -0.60 for

PM2.5 with temperature and wind speed respectively, over Kolkata, India, and

Tran and Mölders, (2011) reported correlation coefficients of -0.58 and -0.33 for

PM2.5, also with temperature and wind speed, over Fairbanks, Alaska.

A full understanding of the impact of aerosol particles in climate and air

quality control strategies requires the retrieval of aerosol amounts and

characteristics over both time and space. However it is difficult to obtain spatial

information from point–based measurements of PM2.5 and meteorological

variables for regional mapping. Advancements in satellite remote sensing during

the last decade have the potential to overcome such limitations, as satellite

measurements of aerosols are spatially much denser and have been used to

estimate PM2.5 mass for areas with no ground-based measurements (Al-Saadi et

al., 2005; Engel-Cox et al., 2004; Gupta et al., 2006b). Spectral aerosol optical


6

depth (AOD) is the most accessible, thus most frequently used parameter(Clarke

et al., 2001, Holben et al., 2001) in statistical models predicting PM2.5 mass

concentrations, especially for areas where ground monitoring stations are not

available. AOD is the measure of extinction (scattering + absorption) of solar

radiance due to aerosol particles (Ångström, 1930; van de Hulst, 1948). Satellite

remote sensing with passive imaging radiometers can provide quantitative

measurements of AOD (Tang et al., 2005) from local to global scales (Kaufman et

al., 2002). Although the spatial variability of AOD can be examined much better

from satellite sensors than from ground stations, satellite retrieval is subject to

four major sources of uncertainty, namely: calibration of the AOD signal, cloud

detection, selection of an appropriate aerosol model and difficulty of

distinguishing between the reflectance from atmospheric aerosols and from the

ground surface (Li et al., 2009).

Over land, the main objective of satellite remote sensing is to retrieve

aerosol optical properties between the top of the atmosphere (TOA) and the

ground surface, as satellite received radiation corresponds to radiation reflected

from the surface affected by atmospheric and aerosol scattering along the path as

well as atmospheric path radiance. Since this path radiance from the atmosphere

corresponds to the optical penetration of light or AOD, the estimation of surface

reflectance (Li et al., 2009; Mishchenko et al., 1999) is an important factor in

developing an aerosol retrieval algorithm over land. Low surface reflectance

values allow good discrimination between the path radiance of aerosols and the

radiance of the land surface. However, high surface reflectance values make it

difficult to accomplish this discrimination as aerosol path radiance may then be

lower than the surface radiance. AOD products over land can be obtained by

sensors such as the Total Ozone Mapping Spectrometer (TOMS, Torres et al.,

2002), the Sea-viewing Wide Field-of-view Sensor (SEAWIFS, Sayer et al., 2012),

the Ozone Monitoring Instrument (OMI, Torres et al., 2007), the Polarization and

Directionality of the Earth’s Reflectances (POLDER, Herman et al., 1997), the

Along Track Scanning Radiometer (ATSR-2, North, 2002), the Geostationary


7

Operational Environmental Satellite (GOES, Prados et al., 2007), the Advanced

Very High Resolution Radiometer (AVHRR, Hauser et al., 2005; Riffler et al.,

2010), the MEdium Resolution Imaging Spectroradiometer (MERIS, Vidot et al.,

2008), and the Multiangle Imaging SpectroRadiometer (MISR, Kahn et al., 2005,

2010). Additionally AOD is available as standard products from the MODerate

Resolution Imaging Spectroradiometer (MODIS) using a deep blue algorithm

(Hsu et al., 2004; Hsu et al., 2006) and a dark target algorithm (Levy et al., 2007a,

2010; Remer et al., 2005, 2008).

Numerous studies have shown the potential of satellite-derived AOD to

give the spatial distribution of PM2.5 (Chu et al., 2003; Engel-Cox et al., 2004;

Gupta and Christopher, 2008, 2009; Gupta et al., 2006b, 2007; Hutchison, 2005;

Koelemeijer el al., 2006; Kumar et al., 2007; Liu et al., 2007; Nichol and Wong,

2009; Tian and Chen, 2010; Wang and Christopher, 2003; Wong et al., 2011).

However, the statistical relationship between AOD and PM2.5 appears to vary

with respect to land cover types, seasons, the AOD retrieval algorithm used, and

its spatial resolutions. For example, Gupta et al. (2006b) reported a correlation of

R = 0.60 in New York, 0.14 in Switzerland, 0.40 in Hong Kong, 0.41 in Delhi and

0.35 in Sydney. Kumar et al. (2007) improved the correlation between AOD and

PM2.5 from 0.67 to 0.87 by improving the spatial resolution from 10 km to 5 km.

The AOD–PM2.5 relationship can also be influenced by local meteorological

variables and these can be used as additional predictors (Gupta and Christopher,

2009; Tian and Chen, 2010). For example, Koelemeijer et al. (2006) and Tsai et

al. (2011) found that the relationship between AOD and PM2.5 was significantly

improved when AOD is divided by mixing layer height.

Monitoring and understanding the behavior of atmospheric aerosols at

local scales over complex urban terrain such as Hong Kong requires aerosol

retrieval algorithms that support high spatial resolution. Thus Li et al. (2005)

modified the DDV algorithm to retrieve AOD from MODIS at a 1 km spatial

resolution over Hong Kong although validation was limited to DDV dark targets.

Wong et al. (2011) retrieved AOD from MODIS at 500 m resolution over Hong


8

Kong using the Minimum Reflectance Technique (MRT) for surface reflectance

estimations. However the MRT computation is time consuming and suffers from

changing surface reflectance. Wang et al. (2012) retrieved AOD from MODIS at

500 m resolution over Mainland China and Hong Kong using Xue and Cracknell's

(1995) surface reflectance estimation method which is based on multiple view-

angle observations and a radiative transfer equation, but validation was limited

to only two months of data. All of these methods make use of a look-up table

(LUT), which includes parameters such as solar and view angles, AOD, single

scattering albedo (SSA), asymmetry factor and aerosol models. The MODIS AOD

retrieved from the LUT process for the current period is based on viewing

geometry and aerosol conditions from earlier observations so that an

independent determination is not made. Therefore, since effective retrieval of

PM2.5 depends on an accurate AOD retrieval, a new, high spatial resolution

satellite aerosol retrieval algorithm based on real viewing geometry and

encompassing a wide range of aerosol conditions and aerosol types ( = 0.80–

1.0) is required to retrieve AOD. Table 1.1 summarizes the previously reported

MODIS aerosol algorithms applied in Hong Kong, as well as new Simplified high

resolution MODIS Aerosol Retrieval Algorithm (SARA) developed in this study.

In Hong Kong, PM2.5 has been monitored using surface meteorological

variables (Shi et al., 2012) as well as MODIS AOD at both 10 km (Gupta et al.,

2006b) and 500 m (Nichol and Wong, 2009; Wong et al., 2011) resolutions

(Table 1.2). These studies are limited, as the influence of meteorological

variables on the relationship between AOD and PM2.5 was not investigated.

Therefore, a detailed study is still needed to investigate the influence of

meteorological variables on the relationship between high resolution AOD (500

m) and PM2.5 in Hong Kong.


9

Table 1.1: A summary of previously reported MODIS aerosol algorithms in Hong Kong, as well as new Simplified high resolution

MDOIS Aerosol Retrieval Algorithm (SARA) developed in this study

Research

Study

Sensor Resolution Look–Up–

Table (LUT)

Surface Reflectance Validation Instrument Time

Period

Li et al.

(2005)

MODIS 1km Yes Dark Dense Vegetation

(DDV) Algorithm

Microtops II Sun Photometer 3 months

Wong et al.

(2011)

MODIS 500 m Yes Minimum Reflectance

Technique (MRT)

AERONET Sun Photometer 1 year

Wang et al.

(2012)

MODIS 500 m Yes Xue and Cracknell's

method

AERONET Sun Photometer 2 months

This Study MODIS 500 m NO MOD09 AERONET, Sky–radiometer and

Microtops II Sun Photometers

3 Years


10

Table 1.2: Description of previously reported PM2.5 studies, as well as this new PM2.5 study in Hong Kong

Research

Study

MODIS

AOD

Air Quality

Stations

Meteorological

Data

Influence of Meteorology

on AOD–PM2.5 Relationship

Statistical

Model

Validation Time

Period

Gupta et al.

(2006)

MOD04

C004

5 No No Linear

Regression

No 1 year

Wong et al.

(2011)

Wong et al.

(2011)

5 No No Linear

Regression

No 1 year

Shi et al.

(2012)

Wong et al.

(2011)

1 Yes No Correlation No 2 year

This Study Bilal et al.

(2013)

5 Yes Yes Deming

Regression

Yes 3 Years


11

1.3 MOD04 C004/C005 aerosol retrieval algorithm

The MODIS operational level–2 aerosol product for land, MOD04 (Terra)

Collection 4 (C004), retrieves AOD from three channels (0.47, 0.66 and 2.12 m)

using a Dense Dark Vegetation (DDV) algorithm (Kaufman et al., 1997a, Remer et

al., 2005). It retrieves AOD over land with spatial resolution of 10 x10 km at

nadir and 20 x 40 km at the edge of the swath. The C004 algorithm makes

fundamental assumptions that (i) aerosols are transparent at 2.12 m, and (ii)

the surface reflectance in the visible channels 0.47 and 0.66 m have a constant

ratio (0.25 and 0.50, respectively) with surface reflectance in the shortwave

infrared channel (2.12 m). For aerosol retrieval, the 500 m pixels within a 20 x

20 (400 pixels) kernel are masked for cloud, snow/ice, and water using NDVI <

0.1. Dark target pixels are then selected for the retrieval using

, and since urban surfaces are not particularly bright in the MODIS SWIR

band, this selection includes both dark vegetated, as well as most urban surfaces.

From the pixels remaining after masking and dark target selection, the darkest

20% (at ) and brightest 50% of pixels are discarded to reduce cloud and

surface contamination. At this stage some urban surfaces may be discarded but

some pixels within the 20 x 20 kernel normally remain for computation of AOD

for that 10 km pixel. The mean measured surface reflectance for each channel is

calculated for the remaining dark target pixels and at least 12 pixels (out of 120

pixels remaining) should be present in the 20 x 20-pixel kernel (Remer et al.,

2005). The C004 AOD has been validated (Chu et al., 2002; Levy et al., 2005;

Remer et al., 2005) with global AErosol RObotic NETwork (AERONET) Sun

photometer data (Holben et al., 1998), and 68% of AOD retrievals globally are

found within an Expected Error (EE) envelope of ± (0.05 + 0.15AODAERONET)

(Remer et al., 2005, 2008; Levy et al., 2010).

Levy et al. (2007a, b) introduced a “second generation” dark target

algorithm, Collection 5 (C005), in the current MOD04 aerosol product. In C005,

the surface reflectance relationship between visible and SWIR channels is


12

presented as a function of vegetation index (NDVISWIR) and scattering angle (Levy

et al., 2007a). The MODIS C005 algorithm reports Quality Assurance (QA) flags

ranging from 3 (good quality) to 0 (poor quality) (Hubanks, 2012; Levy et al.,

2010). Those with QA flag = 3 provide the best match with AERONET AOD and

are strongly recommended for use in quantitative analysis. The C005 dark-target

algorithm has been studied globally and regionally (Hyer et al., 2011; Jethva et

al., 2007; Mi et al., 2007; Levy et al., 2010; Li et al., 2007; Papadimas et al., 2009),

and higher accuracy is reported than for the C004 algorithm. For C005 AOD with

QA flag = 3, 72% of retrievals fall within the Expected Error (EE) envelope of ±

(0.05 + 0.15AODAERONET) on a global scale (Levy et al., 2010) but this varies

regionally due to different surface types and aerosol loadings. The C005 retrieval

is best matched with AERONET AOD where m ~ 0.10–0.15 (moderately

dark) and NDVISWIR ~ 0.30–0.40 (moderately green) but it overestimates and

underestimates AOD (by 0.02 or more) over bright surfaces (m

approaching 0.25 and NDVISWIR < 0.2) and unusually dark surfaces (m < 0.05

or green at NDVISWIR > 0.6), respectively (Levy et al., 2010). The C005 retrieval

has a positive bias in AOD due to cloud fraction which varies from 26% of “Bad

(QA flag = 0)” retrievals to 10% of “Very Good (QA flag = 3)” retrievals (Hyer et

al., 2011). The resulting low (10 km) resolution of the MODIS AOD product is due

to kernel windows required for pixel selection, thus higher quality AOD retrieval

is possible but the spatial resolution is compromised.


13

1.4 Research objectives

The primary objective of this study is to:

Develop a statistical model for detailed monitoring of Fine

Particulate Matter (PM2.5) in Hong Kong and the Pearl River Delta

(PRD) region using high spatial resolution satellite AOD and

meteorological variables.

The secondary objectives of this study are to:

Develop a better and simpler AOD retrieval algorithm for MODIS

images at 500 m resolution for use over mixed surfaces,

Evaluate the MOD04 C005 standard AOD product at 10 km spatial

resolution over the same surfaces for comparison purposes.


14

1.5 Dissertation overview

Chapter 2 discusses the study area and data used including AERONET,

Sky–radiometer, Microtops II Sun photometer, MODIS products, HKO AWS and

the WRF–ARW model.

Chapter 3 describes the spatio–temporal variations and characteristics of

PM2.5 using four years of data from 2005 to 2008 at five air quality stations.

Additionally, the influence of surface–level meteorological variables (STEMP,

SRH, and SWS) is examined.

Chapter 4 provides the research methodology for development of (i)

Simplified Aerosol Retrieval Algorithm (SARA) at 500 m resolution, and (ii) PM2.5

model using SARA–retrieved AOD at 500 m resolution and meteorological

variables obtained from both AWS and the WRF model at 500 m resolution.

Chapter 5 presents the results including validation and sensitivity

analyses of SARA–retrieved AOD at 500 m, evaluation of MOD04 C005 AOD

product at 10 km, estimation of PM2.5 using SARA (500 m) and MOD04 AOD (10

km), and prediction and validation of SARA–retrieved PM2.5 at 500 m.

Chapter 6 discusses the summary and conclusions of this study.

Study Area and Data Used Chapter 2

15

Chapter 2

Study Area and Data Used

2.1 Study Area

The Pearl River Delta (PRD) is one of the world’s fastest developing

regions, located in southern China and covering a land area of 42794 km2 with

population over 40 million. The PRD is facing serious air pollution problems due

to increase in anthropogenic activities (Cao et al., 2003, 2004; Ansmann et al.,

2005; Hagler et al., 2006) including manufacturing, power plants and shipping

(Streets et al., 2006). The PRD covers the major urban areas of Guangdong

Province, as well as Special Administrative Regions of Macau and Hong Kong.

Hong Kong (Fig. 2.1) is situated on complex and hilly terrain on the coast of

southeast China with an area of 1104 km2 and highest elevation of 957 m above

sea level. It has the densest urban areas in the world with population density of

averaging 6540 people km-2 (Census and Statistics Department of Hong Kong,

2010). It has a humid subtropical climate with mean annual rainfall from 1400

mm to 3000 mm. It has been experiencing visibility and air quality problems as

have many other Asian cities (Chan & Yao, 2008) due to the presence of ambient

particulate matter (PM). The four year (2005–2008) annual mean PM2.5 mass

concentration is 40.35 g/m3 which is higher than Hong Kong’s annual AQO (35

g/m3) and 4 times higher than the WHO annual AQS (10 g/m3).


16

Figure 2.1: Study area and the locations of Ground–based Sun Photometers

(circles), air quality stations (triangles) and meteorological stations (round

square) in the complex terrain of Hong Kong

Table 2.1: Latitudes/Longitudes of ground stations

Station Latitude Longitude Station Latitude Longitude

Sun photometers

PolyU 22.303 114.179 CityU 22.340 114.170

Hok Tsui 22.209 114.258 HKIA 22.317 113.917

Air Quality and Automatic Weather Stations

Central 22.282 114.158 Tsuen Wan 22.372 114.115

Tung Chung 22.289 113.944 Yuen Long 22.447 114.039

Tap Mun 22.471 114.361 HKO 22.302 114.174

Tai Mo Shan 22.411 114.124 Sha Lo Wan 22.293 113.904

Wetland Park 22.467 114.109


17

2.2 Data Used

2.2.1 Ground–based PM2.5 Concentration Data

The Hong Kong Environmental Protection Department (EPD) operates an

air quality monitoring network including three roadside and eleven general

stations for measuring hourly gaseous pollutants as well as surface–level

particulate matter including PM2.5 (g/m3). The Tapered–Element Oscillating

Microbalance (TEOM) instrument measures PM2.5 mass under relative humidity

(RH) conditions between 40–50% with an accuracy of ± 1.5 g/m3 for hourly

averages. The hourly surface–level PM2.5 mass concentrations four year (March

2005 to February 2009) were obtained at five urban/suburban and rural air

quality stations (Fig. 2.1, Table 2.1) including Central (city area, commercial area

and urban populated area), Tsuen Wan (city area, commercial area, urban

populated and residential area), Tung Chung (new town, suburban and

residential area), Yuen Long (urban and residential area), and Tap Mun (remote

rural area).

2.2.2 Ground–based AOD Measurements

To monitor and understand the air quality situation, two AERONET

(urban and rural) stations as well as a Sky–radiometer and Microtops II Sun

Photometer have been deployed in Hong Kong. The Urban AERONET station

(Hong_Kong_PolyU) is at the centre of the urban area at the Hong Kong

Polytechnic University (PolyU) and has operated since 2005. The Rural

AERONET station (Hong_Kong_Hok_Tsui) was installed in a remote rural area

near to the coast and operated from 2007 to 2010. AERONET (Holben et al.,

1998, 2001) is a worldwide network of approximately more than 700 well

calibrated Sun photometers. It provides cloud free AOD observations (Smirnov et

al., 2000) in seven channels (0.340–1.020 m) every 15 minutes, with an


18

uncertainty of ~ 0.01–0.02 (Holben et al., 2001), which is three to five times

more accurate than satellite observations (Remer et al., 2009).

The Microtops II Sun photometer (Morys et al., 2001) is a portable

manually operated instrument which was deployed at the Hong Kong

International Airport (HKIA) thought to be the most polluted area of Hong Kong.

The Microtops II Sun photometer measures AOD by direct solar irradiance

measurements in five wavebands (0.380 – 1.020 m) with an uncertainty of ~

0.015–0.02 (Knobelspiesse et al., 2004).

The portable ground-based Sky-radiometer is a scanning spectral

radiometer (Nakajima et al., 1996) deployed at the City University of Hong Kong

(CityU) in a suburban area. It measures AOD in five channels (0.400–1.020 m)

measuring direct and diffuse solar irradiance with an uncertainty of 0.01 – 0.025

(Campanelli et al., 2004). The Sky-radiometer AOD has differences of 0.01 (Che et

al., 2008) to 0.03 (Liu et al., 2011) compared with AERONET AOD.

In this study, the PolyU (urban) AERONET AOD at 0.55 m was used in

the development of the SARA, and for validating the retrievals the Hok Tsui

(rural) AERONET, CityU (suburban) Sky-radiometer, and HKIA (rural) Microtops

II Sun photometers were used. The AOD at 0.55 m for these instruments was

interpolated from other wavelengths using the Ångström exponent (Ångström,

1964). The locations of the AERONET, Sky-radiometer and Microtops II Sun

photometers station in Hong Kong are shown in Fig. 2.1 and their characteristics

are provided in Table 2.1 and 2.2.


19

Table 2.2: Characteristics of ground–based Sun Photometers

Instrument AERONET Sky-radiometer Microtops II

Wavelength (m)

0.340 --- 0.340

0.380 --- ---

0.440 0.440 ---

0.500 0.500 0.500

0.675 0.675 0.675

0.870 0.870 0.870

0.940 --- ---

--- 1.020 1.020

Uncertainty 0.01 – 0.02 0.01 – 0.025 0.015 – 0.02

Field–of–view (degree) 1.2 1.0 2.5

2.2.3 MODIS Satellite Data

MODIS sensors aboard the National Aeronautics and Space

Administration (NASA)’s Earth Observing System (EOS) satellite platforms, Terra

and Aqua, have operated since December 1999 and May 2002 respectively. Terra

and Aqua are polar orbiting satellites at an altitude of approximately 700 km

with equatorial crossing (southward) at around 10:30 a.m. local solar time (LST),

and southward 01:30 p.m. LST, respectively. MODIS has 36 wavelength channels

ranging from 0.41 to 14 m at moderate spatial resolutions (250 m, 500 m and 1

km) and good temporal resolution (1 to 2 days) with a swath width of 2330 km.

Remote sensing of aerosols in Hong Kong is limited by cloud cover, but

cloud–free skies are most common in November, December and January. The

MODIS (Terra) data products (MOD02HKM calibrated radiance, MOD03

geolocation data, MOD09 surface reflectance and MOD04 C005 operational

aerosol product) for Hong Kong were obtained from the GSFC (Goddard Space


20

Flight Center) Level–1 and Atmosphere Archive and Distribution System

(LAADS) (http://ladsweb.nascom.nasa.gov) for the years 2007 to 2009 (Table

2.3).

Table 2.3: A Summary of MODIS Sensor Products

Product Parameter Resolution Purpose

Original Used

MOD02HKM Radiance 500 m 500 m Input

MOD03

Height

1 km

500 m

Input Zenith Angles

Azimuth Angles

MOD09GA Surface Reflectance 500 m 500 m Input

MOD04 C005 AOD 10 km 10 km Comparison

2.2.4 Ground–based Meteorological Measurements

The ground–based hourly meteorological variables including surface

temperature (STEMP), surface relative humidity (SRH), and surface wind speed

(SWS) were collected from the Hong Kong Observatory’s five Automatic Weather

Stations (AWS) nearest to the PM2.5 urban/suburban and rural ground stations,

namely the Hong Kong Observatory (HKO), Tai Mo Shan (TMS), Sha Lo Wan

(SLW), Wetland Park (WLP) and Tap Mun (TM). These AWS are located at

distances of 2.8 km, 5.2 km 4.1 km, 3.3 km and 0.0 km from Central, Tsuen Wan,

Tung Chung, and Yuen Long air quality stations, respectively (Fig. 2.1, Table 2.1).


21

2.2.5 WRF Meteorological Data

The Weather Research and Forecasting (WRF) model (Skamarock et al.,

2005) is a non–hydrostatic, medium–scale numerical prediction model

developed with joint effort between the National Centre for Atmospheric

Research (NCAR), the National Centers for Environmental Prediction and the

Forecast Systems Laboratory of the National Oceanic and Atmospheric

Administration (NCEP, FSL/NOAA), and the larger scientific community. It has

two dynamic cores, namely the Advanced Research WRF (ARW) and the Non-

hydrostatic Mesoscale Model (NMM). In this study version 3.3.1 of WRF–ARW

(Figure 2.2) is used. The WRF–ARW meteorological model is suitable for

research and operational weather prediction applications with scales ranging

from a few hundred meters to thousands of kilometers. It has several options for

nesting, diffusion, lateral boundary conditions, as well as physical

parameterizations including planetary boundary layer, land surface,

microphysics, cumulus convection, and shortwave and longwave radiation. The

previously published research on the WRF model can be found at http://wrf-

model.org/wrfadmin/publications.php.

Figure 2.2: WRF–ARW Modeling System Flowchart


22

The WRF–ARW was simulated for four different domains at spatial

resolutions ranging from 500 m to 13.5 km (d01 = 13.5 km, d02 = 4.5 km, d03 =

1.5 km and d04 = 500 m) with different physics options (Table 2.4). In this study,

only domain–4 (d04) is used to obtain the meteorological variables at 500 m

resolution for the Hong Kong region. The hourly WRF meteorological variables

including 2m temperature (WTEMP), 2m relative humidity (WRH), 2m specific

humidity (WSH), 10m wind speed (WWS), surface pressure (WPSFC), planetary

boundary layer height (WPBLH, Hu et al., 2013; Peña et al., 2013), shortwave

surface radiation (WSW), and longwave surface radiation (WLW) were obtained

for the years 2007 to 2009. As ground observations of the PBLH of the PBLH

were not available for validation of WRF PBLH, the potential uncertainty of the

WRF PBLH estimation was not discussed in this study, and WRF PBLH was

considered accurate over the region. The WRH and WSH are not direct output of

WRF simulation, therefore, the WRH was calculated using WTEMP, WPSFC and

water vapor mixing ratio (Q2), and WSH was calculated using the ratio of water

vapor to unit mass of air (dry air plus water vapor) i.e. Q2 /1+ Q2.

Table 2.4: Physics Schemes for WRF–ARW model Simulations

Physics Option Name

mp_physics 2 Lin et al. scheme

ra_lw_physcis 1 RRTM scheme

ra_sw_physcis 2 Goddard shortwave scheme

sf_sfclay_physics 1 Monin–Obukhov scheme

sf_surface_physics 2 Noah land–surface model

bl_pbl_physics 1 YSU scheme

cu_physics 3 Grell–Devenyi ensemble scheme

Characteristics of PM2.5 and influences of meteorological variables Chapter 3

23

Chapter 3

Characteristics of Particulate Matter (PM2.5) and

Influence of Meteorological Variables

3.1 Spatio–temporal variations in PM2.5

The four year (2005–2008) daily means of PM2.5 concentrations at all air

quality stations were lower than HK's 24-hr AQO of 75 g/m3 in all seasons

except for a very few values in winter at Tung Chung and Yuen Long. These few

daily exceedances (Table 3.1) suggest very good air quality in Hong Kong.

However, the 24–hr air quality conditions at all five stations were almost two to

three times worse than the WHO 24-hr AQS in spring, autumn and winter.

The daily mean PM2.5 concentrations at the five air quality stations for

March 2005 to November 2008 were analyzed for diurnal to seasonal variations

to depict the detailed spatio–temporal variability of air quality over the complex

and rugged terrain of Hong Kong (Figures 3.1 to 3.4). The 4–year (2005-2008)

mean PM2.5 concentration indicated that the winter season has consistently

worse air quality conditions than in other seasons (Table 3.2), and, the 4–year

mean concentrations in summer at all air quality stations were lower than WHO

24–hr AQS which indicated good air quality, as was also reported by other

studies in Hong Kong (Shi et al., 2012; Louie et al., 2005; Lee et al., 2006). In

general, air quality stations in rural areas usually give lower readings than in

urban/suburban areas and have different patterns of temporal variations in

PM2.5 concentration, such as observed in urban and rural areas of Beijing (Zhao

et al., 2009). But in this study PM2.5 concentration at the remote rural air quality

stations (Tap Mun) was not significantly lower than at urban/sub-urban air


24

quality stations. The similar pattern of PM2.5 variations observed suggests the

dominance of regional sources (Fig. 3.1).

Table 3.1: Total number of exceedances of 4–year (2005-2008) daily mean PM2.5

mass concentrations

Site

Time

HK/WHO AQS

Number of Exceedances Above

HK/WHO AQS

Spring Summer Autumn Winter

Central

24-hr

75gm-3/25gm-3

0/92 0/77 2/90 2/92

Tung Chung 0/72 0/12 1/85 6/92

Tap Mun 0/74 0/13 0/85 1/91

Tsuen Wan 01/83 0/26 2/ 87 2/92

Yuen Long 0/79 0/21 2/88 5/92

At all air quality stations a decreasing trend in PM2.5 concentrations was

observed from December to June, while an increasing trend was observed from

June to December (Fig. 3.2). Monthly mean PM2.5 concentrations showed a

unimodal distribution during the study years 2005 and 2008, as was also

observed by Gupta et al. (2006b) for the year 2002 in Hong Kong. The highest

PM2.5 concentrations were observed at Central air quality station fron March to

September, which may be due to extremely dense concentrations of vehicles in

Central. The two air quality stations Tung Chung and Yuen Long show higher

levels of concentrations from October to January, probably due to contribution of

regional pollutants since they are close to the Chinese Mainland.


25

In addition to the daily and mean monthly temporal trends, weekdays and

weekends (Fig. 3.3) as well as hourly concentrations (Fig. 3.4) were examined.

Very similar PM2.5 concentrations were observed during weekdays and

weekends, and it is notable that concentrations at Tap Mun rural station were

from 3 to 9 g/m3 lower but had a similar trend between weekdays and

weekends to other stations which must be attributed to regional emission

sources, as Tap Mun is a remote island in the north–east of Hong Kong without

any local emissions. The urban/suburban stations were more polluted in the

afternoon (or evening) than morning, whereas the maximum hourly means were

observed in the morning at Tap Mun.


26

Table 3.2: Descriptive statistics of 4–year (2005–2008) daily means of PM2.5

Air Quality Stations

N

Season

PM2.5 Concentrations (g/m3)

Min Max Mean (SD)

Central

368 Spring 26.71 68.31 45.81 ± 9.34

368 Summer 20.29 53.99 31.49 ± 6.71

364 Autumn 23.14 78.38 51.15 ± 11.04

268 Winter 34.36 84.19 52.80 ± 9.75

Tung Chung

362 Spring 12.68 70.04 34.41 ± 11.68

364 Summer 7.73 41.36 17.53 ± 7.53

330 Autumn 12.49 82.56 47.37 ± 13.49

271 Winter 34.36 89.28 55.87 ± 12.58

Tap Mun

365 Spring 12.71 58.31 34.69 ± 10.32

368 Summer 6.92 50.38 17.71 ± 8.23

361 Autumn 17.85 72.63 43.20 ± 11.41

271 Winter 24.25 81.06 48.63 ± 10.86

Tsuen Wan

338 Spring 20.86 84.54 38.75 ± 10.55

368 Summer 13.58 47.43 23.58 ± 6.52

363 Autumn 79.96 76.00 47.61 ± 12.38

271 Winter 30.73 79.99 50.64 ± 10.16

Yuen Long

368 Spring 17.43 72.19 38.34 ± 10.91

367 Summer 12.28 61.62 22.42 ± 9.63

364 Autumn 18.17 79.96 51.15 ± 12.38

271 Winter 31.26 86.11 55.59 ± 11.49


27

Figure 3.1: Four year (2005–2008) daily means of PM2.5 concentrations at five air quality stations (Central, Tsuen Wan, Tung

Chung, Yuen Long and Tap Mun)


28

Figure 3.2: Four year (2005–2008) monthly means of PM2.5 at five air quality stations (Central, Tsuen Wan, Tung Chung, Yuen

Long and Tap Mun)


29

Figure 3.3: Four year (2005–2008) weekdays and weekends means of PM2.5 at five air quality stations (Central, Tsuen Wan,

Tung Chung, Yuen Long and Tap Mun)


30

Figure 3.4: Four year (2005–2008) hourly means of PM2.5 at five air quality stations (Central, Tsuen Wan, Tung Chung, Yuen

Long and Tap Mun)


31

3.2 Influence of Seasonal Meteorological Factors

For a better understanding of PM2.5 variability, the influence of

meteorological variables such as STEMP, SRH, and SWS were analyzed at Central,

Tsuen Wan, Tung Chung, Yuen Long, and Tap Mun air quality stations for

February 2007–February 2009 (represented as 2007–2008 in the following

text). The meteorological variables were collected from automatic weather

stations (AWS) at the Hong Kong Observatory (HKO) (Fig. 2.1). Relative humidity

observations from AWS were not available at the remote rural Tap Mun station

during the study period.

The daily mean STEMP, SRH, and SWS (Fig.3.5) were found to be

negatively correlated with daily mean PM2.5 concentrations for all air quality

stations when the four seasons were combined (Table 3.3). Significant positive

correlations were observed for individual seasons particularly during the

summer, autumn and winter seasons for some variables individually e.g. STEMP

(Table 3.3). Correlation coefficients (R) of 0.67 at Yuen Long, 0.64 at Tung Chung,

0.53 at Tap Mun, 0.52 at Tsuen Wan, and 0.29 at Central stations using multiple

linear regression (MLR) were found between PM2.5 concentrations and the

meteorological variables together when the four seasons were combined. The

correlation coefficients significantly increased during the spring and summer

and the meteorological variables together can explain 80% of the variability in

daily mean PM2.5 concentration at Yuen Long (Eq. 3.1), 77 % at Tung Chung (Eq.

3.2), 71% at Tsuen Wan (Eq. 3.3), 69% at Central (Eq. 3.4), , and 67 % at Tap

Mun (Eq. 3.5), and an overall variability of 70% at all five stations (Eq. 3.6).

However, in autumn and winter, the meteorological variables together can

explain only 48% of the variability in PM2.5 concentration at Yuen Long, 37% at

Tung Chung, 37% at Central, 32% at Tsuen Wan, and 8% at Tap Mun air quality

stations. The MLR equations between PM2.5 and meteorological are not discussed

here due to their weak correlation in autumn and winter than in spring and

summer. The PM2.5 concentrations correlated poorly with the meteorological

variables during the winter season in Hong Kong. This is different for the studies


32

reported for Kolkata, India (Gupta et al., 2006a) and for Hanoi, Vietnam (Hien et

al., 2002) during the winter season. This might be due to topographical effect,

atmospheric stability/instability, distribution and major sources of PM2.5

particles. The analyses suggested that STEMP, SRH, and SWS, together, and

STEPM and SRH individually, are useful indicators of PM2.5 concentrations at all

five air quality stations in Hong Kong for spring and summer but may be

unsuitable for autumn and winter.

[Station: Yuen Long (urban), R = 0.80, Study period = 2007–2008 (spring and

summer)]

[Station: Tung Chung (suburban), R = 0.77, Study period = 2007–2008 (spring

and summer)]

[Station: Tsuen Wan (urban), R = 0.71, Study period = 2007–2008 (spring and

summer)]

[Station: Central (urban), R = 0.69, Study period = 2007–2008 (spring and

summer)]


33

[Station: Tap Mun (rural), R = 0.67, Study period = 2007–2008 (spring and

summer)]

[Station: all (five), R = 0.70, Study period = 2007–2008 (spring and summer)]

A previous study of the years 1983–1992 (Cheng and Lam, 1998) shows

that higher values of air pollution in Hong Kong were associated with surface

winds between SSW and ENE which was attributed to power stations, industry

and motor vehicles. In the current study, similar results were observed for

hourly PM2.5 concentrations were plotted as a function of the corresponding

hourly SWD (Fig. 3.6). Fig. 3.6 indicates that the air quality of Hong Kong was

mainly influenced by air masses arriving from PRD region but less affected by air

masses arriving from ocean sides.


34

Table 3.3: Correlation between daily mean PM2.5 and meteorological variables

for years 2007 and 2008

Air Quality Station Period Meteorological Variables

TEMP RH WS

Central

Spring -0.26 -0.35 -0.03

Summer 0.25 -0.22 -0.01

Autumn 0.10 -0.26 -0.40

Winter 0.33 -0.30 -0.16

2007–2008 -0.35 -0.45 -0.06

Tsuen Wan

Spring -0.20 -0.61 -0.32

Summer 0.36 -0.23 -0.12

Autumn -0.05 -0.13 -0.18

Winter 0.17 -0.33 -0.37

2007–2008 -0.42 -0.47 -0.13

Tung Chung

Spring -0.42 -0.33 -0.34

Summer 0.32 -0.27 -0.14

Autumn -0.12 -0.18 -0.19

Winter 0.24 -0.41 -0.29

2007–2008 -0.50 -0.49 -0.30

Yuen Long

Spring -0.36 -0.40 -0.33

Summer 0.31 -0.29 -0.36

Autumn -0.06 -0.29 -0.15

Winter 0.28 -0.35 -0.26

2007–2008 -0.43 -0.56 -0.17

Tap Mun

Spring -0.38 –– 0.00

Summer 0.25 –– -0.06

Autumn -0.30 –– 0.26

Winter 0.30 –– -0.08

2007–2008 -0.53 –– 0.33


35


36


37

Figure 3.5: Temporal variations in daily mean PM2.5 concentrations and meteorological variables (TEMP, RH, and WS) for years

2007–2008 at (a) Central, (b) Tsuen Wan, (c) Tung Chung, (d) Yuen Long, and (e) Tap Mun air quality stations. The Y-axis on left

represents daily mean PM2.5 concentrations (g/m3, dark grey bars), and four Y-axes on right represent TEMP (oC, black), RH (%,

violet), and WS (m/s, green).


38

Figure 3.6: Hourly PM2.5 concentrations from all air quality stations including (a) Central, (b) Tap Mun, (c) Tsuen Wan, (d) Yuen

Long and (e) Tung Chung as a function of corresponding hourly surface wind direction in Hong Kong (2007–2008)

Methodology Chapter 4

39

Chapter 4

Methodology

4.1 Development of a Simplified high resolution MODIS

Aerosol Retrieval Algorithm (SARA)

Because of the uncertainties over mixed surfaces, low resolution and

computational complexities of existing satellite AOD retrieval algorithm, a

Simplified Aerosol Retrieval Algorithm (Fig. 4.1a) was developed that does not

use the common technique of constructing a LUT from the Radiative Transfer

Model (RTM). Instead, RTM equations were applied directly to the MODIS data

products (MOD02, MOD03, and MOD09) to retrieve AOD at high spatial

resolution (500 m). The SARA AOD retrievals have three assumptions; (i) the

surface is Lambertian, (ii) single scattering approximation, and (iii) the single

scattering albedo and asymmetric factor do not vary spatially over the region on

day of retrieval. The SARA requires MOD02HKM data to calculate the TOA

reflectance, MOD03 data to obtain the solar and view angles, MOD09GA to

retrieve the daily surface reflectance and AERONET data to obtain the single

scattering albedo and asymmetric factor for the day of the retrieval. The Hong

Kong Polytechnic University AERONET station is used for this purpose.

The MOD02HKM swath data represent the satellite received TOA radiance

in the visible to mid–infrared wavelengths of the solar spectrum from

0.41 m to 2.16 m (Salomonson et al., 2006). The SARA is based on the satellite

received spectral reflectance ( ), which is a function of measured spectral

radiance ( ), solar zenith angle, earth–sun distance in astronomical unit

and mean solar exoatmospheric radiation (Eq. 4.1):


40

Figure 4.1a: Methodology of the Simplified high resolution Aerosol Retrieval

Algorithm (SARA)


41

where = satellite received TOA spectral reflectance, = satellite

received TOA spectral radiance, = mean solar exoatmospheric radiation

as a function of MODIS band number (Tasumi et al., 2008), = the earth–sun

distance in astronomical unit (Duffie and Beckman, 1991), = cosine of solar

zenith angle.

The satellite received TOA spectral reflectance is defined as a function of

atmospheric path reflectance (scattering of solar radiation within the

atmosphere), and surface function (reflection of the solar radiation from the

surface that is directly transmitted to the TOA). The TOA spectral

reflectance, , a function of solar and view zenith and azimuth angles, can

be estimated using Eq. 4.2 (Tanré et al., 1988; Vermote et al., 1997):

where = solar zenith angle, = view zenith angle, = relative azimuth

angle, = aerosol reflectance resulting from multiple scattering in the

absence of molecules, = Rayleigh reflectance resulting from multiple

scattering in the absence of aerosols, = transmission of the atmosphere on

sun–surface path, = transmission of the atmosphere on the surface–sensor

path, = surface reflectance, = atmospheric backscattering ratios to

account for the multiple reflection between the surface and atmosphere.

and are defined by Eqs. 4.3 and 4.4:


42

The SARA relies on the aerosol reflectance, , which can be

calculated by subtracting the Rayleigh path reflectance, , and surface

function from the satellite measured top of atmosphere reflectance,

(Eq. 4.5):

The correction of the satellite data for Rayleigh scattering (Antoine and

Morel., 1998) depends on determination of the Rayleigh phase function (Lado-

Bordowsky and Naour, 1997; Liang, 2005) and Rayleigh optical depth ( , Eq.

4.6, Liang, 2004):

where = ambient pressure with respect to elevation (mbar), =

1013.25 mbar (pressure at sea–level), z = ground elevation (height) above sea

level in kilometer (km), = wavelength in micrometers.

Surface reflectance is possibly the most important factor in

the retrieval of aerosol optical depth from satellite remote sensing

measurements (Mishchenko et al., 1999; Li et al., 2009). Surface reflectance for


43

SARA is obtained from the MODIS level–2G (MOD09GA) daily surface reflectance

product at 500 m spatial resolution (Vermote and Kotchenova, 2008). This

product employs atmospheric correction (conversion of TOA into surface

reflectance) to estimate the surface reflectance in MODIS bands 1 to 7 (0.470

m–2.13 m). The algorithm used by the MODIS atmospheric products (MOD04

– aerosol optical depth, MOD05 – water vapor, MOD07 – ozone and MOD35 –

cloud mask) and ancillary data (Digital Elevation Model (DEM) and Atmospheric

Pressure) account for aerosol and gaseous scattering and absorption, adjacency

effects of land cover, Bidirectional Reflectance Distribution Function (BRDF)

effect, and contamination by thin cirrus. The accuracy of the surface reflectance

product (MOD09) depends on the accuracy of these input parameters and is

given as 93% and 90% when surface reflectances are 0.3 and 0.1, respectively

(Vermote and Kotchenova, 2008). Our SARA algorithm uses the green

wavelength (0.55m, band 4) of MOD09GA directly.

The Eddington method (Eq. 4.7) can be used as a good approximation of

the atmospheric backscattering ratio ( , Tanré et al., 1979) for the correction of

the surface function.

The remotely sensed aerosol reflectance, , was retrieved for

the 0.55 wavelength by correction for Rayleigh scattering and

the surface function. In the single scattering approximation, satellite measured

aerosol reflectance is proportional to the aerosol optical depth , single

scattering albedo and aerosol scattering phase function (Eq. 4.8, Gordon

and Wang, 1994; Kaufman et al., 1997b; King et al., 1999; Kokhanovsky and de

Leeuw, 2009).


44

Therefore, AOD, , can be retrieved by re–arranging the Eq. 4.8 as Eq. 4.9:

The aerosol scattering phase function represents the angular

distribution of light scattered by particles and can be determined using the

single–term Henyey–Greenstein method (Eq. 4.10, Rahman et al., 1993):

where Θ is the scattering phase angle (Levy et al., 2007b). The asymmetry

parameter indicates the relative dominance of

forward/backward scattering and it remains constant for the most of the aerosol

models (Tanré et al., 1979).

Substituting Eq. 4.5 into Eq. 4.9 yields Eq. 4.11 where the three unknowns

( , , and ) are shown explicitly. The PolyU (urban) AEORNET AOD within

30 minutes of the MODIS local overpass time was used as a on the right

hand side of Eq. 4.11. The values of , and for the day for which the SARA AOD

is being retrieved are determined from a match between SARA AOD ( on the


45

left hand side of Eq. 4.11) as a function of , and and the PolyU (urban)

AERONET AOD (right hand side). This is accomplished empirically by varying

the values of , and until the match is obtained.

For validation, AOD measurements from three ground-based Sun

photometers were used. Due to limited coincident Sun photometer and MODIS

data due to cloud cover and in order to increase the number of statistical

samples and also to consider the spatial variability imposed by atmospheric

motion, the Sun photometer AOD was averaged within 30 minutes (AERONET),

and 60 minutes (Microtops II sun photometers and Sky-radiometer) of the

MODIS overpass time over Hong Kong, and the 500 m retrieved (SARA) AOD was

extracted from a 3x3 spatial subset region (average of 9 pixels) centered on the

ground-based stations (Ichoku et al., 2002a). A total of 42, 32 and 20 data points

of SARA retrieved AOD for 2007 to 2009 were matched with Hok Tsui AERONET,

Sky-radiometer and Microtops II Sun photometers, respectively, due to cloud

cover and unavailability of ground measurements. The MOD04 C005 AOD from

the parameter ‘Optical_Depth_Land_and_Ocean (AOD at 0.55 m for both ocean

(best) and land (corrected) with best quality data, QA flag = 3)’ was extracted for

a 3x3 spatial subset region (average of 9 pixels). For C005 AOD, only 16, 42, and

8 values were found to be matched for comparison against AERONET (Fig. 5.1b),

Sky-radiometer (Fig. 5.3c), and Microtops II (Fig. 5.3d) Sun photometers,

respectively, as most of the time C005 algorithm unable to retrieve AOD over

mixed and complex surfaces of Hong Kong. Finally, the SARA AOD was compared


46

with C005 AOD over nine locations of Hong Kong within different land cover

types (Table 4.2).

4.1.1 Statistical analysis

Statistical indicators were used to evaluate comparisons of the satellite

retrieved AOD with ground-based Sun photometers; namely correlation

coefficient (R), root mean square error (RMSE), mean absolute error (MAE),

expected error (EE) of AOD, and fraction of EE (FOE).

4.1.1.1 Correlation coefficient (R)

The correlation coefficient (R) is a good indicator of agreement between

Sun photometer observed AOD and satellite retrieved AOD with higher values

indicating better agreement. We obtained R from Deming Regression (DR,

Deming, 1943) as opposed to Linear Regression (LR) (LR, Westgard and Hunt,

1973) because DR estimates an unbiased slope by assuming the Gaussian

distribution of errors in both x and y data points (which is typical of our data)

(Deming, 1943; Linnet, 1993, 1998; Cornbleet and Gochman 1979). On the other

hand LR estimates a biased slope by assuming random measurement errors in

the dependent variable (y) and an error free independent variable (x) (Cornbleet

and Gochman 1979; Linnet, 1993; Stöckl et al., 1998), but is inappropriate for use

when significant errors are expected in both variables.


47

4.1.1.2 Root mean square error (RMSE)

The root mean square error (RMSE) is used to measure the differences

between satellite retrieved AOD and Sun photometer measured AOD and it is

sensitive to both systematic and random errors. The RMSE can be defined as

follows:

where is the satellite retrieved AOD and is

the Sun photometer measured AOD.

4.1.1.3 Mean absolute error (MAE)

The mean absolute error (MAE), the most natural measure of mean error

magnitude (Willmott and Matsuura, 2005), is calculated as:

4.1.1.4 Expected error (EE)

Expected error (EE) is used here for the confidence envelopes of MODIS

aerosol retrieval algorithm over land to evaluate the quality of SARA and C005

AOD and is defined as follows (Remer et al., 2008; Levy et al., 2010):


48

Good matches (quality) of satellite–retrieved (SARA and C005) AOD are

reported when the satellite–retrieved AOD falls within the following envelope

(Levy et al., 2010):

where is the absolute value of expected error.

4.1.1.5 Fraction of expected error (FOE)

The fraction of expected error (FOE) is the ratio of satellite Sun

photometer to the absolute value of EE

and can be computed as (Mi et al., 2007):

where < 1, indicates a good match (Levy et al., 2010). Values of FOE

<0 and FOE > 0 represent underestimation and overestimation of the satellite

retrievals, respectively.


49

4.2 Development of a SARA PM2.5 model

The methodology for development of a PM2.5 model for prediction of PM2.5

using SARA–retrieved AOD at 500 m resolution, meteorological variables at 500

m resolution obtained from WRF model, and ground–based variables from AWS

of HKO is presented (Fig. 4.1b) in the following steps.

In Step 1, the relationship between AOD and PM2.5 was explored with

respect to time of day. Previous studies suggest that the correlation between

AOD and PM2.5 increases using hourly averaged PM2.5 concentrations as opposed

to daily means (Gupta and Christopher, 2008; Schaap et al., 2009). Satellite AOD

observations are available around 10:30 a.m. local time, while PM2.5

measurements are obtained throughout the day. In order to understand the

sensitivity of the AOD–PM2.5 relationship with respect to time, the PM2.5 data

were averaged for four time windows 00–24 hr, 08–12 hr, 09–12 hr, and 10–12

hr surrounding the time at which MODIS/Terra passes over Hong Kong. In order

to increase the number of statistical samples and also to account for the spatial

variability imposed by atmospheric motion, the SARA–retrieved AOD at 500 m

resolution (Bilal et al., 2013) was extracted from a 3x3 spatial subset region

(average of 9 pixels) centered on the air quality station. The AOD–PM2.5

relationship according to time of day was examined for individual stations for

the whole of the years 2007–2008 (Table 5.3).

In Step 2, the relationship between PM2.5 and two aerosol products: the

SARA retrieved AOD at 500 m and C005 AOD at 10 km (MOD04) was

investigated to identify the best satellite aerosol product for reliable prediction

of PM2.5 over urban and rural areas of Hong Kong for autumn and winter seasons

(Figs. 5.8 and 5.9 ). Autumn and winter were used because higher correlation

between AOD and PM2.5 was observed in autumn and winter than in spring and

summer (Table 5.3). The relationship of PM2.5 with both AOD products (10 km

and 500 m) for the years 2007 and 2008 was derived using Deming Regression


50

(DR) (Deming, 1943). SARA and MOD04 were extracted from a 3x3 spatial subset

region (average of 9 pixels) centered on the air quality stations.

In Step 3, multiple linear regression (Eqs. 4.17 and 5.1) and DR (Eq. 5.2)

models were developed for prediction of PM2.5 using the best AOD product from

Step 2. Eight meteorological variables including temperature (WTEMP), relative

humidity (WRH), specific humidity (WSH), wind speed (WWS), surface pressure

(WPSFC), planetary boundary layer height (WPBLH), shortwave surface

radiation (WSW), and longwave surface radiation (WLW) were obtained from

WRF model at 500 m resolution, and four ground–based meteorological

variables including temperature (STEMP), relative humidity (SRH), and wind

speed (SWS) were collected from HKO (Fig. 5.5). The remote rural AWS station

(Tap Mun) was not considered in the methodology due to unavailability of SRH

data for the study period. Variables from the WRF model were extracted from a

3x3 spatial subset region (average of 9 pixels) centered on the air quality

stations.

where PM2.5 is the 4–hr mean PM2.5 concentration, the dependent

variable on the left–hand side. The independent variables on the right–hand side

include four ground–based (STEMP, SRH,and SWS) and eight WRF model

(WPBLH, WPSFC, WTEMP, WRH, WSH, WWS, WLW, and WSW) meteorological

variables. The parameter with different subscripts denote the respective

regression coefficients.


51

A correlation coefficient of PM2.5 with SARA AOD (500 m) and

meteorological variables is only acceptable if the P-value is less than 0.05 (Table

5.4 and 5.5). The bin method was used for refinement of SARA–retrieved PM2.5 at

500 m resolution using specific ranges of meteorological variables for autumn

and winter of 2007 and 2008. A total of 90 bins (listed in table 5.6) were

developed using four ground–based and eight WRF meteorological variables.

Available observations of SARA and PM2.5 in the datasets corresponding to each

bin were used to establish a relationship between SARA AOD and PM2.5 for

specific meteorological conditions. Bins for each meteorological variable were

selected when a significant correlation coefficient between SARA AOD and PM2.5,

sufficient number of SARA AOD and PM2.5 observations, and suitable numbers of

measurements of PM2.5 concentrations above HK’s proposed 24–hr AQO were

available in the dataset associated with a particular bin. PM2.5 models (Eqs. 5.3

to 5.15) were developed using DR for bins of each meteorological variable, for

prediction of PM2.5.

In Step 4, PM2.5 models from Step 3 were used to predict PM2.5 at 500 m

spatial resolution at four urban/suburban air quality stations for autumn and

winter for a additional year, 2009 (Fig. 5.12). The SARA–retrieved PM2.5

concentrations at 500 m resolution were validated using the ground–based

observed PM2.5 at the stations using the 2009 data. The best PM2.5 model was

selected based on the statistical parameters slope, root mean square error

(RMSE), mean absolute error (MAE) (Figs. 5.12 and 5.13), mean and standard

deviation (StDev) (Figs. 5.7 and 5.9). Additionally the best PM2.5 model, along

with previously developed PM2.5 models over Hong Kong by Wong et al. (2011)

were evaluated at the Tap Mun air quality station in Hong Kong for autumn and

winter of the years 2007 to 2009 (Fig. 5.14, Table 5.7). Tap Mun in Hong Kong

was selected because it was not considered in the development of the PM2.5

model.


52

Figure 4.1b: Development of SARA PM2.5 model for prediction of PM2.5

Results Chapter 5

53

Chapter 5

Results

5.1 Validation of SARA AOD with AERONET, Sky-radiometer

and Microtops II Sun photometer AOD

Fig. 5.1 shows the scatter plots of SARA 500 m AOD and MOD04 C005

10km AOD with Hok Tsui AERONET AOD used for validation. In Fig. 5.1 the

dashed (green), dotted (orange), and solid (red) lines are the 1:1 line, EE

( ) envelope line, and regression line, respectively.

The SARA 500 m AOD obtained a high correlation coefficient (R = 0.963) and low

values of RMSE (0.044) and MAE (0.037). Fig. 5.1a reveals a close

correspondence between the SARA AOD and AERONET AOD, and the majority of

the observations lie close to the 1:1 line. All the SARA retrieved AOD

observations fall within the confidence envelope

which indicates a good quality of the

retrieved AOD. The standard deviations are 0.17 and 0.16 for the SARA AOD and

AERONET AOD observations, respectively, and the mean difference (SARA –

AERONET) is zero. A good agreement was also observed between MOD04 C005

AOD and Hok Tsui AERONET (R = 0.904, RMSE = 0.082, and MAE = 0.070) AOD

(Fig. 5.1b). However C005 AOD has approximately two times larger RMSE and

MAE than the SARA retrieved AOD. Fig. 5.1b for C005 AOD shows a wide scatter

of points due to underestimation and overestimation. Few observations lie close

to the 1:1 line, and approximately 19% of observations fall outside the

confidence envelope. The standard deviations are 0.19 and 0.19 for the C005

AOD and AERONET AOD observations respectively, and the mean difference

Results Chapter 5

54

(C005 – AERONET) is -0.01 which indicates an overall underestimation of the

C005 AOD.

Figure 5.1: Validation of (a) SARA AOD, and (b) MOD04 C005 AOD against Hok

Tsui (rural) AERONET AOD from 2007 to 2009. The lines are as follow: EE

envelope = dotted orange, 1:1 line = dashed green and regression line = solid red.

Approximately 81% of the fraction of expected error (FOE) values from

C005 AOD and 100% values from SARA AOD fall within the specified range (-1 to

1), indicating better matches (quality) of SARA AOD with AERONET than for

C005 (Fig. 5.2). This implies that SARA AOD data is of good quality and SARA

works well according to the assumptions of the surface, and the single scattering

approximation.

Results Chapter 5

55

Figure 5.2: Fraction of expected error (FOE) in AOD retrievals of SARA and

MOD04 C005 algorithms

Figs. 5.3a and 5.3c show that the SARA and C005 AOD retrievals have

similar correlation coefficients (R = 0.97 and 0.96, respectively) with the Sky-

radiometer, having low RMSE and MAE. Approximately 80% and 67%

respectively of the FOE values from SARA and C005 fall within the EE envelope

for CityU Sky-radiometer respectively (Fig. 5.4). Slopes lower than unity indicate

systematic underestimation by both SARA and C005 AOD retrievals. The mean

differences of SARA AOD and C005 AOD from the Sky-radiometer AOD are -0.07

and -0.09 respectively, and both AOD retrievals have error for high values of

AOD.

Fig. 5.3b shows a higher correlation of the SARA AOD with the Microtops II

AOD than for the C005 AOD observations (Fig. 5.3d). The mean differences

between the SARA AOD and the C005 AOD respectively, and Microtops II AOD

measurements are -0.07 and -0.19. The majority of the SARA retrieved AOD

observations lie close to the 1:1 line, whereas the C005 AOD observations are

very far from the line. Good agreement between satellite– retrieved AOD and the

Sun photometer AOD depends not only on a high correlation coefficient but also

Results Chapter 5

56

on the percentage of data falling within the EE envelope. Thus approximately

78% and 12% the FOE values from SARA AOD and C005 AOD respectively fall

within the EE envelope for HKIA (Fig. 5.4). Fig. 5.4 (and 5.2) shows data gaps in

SARA AOD due to missing PolyU (urban) AERONET AOD measurements. The

SARA AOD has errors for moderate (0.3 – 0.5) AOD observations, whereas C005

AOD has large errors for the whole range (low to high) AOD observations. The

MOD04 C005 AOD retrieval underestimates AOD with R = 0.79 over the Hong

Kong International Airport (HKIA) which is a highly polluted area and a bright

surface (m approaching 0.25).

Figure 5.3: Validation of SARA AOD and MOD04 C005 AOD against (a & c) CityU

Sky-radiometer, and (b & d) HKIA Microtops II AOD measurements. The lines are

as follow: EE envelope = dotted orange, 1:1 line = dashed green and regression

line = solid red.

Results Chapter 5

57

Figure 5.4: Fraction of Expected Error (FOE) in AOD retrievals of SARA and

MOD04 C005 against CityU Sky-radiometer and HKIA Microtops AOD

5.2 Comparison between SARA and MOD04 C005 AOD

The SARA AOD data were also compared directly with C005 AOD at nine

locations representing different cover types in Hong Kong for both high and low

aerosol loading conditions. These locations are the Hong Kong Polytechnic

University (PolyU, urban), Central (CT, urban), Tsuen Wan (TW, urban), Tung

Chung (TC, urban), Yuen Long (YL, suburban), The City University of Hong Kong

(CityU, suburban), Hok Tsui (HT, rural), Hong Kong International Airport (HKIA,

rural), and Tap Mun (TM, rural). High correlations were achieved, with good data

matches between SARA AOD and C005 AOD at all the locations except HKIA,

where the correlation is high but with poor matches (only 50% data within EE =

) because, as previously noted, the C005 retrievals

underestimated AOD over HKIA (Table 5.1). The average correlation coefficient,

RMSE, and MAE are 0.917, 0.087, and 0.072, respectively, between the SARA

retrieved AOD and the C005 AOD observations over Hong Kong. These

comparisons imply that SARA has as good ability to retrieve accurate AOD over

different land cover types during both high and low aerosol loading conditions.

Results Chapter 5

58

Table 5.1: Comparison between SARA AOD and C005 AOD from 2007 to 2009

Location N R Fitted Line

(SARA =)

RMSE MAE Data within EE (%)

PolyU 68 0.901 1.13C005 + 0.04 0.104 0.083 72

CT 62 0.914 1.06C005 + 0.03 0.085 0.068 73

TW 62 0.926 0.98C005 + 0.03 0.067 0.054 85

TC 48 0.925 1.06C005 + 0.03 0.086 0.066 77

YL 62 0.889 1.05C005 – 0.01 0.085 0.063 81

CityU 20 0.941 0.94C005 + 0.01 0.051 0.046 95

HT 10 0.907 1.04C005 – 0.01 0.065 0.064 70

HKIA 06 0.929 1.32C005 + 0.07 0.159 0.137 50

TM 62 0.920 1.00C005 - 0.001 0.085 0.063 86

Results Chapter 5

59

5.3 Sensitivity analysis for SARA methodology

To confirm the robustness of the SARA methodology, a sensitivity analysis

of the SARA AOD was undertaken by increasing / decreasing by 5% the value of

SSA ( ), and by 2% the values of asymmetry parameter ( )

and the surface reflectance ( ). The descriptive statistics used in

statistical analysis are mean, standard error of mean (SE mean), standard

deviation (StDev), minimum, median and maximum of the Hok Tsui AERONET,

CityU Sky-radiometer, and HKIA Microtops II Sun photometers and the SARA

AOD observations (Table 5.2). Table 5.2 shows difference up to 7% between

SARA AOD (real) and ground–based Sun photometers. The ,

and are the new retrieved SARA AOD by increasing /

decreasing the SSA, and values, respectively. The statistics show that the

mean SARA retrieved AOD differences range from -2% to 2%, -2% to 4%, and 5%

to 8% by increasing / decreasing the SSA (5%), (2%) and (2%) values

respectively. These results confirm the robustness of the SARA methodology

over mixed surfaces of Hong Kong as the maximum error (8%) in the new SARA

AOD is almost the same as the error (7%) of the actual SARA AOD.

Table 5.2a: Descriptive statistics for sensitivity of the SARA–retrieved AOD (500

m) for , and at Hok Tsui AERONET AOD

Variable Mean SE Mean StDev Minimum Median Maximum

AERONET 0.383 0.023 0.155 0.170 0.389 0.806

SARA AOD 0.383 0.025 0.165 0.148 0.369 0.841

0.365 0.023 0.157 0.141 0.352 0.801

0.403 0.026 0.174 0.156 0.389 0.886

0.405 0.029 0.195 0.134 0.373 1.128

0.367 0.023 0.157 0.134 0.350 0.685

0.434 0.024 0.164 0.198 0.418 0.876

0.434 0.024 0.164 0.198 0.418 0.876

Results Chapter 5

60

Table 5.2b: Descriptive statistics for sensitivity of the SARA–retrieved AOD (500

m) for , and at CityU Sky-radiometer AOD


Sky-radiometer 0.372 0.032 0.183 0.119 0.325 0.790

SARA AOD 0.297 0.029 0.163 0.090 0.250 0.680

0.283 0.0274 0.155 0.086 0.238 0.648

0.313 0.030 0.172 0.095 0.263 0.716

0.327 0.034 0.191 0.100 0.265 0.730

0.290 0.028 0.161 0.100 0.245 0.680

0.375 0.029 0.165 0.160 0.325 0.740

0.375 0.029 0.165 0.160 0.325 0.740

Table 5.2c: Descriptive statistics for sensitivity of the SARA–retrieved AOD (500

m) for , and at HKIA Microtops II AOD


Microtops II 0.501 0.050 0.223 0.239 0.436 0.968

SARA AOD 0.433 0.057 0.253 0.160 0.335 1.030

0.412 0.054 0.241 0.152 0.319 0.981

0.455 0.060 0.266 0.168 0.353 1.084

0.470 0.069 0.360 0.150 0.350 1.150

0.422 0.054 0.241 0.150 0.350 1.060

0.509 0.056 0.251 0.260 0.415 1.150

0.509 0.056 0.251 0.260 0.415 1.150

Results Chapter 5

61

5.4 Spatio–temporal pattern of AOD

Fig. 5.5a shows the spatial distribution of SARA AOD over Hong Kong and

Pearl River Delta (PRD) region on 30th January 2007. It reveals a high aerosol

loading event (black circle) over PRD region due to emissions from local power

plants and industries and provides detailed information on pollution sources. A

higher value of SARA AOD (~ 0.80) was observed over PRD than over Hong Kong

(AOD ~ 0.65). However, the C005 AOD retrieval was unable to depict the high

pollution event and shows missing pixels over the affected area (Fig. 5.5b),

where it is suspected that high aerosol reflectance resulted in many 500 m pixels

within the 10 km kernel failing the selection criteria in the MODIS visible channel

(0.66m). The SARA AOD is thus judged to be more suitable than C005 AOD to

represent the spatial pattern of aerosols over the complex and hilly terrain of

Hong Kong as well as the industrialized area of PRD. The SARA AOD image

overlaid with the road network of Kowloon and Hong Kong Island (Fig. 5.5c)

shows the ability of the SARA AOD to identify local emission sources especially

over the dense and congested districts of Kowloon Bay (KB), Hung Hom (HH),

Sham Shui Po (SSP), and North Hong Kong Island.

Fig. 5.6 shows a sharp boundary in AOD values between urban and hilly

regions of Hong Kong, indicating greater variations between urban and hilly

areas than between urban and rural areas during both high and low aerosol

loading conditions. Steep mountain slopes create a blocking effect, and this

combined with a boundary layer height often lower than the mountains, traps

pollutants in the lowlands. Therefore AOD has a strong relationship with

elevation and high values of AOD are usually observed at lower elevations (<100

m) which correspond to urban, suburban and rural regions in Hong Kong. The

northwest border of Hong Kong adjacent to the industrial city of Shenzhen on the

Chinese Mainland (arrowed in Fig. 5.6) is a mainly rural, lowland area, and

shows high AOD values on both dates.

Results Chapter 5

62

Figure 5.5: Spatial pattern of SARA (a), and C005 (b) AOD for a high aerosol

loading event (30th January 2007) over the Pearl River Delta (PRD) region and

Hong Kong. Also shown, in panel (c), is the SARA AOD under-laid with road data

over Kowloon and Hong Kong Island.

Figure 5.6: Variations in AOD between hilly and urban regions of Hong Kong

on (a) low (10th November 2008), and (b) high (26th January 2007) aerosol

loading conditions.

Results Chapter 5

63

In spite of large spatial differences, remarkably similar temporal trends of

AOD were observed across urban (PolyU), rural (Tap Mun) and hilly (Tai Mo

Shan) regions of Hong Kong (Fig. 5.7). This suggests that much of the air

pollution in Hong Kong is due to regional factors because if pollution was mainly

local (such as urban vehicle emissions) then urban/rural temporal trends would

be out of phase, unlike the close temporal correspondence as shown in Fig. 5.7.

The figure also indicates that the C005 retrievals overestimate AOD over hilly

area (elevation ~ 957 m) due to their coarse resolution as 10 km pixels cover

both high and low elevations in the same pixel, retrieving the same AOD values

over both. However, the SARA retrievals, because of finer spatial resolution, are

better able to represent AOD values over each distinct region.

Figure 5.7: Temporal trends of (a) MOD04 C005 AOD at 10 km resolution, and

(b) SARA AOD at 500 m resolution from 2007 to 2009

Results Chapter 5

64

5.5 Exploring the AOD–PM2.5 relationship with time of day

The statistical correlation between SARA–retrieved AOD (500 m) and

mean PM2.5 for four time windows (00–24 hr, 08–12 hr, 09–12 hr, and 10–12 hr)

was higher for the 09–12 hr (4–hr) mean PM2.5 than for other time windows at

individual air quality stations for the years 2007 and 2008 (Table 5.3). It was

also observed that the AOD–PM2.5 relationship is sensitive to season, with

significantly higher correlations in autumn and winter than in spring and

summer. Therefore, only the 09–12 hr (4–hr) mean PM2.5 data of autumn and

winter were used in further analysis.

Table 5.3: AOD–PM2.5 relationship with four time windows for 2007 and 2008

Station / Description

Central City area, commercial area, urban populated

AOD–PM2.5 relationship

Years 2007 and 2008 Autumn & Winter

Average Time (Hour) N R Average Time (Hour) N R

00–24 73 0.78 00–24 64 0.83

08–12 73 0.78 08–12 64 0.84

09–12 73 0.78 09–12 64 0.84

10–12 70 0.75 10–12 62 0.81

Tsuen Wan City area, commercial area, urban populated, residential




00–24 69 0.77 00–24 60 0.81

08–12 69 0.77 08–12 60 0.81

09–12 67 0.79 09–12 60 0.82

Results Chapter 5

65

10–12 67 0.80 10–12 59 0.83

Tung Chung New town, residential




00–24 68 0.68 00–24 60 0.73

08–12 68 0.72 08–12 60 0.79

09–12 68 0.73 09–12 60 0.79

10–12 68 0.71 10–12 59 0.78

Yuen Long Urban, residential




00–24 72 0.65 00–24 63 0.72

08–12 72 0.64 08–12 63 0.73

09–12 75 0.67 09–12 63 0.75

10–12 72 0.67 10–12 63 0.75

Tap Mun Remote rural area




00–24 73 0.75 00–24 64 0.81

08–12 73 0.73 08–12 64 0.77

09–12 73 0.73 09–12 64 0.77

10–12 72 0.73 10–12 63 0.76

Results Chapter 5

66

5.6 Relationship between PM2.5 and satellite aerosol products

The number of MOD04 C005 AOD observations varies from one air

quality station to another station in Hong Kong. Only 44 observations of MOD04

were available during autumn and winter of the years 2007 and 2008 due to

cloud cover, and also the MOD04 C005 algorithm was unable to retrieve AOD

during high pollution episodes as well as due to the effect of bright aerosols on

the visible band thresholding, resulting in missing data (Bilal et al., 2013). Both

sets of AOD (500 m and 10 km) observations were plotted against PM2.5

measurements at all five air quality stations (Fig. 5.8) and stronger relationships

were found between PM2.5 and SARA (R ~ 0.78) than between PM2.5 and MOD04

(R ~ 0.64). The substantially higher correlations for SARA than for MOD04 (Fig.

5.9), suggest that the SARA AOD is more able to monitor PM2.5 concentrations

over the mixed surfaces of Hong Kong. Due to the better relationship between

SARA and PM2.5, only SARA was used in the subsequent analyses. The correlation

between PM2.5 and SARA increased from 0.78 to 0.82 when the range 14–136

g/m3 of 4–hr averaged PM2.5 concentrations at five air quality stations during

autumn and winter of the years 2007 and 2008 was divided into 22 bins of 5

g/m3 intervals. The correlation between PM2.5 and SARA also increased from

0.78 to 0.92 using the monthly means of 4–hr averaged PM2.5 concentrations at

five air quality stations during autumn and winter of the years 2007 and 2008

(Fig. 5.10).

Results Chapter 5

67

Results Chapter 5

68

Figure 5.8: Relationship of 4–hr (09–12hr) averaged PM2.5 concentrations

(g/m3) with SARA (500 m) and MOD04 AOD (10 km) at (a) Central, (b) Tsuen

Wan, (c) Tung Chung, (d) Yuen Long, and (e) Tap Mun air quality stations in

Hong Kong for autumn and winter of 2007 and 2008

Results Chapter 5

69

Figure 5.9: Relationship of 4–hr (09–12hr) averaged PM2.5 concentrations

(g/m3) at all five air quality stations in Hong Kong in autumn and winter (2007–

2008) with (a) SARA AOD (500 m), and (b) MOD04 AOD (10 km)

Figure 5.10: Relationship of SARA AOD at all five air quality stations in Hong

Kong for autumn and winter of 2007 and 2008 with (a) bins (interval of 5

g/m3) of 4–hr (09–12 hr) averaged PM2.5, and (b) monthly means of 4–hr (09-

12 hr) averaged PM2.5

Results Chapter 5

70

5.7 Development of PM2.5 model using SARA AOD (500 m),

HKO and WRF (500 m) data

Table 5.4 shows the results of multiple linear regression (MLR) model

(Eq. 4.17) developed to predict PM2.5 using SARA AOD (500 m), four ground–

based meteorological variables from Hong Kong Observatory (HKO), and eight

variables from WRF model at 500 m resolution (Fig. 5.11). The results show that,

while SARA AOD is strongly (and positively) correlated with PM2.5, only WTEMP

and WSH among the 14 meteorological variables are correlated with P–value less

than 0.05 and those are weak negative correlations. To test whether the addition

of WTEMP and WSH could improve the SARA retrieval of PM2.5, another MLR

model (Eq. 5.1) was developed but the contribution of both variables were found

to be insignificant (Table 5.5) as the correlation coefficient only increased from

0.82 to 0.84. Therefore, it is concluded that meteorological variables from

ground stations as well as from the WRF model cannot assist the prediction of

PM2.5 in autumn and winter seasons in Hong Kong. Similar findings were also

presented in Chapter 3, when PM2.5 was compared directly with ground–based

meteorological variables.

[R = 0.84, N = 232, Study period = years 2007–2008 (autumn and winter)]

A linear relationship between SARA AOD and PM2.5 (Eq. 5.2) was also developed

using Deming Regression (DR):

[R = 0.82, N = 232, Study period = years 2007–2008 (autumn and winter)]

Results Chapter 5

71

Table 5.4: Regression coefficients for the prediction of PM2.5 concentrations

using SARA AOD, HKO and WRF meteorological variables at four air quality

stations for autumn and winter of 2007 and 2008

Variables Correlation (R) Estimate ( Standard Error P–value

SARA–AOD 0.820 93.21 4.972 0.000

STEMP -0.089 0.38 0.453 0.401

SRH -0.029 -0.12 0.092 0.200

SWS -0.257 -0.58 0.571 0.310

WPBLH 0.054 -0.01 0.006 0.109

WPSFC 0.075 0.06 0.238 0.806

WTEMP -0.154 2.34 0.831 0.005

WRH -0.131 0.44 0.225 0.052

WSH -0.206 -4.84 1.75 0.006

WWS 0.062 0.40 0.479 0.399

WLW -0.108 0.02 0.080 0.778

WSW 0.030 -0.00 0.013 0.995

Table 5.5: Regression coefficients for the prediction of PM2.5 concentrations

using SARA AOD, WTEMP and WSH at four air quality stations for autumn and

winter of 2007 and 2008

Variables Correlation (R) Estimate ( Standard Error P–value

SARA–AOD 0.820 96.2 4.305 0.000

WTEMP -0.154 1.410 0.390 0.000

WSH -0.206 -2.114 0.459 0.000

Results Chapter 5

Chapter 5

variables Chapter 5

72

Figure 5.11: Descriptive Statistics of surface–level PM2.5, SARA AOD, and meteorological variables from both WRF–ARW and

HKO at four air quality stations for autumn and winter of 2007 and 2008

Results Chapter 5

Chapter 5

variables Chapter 5

73

5.8 Refinement of SARA–retrieved PM2.5 at 500 m resolution

using bins of meteorological variables

A previously developed PM2.5 model over Hong Kong (Wong et al., 2011)

was found to be unable to predict the PM2.5 for subsequent year datasets over the

region (discussed in section 5.10). Therefore, the objective of bin method is to

develop a new PM2.5 model based on SARA–retrieved AOD at 500 m resolution

that can be applied for future years as well as in other regions of the world with

similar air quality conditions. Table 5.6 shows that higher correlation between

AOD and PM2.5 can be achieved for specific meteorological conditions based on

bin method. Twelve bins–a total of 90 (listed in table 5.64) were selected to

develop PM2.5 models using Deming Regression (DR) for prediction of PM2.5.

These PM2.5 models (Eq. 5.3 to 5.15) were developed for specific meteorological

conditions but can be applied for both specific as well as all available

meteorological conditions to predict PM2.5 mass concentrations at 500 m spatial

resolution. PM2.5 models developed using the MLR model, increased the AOD–

PM2.5 correlation by only 2% to 3%. Therefore, PM2.5 models based on the MLR

model are not discussed further.

Results Chapter 5

Chapter 5

variables Chapter 5

74

Table 5.6: AOD–PM2.5 correlation based on bins of meteorological variables from

both HKO and WRF model at five air quality stations in Hong Kong for autumn

and winter of 2007 and 2008

N Bins AOD-PM2.5 Correlation N PM2.5 > HK 24–hr AQO

STEMP (C)

1 4.0–10.9 0.784 15 0

2 11.0–14.9 0.830 39 6

3 15.0–17.9 0.786 56 6

4 18.0–20.9 0.841 64 12

5 23.0–30.9 0.902 34 04

WTEMP (C) at 500 m

6 11.0–14.9: 0.854 26 3

7 15.0–17.9: 0.823 44 6

8 18.0–19.9 0.876 46 10

9 20.0–22.9 0.716 63 3

10 23.0–30.0 0.907 37 5

SRH ()

11 13–39 0.736 43 2

12 30–39 0.704 30 2

13 40–59 0.818 99 16

14 46–69 0.836 139 21

15 47–79 0.843 153 22

16 55–79 0.855 106 13

17 65–79 0.885 47 5

18 70–90 0.852 21 1

WRH () at 500 m

19 20–39: 0.648 37 1

20 40–49: 0.854 44 8

21 50–59: 0.831 67 14

22 55–60: 0.860 40 9

23 55–65: 0.849 71 10

24 66–75: 0.778 32 1

25 71–80: 0.057 21 0

SWS (m/s)

26 0.38–1.99 0.772 56 12

27 1.50–3.00 0.825 96 14

Results Chapter 5

Chapter 5

variables Chapter 5

75

28 2.00–2.99 0.843 62 7

29 2.53–3.40 0.869 51 8

30 3.00–3.99 0.823 45 5

31 4.00–4.99 0.836 26 3

32 5.00–5.99 0.573 16 0

33 6.00–7.9 0.810 19 0

34 8.0–13.9 0.968 8 1

WWS (m/s) at 500 m

35 1.38–3.90 0.805 42 5

36 3.00–4.00 0.843 41 6

37 4.00–4.99 0.851 54 3

38 5.00–5.99 0.796 51 7

39 6.00–6.99 0.842 30 6

40 7.00–13.13 0.839 33 4

WPSFC (Pa) at 500 m

41 996- 1010 0.860 49 5

42 1011–1015 0.819 79 12

43 1016–1120: 0.800 72 8

44 1120–1124: 0.850 42 6

45 1120–1123 0.861 37 6

WSH at 500 m

46 2.30–4.00 0.782 26 1

47 4.00–5.50 0.715 28 2

48 5.00–6.00: 0.875 29 5

49 6.00–7.00: 0.817 35 10

50 7.00–7.99 0.885 29 5

51 8.00–9.00 0.829 20 2

52 9.00–11.00 0.822 40 2

53 11.00–13.00 0.899 30 3

54 12.00–1600 0.921 21 3

WLW (Wm-2) at 500 m

55 250–300 0.813 76 7

56 276–300 0.844 55 6

57 300–320 0.876 50 12

58 321–340 0.608 52 1

59 341–415 0.876 54 8

60 341–400 0.880 48 8

Results Chapter 5

Chapter 5

variables Chapter 5

76

61 350–400 0.917 34 6

WSW (Wm-2) at 500 m

62 340–580 0.875 20 3

63 581–595 0.860 32 7

64 596–610 0.872 33 6

65 601–620 0.356 40 1

66 631–655 0.913 32 5

67 656–699 0.764 34 1

68 700–806 0.855 32 5

WPBLH (m) at 500 m

69 48–199 0.806 49 4

70 200–344 0.829 38 3

71 264–390 0.853 48 7

72 300–390 0.884 35 6

73 300–449 0.875 62 11

74 400–496 0.851 46 10

75 450–546 0.813 51 8

76 500–599 0.751 47 3

77 550–699 0.786 35 4

78 607–799 0.811 31 4

79 700–835 0.533 13 0

SWD (degree)

80 271–90 0.828 148 17

81 91–270 0.822 79 8

82 01–180 0.824 206 22

83 181–360 0.822 26 5

84 0–90 0.824 139 14

85 91–180 0.834 62 06

86 315–45 0.858 39 10

87 46–135 0.789 153 10

88 136–225 0.827 26 04

89 226–314 0.864 09 02

90 226–45 0.850 48 12

Results Chapter 5

Chapter 5

variables Chapter 5

77

[SWD = 226–45, R = 0.85, N = 48, Study period = autumn and winter (2007-

2008)]

[STEMP = 18.0–20.9 C, R = 0.84, N = 64, Study period = autumn and winter

(2007-2008)]

[WTEMP = 18.0–19.9 C, R = 0.88, N = 46, Study period = autumn and winter

(2007-2008)]

[SRH = 47–79, R = 0.84, N = 153, Study period = autumn and winter (2007-

2008)]

[WRH = 40–49%, R = 0.85, N = 39, Study period = autumn and winter (2007-

2008)]

Results Chapter 5

Chapter 5

variables Chapter 5

78

[SWS = 2.53–3.40 m/s, R = 0.87, N = 51, Study period = autumn and winter

(2007-2008)]

[WWS = 3.0–4.0 m/s, R = 0.84, N = 41, Study period = autumn and winter (2007-

2008)]

[WPSFC = 996–110 Pa, R = 0.86, N = 49, Study period = autumn and winter

(2007-2008)]

[WSH = 7.0–7.9%, R = 0.89, N = 29, Study period = autumn and winter (2007-

2008)]

[WLW = 300–320 Wm-2, R = 0.88, N = 50, Study period = autumn and winter

(2007-2008)]

Results Chapter 5

Chapter 5

variables Chapter 5

79

[WSW = 631–655 Wm-2, R = 0.91, N = 32, Study period = autumn and winter

(2007-2008)]

[WPBLH = 300–449 m, R = 0.88, N = 62, Study period = autumn and winter

(2007-2008)]

The AOD–PM2.5 correlation increased from 0.88 to 0.90 when available

AOD and PM2.5 observations in the datasets of Eq. 5.14 was sorted for SRH = 47–

79% (Eq. 5.15). The correlation also increased from 0.88 to 0.99 when AOD and

PM2.5 observations in the datasets of Eq. 5.14 was sorted for SRH = 47–79% and

SWS = 2.53–3.40 m/s (only 10 such measurements were available in the dataset,

therefore Eq. is not given).

[WPBLH = 300 – 449m and SRH = 47 – 79%, R = 0.90, N = 42, Study period =

autumn and winter (2007-2008)].

Results Chapter 5

Chapter 5

variables Chapter 5

80

5.9 Validation of SARA–Retrieved PM2.5 at urban/suburban air

quality stations using data of 2009

PM2.5 models (Eqs. 5.2 to 5.15) are developed at urban/suburban air

quality stations in Hong Kong for specific meteorological conditions using

datasets of 2007 and 2008 (autumn and winter) except Eq. 5.2 which is

developed for all available meteorological conditions. These models were used to

predict PM2.5 at 500 m resolution associated with all available meteorological

conditions at urban/suburban stations using SARA–retrieved AOD at 500 m

resolution for autumn and winter of 2009. Validation results (Figs. 5.12 and

5.13) show a good correlation (R) between predicted and observed PM2.5 mass

concentrations with small to large overestimation by the PM2.5 models. Eq. 5.10

(Fig. 5.12e (left), referred to as SARA PM2.5 model in the discussion below) was

selected as the best of 14 models for prediction of PM2.5 due to having accurate

slope (1.08), low (RMSE ~ 10.38 g/m3 and MAE ~ 9.15 g/m3) errors and

comparable descriptive statistics of image predicted of PM2.5 at ground stations

(Fig. 5.13). The results also suggest that the bin method can improve the

predicting power of satellite–retrieved AOD to monitor PM2.5 mass

concentrations.

Results Chapter 5

Chapter 5

variables Chapter 5

81

Results Chapter 5

Chapter 5

variables Chapter 5

82

Results Chapter 5

Chapter 5

variables Chapter 5

83

Figure 5.12: Validation of SARA–retrieved PM2.5 (g/m3) at 500 m based on (a)

Eqs. 5.2 (left) and 5.3 (right), (b) Eqs. 5.4 (left) and 5.5 (right), (c) Eqs. 5.6 (left)

and 5.7 (right), (d) Eqs. 5.8 (left) and 5.9 (right), (e) Eq. 5.10 (left) and 5.11

(right), (f) Eqs. 5.12 (left) and 5.13 (right) and (g) Eqs. 5.14 (left) and 5.15 (right)

using ground–based observed PM2.5 at four urban/suburban air quality stations

in Hong Kong for autumn and winter of 2009

Results Chapter 5

Chapter 5

variables Chapter 5

84

Figure 5.13: Descriptive Statistics of surface–level observed PM2.5 (g/m3) and SARA–retrieved PM2.5 (g/m3) using 14 PM2.5

models (Eq. 5.2 to 5.15) at four air quality stations for autumn and winter of 2009

Results Chapter 5

Chapter 5

variables Chapter 5

85

5.10 Comparison of SARA with previously developed PM2.5

model for Hong Kong

PM2.5 concentrations were predicted at Tap Mun rural area in Hong Kong

for autumn and winter of the years 2007 to 2009 using the SARA PM2.5 model (Y

= 110.5 X + 12.56, see section 5.8 and 5.9), and as a comparison for a previously

developed PM2.5 model (Y = 63.66 X + 26.56) for Hong Kong by Wong et al.

(2011). Data from the Tap Mun air quality stations were not considered in the

original development of the PM2.5 models due to unavailability of corresponding

SRH data Tap Mun. Air quality at Tap Mun has similar temporal trends as

urban/suburban air quality stations in Hong Kong. Comparisons were conducted

using ground–based observed PM2.5 and results (Fig. 5.14) show that the SARA

predicted PM2.5 concentrations (R = 0.80, slope = 1.05 and N = 90, average

underestimation = 1.78 g/m3) have a better agreement with ground–based

observed PM2.5 concentrations, with slope closer to 1:1 line than by Wong et al.

(R = 0.80, slope = 0.60 and N = 90, average underestimation = 2.53 g/m3).

Descriptive statistics (Table 5.7) show that the SARA predicted PM2.5 is

comparable to the observed PM2.5 at Tap Mun (Hong Kong), whereas Wong et al's

PM2.5 model overestimates the minimum concentrations due to a large intercept

and underestimates the maximum concentrations due to a small slope (Fig.

5.15). Wong et al’s PM2.5 model which is based on linear regression (LR) model

and underestimates the biased slope (Cornbleet and Gochman 1979; Linnet,

1993; Stöckl et al., 1998), is discussed in section 4.1.1.1. Some events of over and

underestimations are marked with red boxes in Fig. 5.15. A lower intercept of the

SARA PM2.5 model enabled it to predict minimum PM2.5 concentrations as low as

24.45g/m3 which is lower than the minimum concentrations 33.41g/m3

predicted by Wong et al. Similarly, the SARA PM2.5 model predicted the maximum

PM2.5 concentrations better than Wong et al. due to relatively large and unbiased

slope which computed using Deming Regression (section 4.1.1.1).

Results Chapter 5

Chapter 5

variables Chapter 5

86

Figure 5.14: Evaluation of (a) SARA and (b) Wong et al. (2011) PM2.5 models at

Tap Mun in Hong Kong.

Results Chapter 5

Chapter 5

variables Chapter 5

87

Table 5.7: Descriptive statistics of predicted PM2.5 by SARA and Wong et al’s

PM2.5 models at Tap Mun air quality station in Hong Kong for autumn and winter

of 2007 to 2009.

Site PM2.5 Min Max Mean SE Mean StDev

Tap Mun Observed 21.91 106.75 49.12 1.91 18.08

SARA PM2.5 Model

Tap Mun Predicted 24.45 105.28 47.34 1.99 18.90

Wong PM2.5 Model

Tap Mun Predicted 33.41 79.98 46.60 1.15 10.89

Results Chapter 5

Chapter 5

variables Chapter 5

88

Figure 5.15: Temporal variations of Observed, SARA, and Wong et al. PM2.5 concentrations at Tap Mun air quality station in Hong

Kong for autumn and winter of the years 2007 to 2009. Red boxes indicate the under and overestimation of Wong et al’s PM2.5

model.

Results Chapter 5

Chapter 5

variables Chapter 5

89

5.11 Spatial Variation of SARA–retrieved PM2.5 (500 m) over

Hong Kong and PRD Region: an example for high pollution

episode

In order to illustrate the applicability of the SARA PM2.5 model to the

whole region, PM2.5 was predicted over Hong Kong and the PRD regions for a

high pollution episode the on 4th December 2007 (Fig.5.15). Fig. 5.15 shows the

higher PM2.5 concentrations over eastern PRD region as well as over HKIA and

the dense urban areas of Hong Kong. The spatial distribution of SARA–retrieved

PM2.5 (500 m) (Fig. 5.15b) suggests higher PM2.5 concentrations are present over

lower elevations (<100 m) of the urban, suburban and rural regions of Hong

Kong. To understand the source of higher PM2.5 concentrations in Hong Kong, a

four-day back-trajectory at three different altitudes (500 m, 2500 m and 5000 m

above ground level (AGL)) was acquired from the NOAA HYSPLIT model (Fig.

5.16a). Fig. 5.16a shows that high level air pollution over Hong Kong was

transported from India (green) and south-east Asia (blue) whereas lower level

pollution was transported from Mainland China. To determine the effect of low

to high level air pollution on Hong Kong air quality, the WD from AWS nearest to

urban/suburban and rural air quality stations, namely HKO, TMS, SLW, WLP and

TM (Table 5.8, Fig. 2.1) as well as 10m WD from WRF at 4.5km resolution (Fig.

5.16b) was analyzed. The WD pattern (Table 5.8 and Fig. 5.16b) suggests that

Hong Kong air quality was more affected by low level air pollution transported

from Mainland China by northeasterly winds than high level air pollution from

elsewhere.

Table 5.8: Wind direction from AWS at HKO, TMS, SLW, WLP and TM during a

high pollution episode (4th December 2007)

AWS HKO TMS SLW WLP TM

WD 94.0 92.5 96.6 131.6 200.4

Results Chapter 5

90

Figure 5.15: Spatial distribution of SARA–retrieved PM2.5 over (a) Pearl River Delta (PRD) region and (b) Hong Kong during a

high pollution episode (4th December 2007)

Results Chapter 5

Chapter 5

variables Chapter 5

91

(a)

(b)

Figure 5.16: (a) HYSPLIT four days back-trajectory at three different altitudes 500 m (red), 2500 m (blue) and 5000 m (green)

AGL, and (b) WRF streamlines showing 10m WD at 4.5 km resolution, during a high pollution episode (4th December 2007)

Summary and Conclusions Chapter 6

92

Chapter 6

Summary and Conclusions

The primary objective of this study was to monitor Fine Particulates

(PM2.5) in Hong Kong and the Pearl River Delta (PRD) region using high

resolution remote sensing. Aerosol optical depth (AOD, Ångström, 1930) from

satellite images has been used for monitoring and understanding the behavior of

PM2.5 from local to global scales. In this study, a Simplified high resolution (500

m) MODIS Aerosol Retrieval Algorithm (SARA) was developed over mixed

surface types of Hong Kong and the PRD region to retrieve AOD by constraining

the local AERONET AOD measurements, assuming Lambertian surfaces, single

scattering approximation, and that single scattering albedo (SSA) and asymmetry

factors are regionally constant for a particular day (Bilal et al., 2013). The SARA

method does not use a look–up–table (LUT) for computation of AOD, as the

algorithm is based on real viewing geometry and encompasses a wide range of

aerosol conditions and aerosol types ( = 0.80–1.0). Deming regression (DR,

Deming, 1943) was used to develop a PM2.5 model based on the SARA–retrieved

AOD at 500 m resolution at four urban/suburban air quality stations in Hong

Kong for spatio-temporal monitoring of PM2.5. The regression coefficients (slope

and intercept) were refined using bins of meteorological variables including

Planetary Boundary Layer Height (PBLH) and surface pressure at 500 m

resolution to enhance the predicting power of the SARA PM2.5 model. The SARA–

retrieved AOD and the SARA PM2.5 model at 500 m resolution were validated

with three different ground–based Sun photometers (AERONET, Sky–radiometer

and Microtops II), and urban/suburban and rural air quality stations

respectively, located in different land cover types of Hong Kong.


93

The MOD04 C005 AOD retrievals are inferior to SARA AOD retrievals and

underestimate AOD over the Hong Kong International Airport (HKIA) which is a

highly polluted area with bright surface (ρ2.21μm approaching 0.25). The C005

retrievals also overestimated the AOD over areas of high elevation. The SARA

AOD because of its higher spatial resolution can identify detailed pollution

sources in the PRD region, whereas the C005 AOD retrieval is unable to depict

pollution sources due to low resolution, as well as the effect of bright aerosols on

the visible band thresholding, resulting in missing data. The correlation

coefficient of the SARA method is higher (R = 0.97) than the previously

developed MODIS aerosol retrieval algorithms based on a LUT at 10 km (R =

0.88, Levy et al., 2007b), 1 km (R = 0.91, Li et al., 2005) and 500 m spatial

resolutions (R = 0.88, Wong et al., 2011; R = 0.83, Wang et al., 2012). The SARA

AOD also achieved a higher correlation (R = 0.97) with the Sky–radiometer than

a previously reported Sky–radiometer study over Hong Kong (R = 0.79, Cheng et

al., 2006).

The degree of correlation between AOD and PM2.5 is sensitive to the

temporal resolution of PM2.5 in Hong Kong as higher correlation of AOD was

observed with 4–hr (09:00–12:00) mean PM2.5 concentrations as opposed to

daily means. Similar results were also obtained by Gupta and Christopher,

(2008) over the southeastern United States and Schaap et al. (2009) over

Cabauw, the Netherlands. It is noted that the slope (110.5 x AOD) between SARA

AOD and PM2.5 estimated by Deming Regression model is much closer to y = x

(1:1 line) than the previous studies in Hong Kong (63.66 x AOD, Wong et al.,

2009), and in Netherlands (124.5 x AOD, Schaap et al., 2009) as they used linear

regression model which estimates the biased slope (Cornbleet and Gochman

1979; Linnet, 1993; Stöckl et al., 1998).

The SARA–retrieved AOD at 500 m resolution has a better correlation (R

= 0.78) than C005 AOD at 10 km resolution (R = 0.64) with the PM2.5

concentrations at five urban/suburban (CT, TW, TC, and YL) and rural (TM) air

quality stations in Hong Kong as the C005 retrieval is unable to retrieve accurate


94

AOD over mixed surfaces due to coarse spatial resolution. The correlation of R =

0.78 achieved for the SARA PM2.5 model is higher than previously reported for

studies in Hong Kong (R = 0.40, Gupta et al., 2006b), in USA (R = 0.63, Engel-Cox

et al., 2004; R = 0.70, Wang and Christopher, 2003; R = 0.62, Gupta and

Christopher, 2008), in southern Sweden (R = 0.77, Glantz et al., 2009), and in the

Netherlands (R = 0.75, Schaap et al., 2009). However the R = 0.64 result for

MOD04 C005 is similar to that for several previous studies. This suggests that the

correlation between AOD and PM2.5 depends on the AOD retrieval algorithm, and

also its spatial resolution, which supports the findings of Kumar et al., 2007.

The meteorological variables are better indicators of PM2.5 in spring and

summer than in autumn and winter as they contributed only 2 % to 3 % to the

correlation of AOD with PM2.5 in autumn and winter. The bin method based on

meteorological variables is very useful to refine the coefficients of the regression

model based on AOD observations for accurate prediction of PM2.5, and

maximum correlation of 0.99 between AOD and PM2.5 was obtained in this study.

The results demonstrate that the SARA PM2.5 model is better than the

existing AOD model for detailed monitoring of PM2.5 concentrations over dense

and congested areas of Hong Kong as well as over the industrialized regions of

the PRD, due to two very important factors: (i) retrieval of high resolution (500

m) MODIS AOD using SARA method by constraining the AERONET, and (ii)

refinement of regression coefficients based on bins of meteorological variables.

Therefore, the SARA PM2.5 model can be applied, and is transferable to any other

region of the world by implementing the two factors mentioned above. The

implementation of this new methodology requires at least two ground–based

Sun photometers, one for the development of the SARA method and the second

for its validation, and this is the limitation of this methodology. The SARA PM2.5

model will help the scientific community to understand the effects of PM2.5 on

climate and air quality as well as on health, and the accuracy of its prediction

depends on the relative contribution of fine and coarse mode particles to the

total AOD.

References

95

REFERENCES

Al-Saadi, J., Szykman, J., Pierce, R.B., Kittaka, C., Neil, D., Chu, D.A., et al. (2005).

Improving National Air Quality Forecasts with Satellite Aerosol

Observations. Bulletin of the American Meteorological Society, 86, 1249-

1261.

Ångström, A. (1930). On the atmospheric transmission of sun radiation II.

Geografiska Annaler, 12, 130-159.

Ansmann, A., Engelmann, R., Althausen, D., Wandinger, U., Hu, M., Zhang, Y., et al.

(2005). High aerosol load over the Pearl River Delta, China, observed with

Raman lidar and Sun photometer. Geophysical Research Letters, 32.

Antoine, D., & Morel, A. (1998). Relative Importance of Multiple Scattering by Air

Molecules and Aerosols in Forming the Atmospheric Path Radiance in the

Visible and Near-Infrared Parts of the Spectrum. Applied Optics, 37, 2245.

Bell, M.L., Ebisu, K., & Belanger, K. (2007). Ambient Air Pollution and Low Birth

Weight in Connecticut and Massachusetts. Environmental health

perspectives, 115, 1118-1124.

Bilal, M., Nichol, J.E., Bleiweiss, M.P., & Dubois, D. (2013). A Simplified high

resolution MODIS Aerosol Retrieval Algorithm (SARA) for use over mixed

surfaces. Remote Sensing of Environment, 136, 135-145.

Campanelli, M., Nakajima, T., & Olivieri, B. (2004). Determination of the Solar

Calibration Constant for a Sun-Sky Radiometer: Proposal of an In-Situ

Procedure. Applied Optics, 43, 651-659.

References

96

Cao, J.J., Lee, S.C., Ho, K.F., Zhang, X.Y., Zou, S.C., Fung, K., et al. (2003).

Characteristics of carbonaceous aerosol in Pearl River Delta Region, China

during 2001 winter period. Atmospheric Environment, 37, 1451-1460

Cao, J.J., Lee, S.C., Ho, K.F., Zou, S.C., Fung, K., Li, Y., et al. (2004). Spatial and

seasonal variations of atmospheric organic carbon and elemental carbon

in Pearl River Delta Region, China. Atmospheric Environment, 38, 4447-

4456.

Census and Statistic Department of Hong Kong, (2010). Retrieved from:

http://www.censtatd.gov.hk/FileManager/EN/Content_803/population.pdf

(19th June 2013).

Chan, C.K., & Yao, X. (2008). Air pollution in mega cities in China. Atmospheric

Environment, 42, 1-42.

Che, H., Shi, G., Uchiyama, A., Yamazaki, A., Chen, H., Goloub, P., et al. (2008).

Intercomparison between aerosol optical properties by a PREDE

skyradiometer and CIMEL sunphotometer over Beijing, China.

Atmospheric Chemistry and Physics, 8, 3199-3214.

Cheng, A.Y.S., Chan, M.H., & Yang, X. (2006). Study of aerosol optical thickness in

Hong Kong, validation, results, and dependence on meteorological

parameters. Atmospheric Environment, 40, 4469-4477.

Cheng, S., & Lam, K. (1998). An analysis of winds affecting air pollution

concentrations in Hong Kong. Atmospheric Environment, 32, 2559-2567.

Chu, D.A., Kaufman, Y.J., Chern, J., Mao, J., Li, C., & Holben, B.N. (2003). Global

monitoring of air pollution over land from the Earth Observing System-

Terra Moderate Resolution Imaging Spectroradiometer (MODIS). Journal

of Geophysical Research, 108.

References

97

Chu, D.A., Kaufman, Y.J., Ichoku, C., Remer, L.A., Tanré, D., & Holben, B.N. (2002).

Validation of MODIS aerosol optical depth retrieval over land. Geophysical

Research Letters, 29.

Chung, C.E., Ramanathan, V., Kim, D., & Podgorny, I.A. (2005). Global

anthropogenic aerosol direct forcing derived from satellite and ground-

based observations. Journal of Geophysical Research, 110.

Clarke, A.D., Collins, W.G., Rasch, P.J., Kapustin, V.N., Moore, K., Howell, S., et al.

(2001). Dust and pollution transport on global scales: Aerosol

measurements and model predictions. Journal of Geophysical Research,

106, 32555.

Cornbleet, P.J., & Gochman, N. (1979). Incorrect least-squares regression

coefficients in method-comparison analysis. Clinical chemistry, 25, 432-

438.

Cruz, C.N., & Pandis, S.N. (1997). A study of the ability of pure secondary organic

aerosol to act as cloud condensation nuclei. Atmospheric Environment, 31,

2205-2214.

Dawson, J.P., Adams, P.J., & Pandis, S.N. (2007). Sensitivity of PM2.5 to climate in

the Eastern US: a modeling case study. Atmospheric Chemistry and Physics,

7, 4295-4309.

DeGaetano, A., & Doherty, O.M. (2004). Temporal, spatial and meteorological

variations in hourly PM2.5 concentration extremes in New York City.

Atmospheric Environment, 38, 1547-1558.

Deming, W.E. (1943). Statistical adjustment of data. New York: Dover

Publications.

Dominici, F., Peng, R.D., Bell, M.L., Pham, L., McDermott, A., Zeger, S.L., et al.

(2006). Fine particulate air pollution and hospital admission for

References

98

cardiovascular and respiratory diseases. JAMA : the journal of the

American Medical Association, 295, 1127-1134.

Donateo, A., Contini, D., & Belosi, F. (2006). Real time measurements of PM2.5

concentrations and vertical turbulent fluxes using an optical detector.


Dubovik, O., Holben, B.N., Eck, T.F., Smirnov, A., Kaufman, Y.J., King, M.D., et al.

(2002). Variability of absorption and optical properties of key aerosol

types observed in worldwide locations. Journal of the Atmospheric

Sciences, 59, 590-608.

Duffie, J.A., & Beckman, W.A. (1991). Solar Engineering of Thermal Processes, 3rd

Edition - John A. Duffie, William A. Beckman. New York: Wiley.

El-Fadel, M., & Hashisho, Z. (2001). Vehicular Emissions in Roadway Tunnels: A

Critical Review. Critical Reviews in Environmental Science and Technology,

31, 125-174.

Elminir, H.K. (2005). Dependence of urban air pollutants on meteorology. The

Science of the total environment, 350, 225-237.

Engel-Cox, J.A., Holloman, C.H., Coutant, B.W., & Hoff, R.M. (2004). Qualitative and

quantitative evaluation of MODIS satellite sensor data for regional and

urban scale air quality. Atmospheric Environment, 38, 2495-2509.

Englert, N. (2004). Fine particles and human health--a review of epidemiological

studies. Toxicology letters, 149, 235-242.

EPD. (2009). Agreement - CE57/2006 (EP) review of the Air Quality Objectives and

development of a long term air quality strategy for Hong Kong - Feasibility

Study (final report – executive summary). Retrieved from: http://www.epd.

gov.hk/epd/english/environmentinhk/air/studyrpts/files/executive_summary_en

.pdf (20th

April 2013).

References

99

Fang, G., Chang, C., Wu, Y., Fu, P.P., Yang, C., Chen, C., et al. (2002). Ambient

suspended particulate matters and related chemical species study in

central Taiwan, Taichung during 1998–2001. Atmospheric Environment,

36, 1921-1928.

Fraser, M.P., Yue, Z.W., & Buzcu, B. (2003). Source apportionment of fine

particulate matter in Houston, TX, using organic molecular markers.


Gao, Y., Chan, E.Y.Y., Li, L.P., He, Q.Q., & Wong, T.W. (2013). Chronic effects of

ambient air pollution on lung function among Chinese children. Archives

of Disease in Childhood, 98, 128-135.

Gent, J.F., Koutrakis, P., Belanger, K., Triche, E., Holford, T.R., Bracken, M.B., et al.

(2009). Symptoms and Medication Use in Children with Asthma and

Traffic-Related Sources of Fine Particle Pollution. Environmental health

perspectives, 117, 1168-1174.

Glantz, P., Kokhanovsky, A., von Hoyningen-Huene, W., & Johansson, C. (2009).

Estimating PM2.5 over southern Sweden using space-borne optical

measurements. Atmospheric Environment, 43, 5838-5846.

Gomišcek, B., Hauck, H., Stopper, S., & Preining, O. (2004). Spatial and temporal

variations of PM1, PM2.5, PM10 and particle number concentration during

the AUPHEP—project. Atmospheric Environment, 38, 3917-3934.

Gordon, H.R., & Wang, M. (1994). Retrieval of water-leaving radiance and aerosol

optical thickness over the oceans with SeaWiFS: a preliminary algorithm.

Applied Optics, 33, 443.

Gotschi, T., Heinrich, J., Sunyer, J., & Kunzli, N. (2008). Long-term effects of

ambient air pollution on lung function: a review. Epidemiology

(Cambridge, Mass.), 19, 690-701.

References

100

Grigg, J. (2009). Particulate matter exposure in children: relevance to chronic

obstructive pulmonary disease. Proceedings of the American Thoracic

Society, 6, 564-569.

Gupta, P., & Christopher, S.A. (2008). Seven year particulate matter air quality

assessment from surface and satellite measurements. Atmospheric

Chemistry and Physics, 8, 3311-3324.

Gupta, P., Christopher, S.A., Box, M.A., & Box, G.P. (2007). Multi year satellite

remote sensing of particulate matter air quality over Sydney, Australia.

International Journal of Remote Sensing, 28, 4483-4498.

Gupta, P., Christopher, S.A., Wang, J., Gehrig, R., Lee, Y., & Kumar, N. (2006b).

Satellite remote sensing of particulate matter and air quality assessment

over global cities. Atmospheric Environment, 40, 5880-5892.

Gupta, A.K., Nag, S., & Mukhopadhyay, U.K. (2006a). Characterisation of PM10,

PM2.5 and Benzene Soluble Organic Fraction of Particulate Matter in an

Urban Area of Kolkata, India. Environmental monitoring and assessment,

115, 205-222.

Gupta, P., & Christopher, S.A. (2009). Particulate matter air quality assessment

using integrated surface, satellite, and meteorological products: Multiple

regression approach. Journal of Geophysical Research, 114.

Hagler, G.S.W., Bergin, M.H., Salmon, L.G., Yu, J.Z., Wan, E.C.H., Zheng, M., et al.

(2006). Source areas and chemical composition of fine particulate matter

in the Pearl River Delta region of China. Atmospheric Environment, 40,

3802-3815.

Hauser, A. (2005). NOAA AVHRR derived aerosol optical depth over land. Journal

of Geophysical Research, 110.

References

101

He, Q., Li, C., Mao, J., Lau, A.K., & Chu, D.A. (2008). Analysis of aerosol vertical

distribution and variability in Hong Kong. Journal of Geophysical Research,

113.

Herman, M., Deuzé, J.L., Devaux, C., Goloub, P., Bréon, F.M., & Tanré, D. (1997).

Remote sensing of aerosols over land surfaces including polarization

measurements and application to POLDER measurements. Journal of

Geophysical Research, 102, 17039.

Hien, P.D., Bac, V.T., Tham, H.C., Nhan, D.D., & Vinh, L.D. (2002). Influence of

meteorological conditions on PM2.5 and PM2.5-10 concentrations during the

monsoon season in Hanoi, Vietnam : Asia. Atmospheric Environment, 36,

3473-3484.

Ho, K.F., Cao, J.J., Lee, S.C., & Chan, C.K. (2006). Source apportionment of PM2.5 in

urban area of Hong Kong. Journal of hazardous materials, 138, 73-85.

Holben, B.N., Eck, T.F., Slutsker, I., Tanré, D., Buis, J.P., Setzer, A., et al. (1998).

AERONET—A Federated Instrument Network and Data Archive for

Aerosol Characterization. Remote Sensing of Environment, 66, 1-16.

Holben, B.N., Tanré, D., Smirnov, A., Eck, T.F., Slutsker, I., Abuhassan, N., et al.

(2001). An emerging ground-based aerosol climatology: Aerosol optical

depth from AERONET. Journal of Geophysical Research, 106, 12067-12097.

Hsu, N.C., Tsay, S.-., King, M.D., & Herman, J.R. (2006). Deep Blue Retrievals of

Asian Aerosol Properties During ACE-Asia. IEEE Transactions on

Geoscience and Remote Sensing, 44, 3180-3195.

Hsu, N.C., Tsay, S.-., King, M.D., & Herman, J.R. (2004). Aerosol Properties Over

Bright-Reflecting Source Regions. IEEE Transactions on Geoscience and

Remote Sensing, 42, 557-569.

References

102

Hu, X.M., Petra, M., & Klein, M. X. (2013). Evaluation of the updated YSU

planetary boundary layer scheme within WRF for wind resource and air

quality assessments. Journal of Geophysical Research: Atmospheres,

118:18, 10,490-10,505.

Hubanks, P., Chu, A., Ridgway, B., Strabala, K., Platnick, S., Mattoo, S., et al. (2012).

MODIS Atmosphere QA Plan for Collection 005 and 051 (Includes Cirrus

Flag & High Cloud Flag (06_CT) Clarification, Deep Blue Aerosol Update,

Aerosol Over Land Update, Water Vapor and Atmosphere Profile Update,

Changes to MOD35 QA Bit Field Documentation) Version 3.10. , 2013.

Hutchison, K.D., Smith, S., & Faruqui, S.J. (2005). Correlating MODIS aerosol

optical thickness data with ground-based PM2.5 observations across Texas

for use in a real-time air quality prediction system. Atmospheric


Hyer, E.J., Reid, J.S., & Zhang, J. (2011). An over-land aerosol optical depth data

set for data assimilation by filtering, correction, and aggregation of MODIS

Collection 5 optical depth retrievals. Atmospheric Measurement

Techniques, 4, 379-408.

Ichoku, C., Chu, D.A., Mattoo, S., Kaufman, Y.J., Remer, L.A., Tanre, D., et al. (2002).

A spatio-temporal approach for global validation and analysis of MODIS

aerosol products. Geophysical Research Letters, 29.

Jethva, H., Satheesh, S.K., & Srinivasan, J. (2007). Assessment of second-

generation MODIS aerosol retrieval (Collection 005) at Kanpur, India.

Geophysical Research Letters, 34.

Kahn, R.A., Gaitley, B., Martonchik, J., Diner, D., Crean, K., & Holben, B. (2005).

Multiangle Imaging Spectroradiometer (MISR) global aerosol optical

depth validation based on Two years of coincident Aerosol Robotic

Network (AERONET) observations. Journal of Geophysical Research, 110.

References

103

Kahn, R.A., Gaitley, B.J., Garay, M.J., Diner, D.J., Eck, T.F., Smirnov, A., et al. (2010).

Multiangle Imaging SpectroRadiometer global aerosol product

assessment by comparison with the Aerosol Robotic Network. Journal of

Geophysical Research, 115.

Kappos, A.D., Bruckmann, P., Eikmann, T., Englert, N., Heinrich, U., Hoppe, P., et al.

(2004). Health effects of particles in ambient air. International journal of

hygiene and environmental health, 207, 399-407.

Kaufman, Y.J., Tanre, D., & Boucher, O. (2002). A satellite view of aerosols in the

climate system. Nature, 419, 215-223.

Kaufman, Y.J., Wald, A.E., Remer, L.A., Gao, B., & Li, R.F., Luke. (1997a). The MODIS

2.1–m Channel—Correlation with Visible Reflectance for Use in Remote

Sensing of Aerosol. IEEE Transactions on Geoscience and Remote Sensing,

35, 1286-1298.

Kaufman, Y.J., Tanré, D., Remer, L.A., Vermote, E.F., Chu, A., & Holben, B.N.

(1997b). Operational remote sensing of tropospheric aerosol over land

from EOS moderate resolution imaging spectroradiometer. Journal of

Geophysical Research, 102, 17051-17067.

Kelley, C.T. (1995). Iterative Methods for Linear and Nonlinear Equations. : Society

for industrial and applied mathematics.

King, M.D., Kaufman, Y.J., Tanré, D., & Nakajima, T. (1999). Remote Sensing of

Tropospheric Aerosols from Space: Past, Present, and Future. Bulletin of

the American Meteorological Society, 80, 2229-2259.

Knobelspiesse, K.D., Pietras, C., Fargion, G.S., Wang, M., Frouin, R., Miller, M.A., et

al. (2004). Maritime aerosol optical thickness measured by handheld sun

photometers. Remote Sensing of Environment, 93, 87-106.

References

104

Koelemeijer, R.B.A., Homan, C.D., & Matthijsen, J. (2006). Comparison of spatial

and temporal variations of aerosol optical thickness and particulate

matter over Europe. Atmospheric Environment, 40, 5304-5315.

Kokhanovsky, A., & Leeuw, G.D. (2009). Satellite Aerosol Remote Sensing Over

Land (Springer Praxis Books / Environmental Sciences). : Springer.

Kulshrestha, A., Bisht, D.S., Masih, J., Massey, D., Tiwari, S., & Taneja, A. (2010).

Chemical characterization of water-soluble aerosols in different

residential environments of semi aridregion of India. Journal of

Atmospheric Chemistry.

Kumar, N., Chu, A., & Foster, A. (2007). An empirical relationship between

PM(2.5) and aerosol optical depth in Delhi Metropolitan. Atmospheric

environment (Oxford, England : 1994), 41, 4492-4503.

Kwok, R.H.F., Fung, J.C.H., Lau, A.K.H., & Fu, J.S. (2010). Numerical study on

seasonal variations of gaseous pollutants and particulate matters in Hong

Kong and Pearl River Delta Region. Journal of Geophysical Research, 115.

Lado-Bordowsky, O., & Naour, I. (1997). Optical paths involved in determining

the scattering angle for the scattering algorithm developed in

LOWTRAN7. International Journal of Infrared and Millimeter Waves, 18,

1689-1696.

Lee, J.H., Kim, Y.P., Moon, K., Kim, H., & Lee, C.B. (2001). Fine particle

measurements at two background sites in Korea between 1996 and 1997.


Lee, S.C., Cheng, Y., Ho, K.F., Cao, J.J., Louie, P.K.-., Chow, J.C., et al. (2006).

PM1.0and PM2.5 Characteristics in the Roadside Environment of Hong

Kong. Aerosol Science and Technology, 40, 157-165.

References

105

Levy, R.C., Remer, L.A., Kleidman, R.G., Mattoo, S., Ichoku, C., Kahn, R., et al.

(2010). Global evaluation of the Collection 5 MODIS dark-target aerosol

products over land. Atmospheric Chemistry and Physics, 10, 10399-10420.

Levy, R.C., Remer, L.A., Martins, J.V., Kaufman, Y.J., Plana-Fattori, A., Redemann, J.,

et al. (2005). Evaluation of the MODIS Aerosol Retrievals over Ocean and

Land during CLAMS. Journal of the Atmospheric Sciences, 62, 974-992.

Levy, R.C., Remer, L.A., Mattoo, S., Vermote, E.F., & Kaufman, Y.J. (2007a). Second-

generation operational algorithm: Retrieval of aerosol properties over

land from inversion of Moderate Resolution Imaging Spectroradiometer

spectral reflectance. Journal of Geophysical Research, 112.

Levy, R.C., Remer, L.A., & Dubovik, O. (2007b). Global aerosol optical properties

and application to Moderate Resolution Imaging Spectroradiometer

aerosol retrieval over land. Journal of Geophysical Research, 112.

Li, C., Lau, A.K.-., Mao, J., & Chu, D.A. (2005). Retrieval, validation, and application

of the 1-km aerosol optical depth from MODIS measurements over Hong

Kong. IEEE Transactions on Geoscience and Remote Sensing, 43, 2650-

2658.

Li, Z., Zhao, X., Kahn, R., Mishchenko, M., Remer, L., Lee, K.-., et al. (2009).

Uncertainties in satellite remote sensing of aerosols and impact on

monitoring its long-term trend: a review and perspective. Annales

Geophysicae, 27, 2755-2770.

Li, Z., Niu, F., Lee, K., Xin, J., Hao, W., Nordgren, B., et al. (2007). Validation and

understanding of Moderate Resolution Imaging Spectroradiometer

aerosol products (C5) using ground-based measurements from the

handheld Sun photometer network in China. Journal of Geophysical

Research, 112.

References

106

Liang, S. (2005). Quantitative Remote Sensing of Land Surfaces. New Jersey: John

Wiley & Sons, Inc.

Linares, B., Guizar, J., Amador, N., Garcia, A., Miranda, V., Perez, J., et al. (2010).

Impact of air pollution on pulmonary function and respiratory symptoms

in children. Longitudinal repeated-measures study. BMC Pulmonary

Medicine, 10, 62.

Linnet, K. (1998). Performance of Deming regression analysis in case of

misspecified analytical error ratio in method comparison studies. Clinical

chemistry, 44, 1024-1031.

Linnet, K. (1993). Evaluation of regression procedures for methods comparison

studies. Clinical chemistry, 39, 424-432.

Liu, Y., Huang, J., Shi, G., Takamura, T., Khatri, P., Bi, J., et al. (2011). Aerosol

optical properties and radiative effect determined from sky-radiometer

over Loess Plateau of Northwest China. Atmospheric Chemistry and

Physics, 11, 11455-11463.

Liu, Y., Franklin, M., Kahn, R., & Koutrakis, P. (2007). Using aerosol optical

thickness to predict ground-level PM2.5 concentrations in the St. Louis

area: A comparison between MISR and MODIS. Remote Sensing of


Louie, P., Watson, J., Chow, J., Chen, A., Sin, D., & Lau, A. (2005). Seasonal

characteristics and regional transport of PM in Hong Kong. Atmospheric

Environment.

Mi, W., Li, Z., Xia, X., Holben, B., Levy, R., Zhao, F., et al. (2007). Evaluation of the

Moderate Resolution Imaging Spectroradiometer aerosol products at two

Aerosol Robotic Network stations in China. Journal of Geophysical

Research, 112.

References

107

Mishchenko, M.I., Geogdzhayev, I.V., Cairns, B., Rossow, W.B., & Lacis, A.A. (1999).

Aerosol retrievals over the ocean by use of channels 1 and 2 AVHRR data:

sensitivity analysis and preliminary results. Applied Optics, 38, 7325.

Morys, M., Mims, F.M., Hagerup, S., Anderson, S.E., Baker, A., Kia, J., et al. (2001).

Design, calibration, and performance of MICROTOPS II handheld ozone

monitor and Sun photometer. Journal of Geophysical Research, 106, 14573

Mysliwiec, M.J., & Kleeman, M.J. (2002). Source Apportionment of Secondary

Airborne Particulate Matter in a Polluted Atmosphere. Environmental

science & technology, 36, 5376-5384.

Nakajima, T., Tonna, G., Rao, R., Boi, P., Kaufman, Y.J., & Holben, B.N. (1996). Use

of sky brightness measurements from ground for remote sensing of

particulate polydispersions. Applied Optics, 35, 2672-2686.

Nichol, J.E., & Wong, M.S. (2009). High Resolution Remote Sensing of Densely

Urbanised Regions: a Case Study of Hong Kong. Sensors, 9, 4695-4708.

Nolte, C.G., Schauer, J.J., Cass, G.R., & Simoneit, B.R.T. (2002). Trimethylsilyl

Derivatives of Organic Compounds in Source Samples and in Atmospheric

Fine Particulate Matter. Environmental science & technology, 36, 4273-

4281.

North, P.R.J. (2002). Estimation of aerosol opacity and land surface bidirectional

reflectance from ATSR-2 dual-angle imagery: Operational method and

validation. Journal of Geophysical Research, 107.

Papadimas, C.D., Hatzianastassiou, N., Mihalopoulos, N., Kanakidou, M., Katsoulis,

B.D., & Vardavas, I. (2009). Assessment of the MODIS Collections C005 and

C004 aerosol optical depth products over the Mediterranean basin.

Atmospheric Chemistry and Physics, 9, 2987-2999.

References

108

Peña, A., Gryning, S.-E., and Hahmann, A. N. (2013). Observations of the

atmospheric boundary layer height under marine upstream flow

conditions at a coastal site. Journal of Geophysical Research:

Atmospheres, 118, 1924–1940, doi:10.1002/jgrd.50175.

Pope, C.A.I., & Dockery, D.W. (2006). Health effects of fine particulate air

pollution: lines that connect. Journal of the Air & Waste Management

Association (1995), 56, 709-742.

Prados, A.I., Kondragunta, S., Ciren, P., & Knapp, K.R. (2007). GOES

Aerosol/Smoke Product (GASP) over North America: Comparisons to

AERONET and MODIS observations. Journal of Geophysical Research, 112.

Rahman, H., Pinty, B., & Verstraete, M.M. (1993). Coupled Surface-Atmosphere

Reflectance (CSAR) Model 2. Semiempirical Surface Model Usable With

NOAA Advanced Very High Resolution Radiometer Data. Journal of

Geophysical Research, 98, 20791-20801.

Remer, L.A., Chin, M., DeCola, P., Feingold, G., Halthore, R., Kahn, R.A., et al.

(2009). Executive summary. In M. Chin, R.A. Kahn, & S.E. Schwartz (Eds.),

Atmospheric Aerosol Properties and Climate Impacts (pp. 1-8). : U.S.

Climate Change Science Program.

Remer, L.A., Kaufman, Y.J., Tanré, D., Mattoo, S., Chu, D.A., Martins, J.V., et al.

(2005). The MODIS Aerosol Algorithm, Products, and Validation. Journal

of the Atmospheric Sciences, 62, 947-973.

Remer, L.A., Kleidman, R.G., Levy, R.C., Kaufman, Y.J., Tanré, D., Mattoo, S., et al.

(2008). Global aerosol climatology from the MODIS satellite sensors.

Journal of Geophysical Research, 113.

References

109

Riffler, M., Popp, C., Hauser, A., Fontana, F., & Wunderle, S. (2010). Validation of a

modified AVHRR aerosol optical depth retrieval algorithm over Central

Europe. Atmospheric Measurement Techniques, 3, 1255-1270.

Salomonson, V.V., Barnes, W., & Masuoka, E.J. (2006). Introduction to MODIS and

an Overview of Associated Activities. Earth Science Satellite Remote

Sensing, 1, 12-32.

Sayer, A.M., Hsu, N.C., Bettenhausen, C., Jeong, M.-., Holben, B.N., & Zhang, J.

(2012). Global and regional evaluation of over-land spectral aerosol

optical depth retrievals from SeaWiFS. Atmospheric Measurement

Techniques, 5, 1761-1778.

Schaap, M., Apituley, A., Timmermans, R.M.A., Koelemeijer, R.B.A., & de Leeuw, G.

(2009). Exploring the relation between aerosol optical depth and PM2.5 at

Cabauw, the Netherlands. Atmospheric Chemistry and Physics, 9, 909-925.

Shen, S., Jaques, P.A., Zhu, Y., Geller, M.D., & Sioutas, C. (2002). Evaluation of the

SMPS–APS system as a continuous monitor for measuring PM2.5, PM10 and

coarse (PM2.5−10) concentrations. Atmospheric Environment, 36, 3939-

3950.

Shi, W., Wong, M.S., Wang, J., & Zhao, Y. (2012). Analysis of airborne particulate

matter (PM2.5) over Hong Kong using remote sensing and GIS. Sensors

(Basel, Switzerland), 12, 6825-6836.

Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Wang W, Powers JG.

2005. A description of the Advanced Research WRF version 2. NCAR

Technical Note TN-468 + STR, 88, available at:

http://www.mmm.ucar.edu/wrf/users/ docs/arw_v2.pdf (19th June 2013).

References

110

Smirnov, A., Holben, B.N., Eck, T.F., Dubovik, O., & Slutsker, I. (2000). Cloud-

Screening and Quality Control Algorithms for the AERONET Database.

Remote Sensing of Environment, 73, 337-349.

Stöckl, D., Dewitte, K., & Thienpont, L.M. (1998). Validity of linear regression in

method comparison studies: is it limited by the statistical model or the

quality of the analytical input data? Clinical chemistry, 44, 2340-2346.

Streets, D.G., Yu, C., Bergin, M.H., Wang, X., & Carmichael, G.R. (2006). Modeling

Study of Air Pollution Due to the Manufacture of Export Goods in China's

Pearl River Delta. Environmental science & technology, 40, 2099-2107.

Sun, J., & Ariya, P. (2006). Atmospheric organic and bio-aerosols as cloud

condensation nuclei (CCN): A review. Atmospheric Environment, 40, 795-

820.

Tang, J., Xue, Y., Yu, T., & Guan, Y. (2005). Aerosol optical thickness determination

by exploiting the synergy of TERRA and AQUA MODIS. Remote Sensing of


Tanré, D., Kaufman, Y.J., Herman, M., & Mattoo, S. (1997). Remote sensing of

aerosol properties over oceans using the MODIS/EOS spectral radiances.

Journal of Geophysical Research, 102, 16971-16988.

Tanré, D., Deschamps, P.Y., Devaux, C., & Herman, M. (1988). Estimation of

Saharan aerosol optical thickness from blurring effects in thematic

mapper data. Journal of Geophysical Research, 93, 15955.

Tasumi, M., Allen, R.G., & Trezza, R. (2008). At-Surface Reflectance and Albedo

from Satellite for Operational Calculation of Land Surface Energy Balance.

Journal of Hydrologic Engineering, 13, 51-63.

Tian, J., & Chen, D. (2010). A semi-empirical model for predicting hourly ground-

level fine particulate matter (PM2.5) concentration in southern Ontario

References

111

from satellite remote sensing and ground-based meteorological

measurements. Remote Sensing of Environment, 114, 221-229.

Torres, O., Bhartia, P.K., Herman, J.R., Sinyuk, A., Ginoux, P., & Holben, B. (2002). A

Long-Term Record of Aerosol Optical Depth from TOMS Observations and

Comparison to AERONET Measurements. Journal of the Atmospheric

Sciences, 59, 398-413.

Torres, O., Tanskanen, A., Veihelmann, B., Ahn, C., Braak, R., Bhartia, P.K., et al.

(2007). Aerosols and surface UV products from Ozone Monitoring

Instrument observations: An overview. Journal of Geophysical Research,

112, D24S47.

Tsai, T., Jeng, Y., Chu, D.A., Chen, J., & Chang, S. (2011). Analysis of the relationship

between MODIS aerosol optical depth and particulate matter from 2006

to 2008. Atmospheric Environment, 45, 4777-4788.

Tsang, H., Kwok, R., & Miguel, A.H. (2008). Pedestrian exposure to ultrafine

particles in hong kong under heavy traffic conditions. Aerosol and Air

Quality Research, 8, 19-27.

van de Hulst, H.C. (1948). Scattering in a Planetary Atmosphere. The

Astrophysical Journal, 107, 220.

Vermote, E.F., Tanre, D., Deuze, J.L., Herman, M., & Morcette, J.-. (1997). Second

Simulation of the Satellite Signal in the Solar Spectrum, 6S: an overview.

IEEE Transactions on Geoscience and Remote Sensing, 35, 675-686.

Vermote, E.F., & Kotchenova, S. (2008). Atmospheric correction for the

monitoring of land surfaces. Journal of Geophysical Research, 113.

Vidot, J., Santer, R., & Aznay, O. (2008). Evaluation of the MERIS aerosol product

over land with AERONET. Atmospheric Chemistry and Physics, 8, 7603-

7617.

References

112

Wallace, J., & Hobbs, P.V. (2006). Atmospheric science: an introductory survey. (pp.

483). : Amsterdam ; Elsevier Academic Press, c2006.

Wang, J., & Christopher, S.A. (2003). Intercomparison between satellite-derived

aerosol optical thickness and PM2.5 mass: Implications for air quality

studies. Geophysical Research Letters, 30.

Wang, Y., Xue, Y., Li, Y., Guang, J., Mei, L., Xu, H., et al. (2012). Prior knowledge-

supported aerosol optical depth retrieval over land surfaces at 500 m

spatial resolution with MODIS data. International Journal of Remote

Sensing, 33, 674-691.

Wang, W., Primbs, T., Tao, S., & Simonich, S.L.M. (2009). Atmospheric Particulate

Matter Pollution during the 2008 Beijing Olympics. Environmental science

& technology, 43, 5314-5320.

Ward, D.J., & Ayres, J.G. (2004). Particulate air pollution and panel studies in

children: a systematic review. Occupational and environmental medicine,

61, e13.

Weinmayr, G., Romeo, E., De Sario, M., Weiland, S.K., & Forastiere, F. (2010).

Short-term effects of PM10 and NO2 on respiratory health among

children with asthma or asthma-like symptoms: a systematic review and

meta-analysis. Environmental health perspectives, 118, 449-457.

Westgard, J.O., & Hunt, M.R. (1973). Use and interpretation of common statistical

tests in method-comparison studies. Clinical chemistry, 19, 49-57.

Willmott, C., & Matsuura, K. (2005). Advantages of the mean absolute error

(MAE) over the root mean square error (RMSE) in assessing average

model performance. Climate Research, 30, 79-82.

References

113

Wise, E., & Comrie, A. (2005). Meteorologically adjusted urban air quality trends

in the Southwestern United States. Atmospheric Environment, 39, 2969-

2980.

Wong, M.S., Nichol, J.E., & Lee, K.H. (2011a). An operational MODIS aerosol

retrieval algorithm at high spatial resolution, and its application over a

complex urban region. Atmospheric Research, 99, 579-589.

Wong, M.S., Nichol, J., Lee, K.H., & Lee, B.Y. (2011b). Monitoring 2.5 µm

particulate matter within urbanized regions using satellite-derived

aerosol optical thickness, a study in Hong Kong. International Journal of

Remote Sensing, 32, 8449-8462.

Xue, Y., & Cracknell, A.P. (1995). Operational bi-angle approach to retrieve the

Earth surface albedo from AVHRR data in the visible band. International

Journal of Remote Sensing, 16, 417-429.

Zhao, X., Zhang, X., Xu, X., Xu, J., Meng, W., & Pu, W. (2009). Seasonal and diurnal

variations of ambient PM2.5 concentration in urban and rural

environments in Beijing. Atmospheric Environment, 43, 2893-2900.

monitoring of fine particulates in hong kong and pearl river delta region using remote sensing

Documents