AND PUBLIC HEALTH: ACOUSTICAL MEASUREMENT AND SOCIAL SURVEY AROUND SYDNEY (KINGSFORD SMITH)

By

Tharit Issarayangyun

Bachelor of Civil Engineering (second honour), Master of Civil Engineering

A thesis submitted in fulfillment of the requirement for the degree of

DOCTOR OF PHILOSOPHY

SCHOOL OF CIVIL AND ENVIRONMENTAL ENGINEERING THE UNIVERSITY OF NEW SOUTH WALES SYDNEY, AUSTRALIA

March, 2005

PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Project Report Sheet

Surname or Family name: ISSARAYANGYUN First name: THARIT Other name/s: Abbreviation for degree as given in the University calendar: PhD School: Civil & Environmental Engineering Faculty: Engineering Title: Aircraft Noise and Public Health: Acoustical Measurement and Social Survey around Sydney (Kingsford Smith) Airport

Abstract 350 words maximum: (PLEASE TYPE)

The development of major commercial promotes the air transport industry and generates positive economic benefits to the airport and to its host economy. However, external costs are associated with these benefits. Any increase in aircraft movement causes negative environmental impacts, especially noise . Governments have reduced aircraft noise levels at their sources, or introduced aircraft noise management strategies; however the problems have never been satisfactorily resolved. This research aims at developing a better understanding of the impacts of aircraft noise on community health and well-being by exploring two core research questions: (1) “Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”; and (2) “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”. The Sydney (Kingsford Smith) Airport has been selected as a case study. The health survey instruments have been developed and piloted, and then translated from English into Greek and Arabic. A postal self-administrative health survey (with follow-up letters) has been implemented in the areas surrounding Sydney Airport (called “aircraft noise exposure group”) and in the matched control group. The total sample size was 1,500 with 47% response rate. This thesis has developed a ‘new’ noise index (named Noise Gap Index, NGI) to describe and assess aircraft noise in such a way that is easily understood by the layperson. Factorial analysis of covariance revealed that “Health related quality of life, in term of physical functioning, general health, vitality, and mental health, of community chronically exposed to high aircraft noise level were worse than the matched control area”. Binary logistic regression analysis found that “Subjects (aged 15 – 87) who have been chronically exposed to high aircraft noise level have the odds of 2.61 of having chronic noise stress. In addition, person who have chronic noise stress have the odds of 2.74 of having hypertension compared with those without chronic noise stress”. Finally, the robust hypotheses of effects of aircraft noise on community health and well-being for future experimental study were proposed.

Declaration relating to disposition of project report/thesis I am fully aware of the policy of the University relating to the retention and use of higher degree project reports and theses, namely that the University retains the copies submitted for examination and is free to allow them to be consulted or borrowed. Subject to the provisions of the Copyright Act 1968, the University may issue a project report or thesis in whole or in part, in Photostat or microfilm or other copying medium.

I also authorise the publication by University Microfilms of a 350 word abstract in Dissertation Abstracts International (applicable to doctorates only).

………………………………………… ……………………………………. …………………………………….. Signature Witness Date

The University recognises that there may e exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing to the Registrar. Requests for a longer period of restriction may be considered in exceptional circumstances if accompanied by a letter of support from the Supervisor or Head of School. Such requests must be submitted with the thesis/project report.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

Registrar and Deputy Principal THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS CERTIFICATE OF ORIGINALITY

I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at UNSW or any other educational institution, excepts where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis.

I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.

(Signed)…………………………………………………………….. Tharit Issarayangyun

ABSTRACT

The development of major commercial airports promotes the air transport industry and generates positive economic benefits to the airport and to its host economy. However, external costs are associated with these benefits. Any increase in aircraft movement causes negative environmental impacts, especially . The public reaction to aircraft noise has been vigorous. Efforts have been made to reduce aircraft noise levels at their source, or to implement aircraft noise management strategy. However, the problems have never been satisfactorily resolved as long as the growth of the air transport market remains high. The issue of community health and well-being has been ignored by aircraft noise management strategies possibly because ‘health’ has been interpreted as just only the absence of disease.

Even though no evidence has been found to support the proposition that aircraft noise is loud enough to significantly deteriorate the hearing of people living around the commercial airports, it is obvious that everyday aircraft noise exposure deteriorates quality of life by disturbing daily activities which can be a cause of chronic emotional stress. Evidence is emerging that appears to associate some forms of health risk with this stress.

This thesis studies association between community health and well-being and aircraft noise exposure by applying the concept of transdisciplinary thinking and a definition of ‘health’ declared by World Health Organisation (WHO) “Health is not only the absence of disease, but also includes a state of complete in physical, mental, and social well- being”. The literature review of possible interconnections of disciplinary knowledge about impacts of aircraft noise on community health helps establish research gaps, leading to two core research questions: firstly, “Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”; and secondly, “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”.

i This thesis selects a descriptive epidemiology of cross-sectional study with a control group as an epidemiological method. A comprehensive health survey instrument for the evaluation of effects of aircraft noise on health related quality of life and adult hypertension has been developed for a postal self-administrative technique. The health survey instruments consist of questionnaire, cover letter, first follow-up letter, second follow-up letter, third follow-up letter, envelop, return envelop, and follow-up procedures. The questionnaire measures seven major characteristics of each subject: 1) health related quality of life; 2) prevalence of hypertension; 3) noise stress; 4) noise sensitivity; 5) noise annoyance; 6) confounding factors; and 7) demographic characteristics. Some scales from the Short Form Health Survey (SF-36v2) have been employed to measure health related quality of life of each subject. The selected scales included Physical Functioning, General Health, Vitality, and Mental Health. The performance of the health survey instruments and the internal reliability of noise stress scale and noise sensitivity scale have been checked by the pilot study.

Residents around Sydney (Kingsford Smith) Airport were selected as a case study. The study population was defined as two groups: aircraft noise exposure group; and a control group. The aircraft noise exposure group includes many suburbs around Sydney Airport where the average annual day of N70 in 2003 was higher than fifty. The control group was located in Western Sydney (approximately 55 kilometres from Sydney Airport) where jet noise is barely detectable, and where the socio-economic status of this area is comparable with the aircraft noise exposure group. The N70 was selected because the current practice of Australian Noise Exposure Forecast is insufficient to reflect people reactions to aircraft noise. The N70 was designed from a concept of transparency in aircraft noise information.

The sample size is 750 subjects per study group, calculated from an expected response rate from the pilot study and the statistical power of SF-36. Every home address (excluding apartments, commercial buildings, addresses for sale or lease, and abandoned addresses) located in the local traffic area in the study population of both aircraft noise exposure area and the control area were observed by the author to minimise the address error. The study sample is randomly chosen by a computer to ensure an equal chance of selection and independence of study sample. The health

ii survey instruments are translated from English into Greek and Arabic to overcome effects from ethnicity bias.

To better understand community responses toward aircraft noise, this thesis aims to expand the ways to describe and assess aircraft noise by undertaking field measurements and developing a ‘new’ noise index (which has been termed Noise Gap Index, NGI) which incorporates the characteristics of background into N70. The NGI was developed based on a research assumption that “People living in different background environmental noise areas might have different responses to the same aircraft noise level”. A preliminary statistical analysis found some form of statistical association between aircraft noise annoyance and NGI. Subjects residing in high and medium background environmental noise areas are more likely to be annoyed by the same aircraft noise exposure level than subjects living in low background environmental noise areas.

A total of 796 questionnaires were returned, of whom 704 filled in the questionnaire and 92 indicated an unwillingness to participate the survey. The number of responses from subjects in the control group (N=332) was a little bit lower than from the noise exposure area (N=372). The total response rate was 47 percent. The percentage of respondents of both selected languages (Greek and Arabic) was very low (1.2%). Subjects who informed in the questionnaire that they have resided in their existing resident for less than 1 year were excluded. Consequently, the total number of samples of aircraft noise exposure group and the control group became 339 and 316, respectively.

Multivariate statistical analysis techniques are employed. Factorial analysis of covariance is applied to explore the first core research question. The analyses are divided into four sub-sections due to the independence of each health related quality of life measures. First, after adjustment by prevalence of hypertension, body mass index, and age, Physical Functioning varied significantly with aircraft noise exposure. Second, after adjustment by hypertension, exercise activity, and age, General Health varied significantly with aircraft noise exposure. Third, after adjustment by employment status, exercise activity, and noise sensitivity, Vitality varied significantly with aircraft noise exposure. Fourth, after adjustment by exercise activity, smoking status, education level,

iii and noise sensitivity, Mental Health varied significantly with aircraft noise exposure. An answer for the first core research question would be: “Health related quality of life, in term of physical functioning, general health, vitality, and mental health, of community chronically exposed to high aircraft noise level were worse than the control area”.

Binary logistic regression analysis is applied to explore the second core research question. The analyses are divided into two sub-sections due to an assumption that “Aircraft noise has indirect impacts to hypertension, it disturbs daily activities and creates chronic noise stress which becomes a mediating factor for hypertension in the future”. The first sub-section focuses on any association between long-term aircraft noise exposure and chronic noise stress. The second sub-section concentrates on any association between chronic noise stress and prevalence of hypertension in adult. First, the analysis was performed to develop a model to predict the presence/absence of chronic noise stress based on aircraft noise exposure and significant confounding factors. According to the Wald criterion, aircraft noise exposure, noise sensitivity, traffic noise annoyance, aircraft noise annoyance, and interaction between traffic noise annoyance and aircraft noise annoyance reliably predicts chronic noise stress. Second, the analysis was performed to develop a model to predict the presence/absence of prevalence of hypertension in adults based on chronic noise stress and significant confounding factors. According to the Wald criterion, chronic noise stress, high cholesterol status, age, and history of hypertensive in parent(s) reliably predicts prevalence of hypertension in adults. The Hosmer-Lemeshow goodness-of-fit statistic reveals that both models are good fits. Finally, an answer for the second core research question would be “Subjects (aged 15 – 87) who have been chronically exposed to high aircraft noise level have the odds of 2.61 of having chronic noise stress compared with the control group. Subjects who suffered from chronic noise stress have the odds of 2.74 of having hypertension compared with those without chronic noise stress”.

While recognising the limitations of descriptive epidemiology of cross-sectional design applied in this research, this thesis does not intend to establish any causality between aircraft noise and disease. The main contribution of this thesis is the establishment of robust hypotheses of effects of aircraft noise on community health and well-being for

iv the future experimental study: first, “long-term aircraft noise exposure deteriorates health related quality of life of community”; and second, “long-term aircraft noise exposure has indirect effects to hypertension in adult via chronic noise stress as a mediating factor”. This thesis proposes that the best way to counter aircraft noise problems is to implement the aircraft noise management strategy in a more sustainable manner. The priority to protect community health and well-being from aircraft noise exposure should be given first before the emerging of knowledge from the future experimental study. By encouraging the policy maker(s) to interpret the meaning of ‘health’ in a more broad way, the effects of aircraft noise on community health and well-being should be considered as a major issue in developing aircraft noise management strategy.

v ACKNOWLEDGEMENTS

I would like to express my heartfelt gratitude to the following persons who gave their support to this thesis. Without them, my thesis would never have been successful.

Thank you very much Professor John Black for your invaluable supervision, general encouragement and comprehensive review of this thesis. You are a great supervisor.

Thank you very much Dr. Stephen Samuels for your generous support, intelligent advice, critical review on Chapter 5, and comprehensive review of this thesis during the examination process. I very much enjoyed your Environmental Noise course.

Thank you very much Associate Professor Deborah Black for your kind support, comprehensible advice and excellent review on Chapters 6 and 7 of this thesis. I would say that you are the best statistics teacher I have seen.

Thank you very much Julie O’Keeffe, Less Brown, and Pattie McLaughlin for supporting me the facilities during my data collection.

Thank you very much David Southgate from Commonwealth Department of Transport and Regional Services, Leigh Kenna, Ken Owen, and Gavan Bennett from Airservices Australia for providing me valuable information about aircraft noise exposure and aircraft noise contour map around Sydney Airport.

Thank you very much to my dear wife, Nisakorn Issarayangyun, for your love and support that I always receive.

Finally, I wish to express my deep gratitude to my father, my mother and my brothers who always love, support, and inspire me for every moments of doing this thesis.

vi LIST OF ABBREVIATIONS

ABS Australian Bureau of Statistics AES Airport Environmental Strategy ANEC Australian Noise Exposure Concept ANEF Australian Noise Exposure Forecast ANEI Australian Noise Exposure Index ANCOVA Analysis of Covariance ANOVA Analysis of Variance ANMS Aircraft Noise Management Strategy AS Australian Standard CD Census Collection District CV Covariate Variable DNL Day-Night Average Sound Level DoTARS Department of Transport and Regional Services Commonwealth DV Dependent Variable EIS Environmental Impact Statement ES Environmental Strategy FAA Ferderal Aviation Authority FICAN Federal Interagency Committee of Aviation Noise GH General Health Scale HRQoL Health Related Quality of Life ICAO International Civil Aviation Organization INM Integrated Noise Model IV Independent Variable LTOP Long Term Operating Plan MH Mental Health Scale N70 Number Above 70 dB(A) NA Number Above Index NAL National Acoustic Laboratory NAP Noise Abatement Procedure NASA National Aeronautics and Space Administration NFPMS Noise and Flight Path Monitoring System NGI Noise Gap Index NHP Nottingham Health Profile NMP Noise Mitigation Procedure NMS Noise Management Strategy NMT Noise Monitoring Terminal NSCF Noise Source Classification Form MANCOVA Multiple Analysis of Covariance MANOVA Multiple Analysis of Variance

vii PF Physical Functioning Scale SACF Sydney Airport Community Forum SANIP Sydney Airport Noise Insulation Program SEIFA2001 2001 Socio-Economic Indexes for Areas SF-36 Short Form Health Survey (36-Items) SIP Sickness Impact Profile SLA Statistical Local Area SPL Sound Pressure Level TNIP Transparency Noise Index Program VT Vitality Scale WHO World Health Organization

viii TABLE OF CONTENTS

ABSTRACT i

ACKNOWLEDGEMENTS vi

LIST OF ABBREVIATIONS vii

TABLE OF CONTENTS ix

LIST OF TABLES xi

LIST OF FIGURES xiii

CHAPTER 1 INTRODUCTION

1.1 Research Problem 1 1.2 Objectives, Scopes, Assumptions, Methodologies, and Limitations 5 1.3 Structure of Thesis 12

CHAPTER 2 LITERATURE REVIEW

2.1 Introduction 14 2.2 Environmental Noise 15 2.3 Aircraft Noise 28 2.4 Health and Well-Being Impacts by Aircraft Noise 48 2.5 Research Gaps 91 2.6 Statistical Techniques 95 2.7 Conclusion 98

CHAPTER 3 AIRCRAFT NOISE MANAGEMENT AND COMMUNITY HEALTH AND WELL-BEING IMPACTS

3.1 Introduction 99 3.2 Aircraft in Australia 100 3.3 Policy for Aircraft Noise Management in Australia 105 3.4 Aircraft Noise Management at Major Commercial Airports in Australia 110 3.5 Aircraft Noise Management in Other Countries 120 3.6 Community Health and Well-Being Issue in 129 Aircraft Noise Management Strategy 3.7 Conclusion 130

CHAPTER 4 HEALTH SURVEY METHODOLOGY, PILOT STUDY, AND CASE STUDY

4.1 Introduction 132 4.2 Epidemiological Method 133 4.3 Proposed Health Survey Procedures 135

ix 4.4 Pilot Study 157 4.5 Study Population 163 4.6 Sample Size 171 4.7 Translation 173 4.8 Conclusions and Discussions 176

CHAPTER 5 DEVELOPMENT OF NOISE GAP INDEX (NGI)

5.1 Introduction 179 5.2 Noise Investigation 180 5.3 Noise Measurement Procedures 181 5.4 Typical Resulting Data 185 5.5 Analysis of Noise Data 187 5.6 The Noise Gap Index (NGI) 195 5.7 A Case Study 196 5.8 Conclusions and Discussions 201

CHAPTER 6 AIRCRAFT NOISE AND HEALTH RELATED QUALITY OF LIFE

6.1 Introduction 203 6.2 Preliminary Data Analysis 204 6.3 Exploring of the First Core Research Question 209 6.4 Summary, Discussions and Concluding Remarks 224

CHAPTER 7 AIRCRAFT NOISE AND HYPERTENSION

7.1 Introduction 230 7.2 Exploring the Second Core Research Question 231 7.3 Summary, Discussions and Concluding Remarks 248

CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS 252

8.1 Conclusions 253 8.2 Recommendations 256

REFERENCES 258

APPENDICES

Appendix A Questionnaire and Contacts Letters A-1 Appendix B Translated Questionnaire and Contacts Letters B-1 Appendix C Description of Variables C-1 Appendix D Data Screening D-1 Appendix E Analysis of Covariance E-1 Appendix F Linearity Test and Test of Homogeneity of Regression of F-1 First Core Research Question Appendix G Logistic Regression Analysis G-1

x LIST OF TABLES

Table 2.1 Relation of Cut-Off Frequencies and Octave Band and 18 1/3 Octave Band Table 2.2 Summary of Previous Studies Relevant to Core Research Questions 94 Table 3.1 Movements at Australian Airports arranged in Alphabetical Orders 112 2003 Calendar Year Totals Table 3.2 Twenty Two Major Commercial Airports in United States 123 Table 3.3 Aircraft Noise Management Procedures of Twenty-Two Major 124 Commercial Airports in United States Table 3.4 Three Busiest Commercial Airports in Canada 125 Table 3.5 Aircraft Noise Management Procedures in Three Busiest Airports 126 in Canada Table 3.6 Three Busiest Commercial Airports in England 127 Table 3.7 Aircraft Noise Management Procedures of Three Busiest Airports 128 in England Table 4.1 Core Concepts and Domains of HRQoL 139 Table 4.2 Major Concepts of HRQoL contained in Six Well-Known 140 Generic Measures Table 4.3 Interpretation of Lowest and Highest Scores of 141 Selected SF-36 Scales Table 4.4 Advantages & Disadvantages of Survey Administrative Techniques 148 Table 4.5 Return Rate of the Proposed Health Survey Procedures 158 Table 4.6 Number of Subject Classified by Completed and Missed Items 159 Table 4.7 Reliability Analysis (Alpha) of Proposed Noise Stress Scale 160 Table 4.8 Reliability Analysis (Alpha) of Proposed Noise Sensitivity Scale 161 Table 4.9 Options of Aircraft Noise Non-Exposure Areas with CD Codes 168 Table 4.10 Mann-Whitney Test of the Comparison of 2001 SEIFA Indexes 170 Between the Aircraft Noise Exposure Area and the Options for the Control Area Table 4.11 Sample Size Needed per Group to Detect Point Difference between 172 Two Non-Experimental Groups, Repeated Measure Design Table 4.12 Three Most Non-English Languages Spoken at Home of 175 Selected Study Population Areas Table 5.1 Noise Stations 181 Table 5.2 Time Average A-Weighted Sound Pressure Level of Background 190 B Environmental Noise (L Aeq,Tk) and Time Interval (Tk) at Each Noise Station A Table 5.3 Time Average A-Weighted Sound Pressure Level (L Aeq,Tk) and 194 Average Hourly Number Above 70 dB(A) (N70) of Aircraft Noise at Each Noise Station Table 6.1 Return Rate by Study Groups 205 Table 6.2 Responded Subjects by Language 205 Table 6.3 Demographic Characteristics and Socioeconomic Status by 206 Study Groups Table 6.4 Descriptive Statistics of Health and Related Measures by 208 Study Groups Table 6.5 Assessing the Significance of Secondary Independent Variables 212

xi for Physical Functioning Table 6.6 Factorial ANCOVA of SQRT(k - PF) and 213 Aircraft Noise Exposure Table 6.7 Assessing the Significance of Secondary IVs for General Health 215 Table 6.8 Factorial ANCOVA of General Health and 215 Aircraft Noise Exposure Table 6.9 Assessing the Significance of Secondary IVs for Vitality 217 Table 6.10 Factorial ANCOVA of SQRT(k – VT) and 219 Aircraft Noise Exposure Table 6.11 Assessing the Significance of Secondary IVs for Mental Health 221 Table 6.12 Factorial ANCOVA of Mental Health and 223 Aircraft Noise Exposure Table 7.1 Categorical Variables Coding 233 Table 7.2 Univariable Logistic Regression Models of the First Analysis 234 Table 7.3 Results of Fitting a Multivariable Model of the First Analysis 235 Table 7.4 Fractional Polynomials for Noise Sensitivity 236 Table 7.5 Fractional Polynomials for Aircraft Noise Annoyance 238 Table 7.6 Fractional Polynomials for Traffic Noise Annoyance 239 Table 7.7 Main Effects Model of the First Analysis 239 Table 7.8 Interaction Test for the Main Effects Model of the First Analysis 240 Table 7.9 Preliminary Final Model of the First Analysis 240 Table 7.10 Contingency Table of Hosmer and Lemeshow Test and 241 Chi-Square Value of the First Analysis Table 7.11 Univariable Logistic Regression Models of the Second Analysis 243 Table 7.12 Results of Fitting a Multivariable Model of the Second Analysis 244 Table 7.13 Adding, Refitting, and Deleting Process of the Second Analysis 245 Table 7.14 Fractional Polynomials for Age 246 Table 7.15 The Main Effects Model of the Second Analysis 246 Table 7.16 Interaction Test for the Main Effects Model of the Second Analysis 246 Table 7.17 Contingency Table of Hosmer and Lemeshow Test and 247 Chi-Square Value of the Second Analysis

xii LIST OF FIGURES

Figure 2.1 How Noise Effects Humans 15 Figure 2.2 The Generation of Sound Waves 16 Figure 2.3 Effects of Wind and Temperature on Sound Rays 23 Figure 2.4 Sources of Adverse Affection by Environmental Noise in Sydney 24 Figure 2.5 Typical Designs and their Applications 29 Figure 2.6 Sound Radiation Patterns of Internal Engine Sources 31 Figure 2.7 Half-Section of Engine Front-End, Showing Relevant Blading 32 Figure 2.8 Example of Typical Noise Time Curve 42 Figure 2.9 Required Noise Measurement Points for Aircraft Noise Certification 43 Figure 2.10 Structures of the Human Ear 49 Figure 2.11 Cross Section of the Cochlea, and Enlargement of Organ of Corti 51 Figure 2.12 Free-Field Equal-Loudness Contours 53 Figure 2.13 The Relationships of Blood Vessels According to Size and the 54 Direction of Blood Flow Figure 2.14 Speech Levels as a Function of Background Noise Level 58 Figure 2.15 Activities Interference by Transportation Noise 60 Figure 2.16 Relationship between Sleep Disturbance and Indoor A-Weighted 67 Sound Exposure Level (ASEL) Figure 2.17 Sleep Disturbance by Aircraft Noise Dosage Response Relationship 69 Figure 2.18 Relation of Noise Induced Permanent Threshold Shift at 4kHz as a 70 Functions of Noise Exposure Level (Leq(8h)) and Duration of Exposure Figure 2.19 Opinion Thermometer 83 Figure 2.20 Dosage-Response Relation of Aircraft, Road Traffic, and 86 Railway Noise Figure 2.21 Relationship between Low-Frequency Aircraft Noise Levels and 86 Prevalence of Annoyance due to Vibrations or Rattling Sounds Figure 2.22 Modifications in Behaviour by Noise Level 91 Figure 2.23 Choosing among Statistical Techniques 96 Figure 3.1 Relationship between ANEF and Community Reactions in 102 Residential Areas Figure 4.1 Correlation among Exposure, Disease, and Confounding Factor 146 Figure 4.2 Proposed Health Survey Procedures with Time Diagram 150 Figure 4.3 Reply-Paid Return Envelope 150 Figure 4.4 Proposed Cover Letter 152 Figure 4.5 Proposed First Follow-Up Letter 153 Figure 4.6 Proposed Second Follow-Up Letter 155 Figure 4.7 Proposed Third Follow-Up Letter 156 Figure 4.8 Health Survey Procedures with Time Diagram 162 Figure 4.9 Comparison of Sydney Airport’s Track Plots Coloured by Height 164 For Jet Arrivals during the Period 2/12/2003 – 8/12/2003 and the Period 2/6/2004 – 8/6/2004 Figure 4.10 Study Population for the Aircraft Noise Exposure Area 165 Figure 4.11 SLA Map of Study Population of Aircraft Noise Exposure Area 166 Figure 4.12 Locations in Sydney of Aircraft Noise Exposure Area and 171 the Control Group Figure 5.1 Noise Source Classification Form 183

xiii Figure 5.2 Typical Noise Time Curve at NS-5 during a Peak Morning Period 186 of Aircraft Landings Figure 5.3 Noise Time Curve at NS-5 during a Morning Period with 187 No Aircraft Over Flights Figure 5.4 Background Environmental Noise Time Curve of Figure 5.2 188 Figure 5.5 Background Environmental Noise Time Curve of Figure 5.3 188 B Figure 5.6 Background Environmental Noise Levels (L Aeq,Tk) for High, 189 Medium, Low, and Control Noise Groups Figure 5.7 Aircraft Noise Time Curve of Figure 5.2 191 A Figure 5.8 Relationship between L Aeq,7am-6pm and N70 for All Data in the 193 High, Medium, and Low Background Environmental Noise Groups Figure 5.9 Relationship between NGI and N70 for High, Medium, and Low 196 Background Environmental Noise Groups Figure 5.10 Daily Average Number of Aircraft Noise Events Louder than 197 70 dB(A) at Sydney Airport During 1 January – 31 December 2003 and Study Population of Aircraft Noise Exposure Area Figure 5.11 Daily Average N70 During 1 January – 31 December 2003 at 198 Study Area and Locations of Noise Station Figure 5.12 Relationship between Aircraft Noise Annoyance Scale and N70 200 Figure 5.13 Relationship between Aircraft Noise Annoyance Scale and NGI 201 Figure 6.1 Population Regression Lines under Two Hypotheses: 207 (a) α 1 = α 2 = … = α a ; and (b) β1 , β2 , …, βa = 0 Figure 7.1 Assumption of the First Core Research Question 230 Figure 7.2 Plot of Estimated Coefficients of Noise Sensitivity vs. 236 the Quartile Midpoints Figure 7.3 Plot of Estimated Coefficients of Aircraft Noise Annoyance vs. 237 the Quartile Midpoints Figure 7.4 Plot of Estimated Coefficients of Traffic Noise Annoyance vs. 238 the Quartile Midpoints Figure 7.5 Plot of Estimated Coefficients of Age vs. the Quartile Midpoints 245

xiv CHAPTER ONE

INTRODUCTION

1.1 RESEARCH PROBLEM

The development of major commercial airports promotes the air transport industry and generates positive economic benefits to the airport and to its host economy. However, external costs are associated with these benefits. Any increase in aircraft movement causes negative environmental impacts, especially noise pollution. Many efforts have been made in order to reduce aircraft noise levels at their sources (NASA, 1999) (ICAO, 1993), or have installed acoustic insulation on houses (and other properties) around airports (see section 3.4 and 3.5). Nevertheless, the problems have never been satisfactorily resolved as long as the growth of the air transport market remains in high demand, as it appears to be at the start of the 21st Century. The abandonment of a commercial airport from the metropolis is not practical as it is an important infrastructure in promoting economy of overall area although Halim Airport, Jakarta, and Kai Tak Airport, Hong Kong, are exception. Jatmika (2001) proposed a technique (called Impact Management System) to reconcile the inherent conflicts between positive economic benefits and negative environmental impacts from the airport development using Sydney (Kingsford Smith) Airport as a case study. The intensive land-use pattern change around the airport has been surveyed, but the assessment of impacts of aircraft noise on community health and well-being around the airport has still not been undertaken.

In general, the negative impacts of aircraft noise on the society can be categorised into two aspects: economic impacts and environmental impacts. While the former aspect concentrates on the depreciation rate of property values (for example, house, land, and building) with aircraft noise, the latter aspect focuses on the impacts of aircraft noise on community health and well-being. This thesis is concerned with the environmental impacts. Specifically, it aims to study the impacts of aircraft noise on community health

1 and well-being. The economic impacts are not addressed in this thesis. This is primarily because previous researches (De Vany (1976) as stated in Kryter (1994, pp.622-624), Pennington et al (1990), Uyeno et al (1993), Levesque (1994)) revealed that the depreciation rate of property values is not perceptibly influenced by aircraft noise alone. Many factors are involved in this issue. For example, the demand of airport workers, people who use the airport on a frequent basis, or workers in certain industries relating to airport operation to live nearby the airport may increase property values. It is also possible that the declining of property values can be influenced by the impacts of aircraft noise on community well-being. Once people feel uncomfortable, or annoyed about aircraft noise, they will typically try to avoid staying or living in these aircraft noise-affected areas.

The current professional practice of developing an aircraft noise management strategy at major commercial airports aims to minimise, as far as practicable, the total number of people in the community exposed to high levels of noise from overflights. It also aims to remedy, as much as possible, the significant aircraft noise exposure in existing noise- sensitive areas. However, the incorporation of the issue of community health and well- being impacts by aircraft noise into the aircraft noise management strategy is not apparent at present. This might reflect the fact that there does not appear to be any good evidence to support the proposition that aircraft noise is loud enough to significantly deteriorate the hearing ability of people living around commercial airports. Furthermore, current policies to protect the environmental conditions from noise pollution relevant to airport operations seem to define health only in terms of the absence of disease.

However, there is some evidence that aircraft noise potentially disturbs (or annoys) the daily activities (such as communication, relaxation, and sleep) of residents living in the vicinity of airports, especially major commercial airports (Bronaft, 1998) (Stansfeld and Matheson, 2003). This particular type of annoyance undermines quality of life and can be a cause of chronic emotional stress. Evidence is emerging that appears to associate some forms of health risk with this stress (Berglund et al., 1999).

In term of physiological effects, there is evidence that exposure to noise from unexpected sources (or unknown characters) or from an unavoidable source may evoke

2 several kinds of reflex responses (WHO, 1980, p.58). These reflex responses cause a stress reaction (called emotional stress) (Spreng, 2000 and 2004). Chronic suffering from stress potentially may lead to several health problems, especially heart and circulatory systems. Berglund et al (1999) also stated that cumulative annoyance from everyday aircraft noise disturbances has long-term effects on the human non-auditory system, especially physiological effects. Noise has been defined as an environmental stressor (Cohen et al., 1986). Exposure to sudden, or uncontrollably intense, noise activates the autonomic and hormonal systems, leading to temporary changes, such as increased blood pressure, increased heart rate and vasoconstriction. After prolonged high noise exposure, susceptible individuals in the general population may develop permanent effects (persistent increase in stress hormone level which leads to blood clotting which is a cause of high blood pressure) which lead to hypertension and cardiovascular diseases. This evidence leads this research to include hypertension as a possible health impact by aircraft.

Furthermore, the World Health Organization (Berglund and Lindvall, 1995) declared that “Health is not only the absence of disease but also including a state of complete physical, mental, and social well-being”. Therefore, it is reasonable to hypothesise that aircraft noise has effects on human health in term of health related quality of life. They went on to argue that policies to prevent environmental conditions from aircraft noise should consider the total aspects of health, as provided by the above WHO definition, and also include the physiological effects from aircraft noise via emotional stress due to noise (called noise stress). While it can be argued that the effects of aircraft noise on community health and well-being would be automatically controlled if the number of people highly exposed to aircraft noise is minimised, the effects of aircraft noise on community health and well-being should not be underestimated (or overlooked) by deficiencies in any interpretation of the meaning of ‘health’. This thesis argues that a highly effective way to encounter aircraft noise problems is to understand fully their effects on the community before devising strategies and counter-measures.

Previous studies have paid much attention to the dose-response relationship and impacts of aircraft noise on the health of school children, especially in term of blood pressure level. Unfortunately, only a few publications have mentioned the impacts of aircraft

3 noise on adults in term of either health related quality of life or blood pressure level. The epidemiological study of adult blood pressure level impacted by aircraft noise by considering stress as a mediating factor is rare, and there is only one study (by Meister and Donatelle, 2000) that examined the impacts of aircraft noise on health related quality of life.

To improve the explanation of community reaction towards aircraft noise, a suitable noise index is also required. The goal of an aircraft noise index (or measurement) is to determine and assess all the dimensions of aircraft noise related to human responses. The current practice of aircraft noise measurement at most of the major commercial airports is based on the concept of the principle of energy equivalence (see Chapter 3). This technique is suitable in developing a noise contour map used for land-use planning and noise management around airports. However, inevitably, this technique rather complicated so that it is less comprehensible and less accessible to members of the community (SSC, 1995).

An alternative way is to use a concept of transparency in aircraft noise information. The transparent aircraft noise concept is based on the concept of the peak level index and ‘everyday talk’ information, such as where the aircraft fly, how often, and at what time. A potential example of the transparent aircraft noise concept is the NA index, or Number of Aircraft Noise Events louder than a certain threshold noise level, which is commonly 70 dB(A) (called N70), over a given time period. The Australian Department of Transport and Regional Services (DoTARS) have declared that the N70 gives a much more realistic picture of aircraft noise to the community than the energy equivalent system (DoTARS, 2002). The N70 contour maps are prepared for many major commercial airports in Australia. The NA has become more widely involved in providing aircraft noise information to the community, such as in the Environmental Impacts Statement of the Second Sydney Airport, than the Australian Noise Exposure Forecast (ANEF) system. However alternative indices need to be explored to suit the aims of the present research.

Aircraft noise is a major environmental problem at airports (especially major commercial airport). The relocation of a commercial airport from the metropolis to

4 prevent environmental impacts from airport noise is usually not practical. Moreover, commercial airports are important contributions to the economy of the metropolis in which they are situated. Many efforts have been paid to alleviate aircraft noise problems in order to balance the negative impacts on environment with the positive economic effects. The conclusions of the introductory remarks in this section are that there are three main research problems that need to be considered:

1. The deficiency in the interpretation of the meaning of ‘health’ as just the absence of disease.

2. The lack of comprehensive epidemiological study of health related quality of life impacts by aircraft noise and study of association between high blood pressure and aircraft noise with emotional stress as a mediating factor.

3. The need for ‘new’ noise index that may be applied to describe and assess aircraft noise in such a way that is easily understood by members of the community.

The following section points out the objectives of this research and the methods adopted to solve the above research problems.

1.2 OBJECTIVES, SCOPE, ASSUMPTIONS, METHODOLOGIES, AND LIMITATIONS

1.2.1 Objectives of the Present Study There are two main research objectives of the present study.

1. To fill the gaps of knowledge by exploring the following two core research questions. • “Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?” • “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”

5 2. To improve the manner in which aircraft noise may be described and analysed by developing a ‘new’ easier-to-interpret aircraft noise index.

1.2.2 Scope of Research In addressing the above objectives, the scope of the research program was set as follows.

1. The research focused on noise generated from aircraft taking off and landing, and excluded airport industry noise generation including that of taxiing aircraft and ground running.

2. The impact of aircraft noise on human health was limited to non-auditory effects.

3. The study samples were located on local traffic roads in metropolitan Sydney. Due allowance was made of the major non-aviation noise sources, such as heavy traffic on main roads.

1.2.3 Research Assumptions In order to progress the research program, two fundamental assumptions were made.

1. Aircraft noise has indirect health impacts to the community. Aircraft noise annoys people by disturbing in their daily activities and then creates stress which finally becomes a mediating factor of health problems in the future.

2. People living in different background environmental noise areas might have different responses to the same aircraft noise level.

1.2.4 Research Methodologies To achieve the research objectives of this present study, a transdisciplinary framework for research problem has been employed. “Transdisciplinary thinking is primarily a process of assembling and mapping the possible interconnections of disciplinary knowledge about any given health problem until the fullest possible understanding of the problem emerges” (Albrecht et al, 2001, p.75). Acquiring knowledge about a

6 substantial transport problem requires a transdisciplinary mode of thinking. The present research has applied this particular framework in order to understand process and change and to create the richest possible description of the context within which the problem – in this case that of aircraft noise – occurs.

Albrecht et al (2001, pp. 80-81) have identified seven key stages when conducting transdisciplinary research.

1. Problem identification.

2. Assemble a group (or network) of researchers with the necessary skills to offer a perspective on the problem.

3. Review existing knowledge on the problem area to exhaust all disciplinary and interdisciplinary conceptualisations and explanations of the problem.

4. Design research enquiry from research gaps identified in (3).

5. Implement research enquiry.

6. Review conceptual understandings and synthesise data sets, including the search for a common conceptual framework that illuminates the problem and provides maximum explanatory power.

7. Specify types of intervention (often with a network of local stakeholders) to resolve the problem.

Based on the above seven key stages, the research methodologies of the present thesis were set up as follows.

1. Identifying the research problems.

As stated in Section 1.1, there are three main research problems addressed in the thesis: (1) the deficiency in the interpretation of the meaning of ‘health’; (2) the lack of comprehensive epidemiological study of community health and well-being

7 impacts by aircraft noise; and (3) the need for ‘new’ noise index that describe and assess aircraft noise in a easy way to understand by community.

2. Assembling a group of researchers.

A small research group was established. Research was undertaken by a doctoral student (the author of this thesis), supervised from the Medical and Engineering Faculties of UNSW and supported by translators from South Sydney Area Health Services. This core research team did not work in isolation from others as the standard review committees established in the School of Civil Engineering for doctoral candidate progress made suggestions on the research proposal, and Faculty Ethics and Occupational Health and Safety Committees approved of details of the survey instrument on behalf of the University of New South Wales.

3. Reviewing existing knowledge on the effects of aircraft noise on community health and well-being.

Studying impacts of aircraft noise on community health and well-being requires an understanding of epidemiology, social surveys, characteristics of environmental noise (especially aircraft noise), and effects of environmental noise on community. It is also important to understand the policies to protect the environmental conditions from noise pollution relevant to airport operations because difference in policies will shape different patterns of noise exposure around the airports.

Basically, aircraft noise disturbs community daily activities (for example, watching TV, listening to radio, sleeping, conversation, or studying). The reactions of people to those disturbances are different. Most people are annoyed by those disturbances. Some of them can habituate (or get use to it), or even avoid, or modify their activities in these noisy places. Unfortunately, in a susceptible group, the intrusion of noise into their house makes them angry and stressful. Moreover, suffering from chronic stress can lead to health problems which can be either physiological or psychological (Berglund and Lindvall,

8 1995). Two core research questions have been proposed by this thesis according to an assessment of the gaps in previous research (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?” and “Does long-term aircraft noise exposure associate with adult high blood pressure level via chronic noise stress as a mediating factor?”).

4. Reviewing the existing aircraft noise management strategies of major commercial airports in Australia and in other countries with particular reference to community health and well-being.

To better understand the process of aircraft noise management in Australia, this thesis has reviewed all the policies and regulations which have been enacted to protect the environmental conditions from noise pollution relevant to airport operations in Australia. The review also included the current practice of aircraft noise management strategies, or plans, of major commercial airports in Australia and in other developed countries, such as United States, Canada, and England. The major sources of data were collected from the official website of each airport.

5. Developing a comprehensive survey instrument for the evaluation of community health and well-being impacts by aircraft noise using a valid and standardised health measure scale and implementing a social survey.

To measure community health and well-being impacts by aircraft noise according to the core research questions, a comprehensive health survey instruments was developed. Health-related quality of life is a subjective health assessment which could be measured effectively by a well-designed questionnaire. The prevalence of adult high blood pressure level could be measured either by a well-designed questionnaire or by automated blood pressure measuring equipment. The latter method seems to provide more accuracy and reliability, but inevitably requires much more time and budget, than the former method. On the other hand result from blood pressure measurements could also be distorted if the individual is on medication that controls blood pressure. By carefully considering the feasibility of both methods, this thesis decided to select a subjective assessment to measure the

9 prevalence of adult high blood pressure level. No medical laboratories, or experimental tests on people, have been undertaken in this research.

This research developed a questionnaire (based on an existing well-established questionnaire instruments) to measure seven major characteristics of each subject: 1) health related quality of life; 2) prevalence of hypertension; 3) noise stress; 4) noise sensitivity; 5) noise annoyance; 6) confounding factors; and 7) demographic characteristics. A specific question that has not been developed previously has been designed by this research. Before the commencement of the main survey, all the survey instruments including the questionnaire, the contact letters, and the administrative techniques were piloted. The pilot study was important because it helped the researcher to estimate the response rate that could used to calculate the sample size required for the main survey and to detect any errors and impractical procedures that needed to be modified so the response rate of the main survey could be maximised.

Effects from ethnicity bias (which usually occurs when undertaking a social survey in a major metropolitan area, like Sydney, where English is not the first language of a significant proportion of respondents) had to be overcome. Therefore, this thesis translated all the health survey instruments from English to the other main languages based on the most Non-English languages spoken at home information provided by 2001 Census of Population and Housing Data, Australian Bureau of Statistics. This research employed a multicultural organisation which is not only fluent in English and the target languages, but which is also knowledgeable about the content area of this research problem to accomplish all the translation processes.

This thesis selected Sydney (Kingsford Smith) Airport as a case study. The study population of the noise exposure group was identified based on long-term aircraft noise contour maps of N70 generated by Airservices Australia. The control group was located in the area where jet noise is not detected and the socio-economic status is comparable with the exposure area.

10 6. Undertaking an extensive acoustic measurement program to measure aircraft noise and background environmental noise.

To encourage the promotion of a transparent aircraft noise concept, this thesis expanded on ways to describe and assess aircraft noise by incorporating the characteristics of background environmental noise into the N70. Based on the research assumption that people living in areas of different background environmental noise may have different reactions to the same aircraft noise level, a ‘new’ noise index, which has been termed the Noise Gap Index (NGI), has been developed. This index was established so it could distinguish between aircraft noise and background environmental noise in a novel manner. To develop the NGI, this thesis implemented an extensive noise monitoring program around Sydney (Kingsford Smith) Airport. Environmental noise data were measured by using a Bruel and Kjaer Type 2236. The measurement procedures were done according to the guidelines provided in Standard Australia Acoustics (SAA 1997). During the measurements, noise sources were classified by the recorders using a Noise Source Classification Form developed specifically for this thesis.

7. Analysing health and well-being data by using a suitable multivariate statistical technique.

To explore the core research questions, knowledge of multivariate statistical analyses was required. Choice of appropriate multivariate statistical analysis techniques depends on the nature of the variables to be included in the model. This thesis selected the most suitable technique in relevance with the nature of each core research question based on the guideline provided by Tabachnick and Fidell (2001). The first core research question, which intends to compare the mean score of health related quality of life between aircraft noise exposure group and the control group, was explored by using factorial Analysis of Covariance. The second core research question, which aims to explore the prediction of the presence/absence of hypertension due to exposure to aircraft noise, was explored by using binary Logistic Regression Analysis.

11 1.2.5 Research Constraints The following points list constraints of this present study.

• Time – This research was conducted as part of requirement for the degree of doctor of philosophy. The restriction of time during the normal period of doctoral candidate is 3 years.

• Budget – This research was supported by the author of this thesis and Research Management Committee, School of Civil and Environmental Engineering, UNSW.

1.3 STRUCTURE OF THESIS

The structure of this thesis is organised into eight chapters.

This chapter (Chapter 1) has introduced the current problem of aircraft noise exposure around the airports, and the required work that shapes the directions of this thesis. Research objectives, scope of research, research assumptions, research methodologies, and research limitations are included in this chapter.

The next chapter, Chapter 2, provides an extensive literature review containing the fundamental concept of sound, characteristics of environmental noise, nature of aircraft noise (especially generation, propagation, qualification, and prediction of aircraft noise), basic knowledge of human auditory system and blood pressure regulation, impacts of aircraft noise on human including auditory effects and non-auditory effects, and responses of humans to noise. This chapter identifies the research gaps, and sets up the core research questions. Also, the techniques to handle multivariate statistical problems are reviewed and selected.

Chapter 3 is a review of the existing aircraft noise measurements in Australia, the policies to manage noise from airport operations in Australia, and the current practice of aircraft noise management strategies of major commercial airports. Similar examples are drawn from the United States, Canada, and England. A discussion of the current

12 aircraft noise management strategies in regard to the issue of community health and well-being impacts by aircraft noise is provided.

Chapter 4 designs the epidemiology study that is regarded as being most appropriate for this research topic. The development processes (including the pilot test) of health survey instruments, including questionnaire, contact letters, and survey administration are described. The chapter discusses ethical aspects of conducting this type of research. Sydney Airport has been selected as a case study. Descriptions, including the identification of study populations, the calculation process of required sample size for the main survey, and the process of translation of the health survey instruments, are provided in this chapter.

Chapter 5 develops the Noise Gap Index (NGI) and applies this index into a case study. The procedures of selecting noise stations, measuring environmental noise, and analysing the collected noise data are provided in this chapter. The chapter also describes the fundamental concepts of noise indices involved in the development of NGI.

Chapter 6 explores the first core research question (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”). The chapter also describes the response rate of the main survey, data screening, and preliminary descriptive statistical analysis. The summary and concluding remarks are provided.

Chapter 7 explores the second core research question (“Does long-term aircraft noise exposure associate with hypertension via noise stress as a mediating factor?”). The summary and concluding remarks are provided.

Chapter 8 concludes and discusses the results, suggests future implications, and recommends the further research directions.

13 CHAPTER TWO

LITERATURE REVIEW

2.1 INTRODUCTION

In accordance with a transdisciplinary framework applied by this thesis, literature relevant to the research problem was examined irrespective of its disciplinary base. The scope of the literature review on aircraft noise and community health and well-being is illustrated in Figure 2.1. The diagram indicates how noise can affect human health. In general, noise is classified into two main groups: occupational noise and environmental noise. The literature on the impacts of occupational noise on worker’s health is not considered. Environmental noise disturbs community daily activities (for example, watching TV, listening radio, sleeping, conversation, or studying). The reactions of people to those disturbances are different. Most people are annoyed by those disturbances. Some of them can habituate (or get use to it), or even avoid, or modify their activities in these noisy places. Unfortunately, in a susceptible group, the intrusion of noise into their house makes them angry and stressful. Moreover, suffering from chronic stress can lead to health problems which can be either physiological or psychological. Whilst Figure 2.1 is applicable to any environmental noise, the literature review of this thesis focuses only on aircraft noise and community health and well- being.

This chapter reviews the literature on the impacts of aircraft noise on humans. The structure of this chapter is organised in accordance with the sequence shown in Figure 2.1. Section 2.2 explains the basic concept of sound generation and propagation, and discusses the general characteristics of environmental noise. Section 2.3 discusses, in detail, the nature of aircraft noise, how to quantify and predict aircraft noise by using different computer software. Section 2.4 begins with the basic introduction to the human auditory system, blood pressure regulation, and then discusses all potential effects of aircraft noise on community. Section 2.5 discusses the research gaps and sets

14 up the core research questions. Section 2.6 discusses a method in choosing statistical techniques that are suitable for each core question.

Noise

Occupational Environmental Noise Noise

Auditory Effects Activities (Physical Impacts) Disturbance (Well-Being Impacts)

Community Responses

Annoyance Nothing

Stress Habituation

Physiological Psychological Effects Effects Figure 2.1: How Noise Effects Humans

2.2 ENVIRONMENTAL NOISE

2.2.1 Basic Acoustics Sound is generated by the physical disturbances of an object(s) in transmittable medium (i.e. air, liquid, or solid) and is received by the ear, or in an electronic device. Subsequent sections are concerned only with air as the transmittable medium and the human ear as the receiver. Figure 2.2 illustrates the generation of sound waves from a

15 source. When the vibrating object moves to the right, the adjacent air layer will be compressed causing an increase in molecule pressure higher than the successive layer. The compressed layer tends to spread away from the object and compresses the next layer creating compression. Rarefaction occurs when the object moves back to the left. The spreading out of the air layer creates variation in air pressure above and below atmospheric pressure in a sinusoidal pattern.

Figure 2.2: The Generation of Sound Waves (source: reproduced from Foreman, 1990, figure.1.1, p.2)

The atmospheric pressure has a value of approximately 105 newtons/m2 (N/m2) or Pascal (Pa) compared with the tiny variation of sound pressure (SP) range from 20µPa - 20Pa that can be detected by a healthy young adult. To represent the average amount of pressure fluctuation created by the sound source, the root-mean-square (rms) is introduced. It is the square root of the mean squared pressure during one period (T). The root-mean-square equation is (Nelson, 1987, p.2.4):

16 1 t p = ∫ p 2 (t)d(t) Pa (N/m2) T t−T where p = root-mean-square pressure (Pa) t = the actual time before the average time T.

As the range of sound pressure that can be detected by the human ear is so wide, it is not practical to identify sound in term of its root-mean-square pressures. The level system (or logarithm scale) was, therefore, introduced. Harris (1992, p.1.8) defined the term “level” as “Level is the logarithm of the ratio of a given quantity to a reference quantity to the same kind”. Initially, the unit of level was expressed by the “Bel”. However, in general, most acoustic levels are presented by the decibel (dB) when 1 Bel is equivalent to 10 decibel. The equations of sound pressure level (Lp or SPL) is presented by equation (2.1) (Harris, 1992, p.1.11). Consequently, the range of sound pressure level that a healthy young adult can be detected ranges from 20 dB to 120 dB.

p 2 p L p (SPL) = 10log( ) = 20log( ) dB (2.1) p0 p0 where po (or pref) is the reference sound pressure (= 20 µPa) or the minimum sound pressure that is heard by the healthy young adult ear in the frequency range where the ear is most sensitive.

The sound that comprises of one unvarying frequency is called a pure tone (or line spectrum) – for example, a sound from one note of a piano. The term frequency ( f ) is defined as a reciprocal of time that is required for one complete cycle of the sinusoidal curve (see Figure 2.2). However, in nature, most of the audible sound is broadband sound (or broadband spectra), which is composed of a combination of more than one tone. The human ear can detect a range of frequency from 20 – 20,000 Hz with different responses to some frequencies in the sound spectrum than to other frequencies. For example, the 80 dB SPL of 1,000 Hz is louder than the same SPL of 100 Hz but a little quieter than the same SPL of 2,500 Hz.

17 Therefore, to encounter the above complexity, the octave band system and the A- weighting method were introduced. The method of the octave band system is to assemble a wide range of sound frequency into interval groups. Each group is represented by its center frequency. The A-weighting method is the system to adjust the SPL in accordance with the human hearing response. Two SPLs that have the same A- weighted level will presumably sound equally loud, or two SPLs with different A- weighted level the SPL with higher A-weighted level will considerably louder than another one. The unit of A-weighted SPL is expressed as dB(A) and will be used throughout this thesis. Table 2.1 tabulates the values of octave band and 1/3 octave band relation to the cut-off frequencies of A-weighting system. From the previous example, the 80 dB SPL of 100, 1,000, and 2,500 Hz after adjusted by Table 2.1 will equal to 60.9, 80, and 81.2 dB(A), respectively.

Table 2.1: Relation of Cut-Off Frequencies and Octave Band and 1/3 Octave Band

(source: reproduced from Kryter, 1994, table 1.2, p.11)

2.2.2 Outdoor Sound Propagation When a sound ray (which is the imaginary line radiating from the sound source and which indicates the direction of sound) propagates from the sound source through the atmosphere to the receiver, the four sound principal properties (reflection, refraction, diffraction and diffusion) occur. Generally, this creates the decrease in sound level

18 (called sound attenuation) with increasing distance between source and receiver. Outdoor sound propagation has been the interest of many researchers for numerous purposes. The main purpose of those interested in this is to accurately understand the behaviour of sound attenuation for sound control purposes. The efficient prediction of sound attenuation will provide sound controllers the effective ways to control the unwanted sound.

The calculation of sound attenuation involves complicated mathematics and this is considered beyond the scope of this thesis. The purpose of this section is to fulfill the basic components of sound (which are sound source, sound path, and sound receiver) and to alert the readers that when considering outdoor sound propagation there are many uncontrollable factors.

The total attenuation (Atotal) is the combination of various attenuation mechanisms, predominantly spreading attenuation (As), atmospheric attenuation (Aa), ground attenuation (Ag), and miscellaneous attenuation (Am) (such as attenuation from building, attenuation from houses, and attenuation from foliage). These types of attenuation should be estimated separately because they were regarded as independent of each other. Then, all of the attenuations are summed to give total attenuation. It is noted that the Ag becomes zero when the sound is obstructed from a barrier between the source and the receiver. The sound ray is also affected by temperature and wind gradients.

In ideal condition where there is not any obstruction by any object, the sound ray is spherically radiated from the source through the atmosphere. The spreading attenuation

(As) depends on the source type and the propagation distance. Basically, the sound sources are classified into three types: plane source, , and point source. The plane source rarely occurs in the environment. It produces the sound wave propagating with the constant wave area per unit of time. An example of a plane source is the moving of a piston in a tube. The line source may be considered as the continuous sound radiation such as a pipeline or the combination of a large number of point sources closely spaced, for instance, a congested highway. A sound source can be considered as a point source if its size is small compared with its wavelength and distance to receiver(s). Examples of point sources in the environment are aircraft, individual

19 vehicles, and industrial plants. The spreading attenuation formula is given as (Sutherland and Daigle, 1996) (as stated in Crocker et al, 1996, p. 341):

r2 As = 20 g log( ) dB (2.2) r1 where g is zero for plane source, g = 0.5 for line source, and g = 1 for point source r2 and r1 are distances between the source and the farthest (i.e., receiver) and the closest positions.

Equation (2.2) implies that As does not depend on frequency of sound and, for example, for the point source the As will increase 6 dB per doubling of distance. Hence, the spreading attenuation is important for short propagation distances. However, for long distance sound propagation such as overflight aircraft noise, the As is less important compared with the atmospheric attenuation (Aa).

During the propagation of sound through the atmosphere, the sound is absorbed by two factors: (1) classical losses by heat conduction and shear viscosity and (2) molecular relaxation losses by oxygen and nitrogen excitations. These losses depend on temperature, atmospheric pressure, humidity content, and sound frequency. The exact prediction of atmosphere attenuation is complicated due to the fluctuation of related parameters. The development of formula to predict air absorption is complicated, and therefore beyond the scope of this study, but more details can be found in ISO 9613- 1:1993 and ANSI S1.26-1978.

Sutherland and Daigle (1996) (as stated in Crocker et al, 1996, pp. 341-343) also provide a good explanation in determining the Aa. They concluded that the Aa depends largely on two factors: sound frequency and propagation distance. Sound with higher frequency will be more absorbed by the air than the lower frequency. The broadband sound will be less absorbed by the air than the pure tone.

When a sound ray is transmitted by a sound source, some (called direct sound) will directly propagate to the receiver but some (called indirect sound) will reflect from the ground and then reach the receiver. This indirect sound is assumed to be propagating

20 from the image source under the ground surface at the same height as the source height. The ground attenuation is a result of the interference between direct and indirect sound rays. The ground attenuation rates depend on the path length difference between the direct and indirect sound, the grazing angle, characteristic of ground, and sound frequency.

The characteristic of the ground is the most complex part of the ground attenuation issue. It depends on many factors, such as the surface type (concrete, grass or plain soil), feature (flat or rough), the porosity and density of the ground, and the characteristic impedance and the flow resistivity of the sound. There have been many attempts to develop the ground attenuation prediction model that involves experimental data and advanced mathematical models. The consensus of ground attenuation research is that the higher sound frequencies have a greater ground attenuation rate than the lower sound frequencies. The greater propagation distances will cause higher ground attenuation. The harder surfaces, with high flow resistance and low porosity (i.e., asphalt or concrete pavement, water, or tamped ground), will make ground attenuation lower than the softer one with low flow resistance and high porosity (i.e., grassing land, farming land, snow, or loose soil). Finally, the sound ray with a grazing angle more than about 15o will not be affected by the ground attenuation and will be called air-to-ground propagation (ANSI S1.26-1978). For example, the propagation of a sound from an overflight aircraft to people under a designated flight path.

The previous three attenuations are considered in a free field condition with no obstruction from objects. In the real environment, the field is combined with a lot of components, such as houses, buildings, forests, bushes and so on. All of these effect sound propagation. The sound attenuation from a building or a vertical construction might be calculated by the ground attenuation concept if reflection is concerned. It also might be calculated by the thick barrier concept if diffraction is concerned. The problem of attenuation due to foliage is predominant when the sound ray propagates through dense forest or farming (or grass) lands. The former can be considered as the attenuation by a barrier and the later can be considered as attenuation by the ground. The roots of the tree also provide some ground attenuation by making the soil porous. However, one tree or light bush gives very little attenuation, and might be neglected. The higher sound

21 frequencies can be more attenuated by the foliage than the lower frequencies. The attenuation due to a housing area such as that encountered when undertake field work for this thesis is depended on the density of housing area (i.e., suburban or urban areas) and unlike the attenuation due to foliage it is independent of sound frequency. More details of the Am can be observed from Harris (1992, pp.3.7-3.10).

2.2.3 Effects of Wind and Temperature Wind speed in general atmospheric condition is not constant: it also gradually increases with the height from the ground because of its viscous property. The wind velocity varies layer by layer. The sound ray speed at the downwind propagation increases as the summation vector between the wind velocity and the velocity of original sound ray. Conversely, the sound ray speed will be decreased when it propagates upwind against the wind velocity. Figure 2.3(a) illustrates the basic feature of sound ray direction refracted at the ground level. The sound direction is pushed forwards and bends downwards to the ground in the downwind direction. However, it will be retarded upward, away from the ground, creating a shadow region when the sound direction is upwind. The shadow region is the area where the sound intensity is reduced. The magnitudes of sound intensity change depend on the rates of change of wind speed. The sound ray direction will not be refracted by wind gradient if the direction path is perpendicular to the wind direction.

In a still uniform atmosphere, during the daytime, temperature decreases with height from the ground. This is called the normal lapse rate. From Figure 2.3(b), when the wind gradient is not considered, the sound ray direction is curved upward from the ground creating a shadow region around the sound source. The shadow region also depends on the strength of temperature gradient. Sometimes, the temperature increases with height from the ground up to the point where it reverts to the normal lapse rate. This is called a temperature inversion. This situation generally occurs in the evening when the sky clears and the ground rapidly cools down. The temperature inversion will bend downward the sound ray direction forward into the ground as shown in Figure 2.3(c). The shadow region does not appear, and the sound level at the point of reception will be very high during this circumstance. Unlike the atmospheric attenuation, the wind and temperature attenuation does not depend on the noise frequency.

22

Figure 2.3: Effects of Wind and Temperature (a) reflection of sound ray by wind, (b) a normal lapse rate, and (c) an inverted lapse rate (source: reproduced from Foreman, 1990, figure 4.17, p.96 and Foreman, 1990, figure 4.19, p.97).

2.2.4 Environmental Noise Characteristics Noise is defined as unwanted sound. Environmental noise can be described as the noise which is generated by a variety of sources and then propagates outdoors through the air to receivers who may be positioned either indoors or outdoors. Environmental noise is mainly classified into four groups: transportation noise, industrial noise, outdoor events noise, and community noise. Transport noise is of direct concern to this research, but other groups are relevant when explaining the ambient noise levels experienced in a residential neighbourhood, and therefore are reviewed here.

2.2.4.1 Transportation Noise The rapid growth of transport demand stimulates the development of transport networks that enhance the frequency and accessibility of the traveller. Transport noise becomes a

23 serious environmental problem (Berglund et al, 1999, p.24) because of its potential to intrude through the living and working environments. This literature review subdivides transport noise into three groups: road traffic noise, train noise, and aircraft noise.

Road traffic creates noise problems for people along road network systems in many areas. Figure 2.4 demonstrates that 73 percent of Sydney people are adversely affected by exposure to traffic noise. Road traffic noise is considered as the aggregation of the noise from individual vehicles in the traffic stream. There are two major sources influencing the road traffic noise (Nelson, 1987). The first one is called vehicle noise source. It is produced from vehicle components: engine; exhaust; tyre/road interaction; and cooling fan. Vehicle aspects simultaneously produce noise depending on the speed of vehicle. Vehicle noise increases when the vehicle speed increases.

Sydney People Adversely Affected by Environmental Noise: 1987

Traffic noise 73%

Rail noise 6% Industrial noise 4%

Aircraft noise 17% Figure 2.4: Sources of Adverse Affection by Environmental Noise in Sydney (source: EPA, 1990)

The second source is called the non-vehicle noise source: of traffic volume; traffic composition; traffic condition; weather and climate conditions; and type and condition of the road pavement. Road traffic noise varies significantly between two roads even though they might have similar traffic volumes because the difference in traffic conditions (i.e., free flow or congested), traffic composition (i.e., high or low truck volumes), or pavement type (i.e., asphaltic concrete, chip seal, or Portland cement concrete). Site conditions and surrounding infrastructure conditions can influence the

24 road traffic noise propagation, which is considered as the ground-to-ground propagation. More details of road traffic noise can be found in Nelson (1987, Part 3).

The general aspects of train noise are similar to road traffic noise in that noise is generated along defined corridors and received by the nearby receivers located by the tracks and stations. However, the total affected area of people from train noise is considerable less than from road traffic noise, because the total train network (or the length of rail track) is much lower than the total road network (or the length of road way). Figure 2.4 supports this idea by illustrating that only 6% of people in Sydney are adversely affected by exposure to train noise.

The sources of train noise are vehicle noise sources and non-vehicle noise sources. The vehicle noise sources of trains are locomotive noise and rolling stock noise. The levels of both locomotive engine noise and rolling stock noise are dependent on the speed of train but in an opposite manner to that of road vehicles. If the train speed increases, on one hand, the locomotive noise level decreases, but rolling stock noise increases. The significant non-vehicle noise sources of train noise are the type of track and the type of ballast. A continuous welded track is considerably quieter than a jointed track. The propagation of train noise also occurs as ground-to-ground transmission. More details of train noise can be found in Nelson (1987, Part 4).

Aircraft noise problems have become the important issues for society since the introduction of large commercial . Aircraft noise disturbs communities surrounding airports, both under and nearby the designated flight paths. The orientation of the runways mainly influences the distribution of aircraft noise through the community because it controls the designation of flight paths. The major sources of aircraft noise are aircraft engines, airframe, reverse thrust, and ground operations. Jet engines generate high noise levels up to 120-140 dB depending on the percentage of thrust power. Ground operation noise creates the vibration of structures due to the low frequency noise levels. Unlike road traffic and rail , aircraft noise propagation can be classified into both air-to-ground (mainly from over flight aircraft) and ground-to- ground (mainly from aircraft ground operations). Aircraft noise is intermittent and varies depending on many factors, which are aircraft type, aircraft operation (landing or

25 taking off), location of receiver, and weather conditions. Further detailed explanations of aircraft noise are provided by section 2.3.

2.2.4.2 Industrial Noise Industry performs as a huge point noise source compared with the other environmental noise sources discussed, and may dictate the in nearby residential areas. Those sources largely vary from a small level size such as water pumps, medium size devices such as compressors and condensers, and large sources like pneumatic piling or mine crushers depending on the nature of the industry (for example, open mine, construction site, various types of manufacturing complexes, or petrochemical plant). By its nature, industrial noise can affect people in a broad area because of high sound power level. Fortunately, industrial locations are controlled by the policy of land use zoning. Most of the industrial activities are separated by distance from the residential area and are limited by their hours of operation. As shown in Figure 2.4, people in Sydney are affected by industrial noise in the lowest proportion compared with the other noise sources. The propagation of industrial noise also performs as ground-to-ground transmission. Ground surface condition, weather and climate conditions influence industrial noise propagation.

2.2.4.3 Community Noise Community noise is mainly divided into neighbourhood noise and appliance noise. Neighbourhood noises are, for example, noise from parties, barking dogs, children playing, arguments/shouting, or T.V/radio. The sources of appliance noise are, for instance, food processors, vacuum cleaner, lawn mowers, laundry machines, water pumps, or air conditioning units. Some community noise, such as the whine of a swimming pool pump which produces considerably low noise level, may create a serious argument between neighbours, especially during the night time when background noise level is very low and people need to sleep. Quieter appliances may be considered as the partial solution of community noise. However, this kind of problem is very sensitive and need great cooperation between neighbours. The propagation of community noise is ground-to-ground.

26 2.2.5 Effects of Noise on Humans The common word used to express the feeling of people suffering from noise is called “annoyance”. Noise annoyance is mostly created by the interference of communication activities or disturbance to rest or sleep. The correlation between annoyance and noise level has been established by a so called dose-response relationship: annoyance tends to increase when the noise level increases. However, in addition, annoyance may be increased as a result of low noise level. The best examples of this are the noise annoyance from a dripping tap or from a ticking clock when getting to sleep.

In addition to annoyance, noise can also cause humans both physiological and psychological harm. The first impact may occur in terms of hearing losses (particularly for people working in a noisy place), blood pressure rises, heart rate and breathing speeding up, muscles tenseness, and stress hormones being released (OECD, 1980, p.25). Most of the physiological effects are considered as the by-product from annoyance, or the other psychological effects, although there is no strong, robust, conclusion about these relations. Noise may be regarded as a stimulus factor for patients with psychiatric illness.

The groups of people affected by noise can be broadly classified into two groups: user group and non-user group. The user group is the noise affected people who directly use (for example, transport users, appliance users) or work with (for example, industrial workers, bus drivers, train drivers, airport workers) each type of noise source. The non- user group are people who do not relate with, or operate, the noise source (for example, pedestrians, neighbourhoods, nearby residences along roads or railway corridors, those under the designated flight paths of aircraft, or surrounding the industrial area). In this review, we are not concerned with the first group of people because they have been considered as the group that are affected by occupation noise. Stansfeld and Matheson (2003) reviewed comprehensively effects of occupation noise on human. More details of non-user group impacts by noise, especially from aircraft noise, will be explained in section 2.4.

27 2.3 AIRCRAFT NOISE

Aircraft noise is central concept to this thesis. The detailed discussion of the nature of aircraft noise (which includes the source of aircraft noise, the generation of aircraft noise, and the propagation of aircraft noise to the receiver), quantification, certification, and prediction of aircraft noise are provided by the following sub-sections.

2.3.1 Sources of Aircraft Noise The primary operations of aircraft that affect surrounding residents in neighbourhoods are takeoff and landing (TOL) operations, taxiing, and the ground running of aircraft following routine maintenance. Aircraft noise can be subdivided into two categories – those generated internally and those generated externally. The major source of internal noise is the aircraft powerplant (aircraft engines), which can be separated into three parts: the fan and compressor noise source; the combustion noise source, and the turbine noise source. The major source of external noise is generated by the jet exhaust mixing of the high-velocity exhaust gas from the nozzle of the engine with the ambient air. The jet noise is the dominant source of aircraft noise during the high-thrust power during takeoff operations, whereas the machinery noise is the dominant source of aircraft noise during the low-thrust power during landing operations. Engine noise and airframe noise both play an important part in aircraft noise, especially during the landing operation.

Since the introduction of jet aircraft, the jet engine has rapidly developed: engine (called pure jet); by-pass jet engine; and jet engine (called high by-pass ratio). Figure 2.5 shows these three typical engine designs and examples of their application, as explained individually in the following paragraphs.

28

Figure 2.5: Typical Jet Engine Designs and their Applications (source: reproduced from Nelson, 1987, figure 18.1, p.18-4).

The three major components of turbojet engine are the compressor, the combustor, and the turbine. The main propulsive thrust is generated by the expulsion of gas at high velocity through the rear single nozzle. Pure jet operation is explained first. Air is drawn into the compression system, mixed with fuel and burnt in the combustion section. Then, the air is rapidly expanded by heat and sent through the turbine section to drive the turbine shaft, which provides the power to drive the compressor. Finally, hot air is expelled through the single rear nozzle providing a propulsive thrust power to move the plane. The exhausted air velocity is approximately 600-700 m/s. This high velocity is the major factor that creates jet engine noise.

29 The turbojet has very powerful thrust generation which is necessary for high-speed flight, usually for military aircraft. However, the turbojet is both a thermodynamically and an economically inefficient system (Nelson, 1987, p.18.3). As a result, the by-pass concept has been introduced to overcome these problems. The low by-pass ratio engine is illustrated in Figure 2.5. In this system, airflow from the low pressure compressor section is separated into two portions, one is drawn into the high pressure compressor section and then sent through the combustor to produce the engine thrust before being expelled into the atmosphere at the core nozzle at high temperatures (called hot gas) with high velocity (around 600 m/s). The remaining airflow (called cold gas) is maintained at a low pressure to ‘by-pass’ the engine core flow in a separated duct. Then it is expelled through the separated nozzle at around two-third of velocity of the hot gas. Generally, the thrust power is equally generated by both hot gas and cold by-pass gas. The word ‘by-pass ratio’ means the proportion of the mass airflow through the front of duct and the mass airflow through the core of the engine.

The high by-pass ratio process has been established as the modern turbofan jet engine that maximises both thermodynamic and economic efficiency. The turbofan jet engine is comprised of four major components: fan, compressor, combustion, and turbine (see Figure 2.5). Typically, the amount of air drawn by the fan is 3-4 times higher than the turbojet process, 15 percent of airflow is sent through the core engine, through both the intermediate and high pressure compressor sections to be compressed to about 40 times atmospheric pressure, then mixed with fuel, burnt at a temperature around 1,500oc in the combustor, then sent into a turbine section, and, finally, expelled through the final nozzle at a velocity around 400-500 m/s. The 85 percent (called cold fan) of the total airflow is drawn passing the separated duct and directly expelled to the atmosphere to create engine thrust. Therefore, the total thrust power on this system mainly comes from the cold fan rather than from the hot core jet. The main role of the hot core jet is to drive the continuous series of engine shafts to move the fan.

2.3.2 Aircraft Engine Noise The aggregation of noise from many components in the aircraft engine described above creates the total level of engine noise. All of these sources create noise in different ways and levels depending on the existing aircraft operation and aircraft engine type and

30 configuration. Engine noise generated by the turbofan engine is only considered because it is the most common aircraft engine type into and out of Sydney Airport (which is the case study of this thesis). A great detail of aircraft engine noise from different engine types can be found in Nelson (1987, pp.18.3-18.34).

Firstly, let us consider the fan noise. The broadband noise is created by the interaction of the tips of the rotating blades with the turbulent flow around the surface of the intake duct. The blade incidence is an important factor that influences this fan noise. If the blade incidence angle is improperly designed, the broadband noise will increase. The fan blade also generates the discrete noise at a blade passing frequency (BPF) and can be obviously heard particularly during the supersonic tip speed conditions. The fan noise normally dominates the engine noise source, particularly during airport approaches or overflight approaching the receiver. This is because fan noise always propagates forward with the aircraft engine, as illustrated in Figure 2.6.

Figure 2.6: Sound Radiation Patterns of Internal Engine Sources (source: reproduced from Nelson, 1987, figure 18.11, p.18-10).

After passing the fan, air is conveyed downwards into two different paths: the fan duct; and the core engine duct (see Figure 2.7). The noise in the fan duct is created by the interaction between the swirling air and a set of the outlet guide vanes (called stators). These stators are a large source of noise. The discrete noise is created as the wakes of air from fan flow slap against the stators at the blade passage frequency (BPF). Moreover, the broadband noise is produced by the interaction of turbulence, which is

31 created by the unsteady airflow from the fan, with the stators. This broadband noise is often heard as a rumbling sound, and is normally propagated rearward of the engine as shown in Figure 2.6.

Once air is pushed into the core engine duct, it is compressed by a set of rotors and stators. The discrete noise is created by the interaction of turbulent flow with adjacent rotating and stationary blade rows. Spacing between rotor and stator is a very important factor to control the compressor noise. If the space is too close, an intense pressure will generate and induce a higher, discrete, noise. There are many efforts to reduce the discrete noise by sustaining this optimal spacing which will create less discrete noise than either wider or closer gaps. The other factors affecting the discrete noise are the number of rotors and stators and their arrangement.

Figure 2.7: Half-Section of Engine Front-End Assembly, Showing Relevant Blading (source: reproduced from Nelson, 1987, figure 18.7, p.18-8).

Naturally, the combustion process generates turbulent flow so the combustion chamber is the source of broadband noise. The combustion chamber is designed to mix the air and fuel rapidly to burn in the combustion zone. Very high pressure and high velocity airflow is, then, mixed with the compressed air at the mixing and dilution zone to sustain a stable flame front. The intense broadband noise is created around the inner combustion chamber surface. Fortunately, according to the size and configuration of the

32 combustor, the combustion noise is not a significant compared with the nearby engine noise sources (compressor noise source and turbine noise source).

The main function of the turbine is to provide the power to drive all the fan and compression stages at the front of the engine. The noise generating mechanisms are identical with those in the fan and the compressor. The burned gas from the combustion chamber is directly sent into the turbine through the elements of the nozzle guide vane (NGVs) and the high-pressure turbine rotor (HPRs) before passing through the final nozzle. Both broadband noise and discrete noise are created in this section of the engine, but the discrete noise dominates more. Although the turbine noise productions are similar to those from the fan, the level of severe noise is lower because the turbine blade numbers are higher than the fan and these all are operated at a very high speed compared with the fan. All the discrete noise frequencies from the turbine section are almost beyond the audible range. Furthermore, the total mass flow through the turbine on a turbofan is very low compared with the fan section. Figure 2.6 illustrates the propagation path of turbine noise as it radiates rearward of the engine. It will be heard only for a short period of time because the turbine noise has to pass through the high temperature and high velocity of the jet exhaust noise so it is immediately refracted down with degree range 100-120 of the intake exit (Nelson, 1987, p.18.9).

2.3.3 Jet Noise Jet noise is categorised as shock noise and jet mixing noise. The shock noise normally occurs on zero/low by-pass jet engine at full power where exhaust velocities are in the region of 600-700 m/s and the jet flow is locally supersonic (Nelson, 1987, p.18.11). It can be heard as well, ahead of an aircraft’s appearance as a harsh tearing noise until the jet mixing noise following the aircraft has passed overhead. This shock sound also happens in the turbofan jet engine, but only during takeoff operations with high thrust power when the fan’s tip speed operates at supersonic conditions. Fortunately, the turbofan jet engine has already reduced this noise problem by lowering the exhaust velocity. It can be concluded that shock noise is not an issue in the modern commercial jet engine.

33 Another important source of jet noise is the jet mixing noise, which occurs in all types of aircraft jet engine. For zero/low by-pass ratio, it dominates for both takeoff and landing operations. For high by-pass ratio, it is significant only during takeoff operations. As implied by its name, the jet mixing noise is generated by the mixing among the hot-core jet, the cold-fan jet, and the ambient air after their expulsing through the final nozzle. The mixing pattern, dimension of the nozzle, and exhaust speed are the important factors creating the jet mixing noise. There are two basic types of jet mixing pattern: unmixed pattern and mixed pattern.

The unmixed engine is the earlier type of by-pass ratio engine. The cold fan gas and the hot core gas mix together after their expulsion from the separated nozzle. A turbulent flow is created, and basically divided into two zones: the small eddies zone; and the large eddies zone. Both zones create the broadband noise. The small eddies zone occurs nearly at the nozzle and creates very high frequency broadband noise. Conversely, at the downstream of the exhaust jet, the large eddies zone is created with a low frequency broadband noise. Normally, the large eddies zone forms at a long distance and results in a very high noise level. It also creates the shock noise at the core of the final nozzle. The modern turbofan jet engine has overcome this problem by an early mix of both cold fan gas and hot core gas, before being expelled out into the atmosphere (called mixed pattern). The large eddies zone length is reduced but, conversely, the small eddies zone is increased. Nevertheless, the high frequency noise is very easily absorbed in the atmosphere before being heard in the far field. The shock noise is also eliminated.

2.3.4 Airframe Noise After the introduction of the turbofan jet engine, the airframe noise became the important issue. Airframe noise can be a significant only when the engines operate at the low power, particularly during the runway approach. The cause of the major elements of airframe noise is the temporary deployment of high-lift devices (flaps and slats) on the wings, and the nose, and main undercarriage assemblies. The other sources are minor such as the passage of air over the wings, , tailplane and powerplant . Normally, the airframe noise is broadband.

34 2.3.5 Propagation of Aircraft Noise Aircraft noise propagation is basically classified into two cases; sound propagation during overflight (air-to-ground) and during ground operations (ground-to-ground). The main attenuations of overflight noise are from spreading attenuation and atmospheric attenuation. Wind and temperature also influence the overflight noise depending on the meteorological conditions. Ground attenuation mostly plays the major role of noise attenuation during ground operations.

Basically, the air-to-ground propagation of aircraft noise has been assumed to be a linear equation (ISO 3891, 1978, pp.53-59; and ICAO, 1993, Annex 16, vol.1, pp.77- 78). ISO 3891 (1978, p.53) has provided a mathematical formula to calculate the atmospheric attenuation of aircraft noise. It is unnecessary for this thesis to present the formula. However, it can be concluded that the total air-to-ground attenuation of aircraft noise depends mainly on aircraft noise frequency and the relaxation of air molecules. High aircraft noise frequencies tend to be absorbed more easily than the low frequencies.

There are some instances of anomalously low attenuation of aircraft noise between the predicted noise level and the measured noise level. This situation comes from the Doppler-shifted frequency and flight Mach number, especially during the overflight procedures as they create noise with the high-speed moving source. The noise propagation is probably considered to be a non-linear equation. This kind of propagation is comprised of complex mathematical knowledge and again is beyond the scope of this thesis. However, there is some research about this non-linear propagation. For example, Morfey and Howell (1981) have developed a model to predict the non- linear propagation of aircraft noise in the atmosphere. They found that the non-linear propagation resulted from the cumulative distortion of the acoustic waveform during the long propagation, especially in the high frequency (5-10 kHz) zone. Yamamoto and Donelson (1993) have developed a model to predict the aircraft en route noise receives on the ground by including both Doppler-shifted frequency and flight Mach number variables. The model reveals a satisfactory result compared with measured data.

35 The ground highly influences the attenuation of sound when it propagates at a low grazing angle. It is comprised by many such factors as groundcover, type of soil, and distance and grazing angle between source and receiver. Moreover, through real world measurements, meteorological conditions (wind and temperature) will sustain the ground effect in some cases (temperature inversion and precipitation ambient). Aircraft noise ground attenuation is classified into two categorises: inside and outside the building. Inside ground operation may simply refer to the routine maintenance and overhaul works of an aircraft. They are normally operated in a covered area, such as a hangar or on the apron. Noise barriers and suppressors are installed to reduce noise levels before they are propagated outside. The natural ground attenuation is minor compared with these artificial barriers. However, after it is emitted through the building, the ground attenuation will perform similarly to the aircraft ground operation located outside the building.

The aircraft noise during taxiing, start of roll for takeoff, or approaching the ground will be attenuated by the ground. It depends on the nearby ground characteristics, for example, the area next to the runway or taxiway pavement (plain concrete or asphalt concrete) covering by grass, rocks, or trees. In some case, the buildings such as the terminal (passenger or cargo), are purposefully arranged to obstruct the aircraft ground operation noise. The natural barrier (for example, hills or mountains) is a very good tool to reduce the aircraft noise if the runway alignment is well planned and it is practical.

Some previous studies developed a model to predict the level of aircraft noise attenuated by the ground. For example, Blumrich and Altmann (1999) investigated the effect of porous, grass-covered ground in a military air-base in Jever-Schortens, Nothern Germany. They found that the attenuation effect will be maximised for sound frequencies in the range 100 – 200 Hz with the attenuation about 20 dB for a distance of 100 metres. This prediction is based on the Weyl-van der Pol equation.

2.3.6 Quantification of Aircraft Noise The goal of aircraft noise quantification is to determine and assess all the dimensions of aircraft noise related to human responses. The manner of human response has various directions depending on many contributing factors of aircraft noise: the frequency of

36 aircraft noise intrusions, the time of the day of these intrusions, and the number of these intrusions over a specific period (Horonjeff and McKelvey, 1999, p.731).

This thesis considers two main concepts of an aircraft noise index development. The first concept, which is the most popular method, is called the principle of energy equivalence. The simple meaning of this method is that human responses are the same for a loud noise (see section 2.4.1.3 for the meaning of loudness) a few times a day and a quieter noise many times a day. This method is considered to have high technical accuracy and completeness which is suitable in developing a noise contour map for land-use planning use and noise management around the airports. However, inevitably, it is considered too complicated an indicator which makes them less comprehensible and less accessible to the layperson (DoTARS, 2002, p.62). Thus, it may not be sufficient to apply this principle to reflect human reactions to aircraft noise.

The example of aircraft noise indices based on the principle of energy equivalence are Day-Night Level (DNL) and Noise Exposure Forecast (NEF) used in United States, Noise and Number Index (NNI) used in United Kingdom, Weighted Equivalent Continuous Perceived Noise Level (WECPNL) used in Italy, Noise Exposure (Kosten unit) (B) used in Netherlands, and Australian Noise Exposure Forecast (ANEF) used in Australia.

The second method to quantify aircraft noise is called the peak level index which is simply referred to a maximum sound pressure level of noise event. This method may be used in accordance with the number of occurrences of a noise event. Subsequently in an attempt to optimise an accurate acoustic concept while sustaining a simplicity of index that can be easily understood by a layperson, the Department of Transport and Regional Services (DoTARS), Australia has introduced the principle of transparency aircraft noise information (DoTARS, 2003a, p.9) based on the concept of the peak level index and ‘everyday talk’ information. Some examples of aircraft noise index based on this principle are Number-Above Index (NA) and People-Event Index (PEI). The thesis places special attention on the NA since it is an important parameter used to develop a Noise Gap Index (NGI) (see chapter 5). The following sections overview briefly the

37 three most popular energy equivalent aircraft noise indices (NEF, ANEF, and DNL) and the NA.

2.3.6.1 Noise Exposure Forecast (NEF) The NEF was developed by the Federal Aviation Authority (FAA) of the United States in 1967. The NEF employs the principle of sound energy equivalence by using the effective perceived noise level (EPNL) as the single-event sound level descriptor with time of the day weighting factors. The EPNL (in EPNdB unit) was primarily developed for aircraft noise certification purpose. The EPNL consists of the measurement of perceived noise level (PNL) with some specific correction factors. As implied by its name, PNL(k) means an instantaneous level of aircraft noise perceived at a given location for a specific overflight. The PNL(k) can be calculated by the following steps: (1) measure an individual aircraft noise level at any time in each 1/3 octave band of SPL(i,k) from 50 to 10,000 Hz and convert them into perceived noisiness, called n(i,k), (2) combine all the n(i,k) to obtain the total perceived noisiness, called N(i,k), and (3) convert the N(i,k) into the PNL(k). After that, the PNL(k) are adjusted for spectral irregularities by the correction factor C(k) to obtain the tone corrected perceived noise levels (PNLT(k)). The maximum tone corrected perceived noise level (PNLTM) which is the maximum values of all PNLT(k) from previous calculation is determined to obtain the correction of time duration (D). Finally, the EPNL is calculated by the summation of the PNLTM with the factor D. More details in calculation of each EPNL’s step can be observed from ISO 3891:1978, p.50.

Equations (2.3) and (2.4) provide the formulae to calculate the total NEF at location (x) on the ground from the overall flight paths. From equation (2.3), the NEFij equals the

EPNLij from aircraft type i using flight path j added with the logarithm of frequency of occurrence and deducted by some scale adjustment factor. From the energy equivalent concept, the measurement of aircraft noise loudness was presented as a logarithm scale. Therefore, the number of aircraft noise events should be presented also in logarithm form. The weighting factor from equation (2.3) also implies that noise sensitivity of humans increases during the night time period. One overflight during this period is equivalent to 16.67 overflights during daytime period.

38 NEFij = EPNLij + 10 log (dij + 16.67nij) – 88 dB (2.3)

i=I j=J NEF NEF = 10 log ∑∑anti log( ij ) dB (2.4) i=11j= 10 where

NEFij is the day/night average Noise Exposure Forecast at location (x) due to aircraft type i on flight path j

EPNLij is the effective perceived noise level at location (x) of aircraft type i on flight path j dij, nij it the number of flight during daytime (7am-10pm) and night time (10pm-7am) of aircraft type i on flight path j 16.67 is a time of the day weighting 88 is an adjustment factor NEF is the total day/night average Noise Exposure Forecast at location (x) due to all aircraft types (i=1,2,…I) on all flight paths (j =1,2,…J).

2.3.6.2 Australian Noise Exposure Forecast (ANEF) After a comprehensive survey of community impact by aircraft noise around five major airports in Australia conducted by Hede and Bullen in 1980, the ANEF has been developed based on that survey results and it has been introduced as the most suitable and acceptable aircraft noise index for the Australian community since 1982. All the procedures and concepts of ANEF are similar to the NEF except in two aspects. Firstly, the ANEF classifies the daytime and night time exposure periods as during 7am to 7 pm and during 7 pm to 7 am, respectively. Secondly, the ANEF uses time of the day weighting equal to 4. Equations (2.5) and (2.6) provide the formulae to calculate the total ANEF at location (x) on the ground from the overall flight paths (Airservices Australia, 1999).

ANEFij = EPNLij + 10 log (dij + 4nij) – 88 dB (2.5)

i=I j=J ANEF ANEF = 10 log ∑∑anti log( ij ) dB (2.6) i=11j= 10

39 where

ANEFij is the day/night average Australian Noise Exposure Forecast at location (x) due to aircraft type i on flight path j

EPNLij is the effective perceived noise level at location (x) of aircraft type i on flight path j dij, nij is the number of flight during daytime (7am-7pm) and night time (7pm-7am) of aircraft type i on flight path j 4 is a time of the day weighting 88 is an adjustment factor ANEF is the total day/night average Australian Noise Exposure Forecast at location (x) due to all aircraft types (i =1,2,…I) on all flight paths (j =1,2,…J).

2.3.6.3 Day-Night Average Sound Level (DNL) According to the computational complexities involved in the above two aircraft noise indicators, the day-night average sound level (DNL) has been introduced by the US Environmental Protection Agency (EPA) in 1973 and was firstly used by FAA in 1976 as the standard aircraft noise descriptor. The DNL also employs the concept of energy sound equivalent but uses the sound exposure level (SEL) as the single-event sound level descriptor with time of the day weighting factors. The SEL is the summation of A- weighted sound pressure over at least the top 10 dB(A) from the maximum value of the aircraft noise event (Horonjeff and McKelvey, 1999, pp.729-731).

The first step in calculating the SEL is to determine the SPL of each frequency band. The second step is to convert all the SPL of each frequency band into the A-weighted scale and then combine all of them to obtain the total A-weighted sound pressure level

(LA) in discrete time (∆t = 0.5 s. for aircraft noise) of an aircraft noise event. Finally, the SEL can be calculated by equation (2.7), following (Horonjeff and McKelvey, 1999, p.730):

N ⎡ 1 LA,i ⎤ SEL (LAE) = 10 log ⎢ ∑(anti log )∆t⎥ dB(A) (2.7) ⎣T0 i=1 10 ⎦ where

To is 1 second to maintain a dimensionless argument for logarithm

LA,i is the instantaneous ith A-weighted sound level measured every 0.5 s

40 To determine the total DNL (called Ldn), the DNLij from all aircraft types (i =1,2,…I) on all flight paths (j =1,2,…J) are averaged over a 24-h period. Equations (2.8) and (2.9) provide the formulae to calculate the total DNL at location (x) on the ground.

DNLij = LAE,ij + 10 log (dij + 10nij) – 49.4 dB(A) (2.8)

i=I j=J DNL DNL = 10 log ∑∑anti log( ij ) dB(A) (2.9) i=11j= 10 where

DNLij is the day/night average sound level at location (x) due to aircraft type i on flight path j

LAE,ij is the sound exposure level at location (x) of aircraft type i on flight path j dij, nij is the number of flight during daytime (7am-10pm) and night time (10pm-7am) of aircraft type i on flight path j 10 is the time of the day weighting 49.4 is an adjustment factor (= 10 log (86,400)) DNL is the total day/night average sound level at location (x) due to all aircraft types (i=1,2,…I) on all flight paths (j =1,2,…J).

2.3.6.4 Number Above Index (NA) The concept for calculating the NA is simple. It is a number of noise events at location (x) where the noise level (dB(A)) exceeds a threshold level over a given time period. For example, from Figure 2.8 (where the y-axis represents the sound pressure level, (dB(A)), the x-axis represents time, (second), and the rectangular symbol represents the aircraft noise), if a threshold level is set as 70 dB(A), and the time period is 20 minutes, the Number Above Index (or N70) of aircraft noise will be nine. However, with the same time period, if a threshold level is 80 dB(A), then the Number Above Index (or N80) of aircraft noise will be six. A practical threshold level is often set at 70 dB(A) because it will be 10 dB(A) attenuated by a structure of house (with open windows) and that 60dB(A) is the indoor sound pressure level of noise event that is likely to interfere with conversation or with listening to the radio or the television (DoTARS, 2002, pp.24- 25).

41 95 90 85 80 75 70 65 60 55 LAeq,1s (dB(A)) LAeq,1s 50 45 40 35

2 2 2 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :0 :0 :0 :02 8 9 0 1 2 3 4 5 6 7 8 9 0 :30 :31 :32 :33 :34 :35 :36 :37 :3 :3 :4 :4 :4 :4 :4 :4 :4 :4 :4 :4 :5 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Figure 2.8: Example of Typical Noise Time Curve (see Chapter 5)

Without the time of the day weighting, the NA is produced in terms of an average annual day or an average monthly day. The NA can also be produced in terms of the sensitive times. For example, the number of events above 60dB(A) for the period 10pm to 6am has been produced by the Environmental Impact Statement (EIS) for the Second Sydney Airport to represent night time noise exposure patterns disturbing sleep (DoTORS, 2002, pp.25, 35).

The DoTARS (2002, pp.23-39, 63-66) provides a great detail on the N70. Some interesting points that should be highlighted will be mentioned here. Firstly, the N70 was used as an important indicator in the EIS for the Second Sydney Airport since it has been proved that it is a better way of communicating with the community instead of the traditional energy equivalent noise indices (such as ANEF and DNL) (see section 3.2). Secondly, the use of the N70 is limited because the calculation of the N70 is not available in many noise programs such the most popular one – INM (see section 2.3.8.1). Fortunately, the DoTARS has developed a computer program called the TNIP (see section 2.3.8.2) to sustain this requirement. Thirdly, even though it represents a noise event at 71 dB(A) as the same as one at 90 dB(A), the N70 is based on the concept that once a noise event exceeds a threshold level it becomes an intrusion and its actual level is not necessarily important. For example, there is no difference between a noise event at 80 dB(A) and 90 dB(A) for a sleeping person since both events are likely to awaken this person. Finally, since there is no formal study to confirm an accuracy of the predicted N70, fieldwork (DoTARS, 2002, table B.1, p.66) shows the agreement of the measured and the predicted N70 is broadly of the order of the mean ±10%.

42 2.3.7 Aircraft Noise Certification Two international institutes provide the worldwide acceptable aircraft noise certification standards: Federal Aviation Authority (FAA) and International Civil Aviation Organization (ICAO). The former is widely applied in the United States. Almost all European countries, Australia and New Zealand, and Asian countries employ the latter one. This research has selected Sydney (Kingsford Smith) Airport as a case study; it is concerned only with the aircraft noise certification standard provided by ICAO. Nevertheless, these two standards are very similar in many aspects (for instance, in noise measurement, measurement locations, maximum noise level, and trade-offs) and these similarities can be considered the same after the conversion of some measured unit (for example, from kilogram to pound, from metre to feet).

ICAO classifies subsonic aircraft into two groups: Chapter 2 and Chapter 3 groups. From ICAO Annex 16 Volume 1, Chapter 2 aircraft means “all subsonic jet aeroplanes for which either the application for certificate of airworthiness for the prototype was accepted or another equivalent prescribed procedure was carried out by the certificating authority before 6 October 1977” and Chapter 3 aircraft means “all subsonic jet aeroplanes for which either the application for certificate of airworthiness for the prototype was accepted or another equivalent prescribed procedure was carried out by the certificating authority, on or after 6 October 1977”. It should be noted that ICAO also includes other types of aeroplanes or driven aeroplanes in both Chapter 2 and 3, but these will not be included in this discussion. Lateral point where noise after leftoff is greatest

450 m for Chapter 2 aircraft 650 m for Chapter 3 aircraft Approach point Flyover point

2,000 m 6,500 m

Threshold of runway or star of takeoff roll Figure 2.9: Required Noise Measurement Points for Aircraft Noise Certification (source: reproduced from Horonjeff and McKelvey, 1999, figure 15.8, p.757).

43 Noise evaluation of both Chapter 2 and 3 aircraft is achieved by means of the effective perceived noise level (EPNL). Methods to measure and calculate the EPNL for both Chapter 2 and 3 aircraft are described in Appendix 1 and 2, respectively, in ICAO Annex 16 Volume 1. Figure 2.9 shows the required noise measurement points for both Chapter 2 and 3 aircraft. The flyover point is located under the flight path 6,500 m away from the start of roll on an extended centre line of the runway. The approach point is located under the approach path 2,000 m from the runway threshold on an extended centre line of the runway. The lateral point is located along the line parallel to, and 650 m (for Chapter 2 aircraft) or 450 m (for Chapter 3 aircraft) from, the runway centre line, where the noise level reaches a maximum level during taking off.

The certification requires all the participating aircraft noise levels to be lower than the maximum noise levels on each required measurement location. These maximum noise levels are mainly based on the maximum gross takeoff weight. ICAO (1993, table 1 and 3, attachment A, pp.111-112) present the formulae to calculate the maximum noise level measured on those three required measurement locations for both Chapter 2 and 3 aircrafts, respectively.

If the maximum noise levels are exceeded at one or two measurement points: (1) the sum of excesses should not be greater than 4 EPNdB (for Chapter 2 aircraft) or 3 EPNdB (for Chapter 3 aircraft), (2) any excess at any single point should not be greater than 2 EPNdB (for both Chapter 2 and 3 aircrafts), and (3) any excesses should be offset by corresponding reductions at the other point or points (for both Chapter 2 and 3 aircrafts).

According to the provision to alleviate the aircraft noise problems, ICAO provided regulations for any noise sensitive airport to consider the banning of certain noisy aircraft operating in its navigator airspace. ICAO suggested that the airport authority start phasing out operations of Chapter 2 aircraft (or which do not comply with the Chapter 3, Volume 1, ICAO Annex 16) from 1 April 1995 and have all of them withdrawn from service by 31 March 2002. Nevertheless, for certain Chapter 2 aircraft groups, which have (1) not completed 25 years of service on 1 April 1995, or (2) been fitted with the quieter (high by-pass ratio) engines, were not immediately affected after

44 1 April 1995 and will be permitted to operate until the final date of withdrawal (ICAO, 2002). However, in some cases, especially for developing countries where the phasing out of Chapter 2 aircraft is either technologically impracticable or economically unfeasible, ICAO also prepared the guidelines (called ICAO Assembly Resolution A33- 7) to sustain the airport authorities with possible ways before start phasing out program. Full details of ICAO Assembly Resolution A33-7 can be observed from .

In June 2001, ICAO adopted a new Chapter 4 aircraft noise standard according to recommendation provided by the fifth meeting of the Committee on Aviation Environmental Protection (CAEP/5). The new aircraft noise standard is more stringent than that contained in the ICAO Chapter 3, and will be commenced on 1 January 2006. The new standard will apply to newly certificated aircraft and to the ICAO Chapter 3 aircraft for which re-certification to Chapter 4 is requested. Note also that at the time of writing this thesis, ICAO was in a process of proposing a Chapter 5 aircraft noise standard which is more stringent than that contained in the ICAO Chapter 4. The aircraft noise certificate continues to be more stringent which reflects the fact that, at present, aircraft noise problem has not been satisfactory resolved. More detail about proposed ICAO Chapter 4 can be observed from .

2.3.8 Prediction of Aircraft Noise The prediction of aircraft noise exposure for a specific point can be done manually (see section 2.3.2) but it is considered time consuming. To manually produce the aircraft noise footprints (imaginary lines through equal noise exposure points on ground) covering a large area of airport vicinity is completely impracticable and unfeasible. Computer software is a way to overcome these problems. Two types of computer software will be overviewed in the following sections. The first program, which is the most internationally acceptable, is called the Integrated Noise Model (INM) developed by the FAA, United States, and the second program, which has been developed by the DoTARS, Australia, is called the Transparency Noise Index Program (TNIP). Airservices Australia uses the TNIP to generate the transparency aircraft noise contours around the major airports in Australia. The TNIP has been developed according to the INM application and operation functions.

45 2.3.8.1 Integrated Noise Model (INM) The INM is a comprehensive software database system for evaluating aircraft noise impacts in the vicinity of airports. It is composed of many analytical uses for assessing changes in noise impact resulting from many airport characteristics or activities changes such as (1) construction of new runways or extension of runways, (2) alteration of routes and airspace structures, (3) changing of traffic demand and fleet mix, and (4) any modifications of the other airport operation procedures (Gulding et al 1999, pp. 2.1- 2.11). The FAA developed the first version of INM in 1978 as a standard tool for determining the predicted noise impact in the vicinity of airports in the United States. Updated versions of INM have been released one after another since 1979 (second version) until 1999 when the sixth version (INM 6.0) was issued.

The major aim of INM is to produce noise exposure contours used for land-use compatibility maps. It also can determine aircraft noise impacts on any specific point such as a hospital, school, or other noise sensitive locations on user’s defined boundaries. In INM 6.0, the model also supports many predefined noise metrics such as NEF and DNL. It also allows for application of a user’s defined noise metric such as the ANEF. The database system of INM 6.0 is very powerful system. It contains, for instance, data on 820 U.S. airports that have 100 operations per day or more, almost 320 standard aircrafts with individual noise profiles based on various operations, and standard aircraft operation profiles (takeoff, landing, touch-and-go, circular). It can also be incorporated or linked with the other database system or information such as terrain information files, census data files, radar tracks files, AutoCAD files, and GIS files.

The core calculation model of INM for producing aircraft noise footprints is based on Standards documents made by the Society of Automotive Engineers (SAE) Aviation Noise Committee (A-21) which are SAE-AIR-1845 (“Procedure for the Calculation of Noise in the Vicinity of Airport”), SAE-AIR-1751 (“Prediction Method for Lateral Attenuation of Airplane Noise During Takeoff and Landing”), and SAE-ARP- 866A (“Standard Values of Atmospheric Absorption as a Function of Temperature and Humidity”). The interested reader is referred to the INM 6.0 User’s Guide (1999). Finally, it can be concluded that the INM is an average value model designed to estimate long- term average impacts in the airport vicinity by using average annual

46 values as input data. Many studies have adopted INM as a tool to simulate aircraft noise level around their study area. The following paragraph provides some examples of INM’s usages.

Garcia et al (1993) employed the INM to generate the contours of NEF around six airports in Spain to develop the relationship between aircraft noise annoyance and aircraft noise level (called dosage-response relation, see section 2.4.6.1.2). Ignaccolo (2000) used the INM to formalise the relationship between the principal factors determining aircraft noise nuisance at Catania-Fontanarossa International Airport, Italy. Wijnen and Visser (2003) applied the INM together with a Geographic Information System and a dynamic trajectory optimisation algorithm to optimise the departure trajectories. The expected result is a new performance index that suggests more practical flight paths that reduce aircraft noise impacts.

2.3.8.2 Transparency Noise Index Program (TNIP) The DoTARS designed the TNIP in a way that is easy to use by a person with non- specialist computer skills. Basically, the TNIP provides the user the options to access five transparency aircraft noise information modules, which are: flight path movements charts; respite charts; N70 contours; measured N70; and single event contours (DoTARS, 2002). The user inputs the provided database files and runs the program to interrogate those databases to produce high quality charts and reports. For more advanced users, TNIP allows the user to create a What-If scenario that can show the changes in noise exposure patterns when the operations are changed in some ways (such as alteration of runway configuration or modification to runway use). It also allows the user to calculate the energy equivalent noise metrics (such as ANEF, DNL) and create a new airport configuration.

There are three main categories of data files contained in TNIP. The first category, which primarily contains aircraft movement data at an airport, is called the Monitoring File. It is recommended that this category should be updated regularly for the purpose of on-going monitoring and reporting of aircraft noise distribution. The second category is called the Configuration File. It contains information, such as airport configuration, aircraft types, population density, flight paths, track file, noise monitoring terminals,

47 and all necessary templates. The updating of this category is required from time to time. The third category contains data derived from detailed grids generated using the INM. More details of those three categories can be found in the TNIP user’s manual version 3.0 (DoTARS, 2003b, pp.44-65).

Finally, it is noted that the TNIP creates the N70 contours using a function derived based on the INM’s noise grid function. The program computes a detailed grid and sums the number of events within the model which register a non zero time above 70 dB(A) (DoTARS, 2002, p.64).

2.4 HEALTH AND WELL-BEING IMPACTS BY AIRCRAFT NOISE

2.4.1 Introduction to the Human Auditory System Before noise impacts on human are reviewed, this section demonstrates how a person receives and responds to the heard sound.

2.4.1.1 Outer, Middle, and Inner Ear Figure 2.10 illustrates a cross section through the human ear. The ear is usually divided into three sections: outer, middle, and inner ear. The outer ear consists of the pinna (or auricle) and the external auditory meatus. The hearing mechanism starts when the sound waves are received and amplified by the pinna and travel along the auditory canal, which is about an inch long, reach the tympanic membrane (or eardrum), which has about the thickness of a piece of paper (Foreman, 1990, p.16). The eardrum then displaces back and forth with various amplitudes according to the sound pressure level. During a normal conversation, for instance, the displacement of the eardrum will be approximately 10-8 cm (Saenz and Stephens, 1986, p.11). The pinna and auditory canal also have the functions to protect the damage of eardrum from any external object or insect attacks.

In the middle ear, the three ossicles called the hammer (or malleus), the anvil (or incus), and the stirrup (or stapes) will transfer forces from the displacements of the eardrum to the base of the stirrup (or the footplate) as a lever system (see Figure 2.10). These three

48 ossicles are situated at the end of the Eustachian tube. This tube is connected with the nasal cavity to enable the air pressure within the middle ear to be adjusted equally to the pressure from ambient air. As the results of the small ossicles vibration, the oval window connected with the footplate will set into motion the fluid (or perilymph) fulfilled in the cochlea located in the inner ear. The main function of the middle ear is to produce an efficient transfer of sound energy from air to fluid inside the cochlea. It also acts as an impedance-matching transformer in the form of “low-pressure, large- amplitude displacement” at the eardrum to “high-pressure, small-amplitude displacement” at the oval window (Foreman, 1990, p.19).

Figure 2.10: Structures of the Human Ear (source: reproduced from Bies and Hansen, figure 2.1, p.45).

49 The middle ear not only transmits the pressure waveform of the sound to the inner ear, but also protects the inner ear from very intense sounds (but not too rapidly) by (1) preventing the transmission of slow, low frequency, pressure waves to the inner ear; (2) rupturing the eardrum until it heals; and (3) attenuating the fairly high-intensity pressure waves by the contraction and stiffness of the small two muscles in the middle ear. This phenomenon is called aural reflex. Further information about aural reflex can be found in Kryter (1994, pp.38-46).

Figure 2.11a illustrates the cross section of the cochlea. It consists of three longitudinal canals called scala vestibuli, scala media (or cochlea duct), and scala tympani. Both scala vestibuli and scala media are divided by Reissner’s membrane. Both scala media and scala tympani are divided by Basilar membrane, which extends along the cochlea’s entire length except for a small gap called the helicotrema at the cochlea’s far end of apex. The movement of the oval window disturbs the fluid in the cochlea forming the travelling wave that starts from the beginning of the cochlea at the oval window through along the upper canal (or scala vestibuli), around the helicotrema, into the lower canal (or scala tympani), and ultimately the round window deflects to accommodate (or absorb) the fluid disturbance.

During the passage of the disturbed fluid along the lower canals, the basilar membrane and the tectorial membrane (see Figure 2.11b) are moved up and down relative to the turbulence. These movements cause the nerve fibers (or hair cells), which are attached between both membranes in the organ of corti, to bend and distort. The hair cells consist of three outer rows and one inner row. There are approximately 25,000 – 30,000 hair cells spread out along the basilar membrane. These hair cells play the major role in the hearing mechanism as they generate the neural-electric impulses from their distortion and then transmit these signals through the individual fibers, which form the auditory nerve, to the auditory centers of the brain stem and brain. It should be noted that the hair cells are very vulnerable structures in the cochlea, and they can be damaged by many factors such as intense sound, metabolic disturbance, lack of oxygen, or drugs. Furthermore, after being destroyed they cannot be regenerated and will cause or lack of ability of hearing sensitive sound (Jones and Chapman, 1984, p.11).

50

Figure 2.11: (a) Cross Section of the Cochlea; (b) Enlargement of Organ of Corti (source: reproduced from Foreman, 1990, figure 2.3a, p.18 and Kryter, 1994, figure 2.3a, p.19).

One important phenomenon that explains a situation when one sound makes the other sound inaudible is called “Masking”. It becomes an important issue because it causes interference of communication either for occupation or for non-occupation activities. Workers in industry may not receive a warning message from nearby friends about possible danger and might be injured. The misunderstanding of command due to the masked noise can become a severe problem in some work places where very accurate operations are required. The masking from environmental noise can disturb nearby communities during leisure or relaxing time such as watching TV, listening to the radio, or speaking with friends by telephone or face to face. It also affects the performance of

51 activities that require mental concentration such as studying and teaching. Kryter (1994, pp.28-36) provides a good explanation of the masking phenomenon.

2.4.1.2 Human Response to Noise Generally, studies concerning the impacts of noise on community aim to focus on the human subjective responses to noise. A person perceives two noises and decides which one is louder than another depending on factors such as sound pressure level, sound frequency, auditory conditions and age of receiver. For a single fixed frequency or narrow band of frequencies, Bies and Hansen (1997, table 2.1, p.53) demonstrate that 50% increasing (or decreasing) of sound energy will change the SPL by only 3 dB and effects on human response is just only perceptible. For instance, when the engine type is the same, the perceived loudness of noise from a two-engine plane will be similar to a four-engine plane. By halving the traffic volume on noisy highway would result in only a 3 dB reduction. Changes in sound energy that make SPL changing for 5, 10, and 20 dB will result in a person’s response being clearly noticeable, a half or twice as loud, and much quieter or louder, respectively.

Figure 2.12 illustrates an experimental result from a study of young healthy auditory people in a free field to compare the loudness of sound with different frequencies. A 1 kHz tone was used as a reference, and a second tone compared with it. The figure shows that the subjective response of humans is both frequency and amplitude dependent. For example, the subject rates a 60 dB with 63 Hz as loud as a 40 dB with 250 Hz, and a 100 dB with 125 Hz as loud as a 95 dB with 500 Hz. For the sake of simplicity, a new unit has been developed (called phons). The units represented on an equal-loudness contour in the Figure 2.12. All tones of the same number of phons were rated by the subject as equally loud. For example, a 500 Hz of 20 phons sounds equally as loud as a 63 Hz of 20 phons, even though the SPL of the later one is approximately 28 dB higher. The lowest dash line is the average threshold of hearing (or minimum audible field, MAF).

52

Figure 2.12: Free-Field Equal-Loudness Contours, in phons (source: reproduced from Bies and Hansen, 1997, figure 2.5, p.55).

The unit of phons does not reflect the subjective response of relative loudness between contours. Therefore, another unit (called sone) has been introduced. A doubling in sones represents a doubling in loudness. The relation between loudness level (phons) and loudness scale (sones) is presented by equation (2.10) (extracted from Bies and Hansen, 1997, p.55). S = 2(P - 40)/10 (2.10) where S is the loudness scale (sones) P is the loudness level (phons)

Additionally, it is noted that the subjective response also depended upon the duration of exposure. The longer the duration of exposure then the higher the subjective loudness is perceived. On the other hand, the shorter the duration of exposure then the less loudness of sound is perceived (Foreman, 1990, p.21).

2.4.2 Introduction to Human Blood Pressure Regulation

2.4.2.1 The Structure and Function of Blood Vessels Blood vessels are classified according to whether they carry blood away from or to the heart, and according to size. They are five mainly type of blood vessels in humans: (1) artery; (2) arteriole; (3) capillary; (4) venule; and (5) vein (see Figure 2.13). Arteries

53 and smaller arterioles carry blood from the heart and to capillaries, which are drained by venules and then larger veins, which return the blood to the heart.

Figure 2.13: The Relationships of Blood Vessels According to Size and the Direction of Blood Flow (source: reproduced from Germann and Stanfield, 2002, figure 13.8, p.412).

Arteries conduct blood away from the heart and toward the body’s tissues; they have relatively large diameters and thick walls. The largest artery has an internal diameter of about 12.5 mm. The smaller arteries have internal diameters ranging from 2 mm to 6 mm and a wall thickness of about 1 mm. Arterial walls contain large amounts of elastic and fibrous tissue, enabling arteries to withstand relatively high blood pressures, which are higher in these vessels than anywhere else in the vasculature. The thickness of arterial walls and its elastic tissue allows arteries both a certain stiffness and the ability to expand and contract as the blood pressure rises and falls with each heartbeat. This combination of stiffness and flexibility enables arteries to reserve blood pressure to ensure a continual, smooth flow of blood through the vasculature.

54 Arterioles are blood vessels that conduct blood from arteries to capillaries. Arterioles are not visible to the unaided eye. The inner diameter of arterioles averages 0.03 mm. The major function of arterioles is to serve as points of control for regulating the flow of blood through the capillary beds. Blood flow is regulated in arterioles by the contraction or relaxation of circular smooth muscle. Contraction of circular smooth muscle cells allows arterioles to constrict, which decreases the flow of blood through them and through the capillaries downstream; relaxation of these smooth muscle cells allows arterioles to dilate, increasing the flow of blood.

Capillaries are the smallest and most numerous blood vessels in the body with diameter range from 5µm to 10 µm. The primary function of capillary is to permit the exchange of materials between cells in tissues and the blood. The thinness of capillary walls permits small molecules (for example, oxygen, carbon dioxide, sugars, amino acids, and water to enter and leave capillaries readily, promoting efficient material exchange. Venules are slightly smaller than arterioles, averaging about 20 µm in diameter. Venules, like capillaries, function in the exchange of materials, and they conduct blood from capillaries to veins.

Veins have roughly the same diameter as arteries. A typical vein has an internal diameter of 5 mm but a wall thickness of only 0.5 mm. The largest veins are even larger in diameter than the largest artery (30 mm, as opposed to 12.5 mm) but have a wall thickness of only 1.5 mm, compared to 2 mm for the largest artery. The relative thinness of the walls of veins reflects the fact that blood pressure in the veins is significantly lower than in arteries. Unlike any other blood vessels in the body, veins are equipped with one-way valves that permit blood to flow toward the heart but prevent it from flowing back toward organs and tissues (see Figure 2.13). More details of blood vessels can be found in German and Stanfield (2000, pp.411-417).

2.4.2.2 Blood Pressure Blood pressure is a result of pumping action of the heart to create enough force to push blood through the arteries, into the arterioles, and finally into the tiny capillaries. Blood pressure is expressed as two numbers (for example, 120/80). The first is the pressure during systole. The systolic pressure measures the force that blood exerts on the artery

55 walls as the heart contracts to pump out the blood. The second number is the pressure at diastole. The diastolic pressure is the measurement of force as the heart relaxes to allow the blood to flow into the heart. Blood pressure is measured in millimetres of mercury (mm Hg). In this example, the systolic pressure equivalent to pressure that produced by a column of mercury 120 mm high.

The pressure of arterial blood is largely dissipated when the blood enters the capillaries. However, the remaining blood pressure is still enough to push a substantial amount of water and some plasma proteins filter through the walls of the capillaries into the tissue space. When blood leaves the capillaries and enters the venules and veins, little pressure remains to force it along. Blood in the veins below the heart is helped back up to the heart by the muscle pump. This is simply the squeezing effect of contracting muscles on the veins running through them. One-way flow to the heart is achieved by valves within the veins. More details of blood pressure can be found in German and Stanfield (2000, pp.417-429).

High blood pressure (or hypertension) is elevated pressure of the blood in the arteries. Two main factors cause hypertension: (1) the heart pumps blood with excessive force; and (2) the arterioles narrow resulting blood flow exerts more pressure against the vessels’ walls. The World Health Organisation defined systolic blood pressure over 160 mmHg or diastolic blood pressure over 90 mmHg as a threshold of high blood pressure.

Hypertension is mainly categorised into two: (1) Primary Hypertension; and (2) Secondary Hypertension. The causes of Primary Hypertension are unknown but are certainly based on complex processes in all major organs and systems, including the heart, blood vessels, nerves hormones, and the kidneys. On the other hand, the cause of Secondary Hypertension, which is a main concern of this thesis, can be identified and treatable. The main causes of Secondary Hypertension are medical conditions (such as kidney disease particularly in older people, sleep apnea or temporary cessation of breathing during sleep, and pregnancy), medications (such as corticosteroids, long-term use of nonsteroidal anti-inflammatory drugs, cold medicines containing pseudoephedrine, and oral contraceptives), alcohol, coffee, cigarettes, and emotional factors (such as mental stress, anxiety, and depression).

56 2.4.3 Interference with Communication by Aircraft Noise One of the basic elements of human well-being is to have ability to clearly form and understand speech. There are two reasons causing human’s auditory incapability to understand speech. Firstly, that speech of interest is masked by some other noise and secondly, the sensorineural hearing system itself does not respond due to damage or fatigue. The latter reason will be briefly discussed in section 2.4.6. The topic of how humans can produce speech and understand the meaning of speech can be found in Saenz and Stephens (1986, pp.249-263).

In a laboratory setting when the background noise is constant, the mean speech levels of adult males and children are higher than adult females for all vocal efforts (i.e., casual, normal, raised, loud, and shout). On average, children speak with the same speech level of adult males in casual, normal, and raised vocal efforts but they are lower during loud and shouting vocal efforts (Kryter, 1994, table 6.1, p.295). For normal speech at normal distance (1 metre) between talkers, the Leq is approximately 57 dB(A) with 0.3∼0.8 kHz. The SPL of various vocal efforts also varies dependent on the distance between talkers (the longer distance, the louder speech level).

In a real world condition, the vocal efforts would be changed involuntarily depending on surrounding factors. Figure 2.14 summaries all of these related factors and draws both calculated and estimated regression lines from plotted data (Pearsons, 1977 as presented in Kryter, 1994, figure 6.6, p.298). The figure is composed of three main axes, the x-axis for background noise, two parallel y-axes for vocal efforts and speech level. The different labels of the plotted data show different locations from where the data was measured.

57

Figure 2.14: Speech Levels as a Function of Background Noise Level (source: reproduced from Kryter, 1994, figure 6.6, p.298).

The background noise level starts from 30 dB(A) in homes and increases up to 80 dB(A) during travelling in a train and aircraft. The speech would be uttered by the talkers at constant level of about 55 dB(A) (called normal vocal effort) during a background noise level of 35-50 dB(A) and then raised 0.5 dB(A) for each 1 dB(A) increment of background noise level, and, finally becomes constant again after background noise level passes 65 dB(A) at 67 dB(A) speech level (called between raised and loud vocal efforts). It should be noted that both regression curves were drawn from the plotted data measured at constant distance between talker and receiver for individual locations (i.e., 0.4 m for train and aircraft, 1 m for house and department store). Any alterations in measured distances will result in modification of the overall feature of both curves.

Speech intelligibility means the ability of the auditory system to receive, transduce, and transmit to the brain of speech signals so they correctly perceive and identify words, phrases, or sentences. Various factors can affect speech intelligibility such as the combination of speech, the level and frequency of masking noise, and the level of vocal effort (see Kryter, 1994, pp.301-311).

58 Berglund et al (1999, pp.56-57) recommend the signal to noise ratio (S/N) (which is the difference between the speech level and the sound pressure level of the interfering noise) should not be below zero dB to ensure speech communication. However, for speech to be intelligible when listening to a complicated message (such as in a classroom or during a conference), the S/N should be at least 15 dB. Therefore, it is strongly recommended that for speech intelligibility the indoor noise level of dwelling and classroom should not exceed 35 dB(A) (Berglund et al, 1999, table 4.1, p.65).

Interference with communication is the most direct impact to a person residing or working around an airport. Aircraft noise has high potential to disturb communication activities either indoor or outdoor. The interference with communication (also with the other types of aircraft noise disturbance) undoubtedly is an important contribution of general aircraft noise annoyance. The patterns of disturbance from aircraft noise are different from other transportation noise sources. While road traffic noise is considered to be a continuous event, aircraft noise is intermittent and with higher peak noise levels. Aircraft noise penetrates though the house from various directions (for example, roof, door, and window) and, therefore, can be equally received by all members no matter where they are in the house. It is obviously different from the road traffic and railway noises which propagate mostly in just one direction and will be initially attenuated by the structure of the house at the front before reaching people inside the house or rear garden. Outdoor noise level from aircraft noise is less attenuated by the structures of house than the other transportation noises (see Kryter, 1996, pp.603-609). Therefore, at the same outdoor noise level, people seem to evaluate aircraft noise as being more annoying than noise from road traffic or railways.

Figure 2.15 supports the previous statement by showing a summary of activity interference due to transportation noises (conducting from European countries and assembled by Schultz, 1978) under the categories of disturbance of conversation and listening to radio or television. The y-axis represents the percentage of people who describe themselves as disturbed by noise and the x-axis represents the average day- night sound level (Ldn). It is obvious that the activity interference due to aircraft noise is higher than the other noise sources. The disturbance due to noise from a street is the lowest. From both figures, on average, the threshold of interference is around Ldn at 46

59 d(B). Grandjean (1980) (as stated in Kryter, 1994, figure 10.8, p. 588) and Miedema (1998) also found that community response to aircraft noise more annoying than to noise from ground transportation noise.

Figure 2.15: Interference by Transportation Noise with (a) Conversation; and (b) Radio/Television Listening (source: reproduced from Schultz, 1978, figure 20, p.388 and Schultz, 1978, figure 21, p.388).

Aircraft noise exposure disturbs teaching activities in schools. It becomes worse especially during the summer when there are opened windows and doors. Crook and Langdon (1973) studied the behavioural change of teachers due to aircraft noise in schools close to , . They found that at least in one of four flyover with peak noise (inside class) at, or above, 70 dB(A) teachers would pause their lesson and quoted themselves as feeling uncomfortable during teaching. However, when dealing with a small group or individual teaching, the nuisance noise level shifts to 75 dB(A). Ko (1979) studied a response behaviour of teacher to aircraft noise in schools around Kai Tak International Airport (which is now closed down), Hong Kong. The results of the study revealed a good correlation between behaviour changes of teachers and the Noise and Number Index (NNI) calculated indoors during school periods. Almost always teachers would pause their lessons to overcome interference with

60 communication during flyovers rather than raise their voice (or shout). Therefore, teachers themselves stated that their teaching was interrupted or discontinuous with frequent pauses and this may result in the loss of concentration by students.

2.4.4 Effects of Aircraft Noise on Performance in School Children In general, the studies of the effects on task performance have been undertaken for occupational noise, either inside or outside of the laboratory. The basic conclusion from those studies is that people who work in high noise areas have a higher deterioration in task performance (i.e., memory, concentration, accuracy, cognition, caution) than those people who work in low noise areas (Smith and Jones, 1992). Nevertheless, other factors are also involved, such as familiarity and unfamiliarity with the noise, types of task, professional abilities, physical conditions, surrounding environments of workplace, age, and gender. Even though noise-induced arousal may produce better performance, it is only for a short period and for a simple task.

For environmental noise, the effect on task performance of children in the classroom is a critical issue. Children have less ability to concentrate on the task than adults and are easily distracted by distracting stimulus. This section spotlights only previous studies of the effects of aircraft noise on performance, especially of children in schools. Further detail of the performance impacts by the other noise sources can be observed from Smith and Jones (1992, pp.1-28).

The obvious effects of aircraft noise on task performance are the effects in schools or places of education. Loud noise from aircraft disturbs both teachers and students during class, or mental activities such as reading and discussing. Flyover noise not only distracts student’s concentration from class but it also forces teachers to raise their voices to sustain a lesson, or, eventually, to pause until the aircraft passes over. This retards the learning development stage of students and strains teachers. Moreover, it might impair both cognitive and motivative performance of student and sometimes can push both teachers and students to move away from the school to another school.

The performance effects on school children have been studied by Cohen et al (1980). School children were selected from (noisy) elementary schools around Los Angeles

61 International Airport, United States, and matched with control (quiet) schools. They found that children from a noisy school have low level of persistence (or high level of helplessness) to finish a given task (puzzle) and more likely to give up before the time to complete the task has elapsed. They also found some evidence that the cognitive performance and noise annoyance of children have not been habituated over time. Increases in length of exposure resulted in an increased negative impact of noise on children.

A follow-up study was conducted by Cohen et al (1981) one year after the first study. The same subjects were tested with similar procedures as with the first study. The second study was intended to identify the longitudinal effects of chronic aircraft noise on children cognitive performance adaptation to confirm the findings of the first study. They found that the results of second study were relatively stable compared with the first study. They concluded that the chronic aircraft noise resulted in cognitive impairment of school children. Berglund et al (1999, p.50) supported these findings by stating that chronic aircraft noise exposure in early childhood appears to impair reading acquisition and reduces motivational capabilities. Longer exposures will result in greater damage.

Haines et al (2001a and 2001b) examined the aircraft noise impacts on cognitive performance of school children around Heathrow Airport, England. Children selected from high aircraft noise exposure schools and control schools were tested in various aspects such as reading comprehension, long-term memory, short-term memory, and motivation. They found that association between chronic aircraft noise exposure and reading (on difficult items) is significant after adjustment for age, household deprivation, and main language spoken. The association between chronic aircraft noise exposure, motivation, and memory (both long-term and short-term) are insignificant. They concluded that chronic aircraft noise exposure impairs difficult cognitive performance of children.

Hygge et al (2002) studied the effects of aircraft noise on cognitive performance in school children around both old and new Munich Airports in Germany. It is a prospective study where there are three periods of data collection: 6 months before, and

62 1 and 2 year after the changeover of airports. The control groups were matched based on sociodemographic characteristics. The subjects were tested on reading, memory, attention, and speech perception. They found that, after the switch of airports, long-term memory, reading, and speech perception of children from the aircraft noise exposure group of the new airport were impaired.

Haines et al (2003) studied qualitative responses of children to environmental noise. The study involved two groups of children: (1) focus group (low effected by aircraft noise) (n=36); and (2) exposure group (highly effected by aircraft noise) (n=18). The impact of noise pollution on everyday activities (for example, schoolwork, homework, and playing) was larger for the children exposed to high levels of aircraft noise compared with the low noise exposed children. The range of coping strategies that children employed to combat noise exposure in their lives was dependent upon the amount of control they had over the noise source. The emotional response of children describing the annoyance reaction to noise was consistent with adult reactions and it would seem that child noise annoyance is the same construct.

Matsui et al (2004) conducted a cross-sectional study to examine the association between aircraft noise exposure and children’s cognitive performance. The sample size was 236 from primary schools around Heathrow Airport in west London. Children were tested on reading comprehension, immediate/delayed recall and sustained attention. It was found that a significant dose-response relationship existed between aircraft noise exposure at home and performance on memory tests of immediate/delayed recall. No strong association was found for the other cognitive outcomes. The study suggested that aircraft noise exposure at home may affect children’s memory.

One interesting study, conducted by Maxwell and Evans (2000), is concerned the effects of noise on pre-school children’s pre-reading skills. The subjects are from the same child care centre located far away from any major external noise source. Therefore, the noise levels concerned by this study were generated by people within the building and a consequence of poor acoustical design. The data collections were implemented before and after the installation of sound absorbent panels. The study

63 found that pre-school children performed better in the quieter conditions on the cognitive measures.

There is strong evidence from previous studies that chronic aircraft noise exposure affects reading, memory, and attention of school children. The above results are very useful, but many factors should be considered carefully before any attempt is made to replicate those results to the other schools. They need to be comparable in various aspects such as the location of classroom or building, the condition of room (insulation equipment or air condition installation), the surrounding environment of school, and the quality standards of teacher and facility.

2.4.5 Sleep Disturbance by Aircraft Noise This section focuses on the impacts of aircraft noise on sleep activity. Previous studies revealed that noise exposure during sleep may increase blood pressure, heart rate and finger pulse amplitude as well as body movements (Spreng, 2004). It is assumed that increases in aircraft noise level lessens the quality of sleep. Loud noise from overflight can make people have difficulty in getting to sleep or they can wake them during the night. These types of disruption cause people to become tired and fatigued due to a lack of good quality of sleep. This might result in declining performance at work on the next day, or getting stressed or frustrated by stimulus that would be easier for people who obtained a good night’s sleep (Hoeger, 2004).

The basic nature of sleep for normal people starts from the lightest stage (or wakefulness), passes through the four labelled of stages (stage I, II, III, IV), and finally reaches the deepest stage (or dream, or rapid-eye-movement (REM)). After that, the stage of sleep returns back to stage II or III and then reaches again the REM stage. This is repeated four to five cycles during the course of a sleeping period until wake up (Nelson, 1987, pp.5.3-5.4). Those changes in stages of sleep can be recorded by the measurement of the changes in electrical brain activity, or electroencephalogram (EEG) activity.

Many factors can affect sleep such as physiological and psychological factors of the sleeper, socioeconomic factors, and environmental factors. One of the physiological

64 factors which plays a major role to influence the nature of sleep is age. In general, a newly born baby needs much more time to sleep than older people. The required numbers of hour for an adequate amount of sleep will decrease with the age of the sleeper. The pattern of sleep can vary between different age groups of person. In the case of a baby, the alteration of sleep and wakefulness occur 24-hours a day. For adult people, except someone who has to sleep during the day according to their work shifts, they will sleep during the night and be awake during daytime. Older people (more than 70 years) will become wakeful during the night but feel sleepy during the day. In general, the frequency of arousal, or awakening, increases with age (Lukas, 1975, p.1233).

Activities or behaviours of people before sleep also affect the structure of sleep. For example, hard exercise causes a sleeper to stay in deep sleep. Anxiety (or pondering over some thoughts) or a new sleeping place, will make it more difficult for a sleeper to get to sleep or to spend long periods in the deep stage of sleep. The consumption of medicine, or the symptoms of insomnia, both affect the structure of sleep. The effects of insomnia causes people to get up early, make it difficult to get to sleep, or frequently awaken during the night. Use of medicine can make it more easy or difficult to get to sleep depending on the medical types. Coffee, or some types of stimulus drinking, can make a person stay awake.

There are four methods to measure the effects of noise on the quality of sleep. The first method is to measure the total number of subjects awaking per night. The second method is to measure the changes in electrical brain activity (or the EEG). The third method is to interview the sleeper’s opinion about the previous night’s sleep. The last method is to record the body movement of the subject. The first three methods are practical for laboratory tests while the last method was developed for the purpose of field observations.

Previous research has studied the effects of noise on sleep both in a laboratory setting and at the subject’s house. Similarly with the dosage-response relationship between noise exposure and annoyance synthesised by Schultz (1978), the dosage-response relationship between noise exposure and sleep disturbance has been synthesised and

65 proposed by many studies of which the two main studies are Finegold et al (1994) and Pearsons et al (1995). Both studies used similar data, but found the final results to be different.

Pearsons et al (1995) found that there are large and systematic differences in sleep disturbance observed between the studies conducted in laboratory and in field settings. The effects of noise on sleep are stronger under laboratory conditions than in a customary sleeping place. They suggested there are many factors influence effects of noise on sleep, such as the nature of noise and response, noise source, background noise level, length of study, and gender of participants which differed from study to study. They also found that although noise disturbs sleep in a field setting to a lower degree than in the laboratory, field settings provide a more reliable result. Finally, they concluded that it is an effect from a habituated response (which is a learned ability to ignore the disturbance of noise intrusions, see Kryter (1994, pp.484-489) for more detail) to noise which plays a major role in any discrepancy between the effects of noise on sleep from laboratory studies and field studies. Pearsons et al (1995) recommended that the results from the laboratory setting should not be used to predict the effect of noise on sleep in community settings.

Finegold et al (1994) agreed with Pearsons et al (1995), but they have implemented further analysis by including some additional field studies of effects on sleep as a consequence of night time aircraft operation. In accordance with the Federal Interagency Committee on Aviation Noise (FICAN), the relation curve between sleep disturbance (% of awakening) and indoor A-weighted sound exposure level (ASEL) has been proposed as presented by Figure 2.16.

66

Figure 2.16: Relationship between Sleep Disturbance and Indoor A-Weighted Sound Exposure Level (ASEL) (source: reproduced from FICAN, 1997, figure 1).

FICAN (1997) recommended Figure 2.16 for predicting awakening due to sound exposure level used in environmental impact analyses until sufficient data from additional field studies are available. It is noted that Figure 2.16 should be used to predict the maximum percent of the exposed population expected to be behaviourally awakened, and the prediction is limited to only adult subjects who have been living in the area for a long time.

Only a few researchers have studied the effects of aircraft noise on sleep in field settings. FICAN (1997) overviewed studies, as reproduced by the following paragraphs. In London, residents residing around four airports were tested by activity meters (for 5,742 subject-nights) and EEG (for 178 subject-nights). The study revealed that even at the high event noise level, once people fall asleep, very few of them have any substantial sleep disturbance.

In Los Angeles, 85 residents near Castle Air Force Base and near Los Angeles International Airport and a control area, were tested for awakening by occurrence of noise intrusions. Participants were instructed to push a button attached by a cable to a palmtop every time they were awakened. Simultaneously with the pushing of buttons by participants, the noise level was recorded as a noise event. It was found that only 16%

67 of total subject nights were awaken by noise, but almost 85% were spontaneously awakened (unassociated with noise events). Conclusively, there is no evidene that even a high level of noise will lead to a bad quality of sleep.

In Denver, a study was conducted near Stapleton International Airport (DEN) and the new Denver International Airport (DIA) in anticipation of the closure of DEN and the opening of DIA. Fifty-seven homes located near the runway ends at both airports were tested (before and after the changeover of the airports) for sleep disturbance by several methods, including button pushes upon awakening and body movement record. The large differences in noise-induced sleep disturbance at either airport could not be detected. The average number of behavioral awakenings per night was 1.8 at DEN and 1.5 at DIA. The number of spontaneous awakening response was 1.5 per night at DEN and 1.3 at DIA.

Based on the three field studies, FICAN (1997) has re-analysed and proposed a dosage- response relationship between aircraft noise exposure and sleep disturbance as presented by Figure 2.17. FICAN suggested and recommended the use of this curve to represent the upper limit of the observed field data for predicting the maximum percent of awakened population by aircraft noise. Similar to Figure 2.16, Figure 2.17 is limited to long-term residents excluding children, because only adults were included in the previous three field studies.

A recent study conducted by Fidell et al (2000) considered the change of sleep behaviour due to the major changes in aircraft noise exposure near DeKalb-Peachtree Airport (PDK), Atlanta, that expected increased night time operations due to the Olympic Games in 1996. Twelve single-family detached homes near the PDK were tested 15 days prior to the opening ceremony of the Olympic Games, continued through the Games, and ended one week after the closing ceremony. It was found that the subjects were not highly disturbed by aircraft noise during night time, and, moreover, they seem well-adapted to noise intrusion. There are no major differences in noise- induced sleep disturbance before and after the Olympic Games.

68

Figure 2.17: Sleep Disturbance by Aircraft Noise Dosage Response Relationship (source: reproduced from FICAN, 1997, figure 2).

2.4.6 Human Auditory Effects by Aircraft Noise For affected people, auditory effects can leave an absence to hear or understand sound which can be well perceived by young healthy people. These effects can be caused by the result of ageing factor (or presbyacusis), the long-term exposure of highly intense sound, and the other auditory disorders by such as hereditary factors, disease, infection, and accidents.

The level of hearing threshold (or hearing sensitivity) deteriorates with increasing age. The hearing threshold is the lowest SPL which can be detected by the ear. The hearing threshold of sound rapidly increases at high frequencies, and the incline is more severe for men than for women (Nelson, 1987, pp.4.7-4.8). Moreover, a person with pathological disorders will have higher threshold shift level than the healthy person.

As previously mentioned in section 2.4.1.1, hair cells are a major factor of the ear to interpret the meaning of signal sound. Their structure is very vulnerable and irreversible once damaged. One of possible causes of destruction of hair cells is the exposure of intense sound. If the ear is exposed to this intense sound for only short period, the threshold of hearing will be temporarily shifted because of the recovery ability of the ear. This ability requires a space of time after quiet. This phenomenon is called noise- induced temporary threshold shift. Sound creating noise-induced temporary threshold

69 shift will not destroy the hair cells because it is considered too short and too moderate. However, more frequent noise-induced temporary threshold shift and/or exposure to very intense sound can result in permanent threshold shift because the demolition of hair cells. This is called noise-induced permanent threshold shift.

Figure 2.18 (developed by ISO DIS 1983 as presented in Nelson, 1987, figure 4.2, p.4.9) illustrates the relation among the noise-induced permanent threshold shift, exposure level, and exposure duration. It implies that noise exposure under 85 dB(A) with 4kHz will not contribute to noise-induced permanent threshold shift. Therefore, the effects of aircraft noise on community in noise-induced permanent threshold shift seem to be insignificant because, in general, there are only few residences located inside this aircraft noise boundary. Study of noise-induced permanent threshold shift normally relates to occupational noise rather than environmental noise. The group of people who can be affected in auditory system by aircraft noise is, for example, airport ground staff (i.e., aircraft mechanical, maintenance staff, or those who work close to aircraft engines or runways).

45 40 35 30 85 dB 90 dB

kHz (dB) 25 4 20 95 dB 100 dB 15 10 NIPTS at at NIPTS 5 0 0 1020304050 Exposure duration (years)

Figure 2.18: Relation of Noise-Induced Permanent Threshold Shift (NIPTS) at 4kHz as

a Functions of Noise Exposure Level (Leq(8h)) and Duration of Exposure (source: reproduced from Nelson, 1987, figure 4.2, p.4.9).

Nevertheless, there are a few studies concerned with the effects of aircraft noise on the auditory system of humans. Recent studies on the auditory effects of aircraft noise on children have showed controversial results. Chen and Chen (1993) found a significant association between aircraft noise exposure and prevalence of noise-induced hearing

70 loss of schoolchildren in Taiwan. However, this conclusion was objected to by Wu et al (1995) who compared the hearing threshold levels of two groups of school children: one from a school located in high aircraft noise exposure area (near CKS International Airport, Taiwan) (N=193) and another from a school located in low aircraft noise exposure area (N=49). They found that there was no statistically significant difference in the hearing ability of school children between two groups. Nevertheless, Chen et al (1997) argued that the size of Wu et al (1995)’s study groups were not comparable leading to the discrepancy in the final result.

Chen et al (1997) studied auditory effects of aircraft noise on a community residing near Kaohsiung Airport, Taiwan. The functions of cochlear and retrocochlear of subjects from aircraft noise exposure group (N=83) and control group (N=93) were measured. After adjusting for necessary demographic characteristics, they found that hearing ability of subjects living near the airport (frequently exposed to aircraft noise) was significantly lower than subjects from the control area. The studies of auditory effects by aircraft noise in Taiwan remain controversial. Most previous studies of aircraft noise focus attention on the non-auditory impacts. Even though there are many difficulties in determining precise correlation between hearing loss and environmental noise, it is worth doing since hearing damage is the most serious consequence of noise exposure.

2.4.7 Human Physiological Effects by Aircraft Noise Noise has been defined as an environmental stressor (Cohen et al., 1986). Exposure to sudden or uncontrollable intense noise activates the autonomic and hormonal systems, leading to temporary changes, such as increased blood pressure, increased heart rate and vasoconstriction. After prolonged high noise exposure, susceptible individuals in the general population may develop permanent effects (persistent increase in stress hormone level and blood pressure) which lead to hypertension and cardiovascular diseases. Stress in the context of this thesis is referred to as “emotional stress” (Berglund et al., 1999).

From a scientific and medical point of view, the correlation between emotional stress and heart disease is weak. However, it is evident that people who live in a chronically

71 stressful condition are more likely to behave in such ways (i.e., smoking, overeating, less exercise, drinking alcohol) that have a direct effect on the development of heart diseases. Moreover, it is well understood that emotional stress causes the blood clotting in coronary arteries leading to reduction of blood flow to the heart muscle, and, finally, may become a major factor in heart attacks.

If annoyance by aircraft noise can stress people, or make it difficult for them to sleep, the assumption that the usage of medicine to release stress or assist sleep is higher in high noise area than in low noise area would be true. Knipschild and Oudshoorn (1977) (as stated in Kryter, 1994, pp.542-543) studied this issue by targeting people around Schiphol Airport, Amsterdam, Netherlands. They found that the rate of use of medicinal drugs (i.e., hypnotics, antacids, cardiovascular drugs, and antihypertensive agents) of people in high noise area (Ldn>64) was higher than of people in low noise area (Ldn<51). They also explained that the rate of medicinal drugs used in low noise area was stable but, conversely, for high noise area this rate was increased progressively by the year of noise exposure. However, the rate of use of sedatives in high aircraft noise area was considered stable by the exposure year because of curfew operation during night time adopted by Schiphol Airport.

Another type of study similar to the study of the rate of use of prescriptive drugs is the study of the rate of physician contact for health problems. Knipschild (1977) (as stated in Kryter, 1994, pp.538-542) found that the rate of physician contact for all four-health problems (psychological problems, psychosomatic problems, cardiovascular disease, and hypertension increase) was higher in high aircraft noise area than in low aircraft noise area. These four-health problems were considered representative of diseases caused by stress from aircraft noise. Knipschild (1977) reported that the possibility of the main conditions of cardiovascular disease occuring (i.e., hypertension, pathological heart shape, taking of cardiovascular drugs) in people in noise areas where Ldn>62.5 was significantly higher than for people in noise area where Ldn<62.5 when adjusted for age, sex, smoking habits, weight, and size of population. Morrell et al (1997, p.228) criticised the studies by Knipschild. They suggested that the response rate of Knipschild’s studies from the whole community sample was low which may lead to

72 some biased results where there is a higher proportion of participation residing in high noise areas.

Another way to understand the effects of aircraft noise on cardiovascular disease due to stroke or cirrhosis of the liver is to measure the mortality rate of people around an airport. After adjusting for age, race, and gender, it was found that aircraft noise has no relationship with the rate of mortality (Nelson, 1987, p.4.5 and Kryter, 1994, p.546). Concern then shifted to the effects of aircraft noise on reproductive outcomes because it was assumed that stress from noise exposure will lead to higher consumption of drugs and alcohol by pregnant women than normal which can result in a fetus with many aspects such as low birth weight, premature birth and fetal abnormality (Morrell et al, 1997, p.230). The most rigorous study by Edmonds et al (1979) found no difference between rate of birth defects between high and low aircraft noise areas in Atlanta which brings to a conclusion that, until there is no further arguments from other more rigorous studies, aircraft noise has no evident effects on reproductive outcomes.

Previous studies have paid many attentions to study the effects of aircraft noise on stress in children. Children are considered as a high risk group vulnerable to noise exposure (or any stimulus). Cohen et al (1980) and Cohen et al (1981) measured blood pressure of school children selected from elementary schools located around Los Angeles International Airport. The control (quiet) schools were selected where socioeconomic status and grade level were comparable with the noise exposure schools. Each child’s resting blood pressure was taken on an automated blood pressure recorder. The study was longitudinal in design with only one year follow-up. At the first step of the study, they found that children from noisy schools have higher blood pressure than those from control schools. Statistical methods reveal that this difference (3 mm Hg in both systolic and diastolic blood pressure) was significant. At the follow-up state (1 year later), however, there was no evident of difference of blood pressure between those two groups. The authors described that it was a situation when noise sensitive children have been relocated from the noisy area. Nevertheless, the study suggested that there is a linage between chronic aircraft noise exposure and increasing in blood pressure of school children.

73 Evan et al (1998) examined the elevation of psycho-physiological stress of children before and after the inauguration of a new Munich International Airport, Germany. Subjects both from aircraft noise exposure areas and matched areas were tested in a sound-attenuated, climate controlled, mobile laboratory. Resting blood pressures were taken by an automated blood pressure monitoring. Twelve-hour overnight urine samples were collected and used to analyse the stress hormone levels (i.e., epinephrine, norepinephrine, and cortisol) of each subjects. After the opening of the new airport, it was found that the resting blood pressures of noise exposure group significantly increased, but changed slightly in the control group. Both epinephrine and norepinephrine increased rapidly in noise exposure group, but increased slightly in the control group. Changes in urinary cortisal were not significant in both groups. The study also demonstrated that the quality of life (physical, psychological, social, and functional daily life) declined significantly in the noise exposure group, but remained relatively stable in the control group. It is noted that there is no evidence that Evan et al (1998)’s study has controlled for any necessary confounding factors. Therefore, the conclusion that aircraft noise has physiological effects on children may be debatable.

Haines et al (2001a and 2001b) studied the effects of long-term aircraft noise exposure on stress responses (which are catecholamines, cortisol, and perceived stress) and mental health outcomes (which are depression, anxiety, strengths and difficulties of children, and self-reported health) of school children randomly selected from primary schools located around Heathrow Airport, England, and control schools in the vicinity that have a low level of aircraft noise. The study is cross-sectional in design. After controlling for potential confounding factors, it was found that chronic aircraft noise exposure was not associated with stress responses. The health outcomes measures revealed no significant difference of mental health scores between both groups.

In Sydney, Australia, Morrel (2003) studied the effects of aircraft noise on children sampled from primary schools within 20 km radius of the Sydney International Airport. The study was longitudinal in design with a baseline study (N=1,230) in 1994 and 1995, and follow-up study in 1997 (N=628). Both systolic and diastolic blood pressures of the subjects were taken using automated blood pressure measuring equipment. After adjusted for necessary confounding factors (i.e., body size, child activity levels, use of

74 salt on food, history of parents of high blood pressure, consumption of breakfast before school, ambient temperature, and background noise level), the study found no consistent statistically significant cross-sectional correlations between mean resting blood pressure level and exposure to aircraft noise either during the baseline study or follow-up.

There has been no robust epidemiological study of aircraft noise dealing with cardiovascular diseases in adults. However, of the traffic noise studies, the Caerphilly (Wales) and Speedwell (England) study conducted by Babisch et al (1993) and Babisch et al (1999) is considered as the most persuasive study dealing with cardiovascular diseases. It was longitudinal and prospective in design with 10 year follow-up. The participants of Caerphilly and Speedwell study were 2,512 and 2,348 men, respectively. The incidence of ischemic heart disease (IHD) (insufficient perfusion of oxygen to the heart muscle) was recorded between the follow-up phases. These events could be, for example, IHD death, definite clinical nonfatal myocardial infarction (MI), electrocardiogram (ECG) and enzyme changes, or ECG-defined-MI. After controlling for potential confounding factors, it was found that there was no convincing evidence to support a causal relationship between traffic noise and IHD. However, they suggested that it was evident that there may be a slight increase in the risk of IHD in people who live in the areas where outdoor traffic noise (LAeq, 06-22) is greater than 65 dB(A).

Some recent studies are concerned with the effects of aircraft noise on hypertension and health measures quality of life. Bronzaft et al (1998) determined health perception and quality of life of people in New York, USA affected by aircraft noise. Two study areas were selected. One (called high aircraft noise area) was located in DNL≥65 while another (called low aircraft noise area) with DNL<65. Both areas were comparable in socioeconomic status. The postal questionnaire was distributed to over 3,000 randomly selected subjects (1,500 for each area). The response rate was 18% with only 9% valid for further calculation. The authors describe that the respondent from both groups were not different in age, gender, race, and education. Chi-square analysis demonstrated that subjects who were bothered by aircraft noise were more likely to complain of sleep difficulties and more likely to perceive themselves to be in poorer health.

75 Miyakita et al (2002) studied the health effects of aircraft noise residing around Kadena and Futenma airfields, Japan, using the Todai Health Index. It consists of 130 questions, 12 scale scores measuring varieties of physical and mental health outcomes. The aircraft noise exposure groups were stratified by DNL. The control group was located in a non- aircraft noise exposure area. The study was cross-sectional in design. The questionnaire was distributed by means of the leave-and-pick-up technique. The total sample size was 8,084 subjects aged 15-75 with 82.8% response rate. After controlling for age, gender, and occupation, multiple logistic regression suggested that residents living around Kadena airfield suffered from both physical and mental effects due to chronically exposure to military aircraft noise and the effects increased with the level of noise exposure as defined by DNL.

A well-defined cross-sectional population-based questionnaire survey on health measure quality of life of aircraft noise has been conducted by Meister and Donatelle (2000) in Metropolitan Minnesota, USA. The questionnaire measured general health outcomes (by MOS-36, see section 4.3.1.1), perceived stress, noise sensitivity (by the Weinstein Instrument, see section 4.3.1.4), noise annoyance, and demographics, and was distributed to over 4,000 residents aged 18-99 in both aircraft noise affected areas and control areas with a 52% response rate. The aircraft noise exposure group was classified by the mean of the annual maximum decibel levels and the frequency of noise loudness. After controlling for potential confounding factors (i.e., marital status, income, education, smoking status, body mass index, gender, age, and sensitivity to noise), the multiple analysis of covariance (MANCOVA) revealed that all health measures (which are general health, sense of vitality, and mental health) were significantly worse in the community areas exposed to aircraft noise. Stress and noise annoyance were also significantly worse for those subjects exposed to high level of aircraft noise.

Rosenlund et al (2001) investigated a relation between residential exposure to aircraft noise and prevalence of hypertension. The study population comprised of two random samples aged 19-80 years, one living in the vicinity of Stockholm Arlanda Airport, Sweden, and another residing in non-aircraft noise exposure areas for the purpose of comparison. The subjects were classified by the time weighted equal energy levels and the maximum aircraft noise levels. The study was cross-sectional in design. The

76 questionnaire measuring history of hypertension (“Have you been diagnosed for hypertension by a physician during the past 5 years?”) and individual living and consuming behaviours were distributed by mail with a 73% response rate. The total sample size was 4875 subjects. After being adjusted for significant confounders, the multivariate logistic regression revealed that the association between aircraft noise and prevalence of hypertension was marginally significant. The authors suggested that long- term aircraft noise exposure may lead to the development of hypertension.

Goto and Kaneko (2002) conducted a cross-sectional study to compare the blood pressure data (N=469 women) of people living around Fukuoka Airport, Japan and data (N=1,177 women) from a quiet area in the same city. The data was collected from the general health examination data around the city. It was found that both diastolic blood pressure (DBP) and systolic blood pressure (SBP) (controlled for variables regarding anti-hypertension treatment, smoking status, food preferences, and alcohol consumption) were not significantly different between the two groups. The authors also observed the changes of blood pressure levels (during 1980 to 1988) of 183 women residing around the same airport. The study samples were classified by Weighted Equivalent Continuous Perceived Noise Level (WECPNL) into three groups: high, low, and control. No significant differences among the three groups were observed. The authors concluded that there was no evidence to support that aircraft noise is a risk factor of increasing in blood pressure level.

Franssen et al (2005) conducted a cross-sectional study to study the impacts of aircraft noise on prevalence of general health status, use of sleep medication, and use of medication for cardiovascular diseases. A postal self-administrative technique was employed. The general health status was measured mainly by two ways: (1) with a single question (“How is your health in general?”); and (2) with a 13-itme questionnaire consisting of a list of health complaints (for example, tiredness, listlessness, and back pain). Note that respondents with six or more symptoms were defined as having a “poor self-rated health”. Questionnaires were sent to 30216 addresses with response rate of 39%. The association between the health indicators and aircraft noise exposure was assessed using a multiple logistic regression model, controlling for potential confounders such as age, sex, education level, country of origin, smoking behaviour,

77 and degree of urbanisation. The study concluded that associations between health indicators (general health status, use of medication for cardiovascular diseases, and use of sleep medication) and the aircraft noise exposure measure (DNL) were significant. None of health indicators were associated with aircraft noise exposure during the night

(LAeq, 23-07h), but use of non-prescribed sleep medication or sedatives was associated with aircraft noise exposure during the late evening (LAeq, 22-23h). Vitality related health complaints such as tiredness and headache were associated with aircraft noise, whereas most other physical complaints were not.

Hatfield et al (2001) examine the possibility that physiological effects of noise may result not only from noise exposure per se, but also from psychological reaction. Residents in each of three areas where aircraft noise exposure was expected to change (higher and lower) or remain the same level due to the opening of the third runway at Sydney Airport were interviewed. The sample size was 1015. The study found that the observed self-reported physiological responses of each area were different significantly regardless of the noise exposure level. Self-reported noise-related physiological health problems were found to be more prevalent in area with worse psychological reaction to noise. For example, Residents of areas with currently low noise exposure reported statistically significantly higher scores for general symptoms, substance use, panic, anxiety, anger and depression if their noise exposure was expected to worsen, than if their noise exposure was expected to remain the same. The study concluded that psychological variables, such as negative attitudes toward noise source, cognitive behavioural, and noise sensitivity, contribute to the adverse physiological effects of noise.

2.4.8 Human Psychological Effects by Aircraft Noise Exposure to noise constitutes a health risk, but evidence that high levels of aircraft noise lead to psychiatric disorders in the community is not well-understood (Stansfeld et al 2000a). To measure the psychological effects from the impacts of aircraft noise, previous researchers have employed the psychiatric hospital admission rate method. The historical data from relevant hospitals were analysed instead of measuring participants directly from laboratory or the field. The hypothesis has been formulated that the rate of psychiatric hospital admission would increase with the level of exposure to aircraft

78 noise. People residing in higher aircraft noise areas would have a higher probability in the occurrence of mental health problems than people in lower aircraft noise areas. The rates of psychologic hospital admission reflect the level of mental health of people around that hospital area.

There are three main procedures with the psychiatric hospital admission rate method. Firstly, the relevant mental hospitals are chosen. The catchment’s area of selected hospitals should cover communities in both high and low aircraft noise areas. The effects from nearby similar hospital(s) should not be significant. Secondly, the patients are classified initially by address based on the level of aircraft noise and then, after being separated into the same aircraft noise zone, they are divided again into smaller groups which have the same characteristics (i.e., age, sex, marital status, and socio- economic status). The psychiatric admission rates then are calculated by the total number of admission in each group divided by the total number of risk-people who have characteristics falling in that group. Finally, regression lines, or basic connected lines, are drawn to obtain the relation between aircraft noise level and psychiatric admission rate. Normally, the first and third steps of work are the same for many studies, but will be different in the second step when various types of classification variables and methods are considered.

Preliminary studies on this issue started with Abey-Wickrama et al (1969), targeting hospitals located in the vicinity of Heathrow Airport, London, and Meecham and Smith (1977) who focused on hospitals around Los Angeles Airport, USA. Both studies reported higher psychiatric admission rates from areas where aircraft noise level was intense (Kryter, 1994, p.526). Gattoni and Tarnopolsky (1973) re-examined Abey- Wickrama et al (1969)’s study by using the same study area and hospital, but updated the new data and also modified some methodologies. However, they also reported similar results as previously done by Abey-Wickrama et al (1969). Higher aircraft noise exposure resulted in higher psychiatric admission rates, but the differential of the psychiatric admission rates between high and low aircraft noise level were not statistically significant (Kryter, 1994, pp.526-527).

79 Jenkins et al. (1981) studied the same data used by Abey-Wickrama et al (1969) and Gattoni and Tarnopolsky (1973), but with more hospitals (three mental hospitals) and a longer study period (four years). They concluded that aircraft noise has no evidence to affect psychiatric admission rates. Moreover they found that the rate of admission from Springfield Hospital, London, shows an unusual trend compared with the other two hospitals, and the previous results found by Abey-Wickrama et al (1969) and Gattoni and Tarnopolsky (1973). The rate of admission of Springfield Hospital actually reduces with increasing noise levels.

Kryter (1990) re-examined Jenkins et al (1981)’s results by using the same data but employed a more comprehensive assessment of the interrelations of all the related socio-economic and aircraft noise variables. It was found that the unusual results for Springfield Hospital occurred because of the unusual large numbers of immigrants that were highly concentrated only in the high aircraft noise zone rather than equally distributed throughout the study area. Kryter (1990) suggested a positive correlation between psychiatric hospital admission rates and the level of aircraft noise.

Tarnopolsky et al. (1980) (as stated in Kryter, 1994, pp.534-535) studied the relation between annoyance and people with psychiatric illness by a screening questionnaire rather than using statistical data (i.e., rate of psychiatric admission). The study found that people with psychiatric illness would be highly annoyed by aircraft noise. However, the number of people with psychiatric illness did not increase with the level of aircraft noise level as the number of annoyed people without psychiatric illness did. The authors concluded that the predominance of annoyance people in high aircraft noise area were people without psychiatric illness. Therefore, most of complaints about aircraft noise were from “normal” people.

Tarnopolsky et al (1978) (as reviewed by Stansfeld et al, 2000b, pp.55-58) found no association between aircraft noise exposure and self-assessed psychiatric morbidity (measured by General Health Questionnaire, GHQ). On the other hand, they found that indeed noise and minor psychiatric disorders were the strongest predictions of annoyance. People with psychiatric morbidity were easier annoyed by noise than normal people. However, not only having psychiatric morbidity leading to annoyance

80 but a wide individual variation in annoyance also occurs in a psychiatrically normal person. This variation can be explained largely in terms of noise sensitivity (Stansfeld et al, 2000b, p.57).

Stansfeld (1992) recommended that noise sensitivity may be an indicator of vulnerability to minor psychiatric disorders. Stansfeld et al (2000b) defined noise sensitivity as a personal vulnerability intervening between noise exposure and individual annoyance responses. General speaking, noise sensitive person feel very annoyed with a certain noise level considered by normal person just to be an acceptable level. Study from Stansfeld et al (1993) confirmed that noise sensitive person were more likely to be annoyed by traffic noise than less noise-sensitive person. Detail description of noise sensitivity can be observed from Stansfeld (1992) and Stansfeld et al (2000b, pp.57-58).

Van Kamp et al (2004) studied the role of noise sensitivity in the relation between aircraft noise and annoyance based on previous data conducted from social survey at three major airports: Sydney Airport, London Heathrow Airport, and Amsterdam Schiphol Airport. The study found that, after adjustment for relevant confounders, noise sensitivity influences noise annoyance irrespective of noise level. At all noise levels, noise sensitive person seems to be more annoyed than the normal person. Noise exposure level and noise sensitivity are important and independent predictors of annoyance.

Hardoy et al (2005) investigated the frequency of mental disorders in a sample living in the immediate surroundings of Elmas Airport (Sardinia, Italy) compared with those from a sample of residents from the same regio who had not been exposed to the risk of aircraft noise. The control sample was drawn from a database of a large community survey, after matching for sex, age and employment status. All subjects were interviewed using a simplified version of the Composite International Diagnostic Interview. It was found that exposed subjects have a higher frequency of generalised anxiety disorder and anxiety disorder not otherwise specified than the control subjects. Stansfeld and Matheson (2003, p.253) suggested that “the effects of noise are strongest for those outcomes that, like annoyance, can be classified under ‘quality of life’ rather

81 than illness”. The following section explains in detail the concept of ‘annoyance’ and community reactions to aircraft noise.

2.4.9 Community Responses toward Aircraft Noise This section describes three potential community responses to aircraft noise. The first response, is considered the most important in this section, is called community annoyance. The second response is called physical against reaction (or complaint reaction), and the last response is called behavioural modification. The last two responses are induced considerably after the occurrence of the first response. More details of these three community responses will be discussed in the following sections.

2.4.9.1 Aircraft Noise Annoyance Aircraft noise disturbs humans in many aspects and levels depending on activity types, individual characteristics, and noise level. Indeed, aircraft noise annoyance can be widespread through large areas where people can hear aircraft noise, or even just observe an overflight. Annoyance from activities disturbance causes emotional stress and accumulated emotional stress is a source of human health disorders. To better understand annoyance, both the behaviour and the nature of annoyance are considered as an important background of this thesis.

2.4.9.1.1 Definition and Assessment of Annoyance Nelson (1987, p.3.7) defined annoyance as “a feeling of displeasure associated with any agent or condition believed to affect adversely an individual or a group”. In this context, aircraft noise acts as a major agent to stimulate community annoyance. The other factors can also arouse community annoyance, such as fear of aircraft crashing or negative attitudes toward the airport or the airlines.

Information about community annoyance by aircraft noise can be obtained by implementing an attitude survey. The two most widely accepted tools used to assess aircraft noise annoyance are (1) the response scale (or called opinion thermometer) as shown in Figure 2.19, and (2) the opinion word description. The major aim of the opinion thermometer is to assist the subjects to rate the reaction to the noise exposure level by some specific questions during the attitude survey, for example, “How much

82 annoyance do you feel when you hear a jet plane passing overhead” (Job et al., 2001). The opinion thermometer is simply scaled by zero to ten where zero represents “none or not at all annoyance” and ten represents “extreme annoyance”.

The second tool used for an attitude survey is the word description rather than a mark description. The example questions of word description are “While you have been at home over the past week, would you say you have been not at all annoyed by aircraft noise, slightly annoyed by aircraft noise, moderately annoyed by aircraft noise, very annoyed by aircraft noise, or extremely annoyed by aircraft noise?” (Job et al, 2001). Some assessments may employ individually mark description or word description or may combine both methods together depending on each study’s purposes.

Figure 2.19: Opinion Thermometer (source: Job et al, 2001, figure 1, p.941).

2.4.9.1.2 Dosage-Response Relationship Community annoyance becomes a major issue for both existing and developing airports. Many researchers tried to develop the relationship between community annoyance and aircraft noise exposure level for airport planning purposes. Schultz (1978) has developed this relationship (called “dosage-effect relationship”) for all types of transportation noise. The single curve (called “synthesis curve”) combines all of the transportation noises (which are air, rail, and road) from both primary and secondary data and translated them into the same unit (the DNL for unit of noise and annoyance for unit of community disturbance). It should be noted that the terminology of

83 disturbance used in attitude surveys could vary from study to study (e.g., “strongly disturbed”, “very often”, “highly affected”). However, for the synthesis curve, community disturbance was represented by the percentage of people highly annoyed (%HA) by noise exposure level with 95% confidence.

The definition of %HA was given by Schultz (1978) as “the percentage of respondents who described annoyance fell within the upper 27%-29% (or cut-off point) of the response scale”. In this case, the cut-off point of Schultz’s synthesis curve will fall within a scale of 7 to 8, which covers “highly” up to “extreme” annoyance. It means that if the subjects rate their response scale from approximately 7.2 to 10, there will be included into the %HA. On the other hand, the respondent scale from 0 to below 7.2 will be excluded from the %HA. Therefore, the cut-off point plays a major role in dosage-response relationship. If the cut-off point is too high, the dosage-effect relationship will underestimate the negative reaction to the noise exposure. Conversely, if the cut-off point is too low, the dosage-effect relationship will overestimate the noise exposure responses.

Fidell et al. (1991) updated the synthesis curve by including more social survey data released after Schultz (1978) and adjusted some errors from Schultz (1978)’s study. The study found that although the number of data points from a new relationship was inferred more than tripled from the original curve, the Schultz (1978)’s synthesis curve still provides a reasonable fit to the data.

However, Kryter (1994) argued that the synthesis curve was very doubtful to combine the relationship between community response and all transportation noises together into one curve. He claimed that the aircraft noise makes more annoyance to the people than the road traffic and railway noises at the same noise exposure level. He also criticised the synthesised curve in many aspects, such as the criteria used for excluding investigations when drawing up the synthesised curve, the definitions of the percentage highly annoyed persons for different investigations, and the methods used to estimate DNL from the data. He explained that fear of aircraft crashing was the major factor that influences people being more annoyed about aircraft noise than other transportation noises. The angle, or pattern, of noise transitions from the source to receiver is also an

84 important factor that distinguishes community annoyance between aircraft noise and the other noise sources (see Kryter, 1994, pp.603-609).

Miedema and Vos (1998) re-developed the synthesised curve following Schultz (1978) and Fidell (1991). They also used the same 72% as the cut-off point as previous studies. The existing and additional dataset were re-examined by the following stringent criteria (a) source specific, (b) noise annoyance question, and (c) consistent cut-off. Finally, they found the consistency results supporting Kryter’s concept that the different types of transportation noise have different dose-response relationships as shown in Figure 2.20, when the y-axis denoted %HA and x-axis denoted DNL. From Figure 2.20, it can be concluded that, for all the dose-response relationship, %HA is treated to become zero when the DNL is around 42 dB(A). It is obvious that aircraft generates the most annoyance (steepest curve) transportation noise source to the community and the railway noise generates the least affect (flattest curve) to community annoyance.

Many attempts have been made to develop the relationship between annoyance and aircraft noise, but most of them only deal with annoyance by overflight noise. Fidell et al (1999) and Fidell et al (2002b), however, studied community annoyance from low- frequency aircraft noise at the vicinity of two US airports: Los Angeles International Airport (LAX) and Minneapolis-St. Paul International Airport (MSP). The example of low-frequency aircraft noise are noise produced behind departing aircraft (called “backblast”), engine run-ups for maintenance purposes, and runway sideline noises (such as taxiing, queuing, acceleration during takeoff, and thrust reverser application on landing). The studies found that the major factors influencing high annoyance in areas (especially area excess of 75 to 80 dB) around runway sideline are rattle and vibration. Figure 2.21 illustrates the correlation between percent of highly annoyance due to rattle or vibration.

85

Figure 2.20: Dosage-Response Relation of (A)Aircraft, (B)Road Traffic, and(C) Railway Noise (source: reproduced from Miedema and Vos, 1998, figure 2, p.3442).

Figure 2.21: Relationship between Low-Frequency Aircraft Noise Levels and Prevalence of Annoyance due to Vibrations or Rattling Sounds (source: Fidell et al., 2002, figure 8, p.1414).

86 Besides using the energy equivalent noise index to develop dosage-response relationship, Rylander and Bjorkman (1997) proposed the concept of number above index. They found that the number of event above 70 dB(A) (N70) is a crucial factor to express community annoyance at the airport where the number of overflights is lower than 70 per day.

In Norway, Gjestland (1994) compared annoyance from noise and annoyance from reference aircraft (Boeing 737-500). He found that, at equal noise levels, the noise from helicopter was judged either more or less annoying than the noise from reference aircraft, depending on the type of helicopter. The noise from a large helicopter type with muti-bladed (named Super Puma AS322L and Sea King/S61N), which are the majority of in Norway, were judged to be similar, or even less annoying than the jet aircraft noise.

Hoeger et al (2002) found that differences between day and night-time annoyance were depended on the noise sources. For the case of road traffic noise no differences between day and night-time annoyance were found. In contrast, annoyance reactions were related to the time of day for railway and aircraft noise. Especially for aircraft noise, above a

Leq of 50dB(A) night-time annoyance rises faster than day-time annoyance.

2.4.9.1.3 Attitudes, Demographic Characteristics, and Non-Acoustic Factors Miedema and Vos (1999) investigated the causes of variation in individual reactions to noise exposure with equal DNL. They used demographic factors (such as sex, age, education level, occupational status, number of persons in household, home ownership, and so on) and attitudinal factors (noise sensitivity and fear of the noise source) as secondary variables to investigate their effects to noise annoyance. The study found that demographic factors were much less related to annoyance. For example, men and women reacted in a similar way to transportation noise, and the numbers of people in the house and occupational status were very low correlates with noise annoyance. However, the study demonstrated that the factors of age and education level have some effect on noise annoyance. Both young and old persons were less annoyed by noise than adults at the same exposure level. Highly educated people were slightly more annoyed by noise than those who have low education level.

87 Attitudinal factors demonstrated stronger correlation to noise annoyance than demographic factors. Respondents who reported themselves as noise sensitive person, expressed fear associated with the activity of noise source, or have negative attitude towards airports or airlines, were expected to be more annoyed by noise than general respondents at the same exposure level. The study concluded that fear influenced annoyance to all modes of transportations at different levels. The fear by railway traffic seems to be the lowest compared with the other modes. Karami and Frost (1999) studied community annoyance caused by aircraft noise in the vicinity of Tehran International Airport, Iran. They also found that fear of crashing is the most significant index influencing aircraft noise annoyance.

Guski (2004) proposed that the existing dosage-response relationship for steady state situations in predicting noise effects in future years is questionable. Noise annoyance reactions of residents may change over the years, and annoyance is affected the change itself. Fidell et al (2002a) studied community response to a step change in aircraft noise exposure associated with the opening of a new runway at Vancouver International Airport. The study found that the proportion of respondents who described themselves as “very” or “extremely” annoyed by aircraft noise in a residential area with increased aircraft noise exposure after the runway opening was markedly greater than that predictable from well-known dosage-response relationships. Analysis suggested that a good part of the “excess” annoyance is attributable to the net influence of non-acoustic factors. Flindell and Witter (1999) studied non-acoustic factors in noise management at Heathrow Airport. They found that by taking non-acoustic factors (which are factors other than noise level alone which contribute to noise annoyance such as local community representative) into account in addition to physical noise levels alone has been considerable beneficially to the noise management plan.

2.4.9.2 Complaint Reactions and Behavioural Modifications Complaints are an activity reaction response from the community about the things that they consider to be undesirable or annoying. Normally, complaint of aircraft noise occurs around the airport area, and under the designated flight paths, and rapidly increase around the time when there are some proposal of airport development or improvement. Complaints have been classified into two types: by activity (such as

88 letter, telephone, or demonstration protest); and by law. There are many ways to ask people to observe the number and characteristics of complaints such as “How annoyed you were by aircraft noise, and whether you had complained to any authorities about the noise?”

However, Kryter (1994, p.618) suggested that “because of a variety of social factors, complaints, or rather lack of complaints, legal and political actions by individuals and citizens are not generally as reliable, or sensitive, a measure of the effects of environmental noise as are attitude surveys”. For example, from study by Mato and Mufuruki (1999), residents surround the Dar es Salaam International Airport (DIA), Tanzania said “NO” when they were asked whether they had taken any action to express their annoyance from aircraft noise. They also gave the reasons of ignoring complaint: firstly, they were afraid of being disregarded; secondly, they feared about being moved away; and thirdly, they are used to the noises. Wong et al (2002) conducted a community surveys toward environmental noise over the territory of Hong Kong through telephone sampling. Even though 40% of sample found traffic noise was not tolerable and caused distraction, many did not complain against the noise. In Pamplona, Spain, Arana and Garcia (1998) implemented a social survey on the effects of environmental noise on the community. They found that even though 91% of the interviewed people considered environmental noise was a very important factor that affected their quality of life, only about 14% of the subjects have complained about noise. It might be only people who have great confidence that they can influence airport authorities who may complain frequently. The conditions in which people feel either there is pollution that they have to accept, or there is no use complaining further because nothing will be done to solve this situation. This, in turn, can lead to feelings of despair and hopelessness, feelings common amongst residents living near airports.

In most major airports (for example, Sydney International Airport, Denver International Airport, Vancouver International Airport, and Amsterdam Schiphol Airport) (see Chapter 3), the residents can complain against aircraft noise via telephone lines or web sites. Nevertheless, most of the complaints are received from people living in low aircraft noise areas. It might be a situation when people in low aircraft noise area

89 (according to the provided aircraft noise contours from airport) expect no aircraft noise in their back yard.

Hume et al (2002) studied the patterns of complaints and frequency of complaining by individuals using data from Manchester Airport. The study found that complaints were steady increased over the week from a low on Monday to a high on Saturday/Sunday. Calculations of the complaints per aircraft movement for each hour of the day showed a striking 24hr pattern with twice as many complaints between 23:00 and 07:00 as the rest of the day (07:00-23:00). However, they concluded that the association between noise exposure level and prevalence of complaint is still not well understood.

Besides any negative reactions to aircraft noise, people might modify their behaviour to confront aircraft noise problems. Patterns of behavioural modification are influenced by economic status of individual household. High-income people can modify their behaviour either by installing any sufficient insulation or relocating from high-noise areas. In general, people may just close windows and/or doors and avoid doing activity in the noisy places of the house. Figure 2.22 gives some examples of these behavioural modifications related to different noise level. It is clear that the percentages of people choosing to close windows as the solution to noise disturbance on watching TV and conversation increase rapidly with noise level. The use of sleeping drug increases from about 12% in Leq = 52 dB(A) to about 20% in Leq = 65 dB(A) and then keeps constant for high noise levels. These results can be well related to previous studies on sleep disturbance by aircraft noise since people trend to be less sensitivity or have a high adaptation to high noise level during sleep. The use of balconies decreases progressively with noise level, while the installation of soundproof increases.

90

Figure 2.22: Modifications in Behaviour by Noise Level (source: reproduced from Nelson, 1987, figure 3.12, p.3.17).

2.5 RESEARCH GAPS

The development of major commercial airports promotes the air transport industry and generates positive economic benefits to the airport and its host economy. However, certain costs are associated with these benefits. Any increase in aircraft movement causes negative environmental impacts, especially noise pollution. Many efforts have been made in order to reduce aircraft noise level at their sources or even installation of acoustic insulation on houses around airports. Nevertheless, the problems have never been satisfactorily resolved as long as the growth of air transport market is still in high demand. Jatmika (2001) proposed the technique (called Impact Management System) to reconcile these inherent conflicts between positive economic benefits and negative environmental impacts from the airport development using Sydney (Kingsford Smith) Airport as a case study. The intensive land-use pattern change survey around the airport has been implemented, but the assessment of impacts on community health and well- being around the airport has still not been undertaken.

91 From the literature, it could be concluded that, aircraft noise disturbs communication activities, especially on school children when concentration is required. Even though aircraft noise seems to have less impact on sleep activity due to the habituation ability, more research is required to reconfirm this delicate conclusion since good quality of sleep is the most desirable activity for every person. Previous studies have paid much attention to developing the dose-response relationship which has concluded that at the same exposure level, noise from aircraft is the most annoying of all transport modes. However, complaints of aircraft noise to authorities is low (or never happen in some places) because the feeling of hopelessness and lack of confidence on the part of residents.

There are few studies on the effects of aircraft noise on the auditory system. It might be the assumption that aircraft noise is not loud enough to significantly deteriorate hearing ability. Many epidemiological studies are concerned with the effects of aircraft noise on children’s blood pressure levels. Even though there is conflicting results, no strong evidence supports that aircraft noise influences children blood pressure levels. There was also no evidence that aircraft noise significantly influences psychological effects.

The literature review found no strong evidence to support that aircraft noise cause any diseases directly to human. However, it cannot be concluded that aircraft noise has no impact to human health because in accordance with Berglund and Lindvall (1995) ‘health’ is not only the absence of disease but also includes a state of completeness in physical, mental, and social well-being. Obviously, aircraft noise deteriorates quality of life of community in the vicinity of airport, especially major commercial airport. Therefore, it might be concluded that aircraft noise has effects on health in term of physical, mental, and social well-being. Only a few research studies have been attempted in this area. Moreover, comparison between studies could not be undertaken rigorously because of different health measure methods and administrative techniques. The epidemiological study of adult blood pressure level influenced by aircraft noise is also rare.

Consequently, this thesis is aimed at filling those two gaps by developing a comprehensive survey instrument for the evaluation of community health and well-

92 being impacts by aircraft noise using a valid and standardised health measure scale. This thesis is seeking to answer two core questions. Firstly, “Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?” Secondly, “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”

Table 2.2 summarises the previous studies that were of considerable relevance to the above core research questions. These selected previous studies were summarised in term of study population, study design, measurement of exposure to noise, measurement of health, health outcome of interest, and main findings. Table 2.2 is beneficial in assisting and guiding the development of research methodologies of the present study.

93 Table 2.2: Summary of Previous Studies Relevant to Core Research Questions.

Reference Study Study Population Sample Measurement Measurement of Administrative Health Measures Non-response Main findings Design size of health exposure to noise Technique analysis Bronaft (1998) Cross- People in New York, 3000 Questionnaire DNL>65dB(A) Postal Self- Health perception No Subjects who were bothered by aircraft Sectional USA, affected by - Community wellness control area located Administration and quality of life noise were more likely to complain of aircraft noise and and health promotion outside DNL contours sleep difficulties and more likely to the control group survey perceive themselves to be in poorer health. Meister and Cross- Residents in Minnesota, 4000 Questionnaire Ranking of community Postal Self- General health, No All health measures were significantly Donatelle Sectional USA, affected by - Short Form - 36 by severity of exposure Administration vitality, and worse in the community areas exposed (2000) aircraft noise and by frequency of loud mental health to aircraft noise. the control area exposure and decibels Rosenlund Cross- Residents around N=266 Questionnaire Equal energy: Postal Self- Prevalence of No Association between aircraft noise (2001) Sectional Stockholm Arlanda Airport, for exposure -single item < 55dB(A), >55 dB(A) Administration hypertension and prevalence of hypertension Sweden, and N=2693 Maximum noise: was marginally significant. the control group for control < 72 dB(A), > 72 dB(A) Miyakita et al Cross- Residents around 8084 Questionnaire Classified by DNL Postal Self- Physical and No Residents effected by aircraft noise (2002) Sectional Kadena and Futenma - Todi Health Index <55 dB(A), 55-60 dB(A), Administration mental health suffered from both physical and airfields, Japan, and 60-65 dB(A), outcomes mental health effects and these effects the control group 65-70 dB(A), >70 dB(A) increased with the level of noise. Goto and Cross- Residents around N=469 Previous data from Divide subjects into Self- Systolic and No No significantly difference of Kaneko Sectional Fukuoka Airport, for exposure doctor three groups: control, Administraion Diastoric blood blood pressure data between (2002) Japan, and N=1177 >75 WECPNL, and pressure study groups. the control group for control >90 WECPNL Morrel (2003) Longitudinal Children from primary N=1230, Automated blood Use INM to produce Self- Systolic and No No consistent statistically significant schools within 20km in 1994 pressure measuring ANEI at the time of the Administration Diastoric blood cross-sectional correlations between mean radius of Sydney Airport N=628, equipment study pressure resting blood pressure level and exposure in 1997 to aircraft noise. Franssen et al Cross- Residents living within 30216 Questionnaire Study area within Postal Self- General health Yes Association between (2005) Sectional a radius of 25km around - single item 20 - 35 Kosten unit, Administration status based on aircraft noise exposure (DNL) and

Schiphol Airport, - multiple item LDEN, LAeq, 23-07 h health complaints health indicators poor general health Netherland. LAeq, 22-23 h status was significant.

94 2.6 STATISTICAL TECHNIQUES

Obviously, there are many variables that need to be measured by a survey and these variables are usually interrelated in highly complex ways which cannot be solved by univariate and bivariate statistics. Therefore, some comment on statistical technique must be provided in this literature review. Therefore, an important research question arises, as to what is the most suitable multivariate technique when there are many possible variable involved and these variables may be interrelated?

Tabachnick and Fidell (2001, pp.17-30) have developed a guideline in choosing statistical techniques for a specific research problem. They suggested that “the most important criterion for choosing a technique is the major research question to be answered by the statistical analysis.” They have categorised the major research questions into four groups: (1) degree of relationship among variables; (2) significance of group differences; (3) prediction of group membership; and (4) time course of events. Recalling the two core research questions that have been formulated after a detailed investigation of the literature, this thesis attempts firstly, to compare the health measures quality of life difference between aircraft noise exposure group and non- exposure group, and secondly, to develop an association between aircraft noise and adult hypertension. Therefore, the first and second core research questions fall into the above second and the third categories of Tabachnick and Fidell (2001), respectively. Figure 2.23 displays a decision tree (reproduced from Tabachnick and Fidell, 2001, table 2.1, pp.27-28) in choosing to statistical technique to answer two major research questions: “Significance of Group Differences” and “Prediction of Group Member”.

95 Major Research Number (Kind) of Number (Kind) of Covariates Analytic Strategy Goal of Analysis Question Dependent Variables Independent Variables None One-way ANOVA Determine One (discrete) Some One-way ANCOVA reliability of mean One (continuous) None Factorial ANOVA group differences. Multiple (discrete) Some Factorial ANCOVA

None One-way MANOVA Create a linear One (discrete) Some One-way MANCOVA combination of DVs Multiple (continuous) Significance None Factorial MANOVA to maximize mean Multiple (discrete) of group Some Factorial MANCOVA group differences. differences Multiple (one discrete Profile analysis of Create linear One (continuous) within S ) repeated measures combination of DVs to maximize mean Multiple (continuous/ group differences and One (discrete) Profile analysis commensurate) differences between levels of Multiple (one discrete Doubly-multivariate within-subjects IVs. Multiple (continuous) within S ) profile analysis

One-way Create a linear None Multiple discriminant function combination of IVs (continuous) Sequential one-way to maximize Some discriminant function group differences.

Create a log-linear Mutiple Multiway frequency combination of IVs One (discrete) analysis (logit) to optimally predict DV. (discrete) Multiple Logistic Create a linear None Prediction (continuous regression combination of the log of group and/or Sequential logistic of the odds of being in Some membership discrete) regression one groups.

Factorial Create a linear None Multiple Multiple discriminant function combination of IVs to (discrete) (continuous) Sequential factorial maximize group Some discriminant function differences (DVs).

Figure 2.23: Choosing among Statistical Techniques (source: reproduced from Tabachnick and Fidell, 2001, table 2.1, p.27).

For the first core research question, this thesis employs some scales of Short Form Health Survey (SF-36) (see section 4.3.1.1) to assess health measures quality of life in many aspects (which are physical functioning, role of physical, general health, sense of vitality, and mental health). Therefore, the number of dependent variables is dependent on how many health scales are involved in the comparison. For example, if research compares just the sense of vitality score of both groups, the number of dependent variable (DV) will be one. Conversely, it will be multiple dependent variables if more than one health scales of both groups are compared at the same time. The number of independent variable (IV) is one if the study considers only aircraft noise as an exposure variable. However, it will involve multiple independent variables if the study

96 finds any other potential confounding factors (such as body mass index and high cholesterol status).

This thesis assumes that the health measures quality of life are affected by some potential covariate variables (CVs) (see Chapter 6). Therefore, from the decision tree, the first core question will be answered by: (1) using one-way analysis of covariance (ANCOVA) if the number of both dependent variable and independent variable is one; (2) using factorial analysis of covariance if the number of dependent variable is one but the number of independent variable is more than one; (3) using one-way multivariate analysis of covariance (MANCOVA) if the number of dependent variable is more than one but the number of independent variable is only one; or (4) using factorial multivariate analysis of covariance if the number of both dependent variable and independent variable is more than one. Nevertheless, at the end of analysis, if it is found that there is no effect of covariate variable on the dependent – independent correlation, the analysis will turn automatically to analysis of variance or multivariate analysis of variance, respectively.

For the second core research question, the analysis is divided into two parts (see section 7.2). The first part develops a model to predict the presence/absence of chronic noise stress due to long-term aircraft noise exposure status. The second part develops a model to predict the presence/absence of hypertension due to the perceived chronic noise stress. In first part, the dependent variable is noise stress (dichotomous variable), and the independent variable s are aircraft noise exposure status plus some potential predictor variables (see section 7.2.1). Then, for the second part, a dichotomous variable of noise stress turn to be the independent variable plus some potential predictor variables (see section 7.2.2), and the dependent variable is a dichotomous variable of hypertension condition. Conclusively, in both parts, the number of dependent variable is one, and the number of independent variables, including some potential predictor variables (mix types of variable), is multiple. This thesis assumes no order of entry of predictors into the model. Therefore, from the decision tree, the solution for the second core question will be achieved by using logistic regression analysis.

97 The selected statistical techniques: Analysis of Covariance, Multiple Analysis of Covariance, and Logistic Regression Analysis have been widely applied in many areas of research. Therefore, it is not uncommon that there are many statistical textbooks consulted on each technique. The thesis selected the most relevant, and up-to-date references, such as Hosmer and Lemeshow (2000), Agresti (2002), Kleinbaum and Klein (2002), Tabachnick and Fidell (2001), Rutherford (2001), and Mickey et al (2004).

2.7 CONCLUSION

This research applied the transdisciplinary concept to study of community health and well-being impacts by aircraft noise. The transdisciplinary framework involves drawing concepts and reviewing literature from a variety of disciplines. The knowledge of epidemiology, social survey, characteristics of environmental noise (especially aircraft noise), management of environmental noise, and effects of environmental noise on community has been reviewed.

This chapter (Chapter 2) has reviewed the up-to-date literatures on characteristics of environmental noise and effects of environmental noise on community. The current practices of aircraft noise management in major commercial airports are reviewed in Chapter 3. The social survey methods and fundamental concept of epidemiology are described in Chapter 4. Chapter 5 describes the procedures of aircraft noise measurement and the development of a ‘new’ noise index. Chapter 6 and 7 provide a detailed discussion in the fundamental concepts and relevant multivariate statistical techniques required in exploring the first and second core research question, respectively.

98 CHAPTER THREE

AIRCRAFT NOISE MANAGEMENT AND COMMUNITY HEALTH AND WELL-BEING IMPACTS

3.1 INTRODUCTION

The previous chapter has reviewed the characteristics of aircraft noise and its impacts on community health and well-being. The review has indicated that even though there is no robust conclusion about the causality between aircraft noise exposure and disease, aircraft noise potentially deteriorates quality of life which is one of the major components of health. Before going into a detailed development of the comprehensive procedures to evaluate community health and well-being impacts by aircraft noise, this chapter reviews the aircraft noise management measures, or strategies, of major commercial airports in Australia and in other countries, such as the United States, Canada, and England. The major sources of data were from the official website of each airport. In addition to discussing noise measurement regulation, the objective of this chapter is to find out whether or not the current practice to manage aircraft noise problem at major commercial airports has taken into account the issue of community health and well-being impacts by aircraft noise.

In general, aircraft noise management measure is divided into two main procedures: noise abatement procedures; and noise mitigation procedures. The noise abatement procedures are measures designed to reduce aircraft noise exposure from noise-sensitive land uses and population. Such measures include changes in runway layout, runway use, or aircraft operational procedures. The noise mitigation procedures are measures designed to remedy significant noise exposure in existing noise-sensitive areas; and to minimise the development of noise-sensitive land uses within areas exposed to significant noise. An example of noise mitigation procedures is the installation of acoustic insulation in noise-sensitive buildings such as dwellings, schools, churches, and hospitals.

99 This chapter is organised as follows. Section 3.2 describes the current practice of aircraft noise measurement in Australia. Section 3.3 describes the current policies relevant to aircraft noise management in Australia. Section 3.4 reviews the current aircraft noise management procedures for the major commercial airports in Australia. Section 3.5 overviews the aircraft noise management procedures of major commercial airports in the other countries. Section 3.6 discusses the current aircraft noise management strategy in regard to the issue of community health and well-being impacts by aircraft noise. Section 3.7 concludes the whole description.

3.2 AIRCRAFT NOISE MEASUREMENT IN AUSTRALIA

In Australia, aircraft noise has been measured by the principle of energy equivalence known as the Australian Noise Exposure Forecast (ANEF) system (the mathematical formula of ANEF is presented in section 2.3.6.2). The ANEF system was developed based on a social survey conducted by the National Acoustic Laboratories in 1982. The survey obtained the reactions of people to aircraft from 3,575 residents living around major commercial airports in Sydney, Adelaide, Perth and Melbourne, and around the Royal Australian Air Force base at Richmond in NSW.

Various subjective reactions (which are dissatisfaction, annoyance, fear, reports of activity disturbance, and complaint) to aircraft noise were represented by a composite parameter, called “general reaction”. The general reaction scale ranges from 0 to 10. Respondents who scored 8 or more were identified as “seriously affected” by aircraft noise, and respondents who scored at least 4 were classified as “moderately affected” by aircraft noise (Hede and Bullen, 1982). The National Acoustic Laboratories also conducted noise measurements at several sites around the study areas. Twenty different noise exposure indices were estimated at the household of each study sample. The analysis found that an energy equivalent aircraft noise index, called the NEF (see section 2.3.6.1), was more highly correlated with community reactions than other types of index, including peak level indices. However, it was found that the original weighting factor (or penalty) for flights during night time (10pm – 7am) of NEF was too high, and that there should be a weighting applied to flights during evening hours

100 (7pm – 10pm). Therefore, the National Acoustic Laboratories decided to change the night time period from between 10pm – 7am to 7pm – 7am, and the noise penalty in the night hours changed from 12 decibels to 6 decibels (Airservices Australia, 1999). The system was then renamed the ANEF system.

The ANEF system consists of three different types of aircraft noise contour maps. The first is called the Australian Noise Exposure Forecast (ANEF). It is a contour map showing aircraft noise levels expected to exist in the future around an airport, generally for 10 years from the date of issue. It is produced based on a forecast of aircraft movement numbers, aircraft types, destinations, and distribution of runway usages and flight tracks for a target year. Airservices Australia recommended a regular review of ANEF every 5 years to ensure its continuing validity. The ANEF has been accepted by the Australian Government, the Australian Standards Association, and the other environmental authorities for assessing average long-term community reactions to aircraft noise for land-use compatibility advice and regulation around commercial and military airports in Australia.

The second contour map of the ANEF system is called the Australian Noise Exposure Index (ANEI). It is a contour map showing aircraft noise levels around an airport for any one year. It is produced based on historical data from a previous year, where all the input data are known. Airservices Australia produces and publishes quarterly the ANEI contour maps of major commercial airports in Australia as benchmarks, or indicators, of change of aircraft noise exposure. Finally, the third type of contour map in the ANEF system is called the Australian Noise Exposure Concept (ANEC). It is a contour map showing aircraft noise levels around an airport based on a hypothetical set of conditions. In general, it will be produced during the planning process of airport development (for example, construction of new runway, or modification of flight paths operation). The ANEC is not intended for land-use planning purposes.

In summary, the ANEF is a noise index that takes into account the following factors of aircraft noise (Airservices Australia, 1999, p.1):

101 • the intensity, duration, tonal content and spectrum of audible frequencies of the noise of aircraft takeoffs, approaches to landing, and reverse thrust after landing (but not including noise generated on the aerodrome from aircraft taxiing and engine running during group maintenance).

• the forecast frequency of aircraft types and movements on the various flight paths, including flight paths used for circuit training.

• the average daily distribution of aircraft takeoff and landing movements in both daytime (7am – 7pm) and night time (7pm – 7am).

The National Acoustic Laboratories has also developed a dosage-response relationship between the ANEF and the community reactions to aircraft noise as shown in Figure 3.1. This figure was used later to derive the land-use compatibility program around Australian’s airports. The Australian Standard (AS 2021-2000) recommended that areas outside the 20 ANEF are acceptable for residential areas, public buildings, schools, universities, hospitals, and nursing homes. From Figure 3.1, the expected percentage of people “seriously affected” by aircraft noise would be less than twelve percent.

Figure 3.1: Relationship between ANEF and Community Reactions in Residential Areas (source: reproduced from Airservices Australia, 1999, Figure A1, p.5).

102 However, due to Figure 3.1, the ANEF has been widely interpreted, especially by some members of the community, as a metric correlated with community annoyance. For instance, people living in areas with less than 20 ANEF would not be at all affected by aircraft noise, and that a zero ANEF means no aircraft noise. The misinterpretation of ANEF has led to many controversies among the federal government and the community in the vicinity of airports, especially after the opening of the third runway (16L/34R) at Sydney Airport in November, 1994. The monthly complaints about aircraft noise dramatically increased from less than 100 in October, 1994, to approximately 5,500 in November, and peaked at around 6,000 in December of the same year. The monthly complaint dropped to around 3,200 in March, 1995, when the Australian Senate established the Select Committee to inquire into aircraft noise at Sydney Airport and to explore possible solutions to the problem.

In November 1995, the Senate Select Committee (SSC) on Aircraft Noise in Sydney published a report called “Falling on Deaf Ears?”, which led to many modifications in the preparation process of Environmental Impact Statements and the measurement of aircraft noise in Australia at the present time. This report identified several limitations of the ANEF system in assessing environmental impact, briefly summarised by the following points (SSC, 1995, pp.185 – 200):

• the ANEF was designed to predict community reactions rather than individual reactions to average aircraft noise exposure. Moreover, the ANEF is based on a logarithmically averaged annual average day aircraft noise energy which is difficult to understand by a layperson. Thus, it provides insufficient information to a person potentially exposed to aircraft noise to judge, on a rational basis, how he/she might be affected by aircraft noise.

• the ANEF does not predict impact in a changing noise environment, especially the one that occurred at Sydney Airport after the opening of the third runway. The aircraft noise exposure level was rapidly increased in areas located on the North/South directions of the airport. The ANEF was developed in 1982 when aircraft noise levels around the airports were stable. The ANEF system lacks ability to deal with short-term responses to aircraft noise exposure.

103 • the ANEF was considered to be outdated since it was developed based on data collected in 1980 when average daily operations at Sydney Airport were approximately 300 (from Airservices Australia’s database, there were 866, 833, 709, and 707 aircraft operations per day at Sydney Airport in year 2000, 2001, 2002, and 2003, respectively), and the most common aircraft types in that time (which were Boeing 727s and Fokker F27s) are considerably noisier than many of the more common aircraft currently flying (which are Boeing 767s and Boeing 747s).

• As the ANEF contour maps were used for land-use advice and regulations, the impacts of aircraft noise on people outside an arbitrarily defined measure of 20 ANEF were disregarded. The nature of the method of computation of ANEF system has two peculiarities that can mislead a member of the community. Firstly, an exposure level of 0 ANEF does not indicate zero aircraft noise exposure. In fact, it would result from 8 aircraft overflights per day at approximately maximum noise level during the overflight of 66 dB(A). Secondly, the doubling/halving in aircraft overflights will increase/decrease only 3 ANEF units. To illustrate, one house exposure to 100 flights a day changing to 200 flights a day would change only 3 ANEF units.

Finally, the Senate Select Committee recommended the re-evaluation of the usefulness, and relevance, of the existing ANEF system as a predictor of long-term community reaction to aircraft noise around Sydney Airport, and the development of indices, or other information, for predicting the noise impact for communities faced with a changing noise environment.

In March 2000, the Australian Department of Transport and Regional Services (DoTARS) released a document (called Expanding Ways to Describe and Assess Aircraft Noise) to promote debate on the development and use of more transparent (or understandable) approaches to describing and assessing aircraft noise around Australian airports. The transparent aircraft noise concept is based on ‘everyday talk’ information, such as where the aircraft fly, how often, and at what time. The feedback on the DoTARS Discussion Paper has led to the outcome of another document released in 2003 (called Guidance Material for Selecting and Providing Aircraft Noise Information)

104 and the development of the Transparent Noise Information Program (see section 2.3.8.2). The guideline supported the previous discussion document by stating that aircraft noise information should be provided in both accurate and comprehensible form to the reader. The term “accurate” is not only the pure technical accuracy, but it also means a clear picture (or facts) of the noise exposure patterns around an airport which the user believes is actually happening. Aircraft information that needs to be interpreted, such as the concept of sound energy equivalent, should be avoided.

A potential example of the transparent aircraft noise concept is the NA index, or Number of Aircraft Noise Events louder than a certain threshold noise level, which is commonly 70 dB(A) (called N70), over a given time period. The outdoor level of 70 dB(A) was chosen because by average 10 dB(A) will be attenuated by the house structure (with open windows), while 60 dB(A) or above is the indoor sound pressure level that is likely to interfere with daily activities, such as conversation, listening to the radio or the television. The DoTARS declared that the N70 gives a much more realistic picture of aircraft noise to the community than the ANEF system. The N70 has been recommended by DoTARS and Environment Australia to be used in several aspects, such as providing a “macro” picture of noise around an airport input into Environmental Impacts Statement and noise assessment reports, providing advice to supplement ANEF contours, and monitoring of aircraft noise levels around an airport. At present, the N70 contour maps are prepared for many major commercial airports in Australia. The N70 has become more widely used in providing aircraft noise information to the community, such as in the case of the Environmental Impacts Statement of the Second Sydney Airport.

3.3 POLICY FOR AIRCRAFT NOISE MANAGEMENT IN AUSTRALIA

The federal government has enacted several acts and regulations to protect the environmental conditions from aircraft noise pollution in the vicinity of airports in Australia. These acts and regulations were designed to control airport noise in different manners which could be classified basically into two groups. The first group deals with noise generated from taking-off, landing, and taxiing of aircraft. These acts and

105 regulations include the banning of noisy aircraft from the air navigation system, the imposing of a levy for noisy arrivals, the banning of either departures or arrivals during night time (or curfew), and the limiting of flight paths over residential areas. The second group of acts and regulations control noise generated in relation to airport operations. These acts and regulations require the airport operator to develop an Environment Strategy which proposes a plan to control and manage the noise sources in relation to airport operations. The following subsections briefly describe those acts and regulations that have direct relevance to this thesis.

3.3.1 Aircraft Noise Levy Act The objective of this policy is to disadvantage airlines that operate noisy aircraft at Australian airports. In accordance with Aircraft Noise Level Act 1995 section (5), the levy is imposed on every landing. A levy is not imposed on a landing if the aircraft making the landing has an assessed noise of less than 265 (see Aircraft Noise Levy Regulations section (5)), or if the landing is made as part of a flight for a purpose that relates to the provision of emergency services (such as rescue of a person, fire-fighting, or medical emergency) or a charitable purpose (such as carrying of goods or people for a charitable body; for example, the provision of emergency relief as part of the recent tsunami disaster in the Indian Ocean). The amount of levy on landing is calculated based on a formula provided in the Aircraft Noise Levy Act 1995 subsection 6(1) and Aircraft Noise Levy Regulations sections (5) and (6). A levy imposed by the above procedures is payable by the operator of the jet aircraft that made the landing to the Minister for Finance, or a person authorized by the Minister for Finance (see Aircraft Noise Levy Regulations sections (9) and (10)).

Aircraft Noise Level Collection Act 1995 Part 4 subsection 6(4) identifies the meaning of a qualifying airport to be a leviable airport at a particular time if at that time there is a public building within the 25 ANEF contour for the area around the airport for a date after that time; or a residence within the 30 ANEF contour for the area around the airport for a date after that time.

106 3.3.2 Airport Curfew Act The objective of this policy is to prevent aircraft noise exposure on the community during the night time period. The federal government has enacted an airport curfew act, specifically, for Sydney Airport called Sydney Airport Curfew Act 1995. A curfew period starts at 11pm on a day and ends at 6am on the next day. During the curfew period, an aircraft must not takeoff from and land at the airport unless it is permitted under Part 3 of the Act. The procedures of controlling the usage of reverse thrust and missed approaches during curfew periods can be observed in Part 2 sections (8) and (9), respectively, of the Act. During the curfew period, an aircraft must operate over Botany Bay, that is takeoffs to the south and landings to the north. The restrictions of runway usage for takeoffs between 10:45pm – 11pm can be observed in Part 2 section (10) of the Act. The restrictions of runway usage for takeoffs and landings on weekends between 6am – 7am and 10pm – 11pm can be observed in Part 2 section (11) of the Act. It is noted that the curfew restrictions do not apply in cases of emergency as defined in Part 3 section (19) of the Act. Finally, the Act provides for fines of up to $550,000 for curfew breaches.

3.3.3 Air Services Act In accordance with Air Services Act 1995 Part 2 subsections 8(1) and 9(2), Airservices Australia has a function to “carrying out activities to protect the environment from the effects of, and the effects associated with, the operation of: commonwealth jurisdiction aircraft, whether in or outside Australia, or other aircraft outside Australia”, and in a manner to perform any functions to “ensure that, as far as is practicable, the environment is protected from the effects of the operation and use of aircraft; and the effects associated with the operation and use of aircraft”, respectively.

3.3.4 Air Navigation (Aircraft Noise) Regulations The Air Navigation (Aircraft Noise) Regulations 1984 are made under the Air Navigation Act 1920. The objective of these regulations is to control the engagement of aircraft in air navigation in Australia in accordance with an aircraft noise certificate (or ICAO noise standards) established for jet and propeller aircraft requirements. A noise certificate should contains the following information: (i) the serial number of the aircraft; (ii) the type and model of the aircraft; (iii) any additional modifications of

107 aircraft for the purpose of compliance with the applicable noise certification standards; (iv) the maximum weights at which compliance with the applicable noise certification standards has been demonstrated; and (v) the noise level and their 90 percent confidence limits at the reference point (see section 2.3.7 of this thesis). The Air Navigation (Aircraft Noise) Regulations 1984 regulation (9) describes a circumstance in which an aircraft may engage in air navigation in Australia. A subsonic jet aircraft will not be allowed to engage in air navigation in Australia if it does not comply with Chapter 3 of a noise certificate (see section 2.3.7 of this thesis). The exceptional case may be granted if it complies with conditions stated in Regulation 9(1). A supersonic aircraft may engage in air navigation in Australia only if permission has been granted under regulation 9AA. The operator of an aircraft that engages in air navigation in contravention of these regulations is guilty of an offence, as detailed stated in Regulation 9(4).

3.3.5 Airport Act In accordance with the Airport Act 1996 Part 6 section (113), each airport in Australia must prepare a final environment strategy which is a draft environment strategy that has been approved by the Commonwealth Minister for Transport and Regional Services. From Part 6 section (116) of the Act, a draft, or final, environment strategy must specify: (1) the sources of environmental impact associated with airport operations; (2) the methods of study, review, and monitoring in connection with the environmental impact associated with airport operations; (3) the time frames for completion of subsection (2); (4) the methods to prevent, control, or reduce the environmental impact associated with airport operations; (5) the time frames for completion of subsection (4); and (6) details of the consultations undertaken in preparing the strategy (including the outcome of the consultations).

Before giving the Minister a draft environment strategy, it has to be published in a newspaper circulating generally in the State or Territory in which the airport is situated. The public is invited to give written comments about the draft to the airport company. Copies of the draft are available for inspection and purchase by members of the public. All the written comments from the public have to be summarised and regarded in preparing the draft strategy. The strategy period of a final environmental strategy is 5

108 years. A new draft environment strategy is required to submit for consideration before the expiry of the original final environmental strategy.

3.3.6 Airport (Environment Protection) Regulations The Airport (Environment Protection) Regulations describe standards and impose requirements that relate to environmental pollution and the emission of noise generated at airports. In conjunction with the Airport Act 1996, the objectives of the Airport (Environment Protection) Regulations 1997 are to establish a Commonwealth system of regulation of activities at airports that generate, or have potential to generate: (i) pollution; or (ii) excessive noise, and to promote improving environmental management practices for activities carried out at airport sites. It is important to note that these regulations do not apply to pollution generated by an aircraft, or noise generated by an aircraft in flight or when landing, taking off or taxiing at an airport.

Airport noises included in these regulations are noise from construction at an airport, noise from road traffic at an airport, noise from rail traffic operated at an airport, noise from ground-based aircraft operations (which means operation of an auxiliary power units of an aircraft, test operations of an engine attached to an aircraft, called ground- based aircraft running, or test-bed running of an aircraft engine removed from an aircraft), and noise from other airport operations (for example, aircraft refuelling, operation of plant or machinery, or fire alarm or warning systems). The acceptable levels of the above airport noises at sensitive receptors and commercial receptors are described in Schedule 4 Part 2 and 3 of the Regulations, respectively. Schedule 4 Part 4 of the Regulations sets out the procedures and standards to be applied by an airport environment officer in measuring the above airport noises. Part 2 Regulation (2.04) of the Regulations defines offensive noise as noise that is generated at a volume, or in a way, or under a circumstance that in the opinion of an airport environment officer offensively intrudes on individuals, communities or commercial amenity.

These regulations also require the airport environmental strategies to: identify sources of environmental impact at airports; propose studies, reviews, and monitoring of environmental impact; and to propose measures for preventing, controlling, or reducing environmental impact associated with airport operations. The regulations emphasise the

109 importance of public involvement in developing any future environmental strategy. Finally, the Regulations require an undertaking by the operator at an airport to take all reasonable and practicable measures (which are the judgment to be made by an airport environment officer) to prevent, or minimise the generation of offensive noise from activities relevant to the airport operations.

3.4 AIRCRAFT NOISE MANAGEMENT AT MAJOR COMMERCIAL AIRPORTS IN AUSTRALIA

It is mandatory in Australia that every airport must prepare an Environmental Strategy which has to be reviewed and approved every five years by the Commonwealth Minister for Transport and Regional Services. The Environmental Strategy is prepared in accordance with the Airport Act 1996 and Airports (Environmental Protection) Regulations 1997. The Environmental Strategy aims to maintain and protect environmental conditions around the airport site. As noise pollution is a major environmental issue of airport operations, some of the major commercial airports in Australia have also developed a Noise Management Strategy for managing airport noise problems in the vicinity of the airport site. Nevertheless, it is important to note that the Noise Management Strategy is dealing with the airport noise generated from the ground-based aircraft operations, such as engine warming up, ground running of aircraft, aircraft servicing, mechanical plant, or servicing equipment. The management of noise from taking off and landing aircraft is a responsibility of Airservices Australia in accordance with Airservices Act 1995 and Air Navigation (Aircraft Noise) Regulations.

Airservices Australia has established environmental principles and procedures for minimising the impact of aircraft noise. It consists of twelve fundamental principles. Each principle aims to minimise the impact of aircraft noise on residential area in the vicinity of airports. For instance, Airservices Australia should develop noise abatement procedures to achieve the lowest possible overall impact of aircraft noise on the community. Aircraft noise should be fairly shared whenever possible, and aircraft noise should be concentrated as much as possible over non-residential areas. Detailed

110 description of these principles can be found in Airservices Australia (2002). Airservices Australia is also responsible for aircraft noise complaints from members of the community. A Noise Enquiry Unit has been established to receive and address aircraft noise complaints via either a confidential telephone line or by internet.

Airservices Australia has permanently installed noise monitoring terminals which are strategically located underneath the main flight paths of Australia’s major airports (which are Adelaide Airport, Brisbane Airport, Cairns Airport, Canberra Airport, Coolangatta Airport, Melbourne Airport, , and Sydney Airport). Basically, the noise monitoring terminal consists of a microphone, atop a mast 6 metres high, and an electronics box. The noise monitoring terminal is a real time noise monitoring system that continuously measures and transmits noise data (both aircraft noise and background noises), via a data line, to the Noise and Flight Path Monitoring System central computer where it is processed and stored for later noise analysis.

The Noise and Flight Path Monitoring System has the capability to distinguish and classify aircraft noise from background noises. When the level and duration of noise from any noise source detected by a noise monitoring terminal exceed the threshold level and duration set for the noise monitoring terminal, a noise event is recorded. The time at which the noise event is recorded at the Noise Monitoring Terminal location is then checked against movement times and radar tracks of aircraft operation. If the time and noise monitoring terminal location of the noise event match the movement time and radar track of an aircraft, the noise event is attributed to that aircraft. Otherwise, it is regarded as part of the background noise (Huynh, et al, 2004). The quarterly reports of Noise and Flight Path Monitoring System for airports where the system is operated are published and available for public scrutiny.

In terms of aircraft noise management, Airservices Australia has declared (Huynh, et al, 2004, p.23) that the Noise and Flight Path Monitoring System is useful in: (1) detecting occurrences of excessive noise levels from aircraft operations; (2) assessing the effects of operational and administrative procedures for and compliance with these procedures; (3) assisting in planning of airspace usage; (4) validating noise

111 forecasts and forecasting techniques; and (5) assisting in answering noise complaints about aircraft operations from the general public.

There are almost 30 commercial airports (where Airservices Australia provides a terminal service) in Australia. Table 3.1 summarises the number of movements at the Australian airports in 2003. The following subsections review the aircraft noise management strategy, or plan, of Australia’s major airports based on the available up- to-date information provided in the official website o f each airport.

Table 3.1: Movements at Australian Airports arranged in Alphabetical Orders – 2003 Calendar Year Totals

(source: www.airservicesaustralia.com.au).

112 3.4.1 Perth International Airport The official website of Perth Airport is . The Perth Airport is currently operated by a private company called Westralia Airports corporation. The airport consists of two operating runways: runway 03/21 (the main runway); and runway 06/24 (the cross runway). It handled almost 94,000 aircraft operations in 2003 (see Table 3.1). Airservices Australia has installed permanently five NMTs around this airport. The airport authorities plan to increase the runway capacity by extending the length of both runways. This proposal is under consideration in 2005.

The current Environment Strategy was approved by the Minister for Transport and Regional Services on 15 July 2004. The Environmental Strategy identified the potential sources of noise according to Airport (Environment Protection) Regulations. The Environmental Strategy stated that the primary impact of noise generated at the airport is the reduction of amenity of users and neighbours of the airport. Noise from ground- based aircraft operations will be monitored and reported quarterly to the Noise Management Strategy Committee. The Committee, which meets quarterly, comprises representatives from Perth Airport, State and Local Government, State and Australian Government departments, Federal Members of Parliament, airlines and community groups.

Besides the monitoring of noise generated in relation with the airport operations, the other principal purpose of the Committee was to contribute to the development of an Aircraft Noise Management Strategy for Perth Airport. The Aircraft Noise Management Strategy was endorsed and submitted to the Minister on May 2000. The Minister subsequently supported the Aircraft Noise Management Strategy on August 2000. Some key issues of Aircraft Noise Management Strategy for noise management and proposed action for implementation are: (1) to determine whether aircraft engine ground running noise is of concern to the community; (2) to produce, monitor, and publish Australian noise exposure contour plans; (3) to enhance the existing noise monitoring program; (4) to inform existing and prospective owners of properties of aircraft noise impacts; (5) to educate the community about airport operations; and (6) to continue community consultation on airport noise. Mention of information about community health and well- being impacts by aircraft noise could not be found on this airport website.

113 3.4.2 Brisbane International Airport The official website of Brisbane Airport is . The Brisbane Airport is currently operated by the private sector, called Brisbane Airport Corporation. The airport’s current runway system configuration consists of two runways: runway 01/19 (the main runway) and runway 14/32 (the cross runway). Runway 01/19 manages most of the domestic and international aircraft. The cross runway is capable only of accommodating smaller aircraft due to length, width, and pavement strength constraints. The total movements of aircraft at this airport were almost 140,000 in year 2003 (see Table 3.1). To increase runway capacity, the airport authorities plan to construct a new, parallel, runway (01L/19R). The proposal is under consideration in 2005. There are five permanent noise monitoring terminal located around this airport.

The current Environment Strategy of Brisbane Airport was officially approved in April 1999 by the Minister for Transport and Regional Services. The renewal of the Environmental Strategy is still in progress. The airport also endorsed the 2003 Airport Noise Management Strategies designed to reduce noise impacts on the community. Noise Abatement Procedures stated in this Noise Strategy include the procedures to direct as much air traffic departing or landing at Brisbane Airport over water (Moreton Bay), or to reduce it as far as possible in noise sensitive areas. The preferred runways used whenever possible for landings and takeoffs, are runway 19 and runway 01 respectively. This reciprocal operation (or “nose to nose” mode) is predominantly utilized at night (10pm – 6am) when meteorological and demand conditions suit its application. If the nose to nose mode becomes unsustainable due to traffic demand or weather conditions, the departures will occur over Moreton Bay and arrivals over Brisbane. The preferred flight paths procedures, and procedures for climb and descent, can also be found in this Noise Strategy. The aircraft engine ground running guidelines are also provided. Complaints regarding noise from engine ground running can be lodged directly to the airport via telephone.

A Technical Noise Working Group has been established to review noise issues at Brisbane Airport. This Group is chaired by Brisbane Airport Corporation and includes representation from Airservices, airlines, DoTARS, the association representing international airlines, the Civil Aviation Safety Authority and an academic from the

114 Queensland University of Technology specializing in noise. Some key focuses of this Group are presented as follows: (1) reviewing and analysing Brisbane Airport noise trends; (2) reviewing and developing noise abatement procedures for Brisbane Airport; (3) ensuring the Aircraft Noise Management Strategy document for Brisbane Airport is relevant and up-to-date; and (4) investigating, documenting and reporting on relevant issues raised by the community, members of the Group, government and other agencies. Information about community health and well-being impacts by aircraft noise could not be found on this airport website.

3.4.3 Canberra International Airport The official website of Canberra Airport is . Canberra Airport has been privatised since 1998 and is currently operated by the Capital Airport Group. The airport consists of two operating runways: runway 17/35 (the main runway); and runway 12/30 (the cross runway). The airport operates 24 hours a day, seven days a week. The total movements of aircraft at this airport were almost 87,500 in year 2003 (see Table 3.1). Airservices Australia has installed three permanent NMTs around this airport.

The current Environment Management Strategy was approved by the Minister for Transport and Regional Services on August, 1999. The noise abatement programs, as stated in the Draft Minor Variation of Master Plan 2020, for the Canberra Airport has been established by Airservices Australia. The noise abatement programs proposed that the aircraft noise abatement areas should comprise three zones: (1) Green zone; (2) Amber zone; and (3) Red zone. The Green zone indicates areas suitable for the development of housing estates. The minimum flying altitude over these areas for all aircraft, other than light aircraft (less than 5,700 kg), is either 1,524 metres (all jets) or 914 metres (propeller aircraft) above ground height. It is possible to protect against the impacts of excessive aircraft noise on residents in these areas. The Amber zone indicates areas where aircraft noise is not generally significant, but it can be still heard by the community. There is no noise abatement of any aircraft type in this zone. The Red zone indicates areas where aircraft noise is excessive. Aircraft take offs and landings to the airport are over these areas at a low level from the terrain. The land in this zone should not be used for housing estate purposes.

115 All operations during night time (11pm – 6am) should be restricted on the cross runway. Aircraft approaching runway 30 should fly via the Kowen Forest, not over rural residential land east of the Ridgeway, Queanbeyan. Information about community health and well-being impacts by aircraft noise could not be found on this airport website.

3.4.4 Coolangatta Airport The official website of Coolangatta Airport (or Goldcoast Airport) is . The Coolangatta Airport is located at the southern end of the Gold Coast, Queensland several kilometres to the north of the Tseed River that separates NSW and Queensland. The airport was sold to Gold Coast Airport Limited in 1998. The airport’s current runway system configuration consists of two runways: runway 14/32 (the main runway) and runway 17/35 (the cross runway). The Coolangatta component of the Noise and Flight Paths Monitoring System has three permanently installed noise monitoring terminals located around the airport. The total movements of aircraft at this airport were almost 87,400 in 2003 (see Table 3.1) predominantly aircraft of under 7,000 kg in weight. The existing Airport Environment Strategy was approved in December, 2004 by the Minister for Transport and Regional Services.

The Airport Noise Abatement Consultative Committee has been established by the airport authority. The Committee meets on a quarterly basis. The Committee consists of members of the community, local councils, State and Federal elected members, Federal regulators, Airservices Australia, airlines and Gold Coast Airport Limited. The Committee has endorsed a policy to minimise aircraft noise exposure level in the vicinity of the airport. Some key issues of the policy are: (1) to operate the full length of the runway (whenever practicable) during taking-off to maximise height over populated areas; (2) to avoid low flying over populated areas; and (3) to respond to community inquiries about noise in a cooperative manner.

This airport operates under a curfew, so that aircraft over 5,700 kg (most jet aircraft) are restricted between 11pm – 6am. There are occasional concessions made to the curfew due to safety considerations or unusual operating conditions, but these are minimal and

116 are recorded for review at the Committee’s meetings. Information about community health and well-being impacts by aircraft noise could not be found on this airport website.

3.4.5 Melbourne International Airport The official website of Melbourne Airport is . The Melbourne Airport is currently operated by Australia Pacific Airports (Melbourne) Corporation Limited. The airport’s current runway system configuration consists of two runways: runway 16/34 (the main runway) and runway 09/27 (the cross runway). The Melbourne component of the Noise and Flight Paths Monitoring System has six permanently installed noise monitoring terminals located around the airport. The total movements of aircraft at this airport were almost 155,800 in 2003 (see Table 3.1) predominantly aircraft of over 7,000 kg in weight. Even though the airport’s Environmental Management System has been accredited to world’s best practice standard, ISO 14001, neither the Environment Strategy nor the Noise Management Strategy is available from the official website of this airport. However, some information relevant to aircraft noise management could be found.

There are four main mechanisms that are used to manage and minimise the noise effects generated by aircraft arriving and departing from the airport. Firstly, the flight paths are designed to minimise, as far as practicable, the noise exposure to residential areas in the vicinity of the airport. Secondly, the noise complaints are monitored on a monthly basis by Airservices Australia. Third, a Noise Abatement Committee has been established. The Committee is chaired by Melbourne Airport and consists of representatives from Airservices Australia, airlines, State EPA, State Department of Infrastructure, DoTARS, and local councils. The Committee’s role is to review the impact of aircraft noise exposure on the surrounding community, and, in a consultative manner, make recommendations to minimise the effect of aircraft noise. The Committee meets on a quarterly basis. Finally, land use controls for the areas around Melbourne Airport are implemented by the State Government to ensure that the operation of the airport is not adversely affected by inappropriate development in noise-affected areas around the airport.

117 Finally, information about community health and well-being impacts by aircraft noise could not be found on this airport website.

3.4.6 Sydney International Airport The official website of Sydney Airport is . The Sydney Airport is currently operated by Sydney Airport Corporation Limited. The airport’s current runway system configuration consists of three runways: runway 16R/34L (the main runway), runway 16L/34R (the parallel runway) and runway 07/25 (the cross runway). Sydney Airport is Australian’s busiest domestic and international airport. The 2003 total movements of aircraft at this airport were 258,206 with almost 68,500 movements from aircraft weight over 136,000 kg (see Table 3.1). The Sydney component of the Noise and Flight Paths Monitoring System has twelve permanently installed noise monitoring terminals which are strategically located around the airport. This system is considered one of the world’s largest and most geographically spread Noise and Flight Paths Monitoring System.

A Preliminary Draft Airport Environment Strategy for the period 2005 – 2010 has been prepared. This strategy will replace the previous 1999 Sydney Airport Environment Strategy. The Airport Environment Strategy describes the current management procedures for both aircraft ground based noise and aircraft noise generated during flying, taking off, landing. The airport has developed Noise Abatement Procedures for ground running of aircraft engines. The objective of these procedures is to provide for safety in aircraft engine ground runs and to minimise the impact of aircraft engine ground running during noise sensitive periods of the day and night, particularly between 9pm – 7am. The detailed description of these procedures can be found in SACL (2000).

The Sydney Airport Community Forum (http://www.sacf.dotars.gov.au) has been established since 1996. One of the main objectives of this Forum is to monitor the implementation of the Long Term Operating Plan. The Sydney Airport Community Forum provides consultation on Sydney Airport flight paths management and aircraft noise impacts on the community. The Forum consists of representatives from the Community, Local Councils, Industry and State and Federal Parliaments. There are a number of aircraft noise mitigation strategies implemented at Sydney Airport to reduce

118 environment and social impact by aircraft noise. The following points briefly outline those strategies.

A Long Term Operating Plan The Long Term Operating Plan has been developed for Sydney Airport, incorporating the Noise Sharing Principle, for the operation of the airport. Currently, there are ten runway modes of operation at Sydney Airport. The objectives of Long Term Operating Plan are: (1) to use all the three runways; (2) to maximise the usage of flight paths over water and non-residential areas; (3) to fairly share noise from overflight; (4) to ensure that the movements at airports do not exceed 80 movement per hour; (5) to maximise the hours per day over residential areas that are either free from overflight, or have a minimum of unavoidable overflights; (6) to ensure that all arrival aircraft are operated with low-power and low-noise operations; and (7) to ensure that residential areas overflown by aircraft arriving should not also be overflown by aircraft departing from the same runway. The detailed description of Long Term Operating Plan is found in Airservices Australia (1996).

The movements of aircraft at Sydney Airport are reported monthly and published by Airservices Australia. This report summarises runway movement, mode utilization, runway end impact, daily mode usage, and complaints at the airport. Aircraft noise information such as Measured Daily N70 values at each noise monitoring terminal is also presented in the report.

Sydney Airport Curfew According to the Sydney Airport Curfew Act 1995 (see section 3.3.2), Sydney Airport is operating under a curfew between 11pm – 6am. During the curfew period, all aircraft movements must be directed to Mode 1 of Long Term Operating Plan (departures on 16R and arrivals on 34L).

Sydney Airport Noise Insulation Program The Commonwealth Government administers the Sydney Airport Noise Amelioration Program, which provides a mechanism for the insulation of homes and public buildings (such as schools, churches and health care facilities) located in the 30 ANEF contour

119 around the airport. Almost $400 million has been spent on this program (approximately the same as the capital cost of the third runway). This has involved the insulation of 4,600 residences and 93 public buildings, as of the beginning of 1995 and end of 1997.

Aircraft Types Only Chapter 3 aircraft (see section 2.3.7) are permitted to operate at Sydney Airport on a regular basis.

Finally, no available information about community health and well-being impacts by aircraft noise can be found on the website of Sydney Airport.

3.5 AIRCRAFT NOISE MANAGEMENT IN OTHER COUNTRIES

This section reviews aircraft noise management strategies, or plans, of major commercial airports in the United States, Canada, and England. Major commercial airports in other countries such as Schiphol Airport (Netherlands), Tokyo International Airport (Japan), Beijing Capital International Airport (China), Frankfurt International Airport (Germany), Charles De Gaulle International Airport (France), and Barajas International Airport (Spain) were excluded because of the language constraint. Even though most of the official websites of these airports have been translated from their first language into English, technical information, such as aircraft noise abatement programs, has been withheld.

Based on data provided in Wikipedia Encyclopedia website , there are almost two-hundred major commercial airports in the United States, and around twenty major commercial airports in both Canada and England. Thus, this thesis arbitrarily selected twenty two major commercial airports in the United States and three busiest airports in both Canada and England as a case study.

3.5.1 United States This subsection reviews aircraft noise abatement/mitigation programs of twenty-two major commercial airports in the United States. Table 3.2 presents the general

120 information of these selected airports. All of them are operated by local/state governments. Note that the United States has many major commercial airports. In 2003, Hartsfield-Jackson Atlanta Airport is the world’s busiest airport in term of passengers annually, and O’Hare International Airport is the world’s busiest airport in term of aircraft movements annually. The new Denver International Airport is the world’s biggest airport by area.

The Federal Aviation Administration (FAA) has authority and responsibility to reduce aircraft noise at the source, implement feasible noise abatement procedures (which are proposed by the airport proprietor and will be approved according to the Federal Aviation Regulation, FAR, Part 150), and encourage compatible land development. The airport proprietors are primarily responsible for planning and implementing actions that manage the effects of aircraft noise within the airport’s environs. The examples of such actions are noise abatement ground procedures; acoustical insulation programs; and proposing noise abatement operational procedures. The State and Local Governments are responsible for compatible land development. The detailed discussion of FAR Part 150 can be found from .

The noise abatement procedures and noise mitigation procedures of these selected airports in the United States were observed from their official website. The results are summarised in Table 3.3. Note that, from Table 3.3, the ‘yes’ symbol means the procedure has been implemented by the airport, and the information of this particular procedure could be found in the website. The ‘no’ symbol means the procedure has not yet been implemented by the airport as it is stated in the website. Finally, the ‘N.A.’ symbol means there is no available information about this procedure (whether it has been implemented or not) provided in the website. Note that, besides the observation of data from these airport website, additional documentation of aircraft noise management procedures of some of these selected airports was provided as part of a benchmarking study for the former Australian Federal Airport Corporations (Black, 1997).

The current aircraft noise measurement system in most of US airports is the Day Night Average Sound Level (DNL) with only some airports using the Community Noise Exposure Level (CNEL) to quantify aircraft noise. From Table 3.3, it might be

121 concluded that most of the major commercial airports in United States have implemented all preliminary procedures of aircraft noise abatement program and aircraft noise mitigation program. These include the preference runway and flight path usage, the noise monitoring and track system, phasing out of noisy aircraft, controlling of ground running noise, land-use compatibility program, and acoustic insulation program. Most of the airports have established community programs to encourage public involvement, and to allow people to complain and make suggestion about aircraft noise problems. Only two airports (Washington National Airport and San Diego International Airport) ban aircraft operation during the night time. Most airports limit the operation of noisy aircraft during the night time. Unlike Sydney Airport, the sound insulation program of airport in United States is a voluntary project. Most of the airports provide acoustic insulation to noise-sensitive public buildings only, such as schools, churches, and hospitals.

There is only one airport (Burbank-Glendale-Pasadena Airport) that has implemented an attitude survey regarding to the aircraft noise management procedures. Nevertheless, it has done so in 1985. Finally, there is no available information about the study of effects of aircraft noise on community health and well-being provided in either the official website or aircraft noise management documents of these major commercial airports in the United States.

122 Table 3.2: Twenty Two Major Commercial Airports in United States

Operations Passenger No. of Name Location Operated by Official Website year 2003 year 2003 Runway Hartsfield-Jackson Atlanta International Airport Atlanta, Georgia Department of Aviation of the City of Atlanta http://www.atlanta-airport.com 911,723 79,086,792 4 O'Hare International Airport Chicago, Illinois Chicago City Department of Aviation http://www.ohare.com/ohare/home/asp 928,691 69,508,672 6 Los Angeles International Airport Los Angeles, California Los Angles World Airports http://www.lawa.org/lax 622,378 54,982,838 4 Dallas-Fort Worth International Airport Dallas/Fort Worth, Texas City of Dallas, City of Fort Worth http://www.dfwairport.com 765,109 53,253,607 5 Denver International Airport Denver, Colorado City & County of Denver Depart of Aviation http://www.flydenver.com 510,275 37,505,138 6 Sky Harbor International Airport Phoenix, Arizona City of Phoenix Aviation Department http://phoenix.gov/aviation/index.html 541,771 37,412,165 2 McCarran International Airport Las Vegas, Nevada Clark County, Nevada http://www.mccarran.com 501,029 36,285,932 4 George Bush Intercontinental Airport Houston, Texas City of Houston Department of Aviation http://iah.houstonairportsystem.org 474,913 34,154,574 5 San Francisco International Airport San Francisco, California City of San Francisco http://www.flysfo.com 334,515 29,313,271 4 Orlando International Airport Orlando, Florida Greater Orlando Aviation Authority http://www.orlandoairports.net 295,542 27,319,223 4 Seattle-Tacoma International Airport Seattle, Washington Port of Seattle http://www.portseattle.org/seatac 354,770 26,755,888 2 Philadelphia International Airport Philadelphia, Pennsylvania City of Philadelphia http://www.phl.org 446,529 24,671,075 4 Portland International Airport Portland, Oregon Port of Portland http://www.portlandairportpdx.com 557,853 12,395,938 3 Washing National Airport Arlington County, Virginia Metropolitan Washington Airports Authority http://www.mwaa.com/national 250,802 14,223,123 3 Washington Dulles International Airport Washington, DC Metropolitan Washington Airports Authority http://www.metwashairports.com/dulles 335,397 16,950,381 3 Baltimore/Washington International Airport Maryland, Washington, DC Maryland Aviation Administration http://www.bwiairport.com 299,300 19,696,130 5 San Diego International Airport San Diego, California San Diego County Regional Airport Authority http://san.org 203,285 15,260,791 3 Burbank-Glendale-Pasadena Airport Burbank, California Burbank-Glendale-Pasadena Airport Authority http://burbankairport.com 178,079 4,729,936 2 Logan International Airport East Boston, Massachusetts Massachusetts Port Authority http://www.massport.com/logan/ 373,304 22,791,1695 San Jose International Airport San Jose, California City of San Jose http://www.sjc.org 235,760 10,700,000 3 Honolulu International Airport Honolulu, Hawai'i Hawai'I State Department of Transportation http://www.hawaii.gov/dot/airports/ 323,726 19,749,905 4 Lambert-St. Louis International Airport St. Louis, Missouri St. Louis Airport Authority http://www.lambert-stlouis.com 324,210 20,431,132 4

123 Table 3.3: Aircraft Noise Management Procedures of Twenty-Two Major Commercial Airports in United States Aircraft Noise Management Procedures Code Airport 123456789101112 Hartsfield-Jackson Atlanta International Airport N.A. N.A. N.A. N.A. yes yes yes N.A. N.A. yes no yes O'Hare International Airport yes N.A. N.A. yes yes yes yes yes yes N.A. no yes Los Angeles International Airport yes N.A. N.A. yes yes N.A. yes N.A. yes N.A. no N.A. Dallas-Fort Worth International Airport yes N.A. N.A. yes yes yes yes yes yes N.A. no N.A. Denver International Airport yes N.A. N.A. N.A. yes yes yes yes yes yes no N.A. Sky Harbor International Airport N.A. N.A. N.A. yes N.A. N.A. N.A. N.A. yes yes no N.A. McCarran International Airport yes N.A. N.A. yes yes N.A. yes yes yes yes no yes George Bush Intercontinental Airport N.A. N.A. N.A. N.A. yes yes yes N.A. N.A. yes no N.A. San Francisco International Airport yes N.A. N.A. yes yes N.A. yes yes N.A. yes no yes Orlando International Airport yes N.A. N.A. N.A. yes yes N.A. N.A. yes yes no N.A. Seattle-Tacoma International Airport N.A. N.A. N.A. yes yes yes yes N.A. N.A. yes no yes Philadelphia International Airport N.A. N.A. N.A. N.A. yes yes yes N.A. yes yes no yes Portland International Airport yes N.A. N.A. yes N.A. N.A. yes yes yes yes no yes Washing National Airport yes N.A. N.A. yes yes yes yes yes yes yes yes yes Washington Dulles International Airport N.A. N.A. N.A. yes yes yes yes N.A. yes N.A. N.A. N.A. Baltimore/Washington International Airport yes N.A. N.A. yes N.A. N.A. yes yes yes yes N.A. yes San Diego International Airport N.A. N.A. N.A. yes yes yes N.A. yes yes yes yes N.A. Burbank-Glendale-Pasadena Airport yes yes N.A. yes yes yes yes yes yes yes no N.A. Logan International Airport yes N.A. N.A. yes yes yes yes yes yes yes no yes San Jose International Airport yes N.A. N.A. yes yes N.A. N.A. N.A. N.A. yes yes yes Honolulu International Airport yes N.A. N.A. N.A. yes N.A. yes yes yes yes N.A. N.A. Lambert-St. Louis International Airport yes N.A. N.A. yes yes yes yes yes yes yes N.A. yes Procedure Code: 5 = Noise Monitoring Terminal 10 = Complaint Unit 1 = Community Program 6 = Noise and Flight Track System 11 = Curfew 2 = Attitude Survey 7 = Preference Runway and Fligth Path Use 12 = Ground Running 3 = Health Impacts study 8 = Aircraft Certification 4 = Sound Insulation Program 9 = Land-Use Compatibility

124 3.5.2 Canada Three busiest airports in Canada in term of passengers annually and aircraft movement annually are Toronto Pearson International Airport, Vancouver International Airport, and Montreal-Pierre Elliott Trudeau International Airport. Table 3.4 tabulates the general information of these airports. All of them are operated by the state government sector. The official websites of each airport are provided in this table.

Table 3.4: Three Busiest Commercial Airports in Canada

Operations Passenger No. of NameLocation Operated by Official Website year 2003 year 2003 Runway Toronto/LB Pearson The Greater Toronto Toronto http://gtaa.com 371,610 24,739,312 5 International Airport Airports Authority Vancouver Vancouver International Vancouver http://www.yvr.ca 290,478 14,321,504 3 International Airport Airports Authority Montreal-Pierre Elliott Trudeau Lontreal Aeroports de Montreal http://www.admtl.com 230,123 8,963,966 3 International Airport

The regulations governing aviation in Canada can be found in the Canadian Aviation Regulations. Information, such as standards for aircraft noise and airport noise abatement, and noise control procedures, are provided in the Canadian Aviation Regulations which can be viewed in full on Transport Canada’s website at . Each airport in Canada has specific Aircraft Noise Abatement Procedures. However, all of them have to be approved and enforced by the national government, Transport Canada.

The noise abatement procedures and noise mitigation procedures of these three selected airports were observed from its official website. The results are summarised in Table 3.5. Note again that, from Table 3.5, the ‘yes’ symbol means the procedure has been implemented by the airport, and the information of this particular procedure can be found in the website. The ‘no’ symbol means the procedure has not yet been implemented by the airport, as stated on the website. Finally, the ‘N.A.’ symbol means there is no available information about this procedure (whether it has been implemented or not) provided in the website.

125 The current aircraft noise measurement system in Canadian airports is the Noise Exposure Forecast (NEF). The aircraft noise management procedures implemented in these three airports are quite similar. For instance, the comprehensive noise monitoring and flight track system has been installed at these three airports. A community program and committee regarding aircraft noise problem has been established. The 24-hr complaint unit about aircraft noise is also available. None of these airports implement curfew procedures, but it is noted that noisy aircraft operations in all airports are restricted during the night time. Unlike the other two airports, Vancouver Airport provides information about attitude surveys of the community, sound insulation programs on noise sensitive buildings (such as schools and hospitals), and reverse thrust monitoring system on its official website. Finally, there is no available information about the study of effects of aircraft noise on community health provided in the official website of these three Canadian airports.

Table 3.5: Aircraft Noise Management Procedures of Three Busiest Airports in Canada Aircraft Noise Management Procedures Code Airport 12345678910111213 Toronto/LB Pearson yes N.A. N.A. N.A. yes yesN.A. yes yes yes yes no yes International Airport Vancouver yes yes N.A. yes yes yes yes yes yes yes yes no yes International Airport Montreal-Pierre Elliott Trudeau yes N.A. N.A. N.A. yes yes N.A. yes N.A. yes yes no yes International Airport

Procedure Code: 1 = Community Program 6 = Noise and Flight Track System 11 = Complaint Unit 2 = Attitude Survey 7 = Reverse Thrust Monitoring System 12 = Curfew 3 = Health Impacts study 8 = Preference Runway and Fligth Path Use 13 = Ground Running 4 = Sound Insulation Program 9 = Land-Use Compatibility Program Procedures 5 = Noise Monitoring Terminal 10 = Aircraft Certification

3.5.3 England The three busiest airports in England in terms of annual passengers and annual aircraft movement are London Heathrow Airport, London , and Manchester International Airport. In 2003, London Heathrow Airport was the third busiest airport in the world, and London Gatwick Airport is the world’s busiest single runway airport. Table 3.6 presents the general information of these airports. All of them are operated by the private sector. The official websites of each airport are provided in the table.

126 Table 3.6: Three Busiest Commercial Airports in England

Operations Passenger No. of Name Location Operated by Official Website year 2003 year 2003 Runway London Heathrow Airport London British Airports http://www.baa.co.uk/main/airports/heathrow 465,000 63,487,136 3 Authority London Gatwick Airport London British Airports http://www.baa.co.uk/main/airports/gatwick 234,899 30,007,021 1 Authority Manchester International Manchester Manchester http://www.manchesterairport.co.uk 189,065 19,901,403 2 Airport Airport Group

The legal responsibility for aircraft noise control in England is mainly based on the Civil Aviation Authority (Civil Aviation Act 1982) and International Civil Aviation Organisation. The national legislations set out by the Department for Transport, United Kingdom (http://www.dft.gov.uk) are also relevant. Additional control may be implemented through locally agreed policies and planning conditions from nearby local councils. Four key elements that have been implemented in many airports in England, to control noise from airports, are recommended by International Civil Aviation Organisation. These are: (1) Reducing Noise at Source – by means of progressive tightening of noise certification standards; (2) Land-Use Planning and Management – to ensure that inappropriate development is discouraged, or prohibited, around airports; (3) Noise Abatement Procedures – to minimise the noise impact from overflights, for example the use where feasible of continuous descent approach; and (4) Operating Restrictions – to limit the operation of aircraft during night time or to phased out of the noisiest aircraft types.

Under the Civil Aviation Act 1982, it is required on a statutory basis that the principal mitigation measure for aircraft noise impacts is the provision of acoustic insulation. In practice, however, all current noise insulation schemes around major airports are provided on a voluntary basis by airport operators supported by local planning agreements. The detailed discussion of aircraft noise management strategies in United Kingdom’s airports could be found on the official website of Department for Transport, Great Britain (http://www.dft.gov.uk).

The aircraft noise management procedures of three selected airports in England are summarised in Table 3.7. The interpretation of the symbols used in Table 3.7 is the

127 same as for Tables 3.3 and 3.5. The current aircraft noise measurement system in

England airports is A-Weighted Equivalent Continuous Sound Pressure Level (LAeq). From Table 3.7, even though the land-use compatibility program is one of the four elements in controlling airport noise in England, the information about this program is not available on the airports’ websites. The noise remedy program (sound insulation) has been implemented in all three airports. Similarly, as with the other major commercial airports, these three airports have installed a comprehensive noise monitoring and flight track system. The usage of flight paths and runways are controlled where feasible to avoid overflight on residential areas. Information of a new arrival procedure called Continuous Descent Approach could be observed from the website of Heathrow Airport. The Heathrow Airport declared that this new procedure results in lower noise and emission than the traditional procedure (called Stepped Approach).

Table 3.7: Aircraft Noise Management Procedures of Three Busiest Airports in England.

Aircraft Noise Management Procedures Code Name 1234567891011121314

London Heathrow Airport yes N.A. N.A. yes yes yes yes yes yes N.A. yes no yes yes

London Gatwick Airport yes N.A. N.A. yes yes yes yes N.A. yes N.A. yes no N.A. N.A.

Manchester International yes N.A. N.A. yes yes yes yes N.A. yes N.A. yes no yes yes Airport

Procedure Code: 6 = Noise and Flight Track System 11 = Complaint Unit 1 = Community Program 7 = Preference Runway and Fligth Path Use 12 = Curfew 2 = Attitude Survey 8 = Continuous Descent Approach 13 = Noise Levy 3 = Health Impacts study 9 = Aircraft Certification 14 = Ground Running 4 = Sound Insulation Program 10 = Land-Use Compatibility 5 = Noise Monitoring Terminal

None of these three airports implement curfew procedures. It was referred by the Department for Transport (2003, paragraph 3.12) that “the government recognises that noise from aircraft operations at night is widely regarded as the least acceptable aspect of aircraft operations … but we must strike a fair balance between local disturbance … and the economic benefits of night flights”. The procedures to control aircraft operations during night time have been developed. Information about financial penalties

128 on noisy aircraft is provided on the website of both Healthrow Airport and Manchester Airport. Finally, there is no available information about the study of effects of aircraft noise on community health and well-being provided in the official website of these three English airports.

3.6 COMMUNITY HEALTH AND WELL-BEING ISSUE IN AIRCRAFT NOISE MANAGEMENT STRATEGY

A documental survey was conducted via the official website of number of airports, to determine current practices in aircraft noise management strategies, at major commercial airports in selected developed countries. It was found that the major concerns of the current aircraft noise management strategies are to minimise, as far as practicable, the total number of people in the community exposed to high levels of noise from overflights, and to remedy, as much as possible, the significant aircraft noise exposure in existing noise-sensitive areas. The issue of community health and well- being impacts by aircraft noise has been ignored by the aircraft noise management strategies. This might reflect the fact that the current policies to protect the environmental conditions from pollution relevant to airport operations defined health only in term of the absence of disease. For instance, the Australian Airport (Environment Protection) Regulation 1997 Part 2 Regulation (2.02) stated that “water pollution has occurred when waters contain a substance or organism that causes … an adverse effect on … public health”, or as stated in Regulation (2.03) that “soil pollution has occurred when land … is contaminated by a substance that causes … harmful to the health or welfare of human beings”.

At present, there is no evidence to support the proposition that aircraft noise is loud enough to significantly deteriorate the hearing ability of people living around the airports. This might be a reason why the effects of aircraft noise on human health have not been mentioned in any acts or regulations to protect environmental conditions from noise pollution due to airport operations. Consequently, as it is not required by law, none of airport operators have considered the effects of aircraft noise on community health as a major issue.

129 Based on previous research (Black, 1997) on environmental management, it was assessed that these were the most advanced airports around the world in terms of dealing with noise and air quality issues. Therefore it was surmised that these airports might have taken initiatives on community health impacts. However, in the case of Australia’s major commercial airports, the website survey mentioned above revealed no information about community health and well-being impacts by aircraft noise. This indicated a requirement to construct a questionnaire survey on this matter.

This thesis argued that policies to prevent environmental conditions from aircraft noise should consider the complete aspects of health, as provided by the WHO definition. While it can be argued that the effects of aircraft noise on community health and well- being would be automatically controlled if the number of people highly exposed to aircraft noise is minimised, the effects of aircraft noise on community health and well- being should not be underestimated (or overlooked) by any deficiency in the interpretation of the meaning of ‘health’.

3.7 CONCLUSIONS

Aviation is one of the world’s fastest growing industries with demand doubling in size over the last decade due to the development of the global economy (Upham, et al, 2003). This economic development is coming at a cost. Aircraft noise pollution is the major environmental issue for airports. Even though new technology is making aircrafts quieter, the rapid growth in air traffic may limit the net reduction in overall noise levels generated by individual airports. Governments, especially in the developed countries, have enacted laws and regulations to control aircraft exposure on communities, so the negative impacts on the environment and the positive impacts on the economy could be fairly balanced. As required by law, the airport authorities have implemented many actions to prevent (called Noise Abatement Programs) and reduce (called Noise Mitigation Programs) aircraft noise exposure on community in the vicinity of airport. This chapter not only reviewed the aircraft noise management procedures in many major commercial airports in Australia, United States, Canada, and England but also

130 aimed to determine whether or not the community health and well-being impact issues have been taken into account into these procedures.

Most of the airports studied implement in parallel both Noise Abatement Programs and Noise Mitigation Programs. The preferences of runway and flight path use, whenever practicable, to avoid overflight on residential area, and the implementation of land-use compatibility around the airport are the most popular methods among the airports. It seems that Sydney (Kingsford Smith) Airport is the one that has spent the most efforts to alleviate aircraft noise exposure on community. This might be a result of the strong opposition that arose from community to the opening of the third runway in 1994.

Even though most of the airports have established a community program to encourage public opinion in solving aircraft noise problems, this research has found no information provided in the airport official website to indicate that the current practice to management aircraft noise problems at major commercial airports has taken into account the issue of community health and well-being. This might be a reason why the aircraft noise problems have never been satisfactorily resolved. This thesis proposes that one way to counter the aircraft noise problems is to fully understand its effects to the community before devising strategies and counter-measures.

It could be argued that the document survey was rather descriptive and included many not available (N.A.) information entries (see Table 3.3, Table 3.5, and Table 3.6). However, these outcomes reflect the fact that most major commercial airports have not provided sufficient information about what has been doing in attempt to alleviate aircraft noise problems to the community. It is accepted that at present the internet can be easily assessable, especially in developed countries (such as Australia, United States, England, and Canada), and has become a major sort of information for the general community. Therefore, it is reasonable to assume that if the community cannot receive information on aircraft noise management strategy from the official website of the airport itself, the community program to corporate community with the airport may not have been effectively established.

131 CHAPTER FOUR

HEALTH SURVEY METHODOLOGY, PILOT STUDY, AND CASE STUDY

4.1 INTRODUCTION

The previous chapter has reviewed the aircraft noise management of major commercial airports in Australia and in other countries, such as United States, Canada, and England. None of the airports have included the effects of aircraft noise on community health and well-being into their aircraft noise management plans, or strategies. Population-based studies of community health impacts to aircraft noise are required, especially in term of health related quality of life. The literature review has also indicated that even though there is no robust conclusion about the causality between aircraft noise exposure and disease, aircraft noise potentially deteriorates quality of life which is one of major components of health. There has been little research into the impacts of aircraft noise on health related quality of life. The review has further indicated that aircraft noise might have association with adult blood pressure with noise stress as a mediating factor. However, it was found that epidemiological studies of adult blood pressure level influenced by aircraft noise are rare. Two core research questions have been established: “Is health related quality of life worse in communities chronically exposed to aircraft noise than in a communities not exposed?”; and “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”.

This chapter proposes a methodology, which will be applied to a major commercial airport as a case study, to explore the core research questions. The Australian aviation sector is chosen as a case study with specific reference to Sydney (Kingsford Smith) Airport. This chapter is organised as follows. Section 4.2 deals with the selection of epidemiological methods suitable for this research investigation. Section 4.3 describes the development of the proposed health survey instruments, including questionnaire,

132 contact letters, survey administration, and also the consideration of ethical aspects of conducting this type of research. The proposed health survey instruments were tested through the pilot study. The results and recommendations for the main survey are discussed in Section 4.4. Section 4.5 describes the selection of study population of both aircraft noise exposure area and a control area. Section 4.6 describes the calculation of the required sample size for the main survey. Section 4.7 describes the translation process of the health survey instruments from English into the other languages. Section 4.8 is a description of the whole process.

4.2 EPIDEMIOLOGICAL METHOD

Epidemiology is a branch of knowledge that is concerned with the distributions and determinants of disease frequency in human populations (Hennekens et al, 1987). The design strategies of the epidemiological method are dependent on the purpose of each investigation. Descriptive epidemiology is suitable for studies that focus on describing the distribution of disease, including considerations of variation of disease rates by time, place (or geographic location), and personal characteristics (for example, age, sex, ethnicity, or socioeconomic status). Analytic epidemiology is concerned with the determinants of a disease, with the goal of judging whether a particular exposure causes or prevents disease. Normally, the analytic study tests the hypothesis of determinants by using information provided by the descriptive study.

At present, there is no evidence to support the causality between aircraft noise and “disease”. The objective of this thesis is to study the impacts of aircraft noise on health in term of physical, mental, and social well-being and to develop any possible association between aircraft noise and the prevalence of hypertension. The concept of subjective health assessment has been employed in this thesis. Thus, the epidemiological design that matches these objectives is a descriptive study. It is also considered that descriptive epidemiology is more feasible for this research than analytical epidemiology. In general, an analytical study requires administration of subject, well-trained staff, a medical laboratory and a follow-up period (normally up to

133 5-year interval). There are two main types of descriptive study that are relevant: an ecological study and a cross-sectional study.

The ecological study uses data from entire populations (not information on an individual) to compare disease frequencies between different groups during the same period of time, or in the same population at different points in time. For example, Armstrong and Doll (1975) (as stated in Hennekens et al, 1987, p.17) found a positive relationship between per capita daily consumption of meat and rates of colon cancer in women by using pre-collected data from various countries. This finding could lead to the proposed hypothesis that meat consumption increases the risk of developing colon cancer. However, the ecological study is limited to data referring to whole populations rather than to individuals. It is not possible to link an exposure to occurrence of disease in any one person. It is not possible to indicate what particular women who have developed colon cancer are the ones who consume a lot of meat. Moreover, the ecological study is limited in controlling for factors that might be different between populations and might affect the disease of concern (in the above case, colon cancer).

The cross-sectional study measures the status of an individual in a defined population with respect to the presence or absence of both exposure and disease at the same point in time (or over a short time period) (Hennekens et al, 1987). It is a useful technique, especially when dealing with public health issues. A basic technique in cross-sectional surveys is to use standardised measures for the studying variables (which are exposure or outcome variables). In epidemiology and health social science, these measures are usually standard questions asking everyone in the sample to indicate their attitudes, beliefs, behaviours, or personal characteristics (Albrecht et al, 2001). As it is an individual measurement, the factors that might affect the disease of interest can be controlled. As both exposure and disease are assessed at the same point in time, a major limitation of cross-sectional study is the lack of distinction whether the exposure preceded the development of disease, or whether the presence of the disease affected the individual’s level of exposure (Hennekens et al, 1987). Nevertheless, this might be overcome by undertaking a repeat measure with suitable time intervals (such as 2, 4, or 6 years) with the same baseline subject. Moreover, if there are the other studies seeking to answer the same research problem by using the equivalent methodology, the

134 similarity in final results will establish a strong hypothesis of determinant, which is very useful for the further analytical analysis. This thesis applied a cross-sectional study with a control group as the epidemiological method.

4.3 PROPOSED HEALTH SURVEY PROCEDURES

This measurement of community health and well-being information by means of a survey can be done by oral (or interview) or by written questioning (or questionnaire). The latter method is administered generally by mail or is handed to the respondent and filled in by him/her with no help from the interviewer (called self-administrative technique). The former method also involves a questionnaire but will be filled in by an interviewer. This method can be administered by face-to-face interviews or telephone interviews. The interview method is time consuming, which is impractical for a study that requires a large sample size, as in this research investigation. A detailed discussion of the selection of administrative technique is presented later in section 4.3.2, but that discussion concludes that a self-administered questionnaire is appropriated because it is easy and cheap to implement with an acceptable response rate.

Regardless of whether the questionnaire is administered by mail or in person, the structure of the questionnaire is similar (Sarantakos, 1998). In general, it should consist of a cover letter, the instructions, and the main body of the questionnaire. The main aims of the cover letter are to introduce the respondents to the research objective, to assure them of anonymity and confidentiality, and to persuade them to participate in the survey. Instructions can be placed either in the cover letter or in the main body of the questionnaire. The main aims of the instructions are to explain to the respondents how to fill in the questionnaire, to remind them that there are no right or wrong answers, and that all questions should be attempted. They can provide additional information, if required, by some specific questions. Finally, they can be instructed about what to do with the completed questionnaire. It is important that instructions should be written briefly in a simple language, and should provide as much information as possible. The main body of the questionnaire includes the questions that need to be answered by the

135 respondents. The information required to explore the core research questions are contained in this section.

4.3.1 Questionnaire The questionnaire for this research has been developed from an international, well- established questionnaire instrument that measures seven major characteristics of each subject: 1) health related quality of life; 2) hypertension condition; 3) noise stress; 4) noise sensitivity; 5) noise annoyance; 6) confounding factors; and 7) demographic characteristics. Since the questionnaire was planned to collect a lot of information, the size of the questionnaire (which is implied by the number of pages in the questionnaire) becomes a major issue. The greater the numbers of pages the respondent perceives, the greater the possibility that the questionnaire will be ignored. Any attempt to decrease the number of pages of the questionnaire by reducing the size of the font is not a good idea. The information on the questionnaire should be easy to read, especially for people with relatively poor eyesight. Therefore, this study has limited the number of pages to six which are printed on both sides of white A4 (21cm×29.7cm) paper (with a margin of 1.27cm, 1.4cm, 1.27cm, and 1.27cm for top, bottom, left, and right, respectively) and are stapled along the spine as a booklet. The font type is “Arial” with a font size “10” in black ink colour.

The survey has been named “Environmental Health Survey”. The cover letter informs the respondent that “the objective of this survey is to study the relationship between your health and environmental noise condition in your neighbourhood”. The layout of the questionnaire is one important issue. The questions begin with those that are the most relevant to the topic of this survey and ending with the questions that are of slightly of less relevance. After a brief instruction of how to complete the questionnaire, the questionnaire starts with a set of questions asking about health related quality of life and then is followed by a set of questions asking about hypertension conditions and related issues (which are history of hypertension of parent(s) and high cholesterol status). The following subsection is a set of questions seeking the environmental noise stress status of respondent, the confounder questions, the noise sensitivity questions, the noise annoyance questions, and, finally, the less relevant to the survey topic but ones that are of most importance for statistical analysis: the demographic questions. At the

136 beginning of each subsection of the questionnaire, the respondent is notified by a bold strip bar with the name of each subsection. A brief instruction of how to complete each subsection, if necessary, is also provided at the beginning of each subsection. At the end of the questionnaire, the respondent is guided by a brief instruction of how to return the completed questionnaire.

The main language used in the questionnaire is English. The questionnaire was designed with simple language and to be easily understood by a layperson. Any language with jargon, slang, or complicated expressions was avoided. The questionnaire was planned to be translated from English into other languages (see section 4.7). This task was done by a professional who is fluent in both English and the target language(s).

Besides the adaptation of well-developed questionnaire instruments, many questions relevant to the research objective were developed: questions seeking information about prevalence of hypertension, prevalence of noise stress; and questions designed specifically to capture potential confounders between aircraft noise and health. The following sections describe the detailed development of each subsection.

4.3.1.1 Health Related Quality of Life Before the detailed discussion of the well-developed health related quality of life measure adopted in this thesis, the following paragraphs briefly explain the meaning and definition of health related quality of life. A detailed description of health related quality of life can be found in Patrick and Erickson (1993).

From the World Health Organization’s (WHO) definition of health as “a state of complete physical, mental, and social well-being, and not merely the absence of disease or infirmity”, knowledge about health related quality of life has emerged. Health related quality of life is a subjective health assessment. Since the measure of health is more than objective measures of morbidity, mortality and activity limitations, the subjective health assessment has become an importance component of contemporary health research (Jenkinson, 1994). The definition of health related quality of life, as stated in Patrick and Erickson (1993, p.22) is “the value assigned to duration of life as modified by the impairments, functional states, perceptions, and social opportunities that are

137 influenced by disease, injury, treatment, or policy.” In general, the applications of health related quality of life are in the areas of: 1) clinical practice; 2) clinical and epidemiologic investigations; 3) program evaluation and policy analysis; and 4) population monitoring.

Patrick and Erickson (1993) derived the health related quality of life to cover five broad concepts that combine the quantity and quality of life on a value scale. These concepts are defined further by “domains”, which are states, behaviours, perceptions, other spheres of action and thought in health related quality of life. Table 4.1 lists the five broad concepts of health related quality of life, domains of these concepts, and indicators, or measures, of the concepts and domains. Table 4.1 assists the questionnaire’s developers in considering which concepts and domains to include into the questionnaire that fit their health assessment objective.

There have been many questionnaires developed to measure health related quality of life. They are various in terms of type of scores produced, range of populations and concepts, and the weighting system (preference method or statistical method) in scoring items and in aggregating different domains (Patrick and Erickson, 1993). The type of scores can be produced in terms of: indicator (a single number obtained from a single item); index (a single number summarising multiple concepts of health related quality of life); profile (multiple numbers on the same metric); and batteries (multiple numbers on different metrics). Health related quality of life measures can also be classified by the range of population to which the measures are applied, and the range of concepts and domains included in the measure. Broadly, there are two groups: generic measures (which are applicable across types of disease, across different medical treatments, and across demographic and cultural subgroups); and specific measure (which are designed to measure specific disease, specific medical treatment, and specific patient populations).

One of the core research questions of this thesis is to monitor, and compare, the health related quality of life of communities between aircraft noise exposure area and a control area. Thus, the health related quality of life measure that is suitable for this thesis is the

138 generic measure. Six well-known generic measures are compared in terms of containing the five core concepts of health related quality of life as shown by Table 4.2.

Table 4.1: Core Concepts and Domains of Health Related Quality of Life

Concepts and Domains Definitions / Indicators Opportunity Social or cultural disadvantage Disadvantage because of health; stigma; societal reaction Resilience Capacity for health; ability to withstand stress; physiologic reserves Health perceptions General health perceptions Self-rating of health; health concern/worry Satisfaction with health Satisfaction with physical, psychological, social function Functional status Social function Limitations in usual roles Acute or chronic limitations in usual social roles (major activities) of child, student, worker Integration Participation in the community Contact Interaction with others Intimacy and sexual function Perceived feelings of closeness; sexual activity and/or problems Psychological function Affective Psychological attitudes and behaviours, including distress and well-being Cognitive Alertness; disorientation; problems in reasoning Physical function Activity restrictions Acute or chronic reduction in physical activity, mobility, self- care, sleep, communication Fitness Performance of activity with vigor and without excessive fatigue Impairment Symptoms/subjective complaints Reports of physical and psychological symptoms, sensations, pain, health problems or feelings not directly observable Signs Physical examination: observable evidence of defect of abnormality Self-reported disease Patient listing of medical conditions or impairments Physiologic measures Laboratory data, records, and their clinical interpretation Tissue alterations Pathological evidence Diagnoses Clinical judgments after "all the evidence" Death and duration of life Mortality; survival; years of life lost

(source: reproduced from Patrick and Erickson, 1993, table 4.1, p.77).

Noting the limitations imposed by the size of the questionnaire discussed above in section 4.3.1, the most relevant core concepts (which are Health perceptions and Functional status) have been included in the questionnaire. From Table 4.2, there are three well-known generic measures that satisfy this primary requirement. The question arises, which one of these three measures is the best for the purposes of this thesis? These three measures are the: Nottingham Health Profile (NHP); the Short Form Health

139 Survey (SF-36); and the Sickness Impact Profile (SIP). They have also been validated by their developers (Jenkinson, 1994).

Table 4.2: Major Concepts of Health Related Quality of Life contained in Six Well- Known Generic Measures

Generic Measure Disability/ Health Nottingham Quality of Short Form Sickness Concept Dimension Distress Utilities Health Well-Being Health Survey Impact Index Index Profile Scale (SF-36) Profile Opportunity + Health perceptions + + + Functional status Social +++++ Psychological + + + + + Physical++++++ Impairment + + + + Death and duration of life + + +

(source: reproduced from Patrick and Erickson, Table 5.6, p.132).

The detailed development of SIP and NHP can be found in Jenkinson (1994). The SIP was developed by Bergner et al (1981) to provide an indicator of perceived health status over time and across groups. The NHP was developed by Hunt et al (1986) using a similar technique with the development of SIP. However, the NHP was designed to overcome the SIP’s problems (Jenkinson, 1994). The NHP was designed to capture what individuals are feeling when they experience different levels of illness and activity limitation. The NHP is short, simple and easily understood, and easier to complete than the SIP (Fayers and Machin, 2000). The SF-36 was developed by Ware and Shebourne (1992). The SF-36 was designed to provide assessments of generic health concepts that are not specific to any age, disease, or treatment group. The SF-36 was found to be more sensitive to small changes in health status than the NHP (Fayers and Machin, 2000).

The SF-36 is a multipurpose scale with 35 questions that assesses eight health concepts: 1) physical functioning; 2) social functioning; 3) role limitations due to physical problems; 4) role limitations due to emotional problems; 5) bodily pain; 6) general mental health; 7) sense of vitality; and 8) general health perceptions, with one question to assess health transition. The SF-36 is practical to be both self-administered by an adult (more than 14 years of age) and administered in person or by telephone. The SF-

140 36 was designed to measure two different recall period: 1) standard measurement with four weeks recall period; and 2) acute measurement with one week recall period. The SF-36 was standardised and validated by its developers, and has been accepted by many studies as a very strong and useful outcome-health-measure tool (Jenkinson et al, 1994). The SF-36 (version 2) was introduced in 1996 with improvement in some scales, and also other aspects, such as layout and scoring method (Ware, 2000).

This thesis has selected the SF-36 (v.2) (standard four-week recall) to measure the health related quality of life because it is the latest measure, most frequently referred by international publications (Ware, 2000), and is widely used in many countries (Fayers and Machin, 2000). The Australian Bureau of Statistics used SF-36 in the 1995 National Health Survey to monitor health related quality of life of people, and to develop the population norm based across the nation. Four scales of SF-36 (v.2) (which are Physical Functioning, General Health, Vitality, and Mental Health) were carefully selected and included into the questionnaire. For each health measure, a summary score in the range of 0 to 100 was obtained with the SF-36 algorithm (Ware, 2000), with a higher score implying a more positive health status. Table 4.3 provides the interpretation of the lowest and the highest scores of those selected SF-36 scales.

Table 4.3: Interpretation of Lowest and Highest Scores of Selected SF-36 Scales

Definition Lowest Possible Score Highest Possible Score Very limited in performing all Performs all types of physical activities Physical Functioning physical activities, including including the most vigorous without (PF) bathing or dressing limitations due to health General Health Evaluates personal health as Evaluates personal health as excellent (GH) poor and believes it is likely to get worse Vitality Feels tired and worn out all Feels full of pep and energy (VT) of the time all of the time Mental Health Feelings of nervousness and Feels peaceful, happy, and (MH) depression all of the time calm all of the time (source: reproduced from Ware and Shebourne, 1992, table.1, p.475).

4.3.1.2 Hypertension and Related Measures To explore the second core research question, a close-end question for assessing the prevalence of hypertension has been developed for this research. “Have you ever been

141 told by a doctor or nurse that you have high blood pressure sometimes called hypertension” (1) Yes (2) Yes, but only temporarily (3) No, and then “If YES, do you currently have high blood pressure? (1) Yes (2) No. A close-response (Yes, but only temporarily) was provided because it might be a situation that the blood pressure level will be temporarily raised due to pregnancy, or for some other reason(s).

It is evident that the history of hypertension of parent(s) and cholesterol level are related to hypertension (Rebbeck et al, 1996; and Chen et al, 1995). Therefore, to prevent the distortion effects from those variables, this research developed the close-end questions for assessing this history of hypertension of parent(s) and high cholesterol status. At any time in the past, have either of your parents ever been told by a doctor or nurse that they have high blood pressure sometimes called hypertension? (1) Yes (2) No (3) Don’t know. Have you ever been told by a doctor or nurse that you have high cholesterol? (1) Yes, and currently have (2) Yes, but already healed (3) No.

4.3.1.3 Noise Stress The literature suggests that human stress response triggers stress hormones which potentially cause changes in heart rate and blood pressure. A sudden, or uncontrollable and intense, noise can be a source of stress hormone release. For acute stress responses, changes in heart rate and blood pressure soon return to normal, but the chronic stress can potentially lead to a persistent increase in stress hormone level and blood pressure which are risk factors in hypertension and circulatory problems in the future. The research hypothesises that long-term aircraft noise exposure is one kind of chronic stress.

There is no well-developed scale to measure the perceived stress due to environmental noise. Therefore, an environmental noise stress scale with 12 months recall period has been developed to measure chronic stress due to noise. The 12 months recall period was selected because it is assumed to represent ‘long-term’ effect (or chronic). The subsection starts with a brief introduction to the meaning of environmental noise pollution and how to complete the questions. The phase “noise pollution (for example, transport noise, industrial noise and community noise)” was used instead of “aircraft noise pollution” to protect any attitudinal bias by the respondent about the noise source.

142 The proposed noise stress scale consists of four close-end items: 1) Have you been upset and/or angered because of noise pollution in your neighbourhood?; 2) Have you felt nervous and “stressed” because of noise pollution in your neighbourhood?; 3) Have you found that you could not cope with noise pollution in your neighbourhood?; and 4) Have you felt noise pollution in your neighbourhood was so high that you could not overcome it?. All items are presented on a 5-point scale: (1) Very often; (2) Fairly often; (3) Sometimes; (4) Almost never; (5) Never. Each item has equal contribution to the scale without any weighting factor. Each item is scored by six minus by the identification number of each response scale. The final score is a summation of all item scores. For example, respondent who selects (1) Very often for every item on the noise stress scale will be obtained a score of 20 (4 items × 5 mark), or a respondent who selects (2) Fairly often for the first two items and (4) Almost never for the least will obtain a score of 12 (2 items × 4 mark plus 2 items × 2 mark). The internal reliability of the proposed environmental noise stress scale was tested in a pilot study.

4.3.1.4 Noise Sensitivity Noise sensitivity is a factor that relates to personal vulnerability and intervenes between noise exposure and noise annoyance. A noise sensitive person might feel very annoyed with a certain noise level considered by non-noise sensitive person as an acceptable level. Theoretically, noise sensitivity has a two-way relation with stress. Noise sensitive people have a low capability to cope (alter behaviour to deal with the stressor) with a noise stimulus leading them to get more stressed than normal. Stress itself makes people less tolerant to an unwanted sound, or more sensitive to noise than people who are more mentally calm. Some health problem(s) are assumed to have either a positive or negative effect to noise sensitivity factor. For example, people who have psychiatric illness are most likely to be more noise sensitive than normal and people who have an auditory deficiency are less sensitive to noise because of the lack of their hearing ability.

Noise sensitivity might underestimate the effects of the noise. For example, noise sensitive person have moved out from the high aircraft noise exposure area, given that remaining person is less sensitive to noise. Noise sensitivity is an important variable in exploring the core research questions.

143 A single-item noise sensitivity question developed by Heinonen-Guzejev (2000) is a global question in measuring noise sensitivity: People experience noise in different ways. Do you experience noise generally as very disturbing, quite disturbing, not especially disturbing, not at all disturbing or can’t say? (1) Very disturbing, (2) Quite disturbing, (3) Not especially disturbing, and (4) Not at all disturbing or can’t say. A respondent who answers either the first or second answers will be classified as a noise sensitive person; otherwise he/she will be classified as a non-noise sensitive person.

However, Zimmer and Ellermeier (1999) found that a single-item scale meets the requirement of psychometric criteria (which are distribution of scores, reliability, and correlations with demographic variables) at a lower level than the multiple-item question. Therefore, this thesis pays attention on the multiple-item question in assessing noise sensitivity rather than a single-item question. The noise sensitivity scale developed by Weinstein (1978) (called Weinstein scale) is one that has been referred to by the most international publications as a validated and accurate noise sensitivity scale (Ohrstrom, 1988).

The Weinstein scale consists of a comprehensive set of twenty-one items asking about how people feel about noise at home, in a library, in cinemas, or in the work place. The original version of the Weinstein scale was considered too lengthy for this research questionnaire, especially when it is combined with the other scales in the questionnaire. Therefore, ten items were carefully chosen as they are considered more functional to the study than the remaining items. For example, 1) I wouldn’t mind living on a noisy street if the apartment I had was nice; 2) I am more aware of noise than I used to be; and 3) Sometimes noise gets on my nerves and gets me irritated. All items are presented on a 6- point scale: (1) Agree strongly; (2) Agree; (3) Agree mildly; (4) Disagree mildly; (5) Disagree; (6) Disagree strongly. The score of noise sensitivity will be obtained by the Weinstein scoring method (Weinstein, 1978). The higher the noise sensitivity score then the more noise sensitive people are. The internal reliability of those selected items was tested by the pilot study (see section 4.4.3).

144 4.3.1.5 Noise Annoyance As mentioned in the literature review, annoyance is a feeling of displeasure associated with any agent, or condition, believed to adversely affect an individual, or a group. Individual noise annoyance depends on many factors such as cultural factors, types of activity at time of noise exposure, attitude to noise source, noise sensitivity, controllability of the stressor, and other individual difference (Berglund et al, 1996). Nevertheless, noise annoyance is not statistically influenced by demographic variables (for example, sex, age, and education level) (Miedema and Vos, 1999). Many efforts have been attempted on developing a scale (either single-item or multiple-item) to measure noise annoyance. They are different in word, pattern and format, and rating scale.

This research adapted standardised noise reaction questions for community noise survey developed by Fields et al (2001). The modified annoyance measurement consists of two sections. Each section consists of two questions assessing annoyance from traffic noise and aircraft noise to protect attitude bias of respondent to the noise source. The first section measures annoyance of subjects from daily activity disturbances: “Thinking about the last 12 months, when you are here at home, how much does noise from aircraft bother, disturb, or annoy you in the following aspects”; and “Thinking about the last 12 months, when you are here at home, how much does noise from road traffic bother, disturb, or annoy you in the following aspects”. Both questions consist of eight items, for example, a) Telephone conversation, b) Listening the radio/TV, c) Ordinary conversation, f) Sleep, except one extra item, i) Fear of air accident or air terrorism, was added into the question of aircraft noise annoyance due to the events in the USA on September 11, 2001. All items are presented on a 5-point scale: (1) Extremely; (2) Very; (3) Moderately; (4) Slightly; (5) Not at all.

The second section starts with a brief instruction “Next is a zero to ten opinion scale for how much noise bothers, disturbs or annoys you when you are here at home. If you are not at all annoyed choose zero, if you are extremely annoyed choose ten, if you are somewhere in between choose a number between zero and ten” and then asks people to consider all items from the first section and rate their overall annoyance from each noise source by circling the chosen number on the provided opinion scales.

145 4.3.1.6 Confounding Factors and Demographic Characteristics Any factor that distorts an association between exposure and disease is called a confounding factor (or confounder). Figure 4.1 illustrates the relation among exposure, disease and confounder factors. One factor will be considered as a confounder if it meets the three following criteria (Hennekens et al, 1987): 1) to be a known risk factor for the result of the disease. 2) to be a factor associated with exposure, but not a result of exposure. 3) to be a factor that is not an intermediate variable between exposure and disease.

Exposure Disease

Confounding Factor Figure 4.1: Correlation among Exposure, Disease, and Confounding Factor

To study an association between exposure and disease, it is necessary that any possible confounders are controlled. However, identification of the confounding factors is not an easy task. Most of the confounding factors are human behaviour, daily activities and living environment which vary vastly between people and society. Previous studies (De Cesaris et al, 1992, Chen et al., 1995, Makoto et al., 2000, and Gregory et al., 2005) have shown that smoking, alcohol intake, exercise activities, and dietary salty intake could be a risk factor of hypertension. A questionnaire that captures all possible potential confounders based on the literature in an association between health and well- being and aircraft noise has been designed by this research. The general confounder questions such as employment status, exercise activities, smoking status, alcohol consumption, and consuming salty food and also the demographic characteristic questions (which are gender, age, weight, height, education, marital status, and a household weekly income) have been adapted from the National Health Survey 2001, Australian Bureau of Statistics (ABS) and 1997 NSW Health Survey Questionnaire.

Previous research has linked large intakes of salt to high blood pressure level. Too much salt (sodium chloride) in one’s diet from salty foods may increase the risk of high blood

146 pressure. It has also been recommended that lifestyle modifications, especially reducing salt intake or performing activities that promote the removal of sodium from the body are most important in treating hypertension (Kikuo, 2004). While the ratio of fish to red meat might relate to cardiovascular health, a set of questions to capture this information is required. It was considered too lengthy for the present research questionnaire (see Section 4.3.1), especially when it is combined with the other scales in the questionnaire.

Some confounder questions have been designed specifically for this research. For instance, a question measuring smoking status of other members of the house has been included to eliminate the impact of passive smoking, “Except you, does anyone else in this house smoke regularly, that is, at least once a day?” (1) Yes, (2) No.

An open-end question asking how long has the respondent lived in his/her house has been developed to satisfy a research assumption that long-term aircraft noise exposure has negative impacts to human health. The analysis was restricted to respondent who had resided at the address for at least one year. A noise confounding question (which is “Have you recently insulated your house from noise?” (1) Yes, (2) No) was included into the questionnaire to eliminate the effect from acoustic insulation (which has been a feature of the noise management plan at Sydney (Kingsford Smith) Airport following the opening of the Third Runway in 1994).

4.3.2 Survey Administration

4.3.2.1 Selection of Administrative Technique Three commonly used techniques, which are face-to-face interviews, phone interviews, and by mail in administering questionnaire surveys, are compared in Table 4.4. First of all, a simple (but essential) criterion in selecting which administration technique is suited to this thesis has been set. As this thesis is a part of PhD project, this research requires an administrative technique that is easy and cheap to implement with an acceptable response rate, because of time and budget constraints that accompany doctoral research.

147 The advantages of face-to-face methods are the abilities: to control questionnaire design; to identify and assist the respondent; and to receive a high response rate. However, the face-to-face method has some disadvantages that are undesirable. This method requires a well-trained interviewer which involves high cost to train and employ. Moreover, the administering of this method takes place, normally in the subject’s home, and to succeed in this access permission is required, which involves time.

Table 4.4: Advantages and Disadvantages of Survey Administrative Techniques Scoring 1 = Poor; 2 = Satisfactory; 3 = Good Face-to-face Phone Mail Mode of delivery Response rates General samples 3 2 2 Specialised samples 3 2 2 Represetative samples Avoidance of refusal bias 3 3 1 Control over who completes questionnaire 3 2 2 Gaining access to selected person 2 3 3 Locating selected person 2 3 3 Effects on questionnaire design Ability to handle: Long questionnaires 3 2 2 Complex questionnaires 3 2 1 Boring questions 3 2 1 Item non-response 3 3 1 Filter questions 3 3 2 Question sequence control 3 3 1 Open-ended questions 3 3 1 Quality of answers Minimise social desirability 1 2 3 Make question order random 1 3 1 Ability to minise distortion due to: Interviewer characteristics 1 2 3 Interviewer opinions 1 2 3 Influence of other people 2 3 1 Allows opportunities to consult 2 1 3 Avoids interviewer subversion 1 3 3 Implementing the survey Ease of obtaining suitable staff 1 2 3 Speed 1 3 2 Cost 1 2 3

(source: reproduced from De Vaus, 2002, table 8.1, p.132).

Almost all of the advantages of the face-to-face method also pertain to the telephone survey. Nevertheless, the telephone method has less ability to handle long and complex

148 questionnaires than the face-to-face method. The telephone survey reduces the bias due to the interviewer which is a weakness of face-to-face method. The telephone survey is easier to implement with less cost and faster than the face-to-face method. However, telephone surveys receive a lower response rate than the face-to-face survey. The disadvantages which are relevant to this research design are the difficulty to obtain a telephone number of subjects in the defined study areas and the requirement of well- trained staff to accomplish the survey. This method may lead to subject bias depending on the time of call. For example, during the day, there is a high probability that housewives or the unemployed will be reached. To minimise this problem, the time of call may be done during the evening or weekend, but it might be difficult to obtain cooperation from the subject who wants to rest and relax. Therefore, to receive a high response rate, several contacts are required if necessary.

The disadvantage of the mail survey is the lack of ability to control the questionnaire design. This method is not recommended for long and complex questionnaire. It also lacks ability to control for both the refusal bias and the influence by the other people. The mail survey is obviously controlled for interviewer bias. This method does not require either appointment or permission from the subject in filling in the questionnaire. Moreover, the mailing questionnaire is the cheapest method and does not require trained staff to implement, which is a great advantage. However, similar to the telephone survey, several follow-ups (in this case, follow-up letter) are required to receive a high response rate.

By carefully considering both advantages and disadvantages of all candidate methods, this thesis decided to employ self-administration by mail. This thesis also attempts to increase the response rate of the mail method by implementing two additional procedures – a follow-up letter; and a pick-up procedure. The proposed procedure, with its time diagram, is presented in Figure 4.2.

149 Cover Letter & First Follow-Up ~1 week ~3 weeks Questionnaire & & First Pick-Up Return Envelope

Second Follow-Up & ~1 week ~1 week Third Follow- Up Second Pick-Up Questionnaire &

Return Envelope Figure 4.2 Proposed Health Survey Procedures with Time Diagram

4.3.2.2 Envelope and Return-Envelope An envelope is a first thing received by a resident who may decide to open it or just throw it away as a junk mail. To limit the later case, this research employed a university pre-paid envelope because it is looks more formal and trustful. A reply-paid return envelop (as presented in Figure 4.3) was enclosed to encourage the respondent to return the completed questionnaire.

Figure 4.3: Reply-Paid Return Envelope.

4.3.2.3 Contact Letters and Pick-Up Procedures The procedures in designing the contact letters including the cover letter and the follow- up letters are based on the guidelines developed by Dillman (2000, pp.156-188). The contact letters deliver the information (such as the objective of survey, the name of

150 research team, or reminding the unresponsive subject to respond) from the administrator to the receivers. The content of each contact letter is different but with the same purpose: that is to increase the response rate. For the purpose of trustworthiness, all the contact letters are printed on a white A4 paper with the symbol of the University of New South Wales on the top-right-hand-side. The format of the contact letters follows the university standard.

Another attempt to increase the response rate is the implementation of the pick-up procedures. The objective of the pick-up procedure is to encourage the respondent, who might feel difficulty in returning the completed questionnaire by mail, to participate in the survey. For unresponsive subjects only, the administrator visits them: first, one week after the first contact letter; and, secondly one week after the second follow-up letter (see Figure 4.2). The administrator introduces himself/herself and informs the objective (just to pick up the completed questionnaire) of the visiting the resident. Actions, such as urging, pestering, or persuading the unwilling participants, are prohibited.

The cover letter (or the first contact letter) was designed to persuade the resident to participate in the survey. Figure 4.4 illustrates the structure and content of the cover letter. The cover letter is the first time the resident receives any information about this survey. Therefore, to make a first impression, it was decided to print in coloured ink. The first message of the cover letter requests the recipients to participate in the survey and explains how, and why, they were selected. The second paragraph persuades the recipients by highlighting the usefulness of the study to community. The third paragraph implies the recipients that this survey is voluntary. It also gives an instruction to the recipients about how to join the survey and how to return the questionnaire back to the administrator. The research team names and addresses are detailed in the forth paragraph. The confidentiality paragraph conveys an ethical commitment to the recipients that all their responses to the questionnaire will be strictly protected by the study and the university. The sixth paragraph provides contact details of the administrators to show the recipients that the research team are willing to answer any questions from them. The final paragraph thanks the recipients for their cooperation.

151 THE UNIVERSITY OF NEW SOUTH WALES

SCHOOL OF CIVIL AND ENVIRONMENTAL ENGINEERING

11 June 2003

RESIDENT Inside address 3299 BRIDGESTHARIT STREET KURNELL 2231 SAMUELS 1111 Dear Resident, S alutation

Environmental Health Survey

The request & We would like you to participate in an environmental health survey. Your home address was randomly how and why you selected from a computer. About 3,000 households have been invited to participate in our study. The were selected? study area covers suburbs in Sydney.

The purpose of our survey is to study the relationship between your health and the noise conditions in Usefulness of your neighbourhood. A better understanding of this relation is required so that we can recommend survey appropriate strategies for the mitigation of environmental noise pollution in Sydney.

Voluntary Please assist us by filling out the attached questionnaire. Anyone in your house can complete this How to do and questionnaire and return it back to us with the enclosed pre-paid envelope by mail. We are also happy to return the pick it up from this address 1 week after your receipt of this questionnaire. You can help us very much by questionnaire spending less than 15 minutes to complete all the questions.

This study is being conducted by researchers from UNSW: Professor John Black, Dr. Stephen Samuels Research team and Tharit Issarayangyun from the School of Civil & Environmental Engineering, and Associate Professor Deborah Black from the School of Public Health & Community Medicine.

This study is strictly anonymous and confidential. Except the above named researchers, no one else will Confidentiality have access to any aspects of information you provide to us in this survey and there is no identification of individual on the questionnaire.

If you have any problems filling out the questionnaire, please contact Ph: 0405158118 or Email: [email protected]. If you need further information about the purpose of this study, please Willingness to contact Ph: 0408625268 or Email: [email protected]. We would be happy to talk with you. answer questions

Thank you very much for helping with this important study. Thank you Sincerely,

Signature

John Black Professor School of Civil & Environmental Engineering UNSW

Figure 4.4: Proposed Cover Letter

The second contact letter (or the first follow-up letter) was designed as either an appreciation of cooperation or as a reminder letter for a late reply. Figure 4.5 illustrates the structure and content of the first follow-up letter. The first message of the first follow up letter states that a questionnaire was mailed to the respondent for what reasons. This message is considered important as for some respondents this is the first

152 time they learn that a questionnaire was sent to them. The second paragraph states the objective of the first follow up letter, thanks those who have responded for cooperation and reminds those who have not responded for late reply and thanks them for cooperation. The third paragraph is an invitation to call for a replacement questionnaire if it was misplaced the first time. It also implicitly invites the respondents who discard the original questionnaire to call for a replacement. And the final paragraph thanks the receipts for their cooperation.

THE UNIVERSITY OF NEW SOUTH WALES

SCHOOL OF CIVIL AND ENVIRONMENTAL ENGINEERING

18 June 2003

RESIDENT Inside address 1099 PRINCE THARIT CHARLES STREET PARADE KURNELL 2231 SAMUELS 1111 Dear Resident, S alutation

Environmental Health Survey

C onnection to Last week a questionnaire seeking the information about your health conditions and the noise conditions the previous mailing in your neighbourhood was mailed to you.

If you have already completed and returned the questionnaire to us, please accept our sincere thanks. If not, please do so today. We are especially grateful for your help. The successful of this study is very Remind & useful so that we can recommend appropriate strategies for the mitigation of environmental noise Thank you pollution in Sydney.

Invite to call for If it was misplaced, could you please call us at Ph: 0405158118 or Email: [email protected] replacement the and we will get another one in the mail to you today. questionnaire Sincerely,

Signature

John Black Professor School of Civil & Environmental Engineering UNSW

Figure 4.5 Proposed First Follow-Up Letter.

The third contact letter (or the second follow-up letter) was designed to elicit a response by encouraging unresponsive respondents to complete and return the questionnaire. Figure 4.6 illustrates the structure and content of the second follow-up letter. The second follow-up letter is considered more insistent in toning and wording. It starts with

153 a strong sentence to tell the receipts that their completed questionnaire has not yet been returned. The second paragraph informs the unresponsive respondents, as a means of encouraging them, that others have completed and returned the questionnaire. The next paragraph reemphasises the usefulness of study which implies that response from everyone is very important. The fourth paragraph reconfirms the confidentiality provided by the study to all participants. The fifth paragraph implies to the recipients that this study is fully voluntary. The next paragraph thanks the recipients for their cooperation. And, finally, the contact details of the administrators are provided to show the recipients that the survey staffs remain willing to answer any questions from them.

The fourth contact letter (or the third follow-up letter) represents the last effort of the administrator to elicit a response. Figure 4.7 illustrates the structure and content of the third follow-up letter. It contains more intensity than those previous follow-up letters, but with softer toning and wording. The intensive insistence occurs because this is the fourth request as stated by the first two paragraphs. The familiar sentences about the usefulness of the study, confidentiality and thank you messages are repeated again. The packaging of this last contact letter should be distinguished from the previous letters to stimulate the respondents to cooperation (Dillman, 2000).

154 THE UNIVERSITY OF NEW SOUTH WALES

SCHOOL OF CIVIL AND ENVIRONMENTAL ENGINEERING

3 July 2003

RESIDENT Ins ide addres s 27099 THARITPRINCE CHARLESSTREET PARADE KURNELL 2231 SAMUELS 1111 Dear Resident, S alutation

Environmental Health Survey

About three weeks ago we sent a questionnaire to you that asked about your health conditions Feedback: We've and the noise conditions in your neighbourhood. To the best of our knowledge, it has not yet not heard from you been returned. However, if you have already completed the questionnaire and are still waiting us to pick it up from your address, then please accept our sincere thanks for your help and forgive us for this letter.

The results of people who have already responded are providing us the better knowledge about the Others have relation between health and noise pollution conditions. responded We are writing again because of the importance that your questionnaire has for helping to get more Usefulness of accurate results. Although we have invited about 3,000 households to participate in our study, it’s only by your response hearing from nearly everyone in the sample that we can be sure that the results are truly representative.

Confidentiality This study is strictly anonymous and confidential. It is very important to us, as well as the university, to protect the confidentiality of people participating this survey.

We hope that you will fill out the attached questionnaire and return it to us soon. If for any reason Voluntary you prefer not to answer the question, please let us know by returning a note or blank questionnaire.

Sincerely,

Signature

John Black Professor School of Civil & Environmental Engineering UNSW

P.S. If you have any problems filling out the questionnaire, please contact Ph: 0405158118 or Email: Willingness to [email protected]. If you need further information about the purpose of this study, please answer questions contact Ph: 0408625268 or Email: [email protected]. We would be happy to talk with you.

Figure 4.6: Proposed Second Follow-Up Letter.

155 THE UNIVERSITY OF NEW SOUTH WALES

SCHOOL OF CIVIL AND ENVIRONMENTAL ENGINEERING 9 July 2003

RESIDENT Inside address 1599 TASMAN THARIT STSTREET KURNELL 2231 SAMUELS 1111 Dear Resident, S alutation

Environmental Health Survey

C onnection to During the last six weeks we have sent you several mailings about an important research study previous mailings & we are conducting on behave of the University of New South Wales. The purpose of our survey is to Usefullness of study study the relationship between your health conditions and the noise conditions in your neighbourhood.

The study is drawing to a close, and this is the last contact that will be made. Time is running out

We are sending this final contact because hearing from everyone in this small sample helps assure that Usefulness of the survey results are as accurate as possible. your response This study is strictly anonymous and confidential. It is very important to us, as well as the university, to Confidentiality protect the confidentiality of people participating this survey.

Finally, we hope that you will fill out the questionnaire and return it to us soon. We appreciate your willingness to consider our request as we conclude this effort to better understand the correlation Thank you between health and noise which is very useful for public.

Sincerely,

Signature

John Black Professor School of Civil & Environmental Engineering UNSW

P.S. If you have any problems filling out the questionnaire, please contact Ph: 0405158118 or Email: Willingness to [email protected]. If you need further information about the purpose of this study, please answer questions contact Ph: 0408625268 or Email: [email protected]. We would be happy to talk with you. Figure 4.7: Proposed Third Follow-Up Letter.

4.3.3 Ethical Consideration It is a rule that all investigations on behalf of the University of New South Wales (UNSW) that involve human participants must comply with the ethics policies at UNSW as stated in . The health survey proposal was approved by the UNSW Human Research Ethics Committee, Human Research Ethics Advisory, Panel H: Science/Engineering, as a project that has low ethical impact on people.

156 In a local community radio broadcast, the confidentiality statement and encouragement to participate in the survey were emphasised. During the period of the survey, the principle research supervisor, Professor John Black, had both been a guest on and a presenter of “The Green City” a weekly broadcast (Tuesday, 5.30 – 6.00pm) on Eastside Community Radio.

4.4 PILOT STUDY

4.4.1 Objective and Method Health survey administrators test the performance of their proposed survey procedures, and identify any possible mistake(s), or unpractical procedure(s), before the commencement of the main survey. The pilot test simulates all the proposed survey procedures but with much smaller sample size than the main one. A pilot study is required as it can increase both reliability and validity of the survey procedure. A small residential suburb (the suburb of Kurnell), located approximately five kilometres to the south of Sydney Airport has been selected as a case study for the pilot test with a sample size of one hundred households. The objectives of this pilot study are to test the performance of the proposed health survey procedures, and to test the internal reliability of both noise stress scale and noise sensitivity scale.

All addresses in Kurnell (excluding apartments, commercial buildings, addresses for sale or lease, and abandoned addresses) were considered as the study population. The addresses located near the Caltex Oil Refinery were excluded to eliminate the effects of major non-aviation noise (in this case, industrial noise). The total number of the study population was 604. The sample size of one hundred was randomly selected from this study population. The first contacts were mailed on 29 May 2003 and the last return was received on 15 July 2003.

4.4.2 Performance of the Proposed Health Survey Procedures Table 4.5 summarises the number of received questionnaires by types of procedure and types of return. The “response” column means the number of subjects who have filled in the questionnaire and returned it (called the responsive respondent). The “refuse”

157 column means the number of subjects who have returned a blank questionnaire. Subjects who refused to participate in the survey during the pick up procedure have not been included into this category. The “unidentifiable” column means the subjects who have returned the questionnaire but the identification number of each subject on the return envelope has been masked or deleted by the respondent. For the main study, this group of respondents is invalid because the lack of ability to identify the group of study area (whether from aircraft noise exposure group or aircraft noise non-exposure group).

Table 4.5: Return Rate of the Proposed Health Survey Procedures Procedures Day Response Refuse Unidentifiable First Contact Letter 29/5/2003 20 0 0 First Pick-Up 9/6/2003 3 0 0 First Follow-Up Letter 5/6/2003 18 3 1 Second Follow-Up Letter 25/6/2003 14 3 1 Second Pick-Up 3/7/2003 3 0 0 Third Follow-Up Letter 9/7/2003 4 0 1 Total Return (subjects) 62 6 3 Total Return (%) 71% Total Not Return (%) 29%

By comparing the return rate between three different methods (which are the first contact letter only, the pick-up procedure, and the follow-up letters), it is obvious that the follow-up letters provided almost twice the return rate (38%) from the first contact letter only (20%). The return rate of the pick-up procedure was only six percent. It was found that 5 of the 6 picked up questionnaires have been already completed done and ready to be mailed back to the administrator, regardless of the pick-up procedure. Moreover, it was also found that 51% of the 84 first-pick-up subjects and 45% of 42 second-pick-up subjects were identified as “Absent”. It might reflect the fact that there was only one visit made by the survey administrator to the target address. All of the visiting was implemented during the daytime. Thus, the percentage of absent households for both pick-up procedures was very high.

The analysis of the valid responses was further undertaken for only the responsive respondents. The criteria have been set by this research design that a responsive subject would be considered as an invalid sample if:

158 (1) The missing item of the Physical Functioning scale (10 items), the Mental Health scale (5 items), the Vitality scale (4 items), the General Health scale (5 items), the noise stress scale (4 items), and the noise sensitive scale (10 items) is higher than a half of total items of each scale; or

(2) One of the following question (which are aircraft noise annoyance opinion scale (1 item), sex (1 item), age (1 item), or hypertension condition (1 item)) has not been answered completely.

Table 4.6 summarises the result of this preliminary analysis. It was found that one respondent ignored all the items of General Health scale, one respondent ignored the question of hypertension condition, and two respondents ignored either the question on sex or age. Therefore, the valid response rate of the proposed health survey procedures was found to be 58 percent.

Table 4.6: Number of Subject Classified by Completed and Missed Items. Physical Functioning subject General Health subject Aircraft Noise Annoyance subject 10 items completed 59 5 items completed 59 Opinion Scale ≤ 5 items missed 3 ≤ 3 items missed 2 1 item completed 62 > 5 items missed 0 > 3 items missed 1 1 item missed 0

Mental Health subject Noise Stress subject Sex&Age subject 5 items completed 62 4 items completed 61 2 items completed 60 ≤ 3 items missed 0 ≤ 2 items missed 11 item missed 2 > 3 items missed 0 > 2 items missed 0

Vitality subject Noise Sensitivity subject Hypertension Condition subject 4 items completed 60 10 items completed 58 1 item completed 61 ≤ 2 items missed 2 ≤ 5 items missed 41 item missed 1 > 2 items missed 0 > 5 items missed 0

4.4.3 Reliability Test of the Proposed Scales Reliability refers to the ability of an instrument to produce consistent results whenever it is repeated in whatever groups of subject, even when conducted by other researchers (Sarantakos, 1998). There are two types of reliability: internal reliability; and external reliability. The internal reliability means consistency of results within the scale, while the external reliability refers to consistency of data across the scales. A scale will be considered as internally consistent if the items in the scale are moderately correlated

159 with each other and each item should correlate with the total scale score (Streinder and Norman, 1995). It means that all of the items should tap different aspects of the same attribute and the summation of items leads to the final result of the attribute.

This research is concerned with only the internal consistency of the proposed scale (i.e., the noise stress scale and the noise sensitivity scale). Both proposed scales were checked for internal consistency by the Cronbach Alpha method1. By using a suitable statistics program (SPSS), the results are presented by Table 4.7 and 4.8 for the noise stress scale and the noise sensitivity scale, respectively. Both tables consist of two sections: Correlation Matrix; and Item-Total Statistics. The correlation matrix provides the correlation value between the items in the scale. The item-total statistic provides the value of Alpha if one item is deleted.

Table 4.7: Reliability Analysis (Alpha) of Proposed Noise Stress Scale

______Correlation Matrix STRESS_A STRESS_B STRESS_C STRESS_D STRESS_A 1.0000 STRESS_B 1.0000 .8231 STRESS_C 1.0000 .8025 .7716 STRESS_D 1.0000 .8262 .6936 .7216

Number of Cases = 63.0

Item-total Statistic Alpha if Item Deleted STRESS_A .9096 STRESS_B .9107 STRESS_C .8978 STRESS_D .9222

Reliability Coefficients 4 items: Alpha = 0.9311 ______

1 Streiner and Norman (1995, pp. 64-65) clearly explain the concept and formula of this method.

160 Table 4.8: Reliability Analysis (Alpha) of Proposed Noise Sensitivity Scale ______Correlation Matrix SEN_A SEN_B SEN_C SEN_D SEN_E SEN_F SEN_G SEN_H SEN_I SEN_J SEN_A 1.000 SEN_B .3702 1.000 SEN_C .4790 .5679 1.000 SEN_D .2261 .1920 .3842 1.000 SEN_E .5333 .3805 .4883 .4436 1.000 SEN_F .4927 .4368 .4941 .3547 .7359 1.000 SEN_G .3771 .2222 .6138 .3870 .5156 .5894 1.000 SEN_H .3633 .2869 .5177 .4398 .4232 .6191 .6729 1.000 SEN_I .5634 .2971 .5014 .2933 .5430 .6571 .6361 .6477 1.000 SEN_J .4151 .4204 .6771 .4786 .5303 .6385 .6897 .6792 .6545 1.000

Number of Cases = 60.0

Item-total Statistic Alpha if Item Deleted Alpha if Item Deleted SEN_A .8993 SEN_F .8869 SEN_B .9067 SEN_G .8899 SEN_C .8895 SEN_H .8904 SEN_D .9047 SNE_I .8891 SEN_E .8911 SEN_J .8844

Reliability Coefficients 10 items Alpha = 0.9030 ______

The correlation value between items in the proposed noise stress scale is relatively high and the Alpha value is also too high. These imply that some items are unnecessary (called item redundancy). The items of noise sensitivity scale moderately correlate with each other. However, the Alpha value is slightly high. Streinder and Norman (1995) recommended that the Alpha value should be above 0.70, but probably not higher than 0.90. Too low Alpha value reflects the inconsistency of scale, while too high Alpha value reflects a high level of item redundancy.

4.4.4 Recommendations for the Main Survey The overall performance of the proposed health survey instruments was found to be satisfactory. No major change was required. Nevertheless, the pilot study revealed some modifications to the proposed instrument for the main survey. The pilot study recommended the withdrawal of the pick-up procedure because it was time-consuming and ineffective in improving the response rate. The instructions of how to return the completed questionnaire by the pick-up procedure, therefore, were deleted from the

161 questionnaire, the cover letter, and the second follow-up letter. Figure 4.8 illustrates the modified health survey procedures with the time diagram for the main survey. In accordance with the withdrawal of pick-up procedures, the research design adjusted the valid response from 58 percent to 52 percent. The expected response rate for the main survey would be 52%.

Second follow- Cover letter up letter & First & Third Questionnaire ~1 week ~2 week ~1 week follow-up Questionnaire follow-up & letter & letter Return envelope Return envelope

Figure 4.8: Health Survey Procedures with Time Diagram.

The response rate from the third contact letter was low. Even though it might reflect the fact that the unresponsive subjects from the previous follow-up letters were unwilling to participate in the survey, the pilot study recommended the usage of a distinguishing package (in colour printing) from the previous ones to stimulate the response rate.

To maintain the alpha value of the noise stress scale and the noise sensitivity scale in the range of 0.7 – 0.9, the pilot study recommended the discarding of some items in both scales. For the noise stress scale, it was decided to ignore two items (which are STRESS_C and STRESS_D) because the high correlation between items in the scale and the high value of Alpha. The re-analysis of the internal reliability revealed that the correlation between STRESS_A and STRESS_B is 0.83 and the Alpha value becomes 0.90. For the noise sensitivity scale, although there are many items in the scale that can be omitted which leads the Alpha value to fall into a suitable range, it was decided to ignore the SEN_J item because its correlation value is relatively high compared with the other items. Thus, the new Alpha value of the noise sensitivity scale becomes 0.88. Appendix A presents the questionnaire and the contact letters for the main health and well-being survey.

162 4.5 STUDY POPULATION

The Sydney (Kingsford Smith) Airport has been selected as a case study. Sydney Airport consists of three main runways: two parallel runways (16R/34L and 16L/34R) and one cross runway (07/25), with average monthly movements of aircraft (excluding helicopters) of 23,000 in year 2004 (Airservices Australia, 2004b). Sydney Airport is considered as the principal gateway and hub for most domestic and international business and leisure visitors to the Sydney Region.

To select a study population, an ideal epidemiological concept is to include every subject exposed by the targeted exposure (which is, in the context of this thesis, aircraft noise) and compares them with a control area (where subjects are not exposed by the targeted exposure). However, the selection of the former case is impractical because it is considered. time- and budget-consuming, especially at Sydney Airport where there are ten runway modes of operation for the main purpose of aircraft noise equally distribution (called Long Term Operating Plan, LTOP) (see section 3.4.6).

The assignment of a runway mode of operation is dependent on the decision of air traffic controllers, and also on the weather conditions. Figure 4.9 provides an impressive example of how the flight paths of arriving aircraft into the Sydney Airport were assigned differently during two periods of time. Figure 4.9 also gives a guideline of how to locate a control group where noise from jet overflights should not be observable. In Figure 4.9, these tracks have been coloured according to the aircraft altitude (ie. Height above the mean sea level): a) red when less than 1000ft; b) orange between 1000ft and 3000ft; c) yellow between 3000 ft and 5000ft; and d) green above 5000ft.

163

Figure 4.9: Comparison of Sydney Airport’s Track Plots Coloured by Height for Jet Arrivals during the Period 2/12/2003 – 8/12/2003 (left) and the Period 2/6/2004 – 8/6/2004 (right) (source: reproduced from Airservices Australia, 2004c, figure 5, p.4 and Airservices Australia, 2004d, figure 5 p.13).

4.5.1 Aircraft Noise Exposure Area The areas exposed to aircraft noise from Sydney Airport are widespread around the Sydney Region. Therefore, only the highly exposed areas where the average annual day of N70 (the number of aircraft noise events that louder than 70 dB(A)) is higher than 50 events per day were selected as the study population for the aircraft noise exposure area. The 2003 daily average N70 contour map around Sydney Airport has been obtained from Airservices Australia, Canberra. The study population for the aircraft noise exposure area has been defined as shown in Figure 4.10. The area includes suburbs of Tempe, Sydenham, St Peters, Marrickville, Stanmore, Petersham, and Mascot.

164

Figure 4.10: Study Population for the Aircraft Noise Exposure Area.

Figure 4.11 is the Statistical Local Area (SLA) Map showing the study population of aircraft noise exposure. The SLAs have been developed by the Australian Bureau of Statistics (ABS) as a general purpose spatial unit. The detailed discussion of SLAs can be found in Edwards (2001, pp.9-13). In general, the SLA map provides the basic information of main road name, roads, unsealed roads, highways, railway lines, bridges, parks or reserves, Census Collection District (CD) boundaries and the number (for example, 1421402, 1421404, and 1421406), and Local Government Area (LGA) boundaries and names (for example, Marrickville, Botany Bay).

165

Figure 4.11: SLA Map of Study Population of Aircraft Noise Exposure Area

The Census Collection District (CD) is the smallest spatial unit used for the collection and dissemination of Population Census data in census years (Edwards, 2001, pp.7-9). In non-census years, CDs are undefined. The aggregation of CDs forms the larger spatial units of SLAs. For the 2001 Census of Population and Housing, there were

166 37,209 CDs defined throughout Australia. In urban areas, an average number of dwellings in each CD is about 220. Socio-economic indices for each CD in the study population of aircraft noise exposure area were used to determine the study population of control areas.

Given that the objective of this research is to study the impacts of aircraft noise on community health and well-being, the main criterion to identify exposure to aircraft noise was based on geographical location (exposed to aircraft and non-exposed to aircraft noise). It could reasonably be argued that health could be affected by socio- economic factors, such as employment status and education level. Highly educated people with white-collar employment may have greater anxiety (which could raise blood pressure) and consequent sensitivity to noise and other nuisances, than people with lower education and blue-collar employment. However, due to the nature of the descriptive epidemiological study employed by the present study, the socio-economic factors were not “selected”. Rather they were “observed” during the course of the study (see Section 4.3.1.6) and thus could possibly represent a “confounding factor” for the research.

Activity patterns of subjects either outside or inside of the household were not studied because the point of the receptor of aircraft noise was defined as the dwelling unit as with studies dealing with community aircraft noise. Therefore it leads to an assumption that subjects from the aircraft noise exposure area, which is defined based on N70 contour map, are similarly exposed by aircraft noise at a high level, and subjects from the control area are not exposed at all by aircraft noise.

4.5.2 Control Area In this thesis, the control area means the area where the socio-economic status is comparable with the exposure area and controlled for the aircraft noise exposure. Therefore, the study population of aircraft noise non-exposure area was selected based on two main criteria. First, this area should be free from jet aircraft noise. Second, the socioeconomic status of the areas subject to aircraft noise exposure and not subject to aircraft noise non-exposure should be similar.

167 The areas located outside of the flight paths were selected by visual inspection from the track plots similar to those in Figure 4.9, but for jet arrivals to and departures from Sydney Airport. In selecting these areas, there was an awareness of the flight paths from the other airports (such as Bankstown Airport, Camden Airport, and Hoxton Park Airport) located in the Sydney Region. Table 4.9 tabulates the options (with CDs code) for the aircraft noise non-exposure area. The next step is to compare the socio-economic status of each option with the study population of the aircraft noise exposure area. The option that has the most similar socio-economic status with the aircraft noise exposure area was selected as the control area.

The Australian Bureau of Statistics has developed the indices (called Socio-Economic Indices for Areas, SEIFA) derived from the 2001 Census of Population and Housing Data to measure different aspects of socio-economic condition by geographic area. The 2001 SEIFA consists of four indices: 1) Index of Relative Socio-Economic Disadvantage; 2) Index of Relative Socio-Economic Advantage/Disadvantage; 3) Index of Economic Resources; and 4) Index of Education and Occupation. All the indices have been constructed so that relatively disadvantaged areas have low index values.

Table 4.9: Options of Aircraft Noise Non-Exposure Areas with CD Codes.

Option AOption B Option C Option D Ashcroft Heckenberg Sadleir South Penrith Dharruk Hebersham Blackett Colyton Oxley Park 1290408 1290401 1290404 1280702 1270602 1270504 1270506 1281404 1281402 1290409 1290402 1290405 1280703 1270607 1270505 1270507 1281405 1281403 1290410 1290403 1290411 1280704 1270608 1270510 1270508 1281406 1281411 1290601 1290406 1290501 1280706 1270610 1270511 1270509 1281407 1281412 1290502 1280707 1270603 1281409 1280708 1270604 1281501 1280709 1270605 1281601 1280801 1270606 1281602 1280804 1270609 1281603 1280805 1281611 1280806 1281612 1282101 1282102 1282103 Census Collection District CODE 1282104 1282107 1282108 1282110

The Index of Relative Socio-Economic Disadvantage is derived to reflect disadvantaged areas. High scores of this index occur when the area has few families of low income and

168 few people with little training and unskilled occupations. On the other hand, low scores of this index occur when the area has many low income families and people with little training and unskilled occupations (Trewin, 2001). The Index of Relative Socio- Economic Advantage/Disadvantage is derived to reflect both advantaged and disadvantaged areas. A high score of this index indicates that an area has a high proportion of people with high incomes or are in skilled jobs, and a low proportion of people with low incomes and unskilled jobs. Conversely, a low score on this index indicates that an area has a high proportion of people with low incomes, more employees in unskilled occupations, and a low proportion of people with high incomes and in skilled occupations (Trewin, 2001).

The Index of Economic Resources is designed to reflect the profile of the economic resources of families within the areas. A high score on this index indicates that the area has a high proportion of families on high income, a low proportion of low income families, and high proportion of households living in large houses (four or more bedrooms). On the other hand, low score of this index implies that the area has high proportion of families on low income, low proportion of high income families, and high proportion of households living in small house (Trewin, 2001).

The Index of Education and Occupation is designed to reflect the educational and occupational structure of communities. An area with a high score on this index will have a high proportion of people with high educational qualifications, or undergoing further education and a high proportion of people employed in skilled occupations. Conversely, a low score on this index indicates an area with a high proportion of people with low educational qualifications and a high proportion of people employed in unskilled occupations or unemployed (Trewin, 2001). A detailed discussion and derivation of each index can be found in Trewin (2001).

The SEIFA indices have been standardised to have a mean of 1,000 and a standard deviation of 100 across all CDs in Australia. Trenwin (2001) emphasised that “the SEIFA indices are ordinal measures and not interval measures. Any arithmetic relationships between index values may not be meaningful”. For example, a CD with an index value of Socio-Economic Disadvantages of 1,100 does not have twice the

169 disadvantage of a CD with an index value of 550. Similarly, the difference between two CDs with index values of Socio-Economic Advantage/Disadvantage of 900 and 1,050 is not necessarily the same as the difference between two CDs with this index’s values of 800 and 950. Therefore, the comparison of SEIFA indices between two groups can be done by the Non-Parametric test. This thesis performed SPSS (version 12) program to complete this task. Table 4.10 summarises the test results.

Table 4.10: Mann-Whitney Test of the Comparison of 2001 SEIFA Indices between the Aircraft Noise Exposure Area and the Options for the Control Area.

Number Advantage and Economic and Education and Disadvantages Mann-Whitney of Disadvantage Resource Occupation Test CD Z Sig.(2-tailed) Z Sig.(2-tailed) Z Sig.(2-tailed) Z Sig.(2-tailed) Exposure Group 30 Option A 13 -5.157 0.000 -5.157 0.000 -5.157 0.000 -5.157 0.000 Option B 18 -0.809 0.418 -3.024 0.002 -2.641 0.008 -4.685 0.000 Option C 17 -4.627 0.000 -5.491 0.000 -5.535 0.000 -5.602 0.000 Option D 15 -3.275 0.001 -5.128 0.000 -5.153 0.000 -5.345 0.000

From Table 4.10, it appears that all 2001 SEIFA indices of Option A, C, and D are statistically different from the aircraft noise exposure area. Even though three aspects of socio-economic measure of Option B are statistically different from the noise exposure area, it is evident that the Index of Relative Socio-Economic Disadvantages of Option B is not statistically different from the noise exposure area. By recognizing that the selection of an ideal control area is sometimes impractical, Option B (suburb of South Penrith) was chosen as a control group. South Penrith is located in the western suburbs of Sydney (approximately 55 km from Sydney Airport), as identified by a ‘star’ symbol in Figure 4.12.

170

Figure 4.12: Locations in Sydney of Aircraft Noise Exposure Area and the Control Group

4.6 SAMPLE SIZE

To estimate the sample size required for the main survey, two factors were considered: the expected response rate; and the statistical power of the health measures. Statistical power refers to the ability of a test statistic to detect a true difference between two or more group. Well-designed test statistics and large sample size usually increase statistical power.

Although the SF-36’s developer recommended sample sizes needed to detect point differences of various study designs, the appropriate design for this research study to detect differences between two quasi-experimental group means was not provided. Two designed studies provided by the SF-36’s developer, that are the most relevant, were considered: (1) study designed to detect differences between two non-experimental

171 groups (repeated measures design); and (2) study designed to detect differences between a group mean and a fixed norm. Based on Ware et al (1993, table 7.6 and 7.8), the former study designed requires a larger sample size than the latter one. Therefore, this research conservatively applied the first design (see Table 4.11). The SF-36’s developer recommended that the smallest point difference (2-point) is suitable for a study designed to measure health outcomes on a population basis. However, a 5-point difference was used for this research as a practical number of point difference to estimate the required sample size, because of research budget constraints. The 5-point difference is appropriate for a study designed to define differences that are clinically and socially relevant.

Table 4.11: Sample Size Needed per Group to Detect Point Difference between Two Non-Experimental Groups, Repeated Measure Design Number of Points Difference Scale 2 5 10 20 Physical Functioning PF 1705 274 69 18 General Health GH 1308 210 53 14 Vitality VT 1385 222 56 15 Mental Health MH 1030 165 42 11 Note: Estimates assume alpha = 0.05, two-tailed t-test, power = 80% and an intertemporal correlation between scores of 0.06 (source: Ware et al, 1993, table 7.6, p. 7.11).

In accordance with the pilot study, the expected valid-response-rate of the health survey procedure (see Figure 4.8) was 52%. Consequently, the required sample sizes per group for the main survey would be 274 (estimated sample size needed to detect 5-point difference) divided by 52% (expected response rate), which yields 527 subjects. Since there are two study groups, the total sample size required for the main survey is therefore 1,054 subjects. Nevertheless, this research improved the statistical power by increasing the sample size for each group to 750 subjects. The total sample size becomes 1,500 subjects.

The study sample (750 subjects per group) was randomly chosen by a computer to ensure an equal chance of selection. Every home address (excluding apartments, commercial buildings, addresses for sale or lease, and abandoned addresses) located in

172 the local traffic area (where the major effects from non-aviation noise sources, such as, industrial noise, railway noise, highway noise, are very low or negligible) in the study population of both aircraft noise exposure area and a control area were observed by the author to minimise the address error. Apartment dwellers were excluded because the acoustical characteristics (for example, layout of room, structure components, and open spaces for activities) of houses and apartments are quite different. One dweller living in one side of apartment facing on to street will be exposed to higher levels of traffic noise than another dweller living in another side of the same apartment. Dwellers who reside on the top floor of apartment may experience louder aircraft noise level than dwellers from the lower floors due to the differences in the structural components of the building, such as the ceiling. On the other hand, residents in a house with a rear courtyard will have more chance of being disturbed by overflight noise while undertaking outdoor activities such as gardening and attending a BBQ than residents in an apartment where open space for outdoor activities is limited.

4.7 TRANSLATION

Streiner and Norman (1996) stated that when undertaking a questionnaire survey in a major metropolitan area, where English is not the first language of a significant proportion of respondents there arises one of two possible alternatives for the questionnaire developer. First, the developer ignores the non-English speaking respondents from the study. However, this raises the possibility that the study sample is not fully representative of the true population (called the ethical bias). The second alternative, which is considered time and cost consuming, is to translate the questionnaire instruments into the languages most commonly spoken within the targeted study area.

Streiner and Norman (1996) also recommend that once the first translation is done, it is necessary that other bilingual persons, who are not associated with the first translation phase and have preferably knowledge about the topic, translates the new version back into English (‘back translation’ process). If the meaning seems to have been lost or

173 changed from the original, then that item should be tracked back through the translation process.

Sydney is a major metropolitan city where approximately, in the 2001 Census of Population and Housing, 33.5% of people aged 5 years or older spoke non-English languages at home. The three most common languages spoken at home, other than English were: Chinese languages (4.9% or 194,603 people); Arabic (including Lebanese) (3.6% or 142,453 people); and Greek (2.1% or 83,915 people) (Trewin, 2002). This study, therefore, aims to translate the questionnaire and the contact letters from English into the others languages to overcome the ethical bias as previously mentioned.

The basic community profiles of every CDs in the study population of both aircraft noise exposure and a control area were downloaded from the official website of 2001 Census and Population and Housing, Australia (http://www.abs.gov.au) as shown in Table 4.12. From the table, overall the three most non-English languages spoken at home are Greek, Arabic (including Macedonian and Lebanese), and Vietnamese, respectively. Since the translation process is complex and costly, this thesis decided to translate the health survey instruments (which are the questionnaire and the contact letters) from English into two other languages most spoken at home which are Greek and Arabic. The expected percentage of non-English subjects that this translation can serve is seven from the total study population.

Translation process itself is complex and requires person who is not only fluent in English and the target languages, but who is also knowledgeable about the content area and is aware of the intent of each item and of the scale of questionnaire as a whole. Even though the study recognises the importance of the ‘back translation’ process, it was impractical to implement because of two main reasons: 1) the budget constraint; and 2) the clinical issue. The health survey instruments proposed by this thesis are the subjective health assessment tool. The possible mistakes from the translation process would not cause any negative effects to the survey participants. This research employed the Multicultural Unit, Prince of Wales Hospital (POWH), South Eastern Sydney Area Health Service (SESAHS) to accomplish all the translation processes. Appendix B

174 presents the translated questionnaire and contact letters for the main health and well- being survey.

Table 4.12: Three Most Non-English Languages Spoken at Home of Selected Study Population Areas

English Only Other Language Totoal (number of people) (number of people) (percent) (percent) People First Second Third Noise Greek Vietnamese Arabic Exposure 14,333 8,016 1,001 747 625 Group 55.9% 7.0% 5.2% 4.4% Control Arabic Greek Italian Group 12,049 10,587 129 86 82 87.9% 1.1% 0.7% 0.7% (source: 2001 Census of Population and Housing, ABS, , accessed November 2004).

In attempt to ensure fair representation of Greek and Arabic language speaking groups, this thesis has considered the possibility to adopt a quota based sample technique in the main health survey. A quota based sample is a type of non-probability sample that includes specified numbers of respondents based on percentages of specific characteristics (in this case, race) that have been predetermined as representative of the target population. To achieve a research criterion that every sample in the study population must have an equal chance to be selected (see Section 4.5) and the nature of postal self-administrative technique (see Section 4.3.2), adopting quota based sample may be done by identifying mailing addresses of all Greek and Arabic speaking resident in the study population by purchasing data from Australian Bureau of Statistics. It can be argued that the study might just ignore the research criteria (which has been setup to prevent selection bias on result) by directly administer resident who speaks Greek or Arabic. This involves cost, for example, to motivate people to participate the survey, to train administration staff (who can speak Greek or Arabic), and to advertise what this survey is doing. It is considered unfeasible for this research.

175 4.8 CONCLUSIONS AND DISCUSSIONS

To explore the two core research questions, this research has proposed a population- based questionnaire survey as an investigation tool. A cross-sectional study was employed as it was considered the best fit to the requirements of this thesis (see Section 4.2). The proposed health survey instruments for the evaluation of the effects of aircraft noise on health related quality of life and the association between aircraft noise and prevalence of hypertension have been carefully developed and piloted in this thesis. The questionnaire measures seven major characteristics of each subject: 1) health related quality of life; 2) hypertension condition; 3) noise stress; 4) noise sensitivity; 5) noise annoyance; 6) confounding factors; and 7) demographic characteristics. No medical laboratories or experimental tests on people have been undertaken in this research.

A postal self-administrative technique to administer the questionnaire is proposed. The cover letter and the follow-up letters were developed to increase the response rate. The proposed instruments have been approved by the UNSW Human Research Ethics Committee as a tool that has low ethical impact on people. The questionnaire will be randomly distributed through the well-defined study population areas that will comprise areas of high aircraft noise exposure (N70 ≥ 50 events per day) and a control area (non- exposure). The total required sample size, which was designed to satisfy the statistical power of the SF-36, is 1,500 subjects. The health survey instruments were translated from English into to Greek and Arabic to minimise the ethical bias from non-English speaking subjects.

The exposure to aircraft noise outside of the house (for example, at workplace or at school) was not considered by this present study. Thus, the effects of exposure outside of the house may cause bias to the result. To minimise this problem at the beginning of questions assessing emotional stress and annoyance due to noise pollution, the respondent was instructed to answer these questions by considering a situation when they are here at their house.

Previous studies, such as Kennedy et al (1998), Robert (1998) and Simon et al (2000), suggested that individuals who have low socio-economic status tend to have poor health

176 outcomes. City person with low socio-economic status tend to live in lower-valued houses which are normally located in noisy areas, such as near highways, train stations, airports, and industries. When comparing health outcomes of resident between high and low noise area, socio-economic status may overestimate the effects of noise. To prevent bias from socio-economic status, the screening process (which was considered an advantage of the present study) was implemented at two different levels. At area level, this research employed Socio-Economic Indices for Areas derived by Australian Bureau of Statistics in selecting the control area where socio-economic characteristics of area was similar with the aircraft noise exposure area. At individual level, the questionnaire measured individual’s socio-economic information of each subject. Data from the individual level was used as input parameter of confounding factor in the multivariate analysis.

Given this thesis hypothesised that community health is detrimentally impacted by long- term aircraft noise exposure, questions measuring noise stress and noise annoyance have been developed with 12-month recall period. Prevalence of hypertension (based on historical data) was measured by two consecutive questions (“Have you ever been told by a doctor or nurse that you have high blood pressure sometimes called hypertension? If Yes, do you currently have high blood pressure?”). Unfortunately, most of the generic health measures to assess health related quality of life have short recall period to prevent recall bias. This thesis has adopted some scales from SF-36, which was considered the best scale for the present study. The SF-36 is only available for Standard measurement with 4-week recall period (which was used by this present study) and Acute measurement with one-week recall period. Therefore, it was inevitable to include possible effects to health related quality of life from other recent incidents that were not related to long-term aircraft noise exposure.

One of the advantages of this research is an attention to control many types of bias. Bias from interviewer was completely eliminated in this research due to the implementation of self-response survey. Selection bias may result if the survey was limited study sample within a certain demographic groups. This research has controlled this bias by randomly select every households in the study population to ensure that each sample had an equal possibility to be selected. In some case, respondent may answer the

177 questions in either negative or positive way if they have attitude toward the thing that is a purpose of the survey (called attitude bias). To prevent attitude bias, the respondents were informed by the contact letters that the objective of this present survey was to study the impacts of environmental noise on community health. The health survey instruments have been design not to inform either intentionally or unintentionally the real objective of the study. The response bias is inevitable with self-response survey, such as postal survey and internet survey. This research minimised the response bias by design and implement pilot test to warrant that the questionnaire was easy to understand, precise, and relatively short. The well-established questionnaires (which are both reliable and validated) to measure health related quality of life, noise annoyance, noise sensitivity, and demographic characteristics were adopted. Recall bias has been minimised by using less than four week recall period on most of questions in the questionnaire. Nevertheless, as previously mentioned by the above paragraph, the need of one year recall period in questions assessing noise stress and noise sensitivity to establish long-term effects from aircraft noise exposure may cause recall bias in the result.

The next chapter will discuss in detail the development of the ‘new’ noise index that describes and assesses aircraft noise in such a way that it is easily understood by a layperson and quantifies relevant aspects of the potential impacts of aircraft noise on community health and well-being. The index is termed the Noise Gap Index (NGI).

178 CHAPTER FIVE

DEVELOPMENT OF NOISE GAP INDEX (NGI)

5.1 INTRODUCTION

Chapter 3 discussed the procedures of aircraft noise measurement and aircraft noise management of airports in Australia. The current practice of aircraft noise management plans of Australian airports relies on a conventional aircraft noise index (called Australian Noise Exposure Forecast (ANEF). Given that ANEF provides insufficient information to people on how to judge, on a rational basis, the effects of aircraft noise (SSC, 1995), Australian Department of Transport and Regional Services (DoTARS, 2002) proposed a more transparent (or understandable) approach to describing and assessing aircraft noise around Australian airports. The transparent aircraft noise concept is based on ‘everyday talk’ information, such as where the aircraft fly, how often, and at what time. A potential example of the transparent aircraft noise concept is the N70. DoTARS (2002) have declared that the N70 gives a much more realistic picture of aircraft noise to the community than the conventional indices.

To encourage the promotion of transparency in aircraft noise concepts, this chapter expands on ways to describe and assess aircraft noise by incorporating the characteristics of background environmental noise into the N70. This thesis has assumed that people living in areas of different background noise may have different reactions to the same aircraft noise level. The ‘new’ noise index, which has been termed the Noise Gap Index (NGI), distinguishes between aircraft noise and background environmental noise in a novel manner. Section 5.2 describes the selection of the locations of the noise investigations. Section 5.3 describes the noise measurement procedures, and the fundamental concepts of noise indices involved in the further calculation. Section 5.4 describes the typical noise data. Section 5.5 explains the detailed analysis of these noise data. Section 5.6 presents the formula of the NGI.

179 Section 5.7 applies the NGI into a case study, and presents the relationship between NGI and aircraft noise annoyance scale. Section 5.8 concludes the whole description.

5.2 NOISE INVESTIGATION

The noise investigation component of this thesis was centred around Sydney (Kingsford Smith) Airport. It involved determining, in a variety of residential areas, both typical ambient noise levels along with the noise levels associated with aircraft over flights. For these noise data to suit the requirements of this research, the effects of non-aviation noise sources had to be minimised. Consequently, noise stations were set up in randomly selected households located in what could be termed local traffic areas.

Households located close to major non-aviation noise sources, such as railway lines, industrial areas, major roads, or highways, were excluded. The selected noise stations were mostly located inside the study population of aircraft noise exposure area which is positioned underneath the main north flight paths of Sydney Airport. Nevertheless, for a variety of noise data, some noise stations were allocated at other locations around Sydney Airport. Three noise stations were also allocated in the study population of aircraft noise non-exposure area for the purpose of background environmental noise comparison. Suburbs included in this noise study were Tempe, Sydenham, St Peters, Marrickville, Stanmore, Petersham, Newtown, Banksia, Mascot, Kurnell, Rosebery, and South Penrith with overall 29 selected noise stations. Table 5.1 summarises the address and the commencement date of the noise survey at each noise station.

180 Table 5.1: Noise Stations

Noise Station Street Suburb Date 1 26 Robinson St Eastlakes 12/10/2003 2 35 Dougherty St Roseberry 12/10/2003 3 43 Frederick St St Peters 23/10/2003 4 70 Terry St Tempe 23/10/2003 5 13 Bruce St Stanmore 24/10/2003 6 64 Westbourne St Petersham 24/10/2003 7 56 Silver St Marrickville 27/10/2003 8 81 Meeks Rd Marrickville 27/10/2003 9 31 Tabrett St Banksia 3/11/2003 10 35 Lywen St Banksia 3/11/2003 11 33 Wellington St Rosebery 5/11/2003 12 37 Alfred St Mascot 5/11/2003 13 16 Horton St Marrickville 11/11/2003 14 23 Perry St Marrickville 11/11/2003 15 205 Australia St Newtown 14/11/2003 16 89 Cavendish St Stanmore 14/11/2003 17 15 Foreman St Tempe 20/11/2003 18 37 Grove St St Peters 20/11/2003 19 10 North St Marrickville 21/10/2004 20 8 Bright St Marrickville 21/10/2004 21 18 Banool Av South Penrith 26/10/2004 22 32 Pindari Dr South Penrith 26/10/2004 23 37 Gladwood Av South Penrith 26/10/2004 24 204 Prince Charles Pd Kurnell 10/11/2004 25 272 Prince Charles Pd Kurnell 10/11/2004 26 AA Marker St Peters St Peters 17/11/2004 27 AA Marker Sydenham Sydenham 17/11/2004 28 146 Corunna Rd Stanmore 19/11/2004 29 9 Hopetoun St Petersham 19/11/2004

5.3 NOISE MEASUREMENT PROCEDURES

5.3.1 Measurement Methods Noise data were collected at each selected noise station from 7am to 6pm on various days from October 2003 to November 2004. Twenty-minute samples per hour were measured using a Bruel and Kjaer sound level meter Type 2236. It is a Type 1, precision sound level meter (SLM) as required for the high accuracy measurements necessary for these measurements (SAA, 1997). The SLM was mounted in front of each residence, 1 m from the nearside lane of traffic and at least 1 m from the façade (SAA, 1997). The Bruel and Kjaer SLM Type 2236 has a capability to measure several environmental noise indices (such as LN%, Leq, SEL, MaxL, and MinL). More detailed

181 specifications of this SLM can be found in Bruel & Kjaer (1993). The SLM was set for the fieldwork as following.

• The frequency weighting was set to A-weighted because it is considered the most suitable frequency weighting system to measure community reactions to aircraft noise.

• The time weighting parameter was set for normal noise measurement (Bruel & Kjaer, 1993).

• The measurement range was set to measure noise between 30 – 110 dB(A). It is important to set measurement range at the most suitable level to prevent signal overload which may cause error in the final result.

• The peak frequency weighting was set for special applications (Bruel & Kjaer, 1993).

• Three percentile levels were set as default (L1, L10, and L90).

• To develop the noise time curve, it is very important to set the auto logging procedure to be automatically logging noise data for every 1 second into the SLM.

In addition to the noise data recorded by the SLM, the noise sources and events heard by the noise surveyor were manually noted into the Noise Source Classification Form (NSCF) developed by this author (see Figure 5.1). The NSCF consists of three main sections. The recorder completes all the necessary information of the noise station in the first section. In the second section, during the measurement, the recorder notes any variation of sound pressure level from the SLM, time, and type of noise source using ID code provided by the third section of NSCF. If the noise source is an aircraft, the recorder is also required to note the types of operation (i.e., takeoff, landing, or other maneuver) and the runway usage.

182 Data stored in the SLM may subsequently be transferred to a computer via the Serial Interface socket at 9600 bit per second. A specific program for this transmission is required as provided by its developer. Microsoft Excel is required to launch the program. The detailed description of program installation and transmission can be found in Bruel & Kjaer (1993). Every-second logged noise data at each site was transferred from the SLM and was used to depict a so called ‘noise time curve’ (see section 5.4). The data from the NSCF were then superimposed onto the noise time curve for further analysis of the NGI.

Study : Health and well-being impacts by aircraft noise Observer : Issarayangyun T Site : Record : Date : S tart time : Stop time : Weather condition : Acoustic Instrumentation : Type 2236 D-009 Sound Level Meter, Bruel & Kjaer

Aircraft Noise Level (ANL) Background Noise Level (BNL) Aircraft IDa Time Aircraft Max. SPL Aircraft R/W Time BNL SPL 1 Boeing 717 a b c (hr/min/sec) ID (dB(A)) Operation Use (hr/min/sec) ID (dB(A)) 2Boeing 737 3 Boeing 747 4 5 6A310 7A320 8SF340 9Others

Operationb TTakeoff LLanding OOthers

BNL Source IDc 1 passenger car 2small truck 3medium truck 4heavy truck 5 motorcycle 6rail 7 community noise 8ambient noise 9bus 10 animal 11 unusual noise

Figure 5.1: Noise Source Classification Form

5.3.2 Noise Parameters The frequency weighting of all noise indices presented here are referred as the A- weighting system. The term ‘background environmental noise’ specifically used in this thesis is different from the term ‘background A-weighted sound pressure level (LA90)’ as defined in the SAA 1997. The LA90, T is the A-weighted sound pressure level, obtained

183 by using time-weighted ‘F’ that is equal to (or exceeded) for 90% of the time interval considered in the absence of the noise under investigation. It has been used mainly to establish noise criteria for new proposals, such as railway lines, highways, industrial plants or something that produces a loud noise to the surrounding community. If the proposed development generates noise that exceeds an acceptable level of the LA90,T, certain action is required to mitigate that excessive noise level.

For the purposes of this thesis, the background environmental noise is the time average B A-weighted sound pressure level (L Aeq,T). This is the value of the A-weighted sound pressure level of a continuous steady sound that, within a measurement time interval (T), has the same mean square sound pressure as a sound (B), in the absence of the aircraft noise, unusual noises, aesthetic sound (see section 5.4), whose level varies with time. Thus, from the noise measurement, the primary index of interest, particularly in quantifying the background environmental noise, was the LAeq,T because it would be B used to determine the L Aeq,T, which would be used subsequently to develop the NGI.

The mathematical formulae of LAeq,T is (SAA, 1997, equation 3, p.6):

2 ⎡ 1 t2 PA(t ) ⎤ LAeq,T = 10log10 ⎢ dt⎥ (5.1) ∫t 2 1 ⎣⎢t 2 − t1 Po ⎦⎥ where

PA(t) is the instantaneous A-weighted sound pressure of the sound signal at time t

Po is the reference sound pressure equal to 20µPa t2, t1 are the start and finish times of the measurement time interval T.

Recall that the NGI is a parameter that differentiates between aircraft noise and background environmental noise in a novel manner. However, comparisons between the background environmental noise and the NA of aircraft noise could not be made directly because of the inherent differences between these two indices. The NA of aircraft noise would need to be transferred into an index measured in dB(A). Consequently, the concept of the day-night average sound level (DNL) was adapted to overcome this problem. The DNL also employs the concept of energy sound equivalent, but uses the sound exposure level (LAE) as the single-event sound level descriptor with time-of-day weighting factors. Nevertheless, the time-of-day weighting was ignored

184 because the measurement periods did not cover the night time. It was also discounted on the basis of occupational and health safety reasons and for the personal security of the noise surveyor. Therefore, the adjustment factor of the time interval was varied depending on the background environmental noise time interval (Tk), as explained subsequently. Equation (5.2) provides the mathematical formulae of LAE (SAA, 1997, equation 4, p.6).

2 ⎡ 1 t2 PA(t ) ⎤ LAE = 10log10 ⎢ dt⎥ (5.2) ∫t 2 1 ⎣⎢t o Po ⎦⎥ where

PA(t) is the instantaneous A-weighted sound pressure at time t

Po is the reference sound pressure equal to 20µPa t2, t1 are the start and finish times of stated time interval long enough to encompass all significant sound of a stated event to is the reference duration of 1s.

5.4 TYPICAL RESULTING DATA

Large amounts of data were collected, and a typical example of these appears in Figure 5.2. It presents a noise time curve at noise station Number 5 in the Suburb of Stanmore during a morning peak period of aircraft arriving into Sydney Airport. The y-axis of noise time curve is represented by LAeq,1s. The x-axis is denoted by time. The peaks were identified by the noise source codes. For example, the ‘PC’, ‘CN’, ‘AN’, ‘UN’ and ‘B747’ denote noises that were generated by a passenger car, from the community, animals, unusual noises, and a Boeing 747, respectively. Note that the frequency of aircraft landing during this particular morning peak period was approximately one landing per two minutes. The other environmental noise sources during this period were observed to be low so the predominant noise source during this period was from aircraft over flights.

185

95 90 B777 B767 SF340 SF340 SF340 SF340 B747 SF340 85 B717 80 PC 75 B737 PC PC 70 PC CN 65 AN 60 55 LAeq,1s (dB(A)) LAeq,1s 50 45 40 AN 35

2 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :02 :0 1 4 :30 :3 :32 :33 :3 :35 :36 :37 :38 :39 :40 :41 :42 :43 :44 :45 :46 :47 :48 :49 :50 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Figure 5.2: Typical Noise Time Curve at NS-5 during a Peak Morning Period of Aircraft Landings.

Figure 5.3 then illustrates a common situation at the same residence during a morning period when there were no aircraft over flights. In this case, the predominant noise source happened to be from a street-cleaning truck. This type of noise source was termed an unusual noise and is one that did not occur typically (or hourly). It may have happened once a day, or perhaps two or three times a week, or less, depending on the type of noise source involved. Typically, these sources included ambulance sirens, fire alarms, garbage trucks, street-cleaning trucks along with noise from community activities such as lawn mowing, children playing, and even animal noise such as dogs barking. Clearly it can be argued that these unusual noises commonly contribute to community noise. However, these noise sources usually operate rather infrequently and for very short time periods. Nevertheless, once unusual noise sources happened, they resulted in very high sound pressure level peaks (see Figure 5.3) which were not representative of the overall background environmental noise. Consequently, the unusual noise sources were excluded from the analysis.

Note also that this thesis considered sound from wild birds as “aesthetic sounds” which could be observed frequently in residential areas located near parks, recreation areas, and natural areas. This thesis also excluded such aesthetic sounds from the background environmental noise analysis because, in general, they were not considered as a noise (or unwanted sound). This thesis assumed that people prefer the sound of wild birds rather than noise from trucks, buses, or aircraft over flights.

186 95 UN 90 UN 85 PC PC 80 PC 75 AN PC PC 70 PC 65 UN 60 AN 55 AN LAeq,1s (dB(A)) LAeq,1s 50 45 40 35

1 1 1 1 1 1 1 1 1 1 :0 :0 :01 :0 :01 :0 :01 :01 :0 :01 :01 :0 :0 :01 :0 :01 :0 :01 :01 :0 :01 1 3 6 8 2 4 7 9 :30 :3 :32 :3 :34 :35 :3 :37 :3 :39 :40 :41 :4 :43 :4 :45 :46 :4 :48 :4 :50 1 1 1 1 1 1 1 1 1 1 1 1 1 11 1 1 11 1 11 1 11 1 11 1 11 1 1 11 1 11 1 11 1

Figure 5.3: Noise Time Curve at NS-5 during a Morning Period with No Aircraft Over Flights.

5.5 ANALYSIS OF NOISE DATA

5.5.1 Analysis of Background Environmental Noise Levels After detailed reviews of the extensive data set such as those of Figures 5.2 and 5.3, it became apparent that the calculation of the background environmental noise levels was not an easy task. The fluctuations of the background environmental noise levels were unpredictable and varied from day to day, or even hour to hour. Therefore, the approach taken was to exclude all the aircraft noise, unusual noises, and aesthetic sound peaks from the noise time curves such as those in Figures 5.2 and 5.3 so that what remained were the background environmental noise time curves as shown by Figures 5.4 and 5.5, B respectively. Each LAeq,1s was then re-calculated by Equation (5.1) to determine L Aeq,Tk B of each background environmental time curve. For example, L Aeq,Tk of Figures 5.4 and 5.5 were 56.1 and 55.5 dB(A), respectively. The background environmental noise time interval (Tk) was varied depending on the characteristics of each noise time curve. The more frequent are aircraft overflights, unusual noise, or aesthetic sound events then the shorter the duration of background environmental noise time intervals. Table 5.2 B summarises the L Aeq,Tk and Tk of each noise stations.

187 95 90 85 80 75 70 65 60 55 LAeq,1s (dB(A)) LAeq,1s 50 45 40 35

2 2 2 2 2 2 2 :02 :0 :02 :02 :0 :02 :02 :0 :02 :02 :0 :02 :02 :0 :02 :02 :0 :02 :02 :0 :02 0 1 2 3 4 5 6 7 8 9 0 2 3 5 6 8 9 :3 :3 :3 :3 :3 :3 :3 :3 :3 :3 :4 :41 :4 :4 :44 :4 :4 :47 :4 :4 :50 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Figure 5.4: Background Environmental Noise Time Curve of Figure 5.2.

95 90 85 80 75 70 65 60 55 LAeq,1s (dB(A)) LAeq,1s 50 45 40 35

1 1 1 1 1 1 1 :0 :01 :01 :0 :01 :01 :0 :01 :01 :0 :01 :01 :0 :01 :01 :0 :01 :01 :0 :01 :01 0 3 6 9 2 5 8 :3 :31 :32 :3 :34 :35 :3 :37 :38 :3 :40 :41 :4 :43 :44 :4 :46 :47 :4 :49 :50 1 1 1 1 1 1 1 11 1 11 11 1 11 11 1 11 11 1 11 11 1 11 11 1 11 11 1 11

Figure 5.5: Background Environmental Noise Time Curve of Figure 5.3.

All of the noise stations were located on local traffic roads, thus the effects of traffic noise from the road itself were low. The noise stations were classified into three groups based on their connection to the other roads. A so called “high” noise group included those noise stations located on roads which were connected with highways or major roads with high traffic volumes. The background environmental noise level in this group was highly influenced by traffic noise from nearby roads. A so called “medium” group included those noise stations located on roads which were linked with alternative roads (or secondary roads). The background environmental noise level in this group was moderately influenced by traffic noise from nearby roads. Finally, the “low” noise station group represented those locations located on roads which were not connected with any highways, major roads, or alternative roads. For this group, the impacts of traffic noise from other roads were either very low or negligible. The noise stations (which are NS21, NS22, and NS23) located in the aircraft noise non-exposure area

188 would not be used for the development of NGI because the aircraft noise level is zero at those areas. They were collected for the purpose of background environmental noise comparison between the aircraft noise exposure area and the control area.

B Figure 5.6 illustrates the average L Aeq,Tk of the background environmental noise levels for the high, medium, and low noise groups, and noise groups in the control area. The subsequent application of the data of Figure 5.6 would be to assist in the classification of the background environmental noise levels at residences shown on aircraft noise contour maps which will be explained in a later section. The long-term time average A- B weighted sound pressure level of the background environmental noise level (L Aeq,(7am-

6pm)) was determined from Figure 5.6 by Equation (5.3) (which is based on SAA 1997, equation 5, p.7).

70

60

50

40 High

Aeq,Tk dB(A) Aeq,Tk 30 Medium B L 20 Low Control 10

0 7am - 8am - 9am - 10am - 11am - 12pm - 1pm - 2pm - 3pm - 4pm - 5pm - 8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pm

B Figure 5.6: Background Environmental Noise Levels (L Aeq,Tk) for High, Medium, Low, and Control Noise Groups.

K B ⎡ 1 B ⎤ LAeq,LT = 10log⎢ ∑(antilog 0.1LAeq,Tk )⎥ (5.3) ⎣ K k=1 ⎦ where B L Aeq,LT is the long-term time average A-weighted sound pressure level of the background environmental noise level K is the number of reference background environmental noise time intervals B L Aeq,Tk is the time average A-weighted sound pressure level of background environmental noise in the kth reference time interval.

189 B Table 5.2: Time Average A-Weighted Sound Pressure Level of Background Environmental Noise (L Aeq,Tk) and Time Interval (Tk) at Each Noise Station.

7am - 8am 8am - 9am 9am - 10am 10am - 11am 11am - 12pm 12pm - 1pm 1pm - 2pm 2pm - 3pm 3pm - 4pm 4pm - 5pm 5pm - 6pm

B B B B B B B B B B B Group NS L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK L Aeq,Tk TK (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) (dB(A) (second) 3 60.1 934 61.1 951 61.7 953 62.8 1017 62.0 1120 62.1 1055 63.0 1169 62.8 1114 64.5 1024 61.7 1007 62.2 1151 4 61.2 940 60.9 956 59.9 1067 58.0 845 58.9 988 60.8 509 58.3 883 58.4 1004 58.7 740 58.7 1087 56.9 933 8 60.8 941 59.6 941 59.5 1036 59.8 946 58.9 1201 58.2 1201 60.0 1012 60.6 1065 63.4 1017 63.5 1025 63.9 943 9 61.8 1201 57.2 1061 59.7 1201 57.0 1201 60.5 858 60.3 796 56.3 1011 59.6 689 61.0 911 59.4 657 62.0 1201 High 15 59.9 1015 59.2 1020 60.6 1025 60.0 1091 57.0 1104 60.0 940 60.1 1066 59.9 1106 58.7 1159 59.3 1064 59.4 887 17 61.3 848 61.4 988 60.7 1034 58.4 981 60.4 851 58.6 891 57.1 1112 56.4 1077 56.8 825 54.4 915 53.8 815 18 60.3 760 58.0 986 58.6 928 57.4 864 60.9 1001 56.6 836 59.2 747 60.7 1085 58.1 600 54.5 631 60.9 834 27 62.3 1987 61.7 2870 60.5 2633 61.2 1211 57.4 838 58.7 857 54.9 994 55.9 1201 56.5 940 57.1 778 57.4 853 2 60.0 1072 57.2 286 59.6 1141 58.9 950 58.0 652 59.7 1143 58.5 1094 58.2 1165 60.7 1201 61.6 1017 58.9 1201 5 56.2 739 58.7 891 54.8 807 54.2 836 55.5 737 51.3 1201 54.0 1164 52.1 1188 54.5 979 57.7 1164 54.5 956 6 60.3 917 66.0 923 65.5 1073 63.2 927 53.5 1153 54.8 1201 53.7 1201 53.9 1201 54.1 1201 57.8 1046 55.9 1019 7 60.8 1201 61.4 837 60.7 974 57.7 974 57.2 1141 56.6 1046 56.1 1201 57.2 873 57.7 904 56.8 1043 57.8 891 10 52.1 888 52.6 975 53.0 1201 48.2 732 48.9 841 50.1 659 50.7 747 53.5 533 53.3 889 54.7 1201 53.2 976 11 55.4 948 56.5 742 55.6 736 53.3 955 55.2 942 53.2 909 50.7 994 53.6 1007 58.3 1201 54.6 936 54.3 792 Medium 12 56.8 951 57.5 738 55.4 916 54.6 668 53.5 974 55.4 846 53.5 985 57.3 934 54.0 884 58.0 874 54.7 903 14 49.5 862 53.8 846 53.9 834 52.2 928 55.1 834 52.9 896 53.7 648 54.5 1126 54.3 1011 55.4 987 56.6 772 16 62.3 980 60.6 1128 55.3 912 54.5 948 54.6 930 55.9 707 53.5 740 53.6 935 59.3 1022 57.7 951 57.9 741 28 63.1 770 63.0 714 58.0 769 58.0 603 59.1 993 59.9 1191 58.1 930 57.1 1133 60.2 972 59.3 1030 60.8 1057 20 52.0 836 58.7 916 55.5 648 53.8 750 56.8 757 54.6 717 53.6 804 54.3 785 58.5 1168 57.3 1201 58.4 1013 24 57.1 2139 56.6 2448 58.3 1960 59.2 959 56.4 704 60.7 962 59.9 1044 57.2 915 59.1 818 60.8 649 59.6 817 25 60.9 2268 58.6 2490 55.5 535 58.8 2180 58.5 1886 56.8 734 59.7 779 52.8 636 57.6 849 54.9 878 56.8 763 1 47.3 803 51.4 889 52.8 1201 47.4 729 48.2 1103 52.1 1079 54.1 1098 52.7 1143 54.4 772 54.1 871 52.4 1202 13 54.7 927 56.4 925 55.9 837 56.6 568 51.8 1201 52.8 1201 57.5 1201 50.4 1142 52.2 1130 54.0 738 55.1 747 Low 19 52.4 775 57.1 906 50.8 715 50.8 633 50.9 905 49.4 664 49.4 901 50.0 763 55.7 1036 55.2 1039 54.3 1048 26 51.3 2412 53.3 1630 52.7 1185 49.4 800 49.2 763 50.4 760 52.5 1015 52.5 797 52.3 708 52.8 637 52.9 700 29 49.3 737 50.0 525 50.8 946 48.2 982 48.6 824 50.8 1145 49.4 1200 50.2 1101 50.4 1185 51.3 898 51.4 980 21 47.6 901 52.2 656 47.1 545 42.8 590 49.1 667 39.9 601 42.6 687 43.6 483 48.0 679 52.9 650 51.4 566 Control 22 50.7 639 54.0 721 53.1 685 48.6 721 49.2 643 49.5 661 42.9 721 49.8 667 51.0 721 51.8 647 55.3 695 23 50.5 681 53.9 550 41.6 646 46.7 724 44.5 496 50.0 721 49.2 721 49.6 580 49.9 650 50.8 627 50.0 721

190 B The L Aeq,(7am-6pm) of high, medium, low, and control noise groups were 59.5, 56.5, 52.0, and 48.8 dB(A), respectively. The background environmental noise levels between the aircraft noise exposure groups and the control group were relatively different. This reflects the fact that the former group is located near the Central Business District and industrial areas where complete control of effects from non-aviation noise sources is restricted. An implication of this observation is that the analysis of health impacts by aircraft noise needs also to consider the effects of background environmental noise between both groups.

5.5.2 Analysis of Aircraft Noise Levels As mentioned previously, the current research adopted the DoTARS concept that outdoor sound levels of 70 dB(A), or more, are likely to interfere with indoor activities such as watching TV and conversing. Therefore, it was assumed that the level of aircraft noise that disturbs indoor activities would also be 70 dB(A), or more, so that aircraft noise events less than 70 dB(A) were discarded from the analysis. The data such as those of Figure 5.2 were revisited by considering only aircraft noise levels that are louder than 70 dB(A) (called aircraft noise time curve) as shown by Figure 5.7. The average hourly N70 of aircraft noise could be determined by extrapolating the value of N70, which was counted from the aircraft noise time curve, into one hour increments. For example, from Figure 5.7, the number of aircraft noise events louder than 70 dB(A) is nine per twenty minutes. Thus the average hourly N70 of aircraft noise becomes nine multiplied by sixty minutes and then divided by twenty minutes which yields twenty- seven events per hour.

95 90 85 80 75 70 65 60 55 LAeq,1s (dB(A)) LAeq,1s 50 45 40 35

2 2 :02 :02 :02 :02 :02 :0 :02 :02 :02 :02 :02 :0 :02 :02 :02 :02 :02 :02 :02 :02 :02 8 :30 :31 :32 :33 :34 :35 :36 :37 :38 :39 :40 :41 :42 :43 :44 :45 :46 :47 :4 :49 :50 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

Figure 5.7: Aircraft Noise Time Curve of Figure 5.2.

191 In addition, the sound exposure levels of aircraft noise (LAEk) equal to, or louder than, 70 dB(A) of each aircraft noise time curve were calculated by Equation (5.2). Once more, in concert with the DNL concept, the LAEk values were then adjusted with a time interval factor (Tk) of the background environmental noise according to Equation (5.4). A Table 5.3 summarises the L Aeq,Tk and the average hourly N70 of each noise stations, except those located in the control areas.

A L Aeq,Tk = LAEk – 10 log (Tk) (5.4) where A L Aeq,Tk is the time average A-weighted sound pressure level of aircraft noise at each noise station (dB(A))

LAEk is the sound exposure level of aircraft noise that equals or exceeds 70 dB(A) during the kth background environmental noise time interval (dB(A)) th Tk is the k background environmental noise time interval (second) (Table 5.2).

From Table 5.3, the long-term time average A-weighted sound pressure levels of aircraft noise at each noise station were then calculated by Equation (5.5) (which is based on AS 1055.1 – 1997, equation 5, p.7). The total N70 (7am – 6pm) of each noise station was determined by simply summing up the N70 value of each hour.

K A ⎡ 1 A ⎤ LAeq,LT = 10log⎢ ∑(antilog 0.1LAeq,Tk )⎥ (5.5) ⎣ K k=1 ⎦ where A L Aeq,LT is the long-term time average A-weighted sound pressure level of the aircraft noise level K is the number of reference background environmental noise time intervals A L Aeq,Tk is the time average A-weighted sound pressure level of aircraft noise in the kth reference background environmental noise time interval.

A Consequently, the relationship between L Aeq,LT and N70 during 7am – 6pm was developed (see Figure 5.8). A non-linear regression analysis of these data resulted in the

192 formula of Equation (5.6), the coefficient of determination (R2) which turned out to be 0.79. At this time, any comparisons between N70 (events) and background environmental noise levels (dB(A)) should be done via Equation (5.6) and Figure 5.6.

A L Aeq,7am-6pm = 18.31 log (N70) + 29.42; N70>0 (5.6)

80 75 70 65

60 2 dB(A) R = 0.79 55 50 45 Aeq,7am-6pm A

L 40 35 30 0 20 40 60 80 100 120 140 160 N70 (events)

A Figure 5.8: Relationship between L Aeq,7am-6pm and N70 for All Data in the High, Medium, and Low Background Environmental Noise Groups.

193 A Table 5.3: Time Average A-Weighted Sound Pressure Level (L Aeq,Tk) and Average Hourly Number Above 70 dB(A) (N70) of Aircraft Noise at Each Noise Station.

7am - 8am 8am - 9am 9am - 10am 10am - 11am 11am - 12pm 12pm - 1pm 1pm - 2pm 2pm - 3pm 3pm - 4pm 4pm - 5pm 5pm - 6pm A A A A A A A A A A A NS L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 L Aeq,Tk N70 (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB(A) (events) (dB (A) (events) (dB(A) (events) 3 67.4 15 65.104 18 64.5 15 64.166 15 - 0 73.868 12 47.0 3 72.299 3 72.6 9 73.25 12 50.4 6 4 66.9 24 66.799 15 66.5 9 63.404 24 73.3 15 76.183 15 68.4 6 70.427 15 64.6 6 66.374 9 65.1 9 8 49.3 6 58.97 9 50.2 6 48.976 3 - 0 - 0 - 0 52.765 6 58.6 12 53.168 9 51.0 6 9 - 0 - 0 - 0 - 0 67.2 18 66.495 12 62.8 6 69.8 24 67.7 12 - 0 - 0 15-0-0-0-0-047.6743-0-0-0-0-0 17 46.8 6 - 0 - 0 - 0 - 0 63.404 6 57.9 6 - 0 66.4 9 65.408 9 65.8 9 18 61.9 12 56.8 6 59.3 9 48.648 3 65.5 12 66.26 6 68.3 15 - 0 71.1 15 74.303 18 54.3 3 27 57.0 3 73.412 11 74.5 16 72.902 10 76.4 6 66.17 12 73.9 6 - 0 49.0 3 69.954 9 74.4 9 2 -0-0-0-0-0-0-0-0-049.1323-0 5 72.9 27 71.35 21 62.2 6 67.724 15 - 0 - 0 - 0 - 0 69.7 6 - 0 67.2 3 6 71.2 15 67.426 18 65.9 12 68.311 18 65.7 3 - 0 - 0 - 0 - 0 57.838 3 55.0 3 7 - 0 58.055 15 48.1 3 51.886 6 49.3 3 - 0 - 0 47.065 9 55.1 6 - 0 56.4 9 10 - 0 - 0 - 0 - 0 64.7 18 69.785 15 65.6 15 68.864 24 61.9 15 - 0 - 0 11 62.7 9 62.781 6 64.0 12 59.547 3 55.7 3 64.895 9 61.0 6 48.187 6 - 0 59.646 12 51.8 3 12 65.5 6 65.339 6 59.7 3 68.957 18 - 0 61.88 3 62.3 6 62.721 9 65.2 15 62.374 6 60.8 6 14 48.2 3 61.347 3 66.5 6 61.558 9 63.7 6 68.789 6 - 0 - 0 67.7 3 58.339 6 69.7 18 16 57.4 3 - 0 58.7 6 62.509 3 57.1 3 66.779 12 70.6 15 48.413 3 68.4 3 70.79 9 61.8 8 28 69.5 24 70.051 24 69.8 18 72.255 24 65.7 6 - 0 64.1 6 - 0 49.5 3 67.575 6 66.2 3 20 - 0 63.282 6 61.2 9 73.921 9 60.9 18 70.014 9 - 0 - 0 - 0 - 0 - 0 24 63.9 13 60.809 7 59.2 6 - 0 - 0 - 0 - 0 - 0 - 0 50.513 3 - 0 25 59.1 13 58.028 9 64.3 16 58.165 12 58.5 10 - 0 47.0 3 54.531 3 - 0 - 0 48.4 3 1 -0-0-0-0-0-0-0-0-0-0-0 13 50.2 3 52.301 3 59.6 6 63.649 9 - 0 - 0 - 0 - 0 - 0 70.143 21 66.2 21 19 - 0 55.301 6 71.5 12 73.632 15 - 0 67.89 3 - 0 - 0 - 0 - 0 58.5 3 26 59.1 6 68.133 17 62.1 5 68.522 15 59.0 6 59.814 9 - 0 54.042 3 61.3 9 64.655 18 64.5 12 29 53.2 9 48.847 3 43.6 3 53.793 6 54.4 6 - 0 - 0 - 0 - 0 68.216 9 61.0 3

194 5.6 THE NOISE GAP INDEX (NGI)

The NGI was defined simply as the difference between aircraft noise and background environmental noise. In the present, case the NGI became the difference between the A B L Aeq,LT and the L Aeq,LT as given in Equation (5.7). It is noted that this formulae is practical for during 7am – 6pm, and when N70 is higher than zero.

A B NGI = L Aeq,LT - L Aeq,LT (5.7)

By substituting Equation (5.6) into Equation (5.7), and also by applying the background environmental noise levels of 59.5, 56.5, and 52.0 dB(A) as determined from Figure 5.6, Equation (5.7) becomes Equation (5.8) below.

NGI = 18.31 log (N70) – Z; N70>0 (5.8) where Z is an adjustment factor = 30.08 in high background environmental noise areas = 27.08 in medium background environmental noise areas = 22.58 in low background environmental noise areas.

Figure 5.9 illustrates the relationship between the NGI and the N70 generated from Equation (5.8). From the figure, the NGI of high, medium, and low noise groups A B become zero (or the L Aeq,LT and the L Aeq,LT is the same) at the N70 of around 44, 30, and 17 events, respectively.

195 25 20 15 10 5 0 -5 0 25 50 75 100 125 150 175 200 -10 NGI (dB(A)) NGI high -15 medium -20 low -25 -30 -35 N70 (events)

Figure 5.9: Relationship between NGI and N70 for High, Medium, and Low Background Environmental Noise Groups

5.7 A CASE STUDY

5.7.1 Calculating the NGI This section applies Equation (5.8) into a study population of aircraft noise exposure area as defined by section 4.5.1. Figure 5.10 illustrates the average annual N70 of aircraft noise around Sydney Airport. The N70 contour maps are regularly produced by Airservices Australia using the TNIP (see section 2.3.8.2). Figure 5.11 illustrates an example of a large scale N70 contour map for the area circled in Figure 5.10 in the suburbs of Stanmore and Petersham. It also displays the locations of the noise stations allocated in these areas.

196

Figure 5.10: Daily Average Number of Aircraft Noise Events Louder than 70 dB(A) at Sydney Airport During 1 January – 31 December 2003 and Study Population of Aircraft Noise Exposure Area. (source: Airservices Australia, 2004a, attachment F)

Originally, the N70 values (see Figure 5.10) are plotted at 0.5 km centres and the colour squares are approximately 0.1666 km2. This scale was considered too coarse for the present research. The large scale N70 map (specifically required from Airservices Australia) was then divided by imaginary grid lines as presented in Figure 5.11. The new colour squares are approximately 0.0185 km2. The N70 values of each new colour square were interpolated from between two original N70 values. For example, the N70 values of B-3 and C-3 were calculated from; 70.121 – (70.121 – 82.069) ÷ 3 = 74.102, and 70.121 – 2 × (70.121 – 82.069) ÷ 3 = 78.086, respectively.

197 A B C D E

1

NS-28

2

NS-6

3

NS-5

4

5

NS-29

Figure 5.11: Daily Average N70 During 1 January – 31 December 2003 at Study Area and Locations of Noise Stations.

From Figure 5.11, the N70 value of each residence could be estimated. The next step was to determine the background environmental noise group of each residence (or the Z value). The types of road are distinguished into three groups: the symbol ‘≡≡≡≡’ for the major roads or the highways; the symbol ‘====’ for the alternative roads; and the single line for the local roads. According to section 5.5.1, roads connected with highways or major roads with high traffic volume would be recognised as high background environmental noise areas. Roads connected with alternative roads (or roads with medium traffic volume) would be recognised as medium background environmental

198 noise areas. Examples of these roads in Figure 5.11 are Westbourne Street, Corunna Road, and Albany Road. Finally, roads without connection with highways, major roads, or alternative roads such as Bent Street, Hopetoun Street, and Temple Street in Figure 5.11 were identified as low background environmental noise areas. By substituting the N70 values from Figure 5.11 and the Z values classified by the types of road where a particular residence is located, the NGI of this residence could be calculated. For example, from Equation (5.8), address 119 Westbourne street Petersham, which is located in medium background environmental noise area inside a colour square B-3, would have a NGI value equal to 7.15 dB(A) (18.31 × log(74.104) – 27.08 = 7.15 ). This process could be repeated for each residence of interest in the study area.

5.7.2 NGI and Aircraft Noise Annoyance Recall that the NGI was developed based on the assumption that people living in different background environmental noise areas might have different responses to the same aircraft noise level. This assumption will now be tested based on some parts of the data collected from the main health survey. It is noted that what now follows in the present section is a descriptive statistical analysis. Potential confounding factors such as noise sensitivity (see section 2.4.5.2), have not been considered. The descriptive statistical analysis of the overall data from the main health survey will be presented in the next chapter.

Data concerning aircraft noise annoyance obtained from the aircraft noise exposure group collected from the main health survey was analysed. These data were quantified by an aircraft noise annoyance scale which is an ordinal variable ranging from zero to ten. A value of zero on this scale means “not at all annoyed” and a value of ten means “highly annoyed” by aircraft noise. The N70 value of each respondent was obtained from the large scale N70 contour map. Then the NGI value of each respondent was calculated based on the procedures described in the above section. Figure 5.12 illustrates the relationship between the aircraft noise annoyance scale and the N70 stratified by the quartile points of the valid samples. For example, there were 25 percent of respondents living in areas where the average N70 value was less than or equal to 78.1 events during 7am – 6pm. The average aircraft noise annoyance score of this group

199 was 5.8. From Figure 5.12, it appears that N70 has a correlation with the aircraft noise annoyance scale: the higher the N70 value then the higher aircraft noise annoyance.

e 10 9 8 7 6 5 4 3 2 1

Aircraft Noise AnnoyanceAircraft Scal 0 47.1 - 78.1 78.1 - 80.6 80.6 - 97.1 97.1 - 138.2 N70 during 7am-6pm (events)

Figure 5.12: Relationship between Aircraft Noise Annoyance Scale and N70.

Figure 5.13 illustrates the relationship between aircraft noise annoyance scale and the NGI stratified by the quartile of points of the valid samples. For example, there was 25 percent of valid respondents located in areas where NGI was between 6.29 and 7.56 dB(A). The average aircraft noise annoyance score of these areas was 6.5. From Figure 5.13, it appears that the average aircraft noise annoyance score of respondents was quite stable (approximately 6.5) in areas with NGI less than 9.7 dB(A). Conversely, the average aircraft noise annoyance score dropped to 5.7 in areas where NGI was higher than 9.7 dB(A). The null hypothesis assumed no difference in the population mean score of aircraft noise annoyance score between the quartile groups. Conversely, the alternative hypothesis assumed a difference in the population mean score of aircraft noise annoyance score between the quartile groups. Analysis of variance (ANOVA) found that (table not shown) aircraft noise annoyance scores varied significantly between the quartile groups with F(3, 312) = 2.748, p-value = 0.043, α = 0.05. The null hypothesis was then rejected.

It also appears that (as indicated in Figure 5.13) the average NGI values of high, medium, and low background environmental noise groups were 5.85, 7.72, and 12.2 dB(A), respectively. These observations tend to support the preliminary conclusion that people living in high and medium background environmental noise areas were more

200 likely to be annoyed by the same aircraft noise exposure level than people living in low background environmental noise areas.

10

e 9 8 7

6 5 4 High Background Medium Background Low Background 3 Environmental Environmental Noise Environmental Noise Group Group Noise Group 2 (NGI=5.85) (NGI=7.72) (NGI=12.2) 1 0

Annoyance Noise Aircraft Scal 0.54 - 6.29 6.29 - 7.56 7.56 - 9.70 9.70 - 13.87 NGI, dB(A)

Figure 5.13: Relationship between Aircraft Noise Annoyance Scale and NGI.

5.8 CONCLUSIONS AND DISCUSSIONS

Current practice of aircraft noise quantification in Australia is deficient in reflecting community responses toward aircraft noise. The current index (called Australian Noise Exposure Forecast, ANEF) was developed based on logarithmically averaged annual average day aircraft noise energy with time-of-day weighting which is never fully understood by a member of the general community. Australian Department of Transport and Regional Services (DoTARS) has proposed more transparent approaches to describing and assessing aircraft noise. The number of aircraft noise events louder than 70 dB(A) (N70) is one alternative index.

This chapter has expanded an index of N70 by incorporating the characteristics of background environmental noise into a ‘new’ noise index which is termed the Noise Gap Index (NGI). This index distinguishes between aircraft noise and background environmental noise in a novel manner based on an assumption that people living in different background environmental noise areas may have different reactions to the

201 same aircraft noise level. The NGI has been developed as an index that is easy to understand by the layperson and that quantifies relevant aspects of the potential impacts of aircraft noise. The preliminary study found some form of statistical association between aircraft noise annoyance and NGI. Subjects residing in high and medium background environmental noise areas appeared to be more likely to be annoyed by the same aircraft noise exposure level than subjects living in low background environmental noise areas. This might reflect the characteristics of people suffering from high level of background environmental noise to be more vulnerable to aircraft noise than people from low background environmental noise areas.

This research intended to use NGI to assist in explaining the effects of aircraft noise on subjects from aircraft noise exposure area (not from the control area). Chapter 7 of this thesis will relate the outcomes of the main health survey to the degree of exposure to aircraft noise, as quantified by the NGI, in order to explore any relationships between long-term aircraft noise exposure and prevalence of hypertension in adults.

202 CHAPTER SIX

AIRCRAFT NOISE AND HEALTH RELATED QUALITY OF LIFE

6.1 INTRODUCTION

Chapter 3 found that the issue of community health and well-being impacts by aircraft noise has not been included in aircraft noise management strategies and policies. Airport authorities interpret that aircraft noise causes no diseases to human health. However, from the literature review, a consensus has been established that aircraft noise does deteriorate the quality of life of the community surrounding the airports, especially major commercial airports. ‘Health’, according to the definition declared by WHO, is not only the absence of disease but also includes a state that is complete in physical, mental, and social well-being. Thus, this thesis hypothesises that aircraft noise should have an effect on human health in terms of health related quality of life and is a topic worthy of further investigation with reference to Sydney Airport – a point supported by a stakeholder workshop on research needs stemming from the New South Wales Government’s Botany Bay Strategy . One international publication by Meister and Donatelle (2000) supported this hypothesis by concluding that the health related quality of life of subjects from aircraft noise exposure areas were significantly worse than the control areas.

This chapter explores the impact of aircraft noise on health related quality of life using data from the survey described in Chapter 4. The content of this chapter is organised as follows. Section 6.2 describes the response rate of the health and well-being survey around Sydney Airport and a control area, and compares the mean of selected variables between these two study groups. Section 6.3 explores the first core research question (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”) by using analysis of covariance technique. Section 6.4 summarises the results and addresses concluding remarks.

203 6.2 PRELIMINARY DATA ANALYSIS

This section describes the preliminary data analysis (descriptive statistical analysis) of data obtained from the main health and well-being survey. The analysis is concerned with variables such as demographic characteristics, socio-economic status, health related quality of life (including physical functioning, general health, sense of vitality, and mental health), noise sensitivity, noise stress, and the other related measures. The exploration of the two core research questions (see section 2.5) will be performed later in section 6.4 and 7.3, respectively.

6.2.1 Response Rate After errors and impractical procedures of the health survey instruments detected during the pilot study had been removed and corrected, the first contact letters were mailed to 1,500 randomly selected subjects on 14 April 2004. The first return questionnaire was received five days later. The follow-up letters were mailed to the non-respondents following the time diagram as illustrated in Figure 4.8. The last return questionnaire was received on the 24 June 2004. Table 6.1 tabulates the response rate by study groups. A total of 796 responses were returned, of whom 704 filled in the questionnaire and 92 indicated unwillingness to participate in the survey. The number of responses from subjects in the control group was a little bit lower than from the noise exposure area. This might reflect the lack of relevance of a questionnaire about aircraft noise in daily life of subjects in the control area.

The total response rate was 47 percent which was slightly lower than the expected response rate of 52 percent. However, the response rate of 47 percent was sufficient, especially when compared with the average response rate (around 40 percent) of this type of survey administration (Dillman, 2000). The total sample sizes of each group were sufficient to detect the 5-point differences in health measures between groups as required by SF-36’s developers (see Table 4.11) at the 5% level of significance with a power of 80%.

204 Table 6.1: Return Rate by Study Groups Noise Exposure Group Control Group Total Response 372 (50%) 332 (44%) 704 (47%) Return Refuse 36 (5%) 56 (7%) 92 (6%) Not Return 342 (46%) 362 (48%) 704 (47%) Total 750 750 1500

The number of respondents classified by language is presented in Table 6.2. The percentage of respondents of both selected languages (Greek and Arabic) was lower than expected (see Table 4.12). In fact, all of the questionnaires from the control group were in English.

Table 6.2: Responded Subjects by Language Noise exposure group Control group Language number % number % English 362 97% 332 100% Greek 7 2% 0 0% Arabic 3 1% 0 0% Total 372 100% 332 100%

6.2.2 Demographic and Socioeconomic Status of Samples It is important to note that this research has assumed that long-term aircraft noise has indirect negative community health and well-being impacts. Consequently, subjects who have resided in their existing residence for less than 1 year are excluded from the study. In the total sample, there were 33 (8.9%) of 372 from the noise exposure group and 16 (4.8%) of 332 from the control group who have lived in their existing residence for less than one year. These subjects were, therefore, excluded from the study. Thus, the total sample size becomes 339 for the noise exposure group and 316 for the control group.

Table 6.3 compares the demographic characteristics and socioeconomic status of both study groups. In the total sample, age ranges from 15 to 87. The distributions of age were considered normal in both groups with no outliers. The mean age of the control group was approximately four years higher than the noise exposure group. From student t-test, it was found that this difference was statistically significant (p-value = 0.001). This might reflect the higher percent of “not in labour force” of subjects in the control

205 group. In the control group, 66.1 percent of the sample was female, which is 10.1 percent higher than in the noise exposure group. And a chi-square test revealed that this difference was significant (p-value = 0.009).

Table 6.3: Demographic Characteristics and Socioeconomic Status by Study Groups

Variable Noise Exposure Group Control Group p-value Mean age (year) 46.63 (SD=15.57) 50.85 (SD=15.22) 0.001 Sex (% female) 190 (56.0%) 209 (66.1%) 0.009 Body Mass Index 0.006 Obesity 55 (16.2%) 81 (25.6%) Overweight 104 (30.7%) 90 (28.5%) Acceptable weight 132 (38.9%) 91 (28.8%) Underweight 25 (7.4%) 18 (5.7%) Education <0.001 Bachelor degree or higher 118 (34.8%) 37 (11.7%) Certificate - Diploma 106 (31.3%) 144 (45.6%) High school or lower 108 (31.9%) 131 (41.5%) Employment status 0.003 White collar 167 (49.3%) 118 (37.3%) Blue collar 50 (14.7%) 64 (20.3%) Unemployed 19 (5.6%) 13 (4.1%) Not in labour force 94 (27.7%) 119 (37.7%) Marital status <0.001 Married/De facto 186 (54.9%) 219 (69.3%) Widowed/Divorded/Separated 55 (16.2%) 64 (20.3%) Never married 93 (27.4%) 29 (9.2%) Smoking status 0.014 Current smoker 76 (22.4%) 44 (13.9%) Ex-smoker 98 (28.9%) 108 (34.2%) Never smoke 149 (44.0%) 153 (48.4%) Alcohol consumption 0.623 High 54 (15.9%) 43 (13.6%) Low 186 (54.9%) 183 (57.9%) None 68 (20.1%) 68 (21.5%) Exercise activity level 0.034 High exercise 59 (17.4%) 40 (12.7%) Moderate exercise 77 (22.7%) 66 (20.9%) Low exercise 134 (39.5%) 122 (38.6%) Sedentary 58 (17.1%) 82 (25.9%) Household weekly income 0.451 Over AUD$2,000 33 (9.7%) 22 (7.0%) AUD$401 - AUD$1,999 225 (66.4%) 214 (67.7%) Under AUD$400 68 (20.1%) 66 (20.9%) Acoustic Insulation <0.001 Yes 126 (37.2%) 9 (2.8%) No 198 (58.4%) 286 (90.5%) Nutrition 0.135 Salty food 54 (15.9%) 65 (20.6%) No salty food 277 (81.7%) 246 (77.8%)

In terms of socioeconomic status, subjects in the noise exposure group seem to have a higher education (p-value < 0.001) and better employment status (p-value = 0.003) than the control group. However, both groups were similar in term of household income (p-

206 value = 0.451). The consumption of alcohol (p-value = 0.623) and salty food (p-value = 0.135) of subjects from both groups were not significantly different. Subjects in the noise exposure group seem to smoke more tobacco (p-value = 0.014) than the control group. In the control area, subjects took less exercise (p-value = 0.034) than in the noise exposure area. Therefore, the percentage of obesity in the control area was considerably higher (p-value = 0.006). Other researches have noted the problem of obesity in suburban life styles (Vandegrift and Yoked, 2004). The marital status between both groups was significantly different (p-value < 0.001).

Not surprisingly, there was only 2.8 percent of the sample in the control group that has insulated their house from noise. Around 37% of houses in the noise exposure group have been insulated from noise. This is a result of a noise management plan (called Sydney Airport Noise Insulation Project, SANIP) at Sydney Airport following the opening of the Third Runway in 1994.

6.2.3 Health and Related Measures Table 6.4 compares statistically health and related measures of subjects between study groups. From Table 6.4, it appears that most of the health measures of the noise exposure group were lower than the control group, implying that the health related quality of life of subjects from noise exposure group was worse than the control group. However, without any control for covariates, analysis of variance revealed almost all of these differences (except Mental Health Score) were not statistically significant. The proportion of people with hypertension in the control group was slightly higher than in the noise exposure group. However, this difference is not statistically significant (p- value = 0.450). The proportion of hypertensive in parent(s) and high cholesterol level in the noise exposure group was higher than the control group. However, these differences were also not statistically significant.

Undoubtedly, subjects in the noise exposure group were more sharply annoyed by aircraft noise (p-value < 0.001) than the control group. Even though, the level of traffic noise annoyance between both groups was slightly different, it was statistically B significant (p-value = 0.001). This might reflect the fact that the L Aeq,(7am-6pm) of the control group was lower than the noise exposure group (see Figure 5.6). The level of

207 noise sensitivity between both groups was not significantly different (p-value = 0.193). Thus, subjects from the noise exposure area have a high level of noise stress (p-value < 0.001) than the control area. Apparently, this noise stress is largely due to the exposure of aircraft noise.

Table 6.4: Descriptive Statistics of Health and Related Measures by Study Groups Variable Noise Exposure Group Control Group p-value Mean Physical Functioning Score 79.09 79.23 0.941 Mean General Health Score 64.49 66.08 0.370 Mean Vitality Score 54.58 57.02 0.128 Mean Mental Health Score 68.02 73.53 <0.001 Hypertension 51 (15.0%) 55 (17.4%) 0.450 Hypertension in Parent(s) 154 (45.4%) 132 (41.8%) 0.297 High Cholesterol Level 62 (18.3%) 47 (14.9%) 0.215 Mean Noise Stress Score 6.44 (SD=2.31) 4.25 (SD=1.93) <0.001 Mean Noise Sensitivity Score 27.76 (SD=7.92) 26.97 (SD=7.38) 0.193 Mean Aircraft Noise Annoyance 6.27 (SD=3.04) 1.03 (SD=2.01) <0.001 Mean Traffic Noise Annoyance 2.61 (SD=2.57) 1.96 (SD=2.31) 0.001

Recall that noise sensitivity score ranges from 0 to 45, there is no formal cutoff point on the noise sensitivity scale. The higher noise sensitivity score means the more susceptible to noise. Based on data collected by the main health survey of this thesis, there were 16% and 6% of the total subject who reported noise sensitivity score higher than 36 and 40, respectively. Note also that, there was 1% of the total subject (7 from 637) who reported noise sensitivity score equals to the maximum value. If Sydney’s population is 4 million, it is reasonable to assume that there are approximately 40,000 people who are very “susceptible” to noise.

Descriptive statistical analysis found many different aspects between these two study groups, which need to be carefully controlled when drawing inferences about health and well-being. The result of this section should not be interpreted as a final conclusion because some potential covariates (such as age and noise sensitivity) and confounders (for example, obesity, employment status, and exercise activity) have not been taken into account. The approach required for this situation is called the multivariate statistical analysis (see section 2.6).

The full description of variables (including variable coding) involved in the analysis of the two core research questions is provided in Appendix C. Prior to the commencing of

208 the main analyses, preliminary data screening has been implemented and the outliers (especially in continuous variables) have been discarded as presented in Appendix D. Appendix E explains, in detail, the general purposes, fundamental concepts, and example analysis of analysis of covariance. The following section explores the first core research question.

6.3 EXPLORING OF THE FIRST CORE RESEARCH QUESTION

Appendix E describes how to implement the one-way analysis of covariance with completely randomised design, and provided an example of a hand-calculation of this technique. Obviously, those sections were useful in assisting the analyst to understand the process of analysis of covariance. However, a suitable computer program is required when dealing with more complex designs (such as unequal sample sizes and multiple covariate variables and independent variables).

This section explores the first core research question (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”) by using data collected from the population-based health and well-being survey around Sydney Airport and a well-defined control group (see chapter 4).

The dependent variables are health measure scales which are Physical Functioning, General Health, Sense of Vitality, and Mental Health. The interpretation of each scale has been presented in Table 4.3. Note that, each health measure scale assesses different, and independent, aspects of health related quality of life of an individual. Consequently, the analysis is divided into four sections relevant to each of the health related quality of life aspects. The main effect variable for each section is long-term aircraft noise exposure. The assumptions required by analysis of covariance are evaluated in the following subsection.

6.3.1 Evaluation of Assumptions The covariate variable must be a continuous variable so the linear relationship between covariate variable and dependent variable can be established. As required by one of the

209 main assumptions of analysis covariance, the covariate variable and the independent variable(s) should be independent with each other. Aircraft noise annoyance score was therefore ineligible to be a covariate variable in this analysis. Since the covariate variable should be measurable on an interval scale (Mickey et al, 2004), the study decided to select age and noise sensitivity as candidates for covariate variable.

The pair-scatter plots of the covariate variables (age, noise sensitivity) and the dependent variables (Physical Functioning, General Health, Vitality, Mental Health) classified by aircraft noise exposure are presented in Figure F.1 in Appendix F. From this figure, it was found that the relationships between covariate variable and dependent variable by study groups that appear to fit both linearity and homogeneity of regression assumptions were age & Physical Functioning, age & General Health, age & Vitality, noise sensitivity & Vitality, and noise sensitivity & Mental Health. These covariate variables would be checked for significance in the model by the Type III sum of squares method.

The survey was designed as strictly anonymous and confidential, and the subjects were informed of this by both the cover letter and contact letters. Thus, it was assumed that subjects are reasonable consistent in reporting the information about their age (“How old were you on your last birthday”). High reliability of age is expected. The reliability of noise sensitivity was checked and adjusted by the pilot study which resulted in an Alpha value equal to 0.88 (see section 4.4.4). Therefore, the study concluded that the assumption of the reliability of covariates in analysis of covariance was satisfactory.

As all the study samples were randomly selected from a computer (see section 4.5), an equal chance of selection was assured. Thus, independence of the study sample is assumed in this analysis. The SF-36’s developer declared that when dealing with a large sample size, the distribution of each SF-36 scale is considered Normal (Ware and Sherbourne, 1992). An assumption of normality was, therefore, assumed. An assumption of homogeneity of variance for each study sections would be checked by Levene’s test during the analysis processes.

210 6.3.2 Physical Functioning and Aircraft Noise Exposure The hypothesis for the first analysis has been set up as:

H0: There is no difference in the population mean Physical Functioning score for aircraft noise exposure group and population mean Physical Functioning score for

the control group (or α 1 = α 2).

Ha: There is a difference in the population mean Physical Functioning score for aircraft noise exposure group and population mean Physical Functioning score for the

control group (or α 1 ≠ α 2).

From the scatter plot of covariate variables and Physical Functioning, a suitable covariate variable for this first analysis was age variable. Besides the main effect independent variable, other secondary independent variables that have been considered to have clinically meaningful relevance to the dependent variable were selected and will be assessed for significance by carefully considering the meaning of Physical Functioning scale provided in Table 4.3. It is a subjective measure of the physical functioning of an individual. Subjects with a higher score of Physical Functioning imply a better performance in all types of physical activities, including the most vigorous, without limitations due to health in comparison to a person with a lower score of Physical Functioning. Prevalence of hypertension (HY), high cholesterol status (CHOL), and body mass index (BMI) were selected as candidates for the secondary independent variable for this analysis.

For the next step, those selected secondary independent variables were added (one-by- one) to the model (with aircraft noise exposure as a main effect independent variable and age as a covariate variable). The significance of secondary independent variables were checked by performing SPSS – ANALYSE - GENERAL LINEAR MODEL – UNIVARIATE ANALYSIS (or ANCOVA). Based on Tabachnick and Fidell (2001), Type I sums of squares adjusted for all previous but not following effects are used to evaluate independent variables. Table 6.5 shows the result of this process. Any secondary independent variable that provides p-value less than 0.05 will be discarded from the analysis.

211 From Table 6.5, all of the selected secondary independent variables were significant. The next step is to perform the multiple-ways (or factorial) analysis of covariance by pooling these three significant independent variable into the model (which contains aircraft noise exposure as a main effect independent variable, and age as a covariate variable). The model that best fits should consist of all significant p-values of independent variable (tested by Type I sum of squares) and covariate variable (tested by Type III sum of squares). The homogeneity of variance is also checked.

Table 6.5: Assessing the Significance of Secondary Independent Variables for Physical Functioning Type I Sum Source df Mean Square F Sig. of Squares HY 3494.09 1 3494.09 7.139 0.008 GROUP 1866.25 1 1866.25 3.813 0.051 CHOL 2001.44 1 2001.44 4.083 0.044 GROUP 1526.41 1 1526.41 3.114 0.078 BMI 6938.11 3 2312.70 4.792 0.003 GROUP 2617.99 1 2617.99 5.424 0.020

The first trial of factorial analysis of covariance (Table not shown) showed of high cholesterol status was not statistically significant in the model. The model was then refitted again after excluding the high cholesterol status. Unfortunately, even though the second trial of factorial analysis of covariance (Table not shown) was found significant for all independent variables in the model, an assumption of homogeneity of variance was violated. The p-value of Levene’s test is less than 0.05. To encounter this problem, the study decided to transform the dependent variable scores based on procedures provided by Tabachnick and Fidell (2001, table 4.3, p. 83). Refer to Table D.1, the distribution of Physical Functioning score was moderately negative. Thus, the Physical Functioning (PF) was transformed into square root format, SQRT(k - PF). The term k is a constant from which each score is subtracted so that the smallest score is one (usually k equals to the largest score plus one). In this case, the maximum value of Physical Functioning is 100, so k equals to 101.

The third trial of factorial analysis of covariance of transformed Physical Functioning was performed. This time, the significance of independent variables were met, and the assumption of homogeneity of variance was satisfactory as shown in Table 6.6. From

212 the table, there was a significant interaction between aircraft noise exposure and highest education level. However, the strength of this association was weak (η2 = 1.3%). Additional assessment of significance of covariate variable was implemented (Table not shown) by using Type III sum of squares. It was found that age is a significant covariate variable (F(1, 573) = 140.7, p-value < 0.001) of this model. The null hypothesis that there is no effect from covariate variable (slope of the regression line (β) between the dependent variable and the covariate variable equals to zero) was, therefore, rejected.

Table 6.6: Factorial Analysis of Covariance of SQRT(k - PF) and Aircraft Noise Exposure Type I Sum Eta Source df Mean Square F Sig. of Squares Squared Corrected Model 1207.50 16 75.47 15.683 0.000 0.305 Intercept 8944.39 1 8944.39 1858.765 0.000 0.764 AGE 1018.60 1 1018.60 211.679 0.000 0.270 HY 38.65 1 38.65 8.032 0.005 0.014 BMI 77.62 3 25.87 5.377 0.001 0.027 GROUP 28.72 1 28.72 5.968 0.015 0.010 HY * BMI 10.40 3 3.47 0.721 0.540 0.004 HY * GROUP 2.60 1 2.60 0.540 0.463 0.001 BMI * GROUP 14.86 3 4.95 1.029 0.379 0.005 HY * BMI * GROUP 16.06 3 5.35 1.113 0.343 0.006 Error 2757.28 573 4.81 Total 12909.17 590 Corrected Total 3964.78 589

Levene's Test of Equality of Error Variances Dependent Variable: SQRTPF F df1 df2 Sig. 1.653 15 574 0.056

The adjusted mean scores of Physical Functioning evaluated at age equals 48.33 (Table not shown) were 84.88 for the control group and 83.05 for noise exposure group. After the adjustment, the difference of Physical Functioning mean score between both groups was slightly increased from (see also Table 6.4) 0.14 (79.23 – 79.09) to 1.83 (84.88 – 83.05).

It is important to recognise that although there is a significant association between aircraft noise exposure and SQRT(k - PF), the strength of association is weak (η2 = 1.0%) (as shown in the last column, or Eta Squared, of Table 6.6). Only 1 percent of the variance in the adjusted transformed Physical Functioning score was associated with

213 aircraft noise exposure level. The study rejected the null hypothesis, and concluded that when removing the linear effects of age on Physical Functioning and controlling potential confounding factors (which are prevalence of hypertension and body mass index), the mean score of Physical Functioning of the noise exposure group was significantly lower than the control group. Health related quality of life in term of physical functioning of subject from the aircraft noise exposure group was worse than the subject from the control group.

6.3.3 General Health and Aircraft Noise Exposure The hypothesis for the second analysis has been set up as:

H0: There is no difference in the population mean General Health score for aircraft noise exposure group and population mean General Health score for the control

group (or α 1 = α 2).

Ha: There is a difference in the population mean General Health score for aircraft noise exposure group and population mean General Health score for the control group

(or α 1 ≠ α 2).

From the scatter plot of covariate variables and General Health, a suitable covariate variable for this second analysis was age variable. The General Health scale provided in Table 4.3 is a subjective evaluation of the personal health of an individual. Subjects with very low scores of General Health believe that his/her health is poor and likely to get worse. The study decided to select the prevalence of hypertension (HY), high cholesterol status (CHOL), exercise activity levels (EXER), smoking status (SMK), alcohol consumption (ALC) and body mass index (BMI) as candidates for the secondary independent variable for this analysis.

Similar to the first analysis, the secondary independent variables were assessed for significance. The process of adding and refitting was implemented. Any secondary independent variable that provided p-value less than 0.05 were discarded from the analysis as marked by the ‘X’ symbol in Table 6.7.

214 Table 6.7: Assessing the Significance of Secondary Independent Variables for General Health Type I Sum Source df Mean Square F Sig. of Squares HY 10630.26 1 10630.26 24.070 0.000 GROUP 1741.50 1 1741.50 3.943 0.047 CHOL 1564.34 1 1564.34 3.443 0.064 X GROUP 1527.73 1 1527.73 3.393 0.066 X EXER 8225.55 3 2741.85 6.092 0.000 GROUP 3238.73 1 3238.73 7.196 0.007 SMK 9717.82 2 4858.91 10.800 0.000 GROUP 1785.59 1 1785.59 3.969 0.047 ALC 2568.28 2 1284.14 2.769 0.064 X GROUP 1600.14 1 1600.14 3.450 0.064 X BMI 5991.64 3 1997.22 4.470 0.004 GROUP 2089.92 1 2089.92 4.677 0.031

Similar to the first analysis, the first trial factorial analysis of covariance was carried out by SPSS. Any secondary independent variable that was not significant in this model would be excluded. The first trial of factorial analysis of covariance (Table not shown) revealed that the smoking status variable (F(2, 427) = 2.867, p-value = 0.058) and body mass index variable (F(3, 427) = 2.34, p-value = 0.073) were not significant. After excluding these two variables from the model, the second trial of factorial analysis of covariance was then executed, as presented in Table 6.8.

Table 6.8: Factorial Analysis of Covariance of General Health and Aircraft Noise Exposure Type I Sum Eta Source df Mean Square F Sig. of Squares Squared Corrected Model 60390.98 16 3774.44 8.827 0.000 0.187 Intercept 2685703.27 1 2685703.27 6280.604 0.000 0.911 AGE 32227.14 1 32227.14 75.364 0.000 0.109 HY 11053.16 1 11053.16 25.848 0.000 0.040 EXER 7510.84 3 2503.61 5.855 0.001 0.028 GROUP 2820.52 1 2820.52 6.596 0.010 0.011 HY * EXER 1341.83 3 447.28 1.046 0.372 0.005 HY * GROUP 37.62 1 37.62 0.088 0.767 0.000 EXER * GROUP 828.52 3 276.17 0.646 0.586 0.003 HY * EXER * GROUP 4571.35 3 1523.79 3.563 0.014 0.017 Error 262985.46 615 427.62 Total 3009079.72 632 Corrected Total 323376.44 631

Levene's Test of Equality of Error Variances Dependent Variable: GH Fdf1df2Sig. 1.195 15 616 0.270

215 From Table 6.8, the main effect independent variable and the secondary independent variables are significant. There is significant interaction among prevalence of hypertension, exercise activity level, and aircraft noise exposure in the model. This interaction explained 1.7 percent of variance in the adjusted General Health score. The Levene’s test found no significant difference of variance between groups. Thus, an assumption of homogeneity of variance is satisfactory. Moreover, an additional analysis of covariate variable performed by SPSS using Type III sum of squares (Table not shown) revealed that age variable was a significant covariate variable (F(1, 615) = 33.647, p-value < 0.001) of this model. The null hypothesis that there is no effect from covariate variable (slope of the regression line (β) between the dependent variable and the covariate variable equals to zero) was therefore rejected.

Adjusted mean scores (Table not shown) of General Health were 63.41 for the control group and 60.24 for noise exposure group. After the adjustment, the difference of General Health mean score between both groups was increased from (see also Table 6.4) 1.59 (66.08 – 64.49) to 3.17 (63.41 – 60.24). The difference between adjusted means becomes significant. This leads to the conclusion that after removing the linear effects of age on General Health and controlling for significant effects of secondary covariate variables (which are prevalence of hypertension and exercise activity levels), the difference in mean scores of general health was due to the effects of long-term aircraft noise exposure. The study therefore rejected the null hypothesis, and concluded that health related quality of life in term of general health of the subject from the aircraft noise exposure group was worse than the subject from the control group.

Nevertheless, similar to the first analysis, it is noted that although there is a significant association between long-term aircraft noise exposure and general health, the strength of association is weak (η2 = 1.1%), implying that there is only 1.1 percent of the variance in the adjusted General Health score that was associated with aircraft noise exposure level.

6.3.4 Vitality and Aircraft Noise Exposure The hypothesis for the third analysis has been set up as:

216 H0: There is no difference in the population mean Vitality score for aircraft noise

exposure group and population mean Vitality score for the control group (or α1 =

α2).

Ha: There is a difference in the population mean Vitality score for aircraft noise

exposure group and population mean Vitality score for the control group (or α1 ≠

α2).

From the scatter plot of covariate variables and Vitality, two suitable covariate variables for this analysis were found, which are age variable and noise sensitivity variable. From Table 4.3, Vitality is a subjective measure of the state of an individual being strong and active and full of energy. Subjects with very high scores of Vitality are people who feel full of pep and energy all of the time. On the other hand, subjects with very low scores of Vitality feel tired and worn out all of the time. The study decided to select the secondary independent variable that comparatively reflects the above aspects. The variables included exercise activity levels (EXER), smoking status (SMK), alcohol consumption (ALC), employment status (EMP), marital status (MAR), and weekly household income (INC). The process of assessing the significance of these selected secondary independent variables was implemented as summarised in Table 6.9. The secondary independent variables with p-value less than 0.05 were discarded from the analysis as marked by the ‘X’ symbol.

Table 6.9: Assessing the Significance of Secondary Independent Variables for Vitality

Type I Sum Source df Mean Square F Sig. of Squares EXER 12968.06 3 4322.69 11.653 0.000 GROUP 2638.36 1 2638.36 7.113 0.008 SMK 855.66 2 427.83 1.774 0.171 X GROUP 871.17 1 871.17 2.276 0.132 X ALC 759.12 2 379.56 0.956 0.385 X GROUP 1105.40 1 1105.40 2.783 0.096 X EMP 14873.93 3 4957.98 13.250 0.000 GROUP 947.58 1 947.58 2.532 0.112 EDU 1559.21 2 779.60 1.994 0.137 X GROUP 1112.38 1 1112.38 2.846 0.092 X MAR 1272.65 2 636.33 1.619 0.199 X GROUP 1390.85 1 1390.85 3.539 0.060 X INC 7070.44 2 3535.22 9.134 0.000 GROUP 934.06 1 934.06 2.413 0.121

217 The first trial of factorial analysis of covariance (including aircraft noise exposure as a main effect independent variable, age variable and noise sensitivity variable as covariate variables, and significant secondary independent variables from Table 6.9) were executed by SPSS. It was found that (Table not shown) a variable of weekly household income was not significant (F(2, 536) = 1.042, p-value = 0.354), and the assumption of homogeneity of variance was also violated (p-value < 0.05). Referring to Table D.1, the distribution of Vitality score was slightly negative in both study groups: skewness = - 0.465 for noise exposure group; and skewness = -0.504 for the control group. Thus, the study decided to transform the Vitality (VT) to the square-root form, SQRT(k – VT) (Tabachnick and Fidell, 2001, table 4.3, p. 83). Similar to the first analysis, the maximum value of VT is 100, so k was replaced by 101.

After excluding weekly household income from the model and transforming the Vitality, the second trial of factorial analysis of covariance was performed. It was found that (Table not shown) all the independent variables were significant and the distribution variances were equal across the groups. However, an additional analysis of covariate variable (using Type III sum of squares) revealed that (Table not shown) age was not a significant covariate variable (F(1, 585) = 0.008, p-value = 0.928) of this model. Subsequently, the study excluded age variable from the model, and performed the third trial of factorial analysis of covariance as shown in Table 6.10.

The significance of covariate variable was re-checked. It became significant (F(1, 589) = 20.882, p-value < 0.001). The null hypothesis that there is no effect from covariate variable (slope of the regression line (β) between the dependent variable and the covariate variable equals to zero) could be rejected. The possible interactions between independent variables in the model have been checked in Table 6.10. No significant interaction between independent variables was found.

218 Table 6.10: Factorial Analysis of Covariance of SQRT(k - VT) and Aircraft Noise Exposure Type I Sum Eta Source df Mean Square FSig. of Squares Squared Corrected Model 266.23 32 8.32 3.885 0.000 0.174 Intercept 26665.44 1 26665.44 12452.720 0.000 0.955 SEN 57.52 1 57.52 26.861 0.000 0.044 EMP 106.45 3 35.49 16.571 0.000 0.078 EXER 65.38 3 21.79 10.177 0.000 0.049 GROUP 13.06 1 13.06 6.097 0.014 0.010 EMP * EXER 7.45 9 0.83 0.387 0.942 0.006 EMP * GROUP 2.20 3 0.73 0.343 0.794 0.002 EXER * GROUP 4.11 3 1.37 0.640 0.590 0.003 EMP * EXER * GROUP 10.06 9 1.12 0.522 0.859 0.008 Error 1261.25 589 2.14 Total 28192.92 622 Corrected Total 1527.48 621

Levene's Test of Equality of Error Variances Dependent Variable: SQRTVT Fdf1df2Sig. 1.03 31.00 590 0.424

Adjusted mean scores (Table not shown) of Vitality were 59.85 for the control group and 54.90 for the noise exposure group. After the adjustment, the difference in Vitality mean score between both groups was increased from (see also Table 6.4) 2.44 (57.02 – 54.58) to 4.95. The difference in adjusted mean score of Vitality between the noise exposure group and the control group becomes significant. The null hypothesis was therefore rejected. Subsequently, the study concluded that, after removing the effects from potential covariate variable and secondary independent variables, the difference in mean scores of vitality between the study groups is due to the effects of long-term aircraft noise exposure. The health related quality of life in term of vitality of subjects from the aircraft noise exposure group was worse than for the control group.

Nevertheless, similar to the previous analyses, it should be emphasised that although there is a significant association between long-term aircraft noise exposure and vitality, the strength of this association is quite weak (η2 = 1.0%), implying that there is only one percent of the variance in the adjusted Vitality score was associated with aircraft noise exposure.

219 6.3.5 Mental Health and Aircraft Noise Exposure Unlike the previous health measure scales, the unadjusted mean score of Mental Health of the noise exposure group was lower relative to the control group. Without any control for covariance effects, analysis of variance table (see Table 6.4) revealed that this difference was statistically significant (p-value < 0.001). In general, it is quite reasonable to say that long-term aircraft noise exposure has significant association with mental health. The mental health condition of a subject from the noise exposure group was worse than for the control group. However, this study still decided to apply analysis of covariance into this association because: firstly, a variable of noise sensitivity appears to have a linear relation with Mental Health (although the relationship was quite weak, see Figure F.1); and secondly, the descriptive statistic found many different aspects of demographic and socioeconomic status between both study groups. Therefore, it was reasonable that these covariance and confounding factors be controlled before drawing any inferences about aircraft noise and Mental Health.

The hypothesis for the fourth analysis has been set up as:

H0: There is no difference in the population mean Mental Health score for aircraft noise exposure group and population mean Mental Health score for the control

group (or α 1 = α 2).

Ha: There is a difference in the population mean Mental Health score for aircraft noise exposure group and population mean Mental Health score for the control group (or

α 1 ≠ α 2).

From the scatter plot between covariate variables and dependent variable, an appropriate covariate variable for this analysis was noise sensitivity variable. From Table 4.3, Mental Health is a subjective measure of the general mental health status. The highest possible score of Mental Health denotes a person who feels peaceful, happy, and calm all of the time. On the other hand, the lowest possible score of Mental Health indicates a person who feels nervous and depressed all of the time. The study decided to select secondary independent variables that comparatively reflect these above aspects. The variables include exercise activity levels (EXER), smoking status (SMK), alcohol

220 consumption (ALC), employment status (EMP), highest education level (EDU), marital status (MAR), and weekly household income (INC). Similar to the previous analyses, the process of assessing the significance of these selected secondary independent variables was implemented as summarised in Table 6.11.

Table 6.11: Assessing the Significance of Secondary Independent Variables for Mental Health Type I Sum Source df Mean Square F Sig. of Squares EXER 2594.39 3 864.80 2.923 0.033 GROUP 4842.66 1 4842.66 16.367 0.000 SMK 6629.77 2 3314.89 11.602 0.000 GROUP 3167.87 1 3167.87 11.088 0.001 ALC 2194.96 2 1097.48 3.598 0.028 GROUP 3178.24 1 3178.24 10.420 0.001 EMP 10772.71 3 3590.90 12.683 0.000 GROUP 4499.99 1 4499.99 15.893 0.000 EDU 3783.51 2 1891.75 6.543 0.002 GROUP 5049.93 1 5049.93 17.466 0.000 MAR 2121.70 2 1060.85 3.552 0.029 GROUP 3972.77 1 3972.77 13.300 0.000 INC 5128.55 2 2564.27 8.653 0.000 GROUP 3548.37 1 3548.37 11.973 0.001

As previously mentioned, the mean score of Mental Health was already significantly different between two study groups. Therefore, one of two purposes of this analysis is to check whether this difference was significantly influenced by the other secondary independent variables (such as demographic and socioeconomic status) or not. Another purpose is to find the best composition of independent variables to explain the difference of Mental Health between both study groups. The results from Table 6.11 show that each of the selected secondary independent variables were significant. However, the main effect independent variable (which is aircraft noise exposure) of each model were also significant. This might reflect a situation that not only the main effect independent variable has association with Mental Health but the composition of independent variables also has an association. The next step is to clarify the best composition of independent variables to explain Mental Health.

The trials of factorial analysis of covariance which included aircraft noise exposure as the main effect independent variable, noise sensitivity variable as the covariate variable, and a series of tree-diagram of secondary independent variables (for instance, EXER +

221 SMK, EXER + ALC, EMP + EDU + MAR, EMP + EDU + INC, EXER + SMK + ALC + EMP, EXER + SMK + ALC + EDU, and EXER + SMK + ALC + EMP + EDU) were performed by SPSS (Tables not shown). A model would be selected if it provided: (1) no violation of all analysis of covariance’s assumptions; (2) significance of independent variables and covariate variable; and (3) the highest strength of association between aircraft noise exposure and Mental Health.

Table 6.12 presents the analysis of covariance table of the model that achieved the above criteria. The model consists of aircraft noise exposure as the main effect independent variable, noise sensitivity as the covariate variable, and exercise activity levels, smoking status, and highest education level as the secondary independent variables. The distribution of variance was equal across the group (F(66, 538) = 1.246, p- value = 0.101). Thus, the assumption of homogeneity of variance was accomplished. No significant interaction between independent variables was found. Additional assessment of the significance of covariate variable was performed (Table not shown) and found noise sensitivity variable as a significant covariate variable (F(1, 537) = 30.14, p-value < 0.001) variable of this model. The null hypothesis that there is no effect from covariate variable was rejected.

Adjusted mean scores of Mental Health evaluated at noise sensitivity equal 27.45 (Table not shown) were 71.42 for the control group and 67.37 for the noise exposure group. After the adjustment, the difference of Mental Health mean score between both groups was decreased from (see also Table 6.4) 5.51 (73.53 – 68.02) to 4.05 (71.42 – 67.37). The gap of Mental Health due to the main effect independent variable was reduced because the effects from the secondary independent variables. Nevertheless, the difference of Mental Health due to aircraft noise exposure was still significant.

222 Table 6.12: Factorial Analysis of Covariance of Mental Health and Aircraft Noise Exposure Type I Sum Eta Source df Mean Square F Sig. of Squares Squared Corrected Model 46661.34 67 696.44 2.550 0.000 0.241 Intercept 3027304.16 1 3027304.16 11083.852 0.000 0.954 SEN 9510.34 1 9510.34 34.820 0.000 0.061 EXER 2423.90 3 807.97 2.958 0.032 0.016 SMK 7053.19 2 3526.60 12.912 0.000 0.046 EDU 2357.47 2 1178.74 4.316 0.014 0.016 GROUP 4148.78 1 4148.78 15.190 0.000 0.028 EXER * SMK 2916.06 6 486.01 1.779 0.101 0.019 EXER * EDU 976.31 6 162.72 0.596 0.734 0.007 SMK * EDU 1490.00 4 372.50 1.364 0.245 0.010 EXER * SMK * EDU 4686.91 12 390.58 1.430 0.148 0.031 EXER * GROUP 407.71 3 135.90 0.498 0.684 0.003 SMK * GROUP 1598.45 2 799.22 2.926 0.054 0.011 EXER * SMK * GROUP 2704.06 6 450.68 1.650 0.131 0.018 EDU * GROUP 1952.74 2 976.37 3.575 0.029 0.013 EXER * EDU * GROUP 914.86 6 152.48 0.558 0.764 0.006 SMK * EDU * GROUP 861.63 3 287.21 1.052 0.369 0.006 EXER * SMK * EDU * GROUP 2658.95 8 332.37 1.217 0.287 0.018 Error 146669.44 537 273.13 Total 3220634.94 605 Corrected Total 193330.78 604

Levene's Test of Equality of Error Variances Dependent Variable: MH Fdf1df2Sig. 1.246 66 538 0.101

Subsequently, it was decided to reject the null hypothesis, and concluded that, after removing the linear effect of noise sensitivity and taking into account the effects from exercise activity levels, smoking status, and highest education level, the difference of mean scores of mental health between study groups is due to the effects of long-term aircraft noise exposure. The health related quality of life in term of mental health condition of subjects from the aircraft noise exposure group was worse than for the control group.

Finally, similar to the previous analyses, it should be recognised that although there is a significant association between long-term aircraft noise exposure and mental health, the strength of this association is still weak (η2 = 2.8%), implying that there is only 2.8 percent of the variance in the adjusted Mental Health score was associated with aircraft noise exposure.

223 6.4 SUMMARY, DISCUSSIONS AND CONCLUDING REMARKS

This chapter has explored the first core research question (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”) based on the health and well-being data surveyed from the community surrounding Sydney (Kingsford Smith) Airport and the control community. The questionnaire measured four different aspects of health related quality of life by using a well-developed questionnaire (called the Short Form Health Survey, SF-36) developed by Ware and Sherbourne (1992).

The SF-36 is a multipurpose scale consisting of eight health measure scales, which are Physical Functioning, Role Physical, Body Plain, General Health, Vitality, Social Functioning, Role Emotional, and Mental Health (see section 4.3.1.1). These eight scales are summarised by two composite scales known as: Physical Health Composite Scale; and Mental Health Composite Scale. The former scale aggregates Physical Functioning, Role Physical, Body Plain, and General Health, while the latter combines Vitality, Social Functioning, Role Emotional, and Mental Health. Due to the limitation of questionnaire length and the relevance of scale to the context of the research topic, this thesis adapted only four scales of SF-36 which are Physical Functioning, General Health, Vitality, and Mental Health.

The composite scale of physical health and mental health could not be calculated. Thus, the study assumed that these four selected SF-36 scales measures independently, aspects of health related quality of life. In this case, the application of multivariate analysis of covariance is not practical. This is because it is suitable to a situation in which there are several dependent variables, and each dependent variable measures different dimension which attributes to the target variable. Therefore, it was decided to separately analyse the mean difference of selected SF-36 scales between study groups using the analysis of covariance (two-tailed test) technique. All the analyses were performed by SPSS version 12. While it was recognised that a one-tailed test of analysis of covariance could be used given the hypothesis relating to the detrimental effect of aircraft noise, a two- tailed test was selected because it is considered more conservative in term of statistical significance. Given that all the analysis of covariance’s results was statistically

224 significant and the same direction hypothesised, they would have also been significant in one tailed test.

The practical limitations (or assumptions) of analysis of covariance (see Appendix E) were evaluated before the beginning of analyses. All the assumptions were satisfactory (see section 6.3.1). The assumption of homogeneity of variance was tested during the analysis process of each sub-section. The null hypotheses of each sub-section assumed no difference in the population mean score of each health related quality of life measure between aircraft noise exposure group and the control group. The main effect independent variable was aircraft noise exposure.

For the first sub-section (see section 6.3.2), factorial analysis of covariance was performed on Physical Functioning score (PF). The potential secondary independent variables (or confounding factors) consisted of prevalence of hypertension and body mass index. The significant covariate was age. The assumption of homogeneity of variance was violated. Thus, the Physical Functioning was transformed into SQRT(k – PF) format. After adjustment by potential confounding factors and significant covariates, Physical Functioning varied significantly with aircraft noise exposure, as summarised in Table 6.6, with F(1, 573) = 5.97, p-value = 0.015, η2 = 1.0%. The adjusted mean scores of Physical Functioning were 84.88 for the control group and 83.05 for noise exposure group. The study rejected the null hypothesis, and concluded that when removing the linear effects of age on Physical Functioning and controlling for potential confounding effects of prevalence of hypertension and body mass index, the adjusted mean score of Physical Functioning of the aircraft noise exposure group was significantly lower than the control group. This implies that health related quality of life in term of the physical functioning of subject from aircraft noise exposure group was worse than the control group.

For the second sub-section (see section 6.3.3), factorial analysis of covariance was performed on General Health score. The potential confounding factors consisted of prevalence of hypertension and exercise activity levels. The significant covariate was age. The assumption of homogeneity of variance was satisfactory. After adjustment by potential confounding factors and significant covariates, General Health varied

225 significantly with aircraft noise exposure, as summarised in Table 6.8, with F(1, 615) = 6.60, p-value = 0.010, η2 = 1.1%. Adjusted mean scores of General Health were 63.41 for the control group and 60.24 for noise exposure group. The study rejected the null hypothesis, and concluded that when removing the linear effects of age on General Health and controlling for potential confounding effects of prevalence of hypertension and exercise activity levels, the adjusted mean score of General Health of aircraft noise exposure group was significantly lower than the control group. This implies that health related quality of life in term of the general health of subject from aircraft noise exposure group was worse than the control group.

For the third sub-section (see section 6.3.4), factorial analysis of covariance was performed on Vitality score (VT). The potential confounding factors consisted of employment status and exercise activity levels. The significant covariate was noise sensitivity. The assumption of homogeneity of variance was violated. Thus, the Vitality was transformed into SQRT(k – VT) format. After adjustment by potential confounding factors and significant covariates, Vitality varied significantly with aircraft noise exposure, as summarised in Table 6.10, with F(1, 573) = 6.10, p-value = 0.014, η2 = 1.0%. Adjusted mean scores of Vitality were 59.85 for the control group and 54.90 for the noise exposure group. The study rejected the null hypothesis, and concluded that when removing the linear effects of noise sensitivity on Vitality and controlling for potential confounding effects of employment status and exercise activity level, the adjusted mean score of Vitality of aircraft noise exposure group was significantly lower than the control group. This implies that health related quality of life in term of the sense of vitality of subject from aircraft noise exposure group was worse than the control group.

For the fourth sub-section (see section 6.3.5), factorial analysis of covariance was performed on Mental Health score. The potential confounding factors consisted of exercise activity levels, smoking status, highest education level. The significant covariate was noise sensitivity. The assumption of homogeneity of variance was satisfied. After adjustment by potential confounding factors and significant covariates, Mental Health varied significantly with aircraft noise exposure, as summarised in Table 6.12, with F(1, 537) = 4.32, p-value < 0.001, η2 = 2.8%. Adjusted mean scores of

226 Mental Health were 71.42 for the control group and 67.37 for the noise exposure group. The study rejected the null hypothesis, and concluded that when removing the linear effects of noise sensitivity on Mental Health and controlling for potential confounding effects of exercise activity levels, smoking status, highest education level, the adjusted mean score of Mental Health of aircraft noise exposure group was significantly lower than the control group. This implies that health related quality of life in term of the mental health of subject from aircraft noise exposure group was worse than the control group.

On the basis of what was summarised above, the concluding remarks were drawn as following:

• After controlling for age, prevalence of hypertension, and body mass index, health related quality of life in term of Physical Functioning of subjects from the aircraft noise exposure area was worse than the control group.

• After controlling for age, prevalence of hypertension, and exercise activity levels, health related quality of life in term of General Health of subjects from the aircraft noise exposure area was worse than the control group.

• After controlling for noise sensitivity, employment status, and exercise activity level, health related quality of life in term of Vitality of subjects from the aircraft noise exposure area was worse than the control group.

• After controlling noise sensitivity, exercise activity levels, smoking status, and highest education level, health related quality of life in term of Mental Health of subjects from the aircraft noise exposure area was worse than the control group.

At the time of writing this thesis, there was only one published article by Meister and Donatelle (2000) study the impacts of aircraft noise on health related of life by using an equivalent technique to that employed in this thesis. Even though the detail analysis of covariate variable and confounding factor has not been clearly provided, Meister and Donatelle (2000) concluded that after control for potential confounding factors all

227 generic health measures (which were General Health, Vitality, and Mental Health) were significantly worse in the neighbourhoods exposed to commercial-aircraft noise. The present study improved the study of Meister and Donatelle (2000) by (1) including more potential confounding factors into the analysis, such as exercise activity, nutrition, employment status, and prevalence of hypertension; (2) selecting the control area based on standardised socio-economic status indices; (3) studying the impacts of aircraft noise on Physical Functioning; and (4) preventing attitude bias toward noise source. The results of the analysis described in this thesis strongly support what was found by Meister and Donatelle (2000).

Recently studies by Bronaft (1998), Miyakita et al (2002), and Franssen et al (2005) appear to support the result of this thesis and Meister and Donatelle (2000) even though the analysis methods and health indicators are different. Bronaft (1998) found that subjects who were bothered by aircraft noise were more likely to perceive themselves to be in poorer health status. Miyakita et al (2002) found that residents living around military airfields were suffered from physical (vague complaints, respiratory, and digestive) and mental (mental instability, depression and nervousness) effects as a result of exposure to military aircraft noise and that such responses increase with the level of noise exposure. Franssen et al (2005) found that vitality related health complaints, such as tiredness and headache, and general health status (“How is your health in general?”) were associated with aircraft noise exposure.

A disadvantage of this study may be the lack of non-response analysis, especially in Non-English speaking sample (Greek and Arabic). The return rate of questionnaire of Non-English languages was very low (see Table 6.2). This might due to many factors. For example, the less concern (or less annoyance) about noise pollution of Non-English speaking people. They may also suspicious about filling in such questionnaires. Non- English speaking at home may prefer to answer the questionnaire in English rather than the provided Non-English languages. The contribution from some mistake, or error, in the translated health survey instruments due to the lack of ‘back translation’ process. Two constraints that hampered the non-response analysis were research budget and the confidentiality that the present research has committed to the subjects and the UNSW Human Research Ethics Committee that “Anyone in this house can participate the

228 survey. There is no identification of individual on the questionnaire” (see Figure 4.4). Nonetheless there was no sign of major distortion effects on the research results due to the lack of non-response analysis. Moreover from Table 2.2, most of the previous studies did not implement a non-response analysis except Franssen et al (2005).

Finally, an answer for the first core research question would be: “Health related quality of life, in term of physical functioning, general health, vitality, and mental health, of community chronically exposed to high aircraft noise level were worse than the control area”.

229 CHAPTER SEVEN

AIRCRAFT NOISE AND HYPERTENSION

7.1 INTRODUCTION

This chapter is concerned with the impacts of aircraft noise on adult blood pressure level. The main objective of this chapter is to explore the second core research question: “Does long-term aircraft noise exposure associate with hypertension via chronic noise stress as a mediating factor?”, as has been set up by section 2.6. To achieve this objective, this research has established a major assumption stating that “Long-term aircraft noise exposure has an indirect impact on blood pressure level. Aircraft noise annoys people by disturbing their daily activities and then creating chronic noise stress which after control for noise sensitivity and some potential confounding factors finally becomes a mediating factor of high blood pressure level in the future” This assumption could be illustrated by Figure 7.1 as follows:

Confounder(s)

Background Noise Daily activities Aircraft Noise High Blood Annoyance Noise disturbance Stress Pressure

Noise Sensitivity

Figure 7.1: Assumption of the Second Core Research Question

In Figure 7.1, it is suggested that aircraft noise potentially disturbs community daily activities. Everyday disturbance from aircraft noise causes annoyance. Differences in background environmental noise conditions that might influence community response toward aircraft noise are controlled by applying the Noise Gap Index. It has been shown by Weinstein (1978), Stansfeld (1992), and van Kamp et al (2004) that levels of

230 annoyance are influenced by personal attitude to noise, or noise sensitivity. Noise sensitive persons will be easily annoyed with a certain noise level considered by normal person just to be an acceptable level. Annoyance from unavoidable noise such as aircraft noise can be a good cause of emotional stress. The present research assumed that stress due to noise has two-way correlation with noise sensitivity. A person who is susceptible to noise might have a low capability to cope with a noise stimulus leading them to become more stressed than normal. On the other hand, person who has been suffered by stress due to noise may become less tolerable to unwanted sound than person who is more mentally calm.

As mentioned previously in this thesis, it has been shown that emotional stress can cause high blood pressure. This thesis also assumed that hypertensive people themselves may be easily stressed by noise than “normal” people. Moreover, this thesis assumed that persons who have high blood pressure levels are more sensitive to noise than normal. The concept of confounder has been explained by Section 4.3.1.6.

The description of variables (including variable coding) involved in the analysis of the two core research questions is provided in Appendix C. Appendix G explains in detail the general purposes and fundamental concepts of the logistic regression model. It also provides an example of logistic regression analysis, the modelling strategy, and the method to assess the fit of the model. Section 7.2 explores the second core research question by using logistic regression analysis. Section 7.3 summarises the results and addresses concluding remarks.

7.2 EXPLORING THE SECOND CORE RESEARCH QUESTION

Appendix G explained how to implement logistic regression analysis, guiding the strategy to select the best variable to the model, and assesses the fit of the model. This section assesses an association between aircraft noise and hypertension based on all the techniques mentioned by the above sections. Referring back to the set up of alternative research assumption that “Aircraft noise has indirect impacts to hypertension, it disturbs daily activities and creates noise stress which becomes a mediating factor for

231 hypertension in the future”, the study therefore divides logistic regression analysis into two parts.

First of all, it is important to note that the distribution of noise stress is originally a continuous variable (which is represented by STR). However, this analysis is dealing with logistic regression analysis when the dependent variable (or the disease variable) must be a dichotomous variable (presence or absence). Thus, the study decided to transform the STR to a dichotomous variable. From Figure D.1 in Appendix D, most of the subjects expressed their noise stress level in range of two to six. Only a few subjects were beyond this range (STR > 6). Since it was controlled for aircraft noise and other major non-aviation noise sources, subjects from the control group were expected to have a low level of noise stress. Thus, this study assumed that subjects who were not highly stressed by noise would select STR lower than seven. Consequently, STR equals seven was selected as a cut-off point for a dichotomous variable of chronic noise stress.

The first part analyses association between aircraft noise exposure and chronic noise stress. The chronic noise stress (which is denoted by STR_7) has been set as a dependent variable. It is dichotomous variable which ‘0’ represents the subject who has no chronic noise stress (STR < 7) and ‘1’ represents the subject who has chronic noise stress (STR ≥ 7). The aircraft noise exposure (GROUP) variable has been set as an exposure variable (E) which ‘1’ represents the subject from the aircraft noise exposure areas and ‘0’ represents the subject from the aircraft noise non-exposure area (or the control area). The hypothesis of the first analysis has been set up as:

H0: There is no association between aircraft noise exposure and chronic noise stress.

Ha: There is an association between aircraft noise exposure and chronic noise stress.

The second part of the core question explores concerns with the correlation between chronic noise stress and prevalence of hypertension. The prevalence of hypertension has been set as a dependent variable when ‘1’ represents hypertension and ‘0’ represents normotensive, and the chronic noise stress becomes an exposure variable. Similarly with the first analysis, the hypothesis of the second analysis has been set up as:

232 H0: There is no association between chronic noise stress and prevalence of hypertension.

Ha: There is an association between chronic noise stress and prevalence of hypertension.

All the categorical variables have been recoded into a series of dichotomous variables as shown by Table 7.1.

Table 7.1: Categorical Variables Coding Parameter coding Variable Measure (1) (2) (3) current smoker 1 0 SMK ex-smoker 0 1 never smoked 0 0 obesity 1 0 0 overweight 0 1 0 BMI acceptable weight 0 0 1 underweight 0 0 0 high exercise 1 0 0 moderate exercise 0 1 0 EXER low exercise 001 sedentary 0 0 0 white collar 1 0 0 blue collar 0 1 0 EMP unemployed 0 0 1 not in labour force 0 0 0 high education 1 0 EDU medium education 0 1 low education 0 0 married/de facto 1 0 MAR widowed/divorded/separated 0 1 never married 0 0 high income 1 0 INC medium income 0 1 low income 0 0 high alcohol 1 0 ALC low alcohol 0 1 never consumed 0 0

7.2.1 Exploring the First Hypothesis: Aircraft Noise and Noise Stress

7.2.1.1 Modelling and Assessing the Fit The first step of analysis begins with fitting the univariable logistic regression models as shown by Table 7.2. Any independent variable with p-value > 0.25, and not clinically

233 meaningful, will be discarded from the multivariate analysis. Those variables are marked by the ‘X’ symbol as shown in Table 7.2.

Table 7.2: Univariable Logistic Regression Models of the First Analysis

Odds 95% CI Symbol Coeff. Std. Err. G p-value Ratio Lower Upper GROUP 1.923 0.217 6.844 4.475 10.465 97.376 <0.001 PARHY 0.076 0.177 1.079 0.763 1.527 0.187 0.666 X CHOL 0.281 0.227 1.325 0.849 2.069 1.498 0.221 EMP EMP(1) -0.074 0.199 0.929 0.629 1.372 5.184 0.159 EMP(2) -0.586 0.280 0.557 0.321 0.965 EMP(3) 0.057 0.410 1.058 0.474 2.362 EXER EXER(1) -0.336 0.306 0.714 0.392 1.300 2.068 0.558 X EXER(2) 0.082 0.261 1.086 0.650 1.813 X EXER(3) -0.061 0.234 0.941 0.595 1.488 X PARSMK 0.226 0.214 1.254 0.824 1.908 1.097 0.295 X SMK SMK(1) 0.711 0.237 2.037 1.281 3.240 10.178 0.006 SMK(2) 0.445 0.207 1.561 1.041 2.341 ALCO ALCO(1) 0.367 0.292 1.443 0.814 2.558 1.948 0.378 X ALCO(2) 0.043 0.229 1.044 0.667 1.637 X NUTRI 0.457 0.217 1.579 1.033 2.414 4.322 0.038 SEN 0.103 0.014 1.109 1.079 1.139 65.932 <0.001 ANNOAIR 0.418 0.034 1.519 1.422 1.622 226.827 <0.001 ANNOTRAF 0.331 0.038 1.393 1.293 1.500 86.090 <0.001 NGI 0.147 0.020 1.158 1.113 1.204 56.277 <0.001 INSUL 0.893 0.206 2.442 1.632 3.655 18.391 <0.001 SEX 0.129 0.180 1.138 0.799 1.620 0.510 0.475 X AGE -0.004 0.006 0.996 0.985 1.008 0.420 0.517 X BMI BMI(1) -0.683 0.365 0.505 0.247 1.033 4.292 0.232 BMI(2) -0.715 0.350 0.489 0.246 0.971 BMI(3) -0.636 0.344 0.530 0.270 1.039 EDU EDU(1) 0.309 0.222 1.362 0.881 2.105 7.694 0.021 EDU(2) -0.317 0.209 0.728 0.483 1.098 MAR MAR(1) -0.211 0.226 0.810 0.520 1.261 1.891 0.389 X MAR(2) -0.398 0.292 0.671 0.379 1.190 X INC INC(1) -0.613 0.385 0.542 0.255 1.153 2.715 0.257 X INC(2) -0.192 0.215 0.825 0.541 1.258 X

Table 7.3 shows the results of fitting a multivariable model containing the covariates significant at p-value 0.25 level in Table 7.2. In this step, the covariates with a p-value less than 0.05 and not clinically meaningful will be discarded (as marked by the ‘X’ symbol). Consequently, there are three covariances (which are noise sensitivity (SEN), aircraft noise annoyance (ANNOAIR) and traffic noise annoyance (ANNOTRAF)) plus one exposure variable (GROUP) remaining in the model. Consequently, the multivariable model has been refitted and currently provides the deviance (called -2 log- likelihood value) of 489.922 with 5 degree of freedom (Table not shown). The variables not selected for the original multivariable model are adding (one-by-one), refitting, and deleting until the model contains the most essential (and significant) variable and the best fit. No additional significant variable was found (Table not shown).

234 Table 7.3: Results of Fitting a Multivariable Model of the First Analysis

Odds 95% CI Variable Coeff. Std. Err. Wald df Sig Ratio Lower Upper GROUP 0.546 0.581 0.883 1.000 0.347 1.726 0.553 5.391 CHOL -0.555 0.365 2.316 1.000 0.128 0.574 0.281 1.173 X EMP 3.469 3.000 0.325 X EMP(1) -0.084 0.326 0.067 1.000 0.796 0.919 0.486 1.74 X EMP(2) -0.741 0.415 3.193 1.000 0.074 0.477 0.211 1.074 X EMP(3) -0.272 0.606 0.202 1.000 0.653 0.762 0.232 2.499 X SMK 6.327 2.000 0.042 X SMK(1) 0.760 0.335 5.144 1.000 0.023 2.138 1.109 4.123 X SMK(2) 0.540 0.293 3.402 1.000 0.065 1.716 0.967 3.045 X NUTR 0.249 0.320 0.603 1.000 0.437 1.282 0.685 2.401 X SEN 0.072 0.019 14.455 1.000 0.000 1.075 1.036 1.115 ANNOTRAF 0.225 0.055 16.645 1.000 0.000 1.252 1.124 1.394 ANNOAIR 0.295 0.054 30.109 1.000 0.000 1.343 1.209 1.492 NGI -0.025 0.054 0.209 1.000 0.647 0.976 0.878 1.084 X INSULA 0.304 0.307 0.981 1.000 0.322 1.356 0.742 2.475 X BMI 1.958 3.000 0.581 X BMI(1) -0.498 0.503 0.982 1.000 0.322 0.608 0.227 1.627 X BMI(2) -0.672 0.482 1.948 1.000 0.163 0.510 0.198 1.313 X BMI(3) -0.481 0.472 1.041 1.000 0.308 0.618 0.245 1.558 X EDU 1.867 2.000 0.393 X EDU(1) -0.294 0.379 0.601 1.000 0.438 0.746 0.355 1.566 X EDU(2) -0.447 0.328 1.861 1.000 0.173 0.640 0.336 1.216 X Constant -4.763 0.776 37.697 1.000 0.000 0.009

All the covariates appearing in the preliminary main effects model are considered continuous variables. The linearity of noise sensitivity, aircraft noise annoyance, and traffic noise annoyance variable are tested by two methods (the graphical method and the analytical method) as presented by the following paragraphs.

The first, second, and third quartile of noise sensitivity are 22, 27, and 33, respectively. The variable of noise sensitivity ranges from 4 to 45. Thus, the quartile midpoints of noise sensitivity are 13, 24.5, 30, and 39. The variable of noise sensitivity is then transformed into a categorical variable based on the quartile values. The preliminary main effects model replacing the continuous variable of noise sensitivity with those newly created categorical variables is refitted. The estimated coefficients of the categorical variable of noise sensitivity are plotted against the midpoint of each group as presented by Figure 7.2.

235 1.2

1.0

0.8 f 0.6 Coef 0.4

0.2

0.0 10 15 20 25 30 35 40 45 SENSITIVITY

Figure 7.2: Plot of Estimated Coefficients of Noise Sensitivity vs. the Quartile Midpoints

Figure 7.2 shows a gradual increase followed by a rapid increase of the coefficient of noise sensitivity. The results seem to support linearity in the logit for noise sensitivity. However, for a more confident conclusion, the fractional polynomial analysis (called an analytical technique) was therefore applied and presented in Table 7.4. It is obvious that the best non-linear transformations of noise sensitivity are not significantly different from the linear model and thus the fractional polynomial analysis supports treating the sensitivity score as linear in the logit.

Table 7.4: Fractional Polynomials for Noise Sensitivity df Deviance G for Model vs. Linear p-value Powers Not in model 0 508.22 Linear 1 489.92 0.000* 1 J = 1 2 488.66 1.26 0.261** 3 J = 2 4 487.77 2.15 0.542*** -2, 3 * Compare the linear model to the model without SEN ** Compare the J = 1 model to the linear model *** Compare the J = 2 model to the linear model

The first, second, and third quartile of aircraft noise annoyance scale are 0, 3, and 7, respectively. The aircraft noise annoyance variable ranges from 0 to 10. Thus, the quartile midpoints of aircraft noise annoyance are 0, 1.5, 5, and 8.5. The aircraft noise annoyance variable is then transformed into a categorical variable based on the quartile values. The preliminary main effects model replacing the continuous variable of aircraft noise annoyance with those new created categorical variables is refitted. The estimated

236 coefficients of the categorical variable of aircraft noise annoyance are plotted against the midpoint of each group as presented by Figure 7.3.

2.0

1.5

1.0 f 0.5 Coef 0.0 0246810 -0.5

-1.0 ANNOAIR

Figure 7.3: Plot of Estimated Coefficients of Aircraft Noise Annoyance (ANNOAIR) vs. the Quartile Midpoints

Figure 7.3 shows a decrease in slope from the first to the second quartile, but followed by a constant increase of the aircraft noise annoyance coefficient. The results do not seem linear in the logit for aircraft noise annoyance. Moreover, the fractional polynomial analysis (see Table 7.5) reveals that the best non-linear transformations of aircraft noise annoyance are significantly different (p-value < 0.05) from the linear model. Therefore, it was decided to transform the continuous variable of aircraft noise annoyance to a dichotomous variable. The question arose of what is the most suitable cut-off point? From Figure D.1 in Appendix D, the distribution of aircraft noise annoyance in the noise exposure group was noticeably divided into two parts: aircraft noise annoyance score less than seven and aircraft noise annoyance score higher or equal seven. It appears that subjects who felt not highly annoyed by aircraft noise would select a range of aircraft noise annoyance from zero to less than seven, and subjects who were highly annoyed by aircraft noise would choose aircraft noise annoyance scale from seven to ten. Thus from this evidence, it was decided to use aircraft noise annoyance score equal seven as a cut-off point. This transformed dichotomous variable is denoted by ANNOAIR_7.

237 Table 7.5: Fractional Polynomials for Aircraft Noise Annoyance (ANNOAIR) df Deviance G for Model vs. Linear p-value Powers Not in model 0 535.10 Linear 1 489.92 0.000* 1 J = 1 2 481.12 8.80 0.003** 2 J = 2 4 479.84 10.09 0.018*** 0.5, 2 * Compare the linear model to the model without ANNOAIR ** Compare the J = 1 model to the linear model *** Compare the J = 2 model to the linear model

The first, second, and third quartile of traffic noise annoyance are 0, 1, and 4, respectively. The variable of traffic noise annoyance ranges from 0 to 10. Thus, the quartile midpoints of traffic noise annoyance are 0, 0.5, 2.5, and 7. The variable of traffic noise annoyance is then transformed into a categorical variable based on the quartile values. The preliminary main effects model replacing the continuous variable of traffic noise annoyance with those new created categorical variables is refitted. The estimated coefficients of categorical variable of traffic noise annoyance are plotted against the midpoint of each group as presented by Figure 7.4.

1.2

1.0

0.8 f 0.6 Coef 0.4

0.2

0.0 0246810 ANNOTRAF

Figure 7.4: Plot of Estimated Coefficients of Traffic Noise Annoyance (ANNOTRAF) vs. the Quartile Midpoints

Figure 7.4 shows a constant increase of slope of the traffic noise annoyance coefficient. The plot is considered linear. The fractional polynomial analysis (see Table 7.6) of traffic noise annoyance demonstrates that the best non-linear transformations of traffic noise annoyance are not significantly different from the linear model. It was decided to keep the traffic annoyance score linear in the logit.

238 Table 7.6: Fractional Polynomials for Traffic Noise Annoyance df Deviance G for Model vs. Linear p-value Powers Not in model 0 506.69 Linear 1 489.92 0.000* 1 J = 1 2 489.83 0.09 0.765 2 J = 2 4 488.67 1.25 0.741 2, 3 * Compare the linear model to the model without ANNOTRAF ** Compare the J = 1 model to the linear model *** Compare the J = 2 model to the linear model

The new model is refitted (which yields deviance value equals 484.583 with df = 6) as shown in Table 7.7. This model is now referred to as the main effects model. It is important to compare both the coefficient and significance of the Wald test for all variables before and after the alteration of aircraft noise annoyance variable. Note that, there are no noticeable changes to the coefficient and significance of the Wald test of all variables, except aircraft noise exposure. The aircraft noise exposure coefficient was doubled and the Wald test became significant. The transformation of aircraft noise annoyance improves the significance of aircraft noise exposure.

Now it is time to check for any possible interaction between variables in the main effects model. Each possible interaction (as presented by the first column of Table 7.8) was added into the model one at a time. The deviance, G statistic, degrees of freedom, p-value, and Wald statistic (z) of each possible interaction were calculated. There is only one interaction (ANNOAIR_7 × ANNOTRAF) that is significant in both p-value and Wald statistic (z). Therefore, this significant interaction was added into the main effects model. The fit of this model yields significance in the Wald statistic (z) for all variables. The final preliminary model can be established as presented by Table 7.9. The model is now ready for the goodness-of-fit test.

Table 7.7: Main Effects Model of the First Analysis

Odds 95% CI Variable Coeff. Std. Err. Wald df Sig Ratio Lower Upper GROUP 0.921 0.305 9.129 1 0.003 2.512 1.382 4.567 SEN 0.065 0.017 14.954 1 0.000 1.067 1.032 1.103 ANNOTRAF 0.224 0.047 22.477 1 0.000 1.251 1.14 1.372 ANNOAIR_7 1.918 0.282 46.214 1 0.000 6.809 3.916 11.837 Constant -4.830 0.561 74.035 1 0.000 0.008

239 Table 7.8: Interaction Test for the Main Effects Model of the First Analysis

Wald Interaction Deviance G df p-value statistic Main Effects Model 484.583 5 GROUP SEN 484.519 0.064 6 0.800 0.800 GROUP ANNOAIR_7 482.673 1.910 6 0.167 0.169 GROUP ANNOTRAF 480.994 3.589 6 0.058 0.061 SEN ANNOAIR_7 484.543 0.040 6 0.841 0.840 SEN ANNOTRAF 484.571 0.012 6 0.913 0.909 ANNOAIR_7 ANNOTRAF 476.872 7.711 6 0.005 0.005

Table 7.9: Preliminary Final Model of the First Analysis

Odds 95% CI Variable Coeff. Std. Err. Wald df Sig Ratio Lower Upper GROUP 0.958 0.312 9.455 1 0.002 2.608 1.416 4.804 SEN 0.057 0.017 11.347 1 0.001 1.059 1.024 1.094 ANNOTRAF 0.376 0.073 26.537 1 0.000 1.456 1.262 1.680 ANNOAIR_7 2.699 0.408 43.799 1 0.000 14.872 6.686 33.079 ANNOAIR_7 by ANNOTRAF -0.257 0.092 7.757 1 0.005 0.773 0.645 0.927 Constant -5.062 0.583 75.423 1 0.000 0.006

The value of the Hosmer-Lemeshow goodness-of-fit statistic computed from the contingency table in Table 7.10 is 3.92 and the p-value computed from the chi-square distribution with 8 degree of freedom is 0.865. This indicates that the model is a good fit. A comparison of the observed and expected frequencies in each cell in Table 7.10 also supports this conclusion. Tabachnick and Fidell (2001, p.538-539) have suggested that a well-fitted logistic regression model should have high observed frequencies with outcome 1 in the high deciles of risk and most with outcome 0 in the low deciles of risk. Table 7.10 satisfies this recommendation. Finally, the fit model of the logit of association between aircraft noise and noise stress can be written as in equation (7.1). This model could be referred as the final model.

240 Table 7.10: Contingency Table of Hosmer and Lemeshow Test and Chi-Square Value of the First Analysis Decile of STR_7 = no STR_7 = yes Total Risk Observed Expected Observed Expected 1 63 62.82 1 1.18 64 2 58 59.17 3 1.83 61 3 60 60.31 3 2.69 63 4 61 59.34 2 3.66 63 5 59 57.55 4 5.45 63 6 53 53.39 10 9.61 63 7 42 43.17 21 19.83 63 8 24 27.04 39 35.96 63 9 24 20.20 40 43.80 64 10 13 14.00 51 50.01 64

Chi-square df Sig. 3.92 8 0.865

^ g(x) = (-5.062) + (0.958)GROUP + (0.057)SEN + (0.376)ANNOTRAF + (2.699)ANNOAIR_7 + (-0.257)(ANNOAIR_7)(ANNOTRAF) (7.1) where GROUP is an aircraft noise exposure variable (1 = aircraft noise exposure group, 0 = control group) SEN is a noise sensitivity variable (range of 0 – 45 point) ANNOTRAF is a traffic noise annoyance variable (range of 0 – 10 point) ANNOAIR_7 is an aircraft noise annoyance variable higher or equal seven (1 = yes, 0 = no).

7.2.1.2 Calculation of Odds Ratio and Confidence Interval

Since there is no potential effect modifier (Wj) in the final model, the adjusted odds ratio when Wj=0 is simply a natural logarithm of the estimated coefficient of exposure variable which can be expressed as follows:

^ Adjusted odds ratio = exp ( β1 ) = exp (0.958) = 2.61

^ ^ ^ The 95% confidence interval can be estimated as: exp[ β i ± Z1−α / 2 SE(β i ) ] = exp[ 0.958 ± 1.96 (0.312)], which yields (1.42, 4.80). The odds ratio adjusted for significant

241 covariate variables and interaction is significant because the 95% confidence interval does not include one in its range.

7.2.1.3 Interpretation It is reasonable to conclude that there is some significant association between long-term aircraft noise exposure and chronic noise stress. The established logistic regression model between aircraft noise and chronic noise stress is a good fit. The study rejected the null hypothesis and accepted the alternative hypothesis that long-term aircraft noise exposure has a significant association with chronic noise stress. After controlling for noise sensitivity, traffic noise annoyance, aircraft noise annoyance, and interaction between traffic noise annoyance and aircraft noise annoyance, subjects from the aircraft noise exposure area have the odds of 2.61 of having chronic noise stress compared with subjects from the control area.

7.2.2 Exploring the Second Hypothesis: Noise Stress and Hypertension

7.2.2.1 Modelling and Assessing the Fit As with the first analysis, the first step of the second analysis begins with fitting the univariable logistic regression models as shown by Table 7.11. Any independent variable with a p-value less than 0.25, and not clinically meaningful, will be discarded from the multivariate analysis. Those variables are history of hypertension in parent(s) (PARHY), exercise activity levels (EXER), parent smoking (PARSMK), smoking status (SMK), aircraft noise annoyance (ANNOAIR), traffic noise annoyance (ANNOTRAF), noise gap index (NGI), and insulation (INSUL). Note that although the variable of aircraft noise exposure (GROUP) was not statistically significant, it was maintained because it will be used to control the effect of aircraft noise on chronic noise stress. More details will be discussed later.

242 Table 7.11: Univariable Logistic Regression Models of the Second Analysis

Odds 95% CI Symbol Coeff. Std. Err. G p-value Ratio Lower Upper PARHY 0.104 0.213 1.11 0.73 1.69 0.239 0.625 X CHOL 1.739 0.238 5.69 3.57 9.08 50.489 <0.001 GROUP -0.160 0.213 0.85 0.56 1.29 0.569 0.451 STR7 0.680 0.222 1.97 1.28 3.05 9.046 0.003 EMP EMP(1) -1.137 0.254 0.32 0.20 0.53 28.589 <0.001 EMP(2) -1.050 0.344 0.35 0.18 0.69 EMP(3) 0.291 0.413 1.34 0.60 3.01 EXER EXER(1) -0.470 0.356 0.63 0.31 1.26 3.345 0.341 X EXER(2) -0.543 0.323 0.58 0.31 1.09 X EXER(3) -0.324 0.269 0.72 0.43 1.23 X PARSMK -0.292 0.287 0.75 0.43 1.31 1.086 0.297 X SMK SMK(1) 0.143 0.288 1.26 0.69 2.31 0.243 0.886 X SMK(2) 0.038 0.248 0.41 0.05 3.16 X ALCO ALCO(1) -0.011 0.322 0.99 0.53 1.86 10.204 0.006 ALCO(2) 0.725 0.262 0.48 0.29 0.81 NUTRI 0.354 0.259 1.43 0.86 2.37 1.790 0.181 SEN 0.027 0.014 1.03 1.00 1.06 3.501 0.061 ANNOAIR 0.025 0.029 1.03 0.97 1.09 0.767 0.381 X ANNOTRAF 0.043 0.042 1.04 0.96 1.13 1.034 0.309 X NGI -0.022 0.024 0.98 0.93 1.03 0.863 0.353 X INSUL 0.009 0.264 1.01 0.60 1.69 0.001 0.974 X SEX 0.438 0.214 1.55 1.02 2.36 4.151 0.042 AGE 0.067 0.008 1.07 1.05 1.09 81.264 <0.001 BMI BMI(1) 1.017 0.514 2.76 1.01 7.57 24.594 <0.001 BMI(2) 0.629 0.509 1.88 0.69 5.09 BMI(3) -0.400 0.536 0.67 0.24 1.92 EDU EDU(1) -1.395 0.336 0.25 0.13 0.48 24.885 <0.001 EDU(2) -0.843 0.241 0.43 0.27 0.69 MAR MAR(1) 1.434 0.438 4.20 1.78 9.90 18.251 <0.001 MAR(2) 1.679 0.474 5.36 2.12 13.57 INC INC(1) -1.852 0.551 0.16 0.05 0.46 32.818 <0.001 INC(2) -1.285 0.236 0.28 0.17 0.44

Table 7.12 tabulates the results of fitting a multivariable model containing the covariates significant at p-value 0.25 level in Table 7.11. The covariates with a p-value less than 0.05, and not clinically meaningful, will be discarded by this step (as marked by the ‘X’ symbol). Consequently, there is one exposure variable (chronic noise stress, STR_7) and three covariate variables (which are aircraft noise exposure, high cholesterol status, and age) remaining in the model.

243 Table 7.12: Results of Fitting a Multivariable Model of the Second Analysis Odds 95% CL Variable Coeff. Std. Err. Wald df Sig Ratio Lower Upper STR_7 1.005 0.355 8.017 1 0.005 2.732 1.363 5.479 GROUP -0.525 0.346 2.305 1 0.129 0.591 0.300 1.165 CHOL 1.419 0.315 20.300 1 0.000 4.132 2.229 7.659 EMP 2.679 3 0.444 X EMP(1) 0.396 0.444 0.795 1 0.373 1.486 0.622 3.549 X EMP(2) 0.004 0.560 0.000 1 0.995 1.004 0.335 3.009 X EMP(3) 0.904 0.674 1.797 1 0.180 2.470 0.659 9.262 X ALC 1.683 2 0.431 X ALC(1) 0.530 0.448 1.399 1 0.237 1.698 0.706 4.086 X ALC(2) 0.103 0.374 0.075 1 0.784 1.108 0.532 2.308 X NUTR 0.036 0.332 0.012 1 0.913 1.037 0.541 1.986 X SEN -0.008 0.019 0.197 1 0.658 0.992 0.956 1.029 X SEX 0.303 0.312 0.943 1 0.332 1.354 0.734 2.495 X AGE 0.053 0.015 13.016 1 0.000 1.054 1.024 1.085 BMI 6.008 3 0.111 X BMI(1) 0.819 0.721 1.289 1 0.256 2.268 0.552 9.320 X BMI(2) 0.619 0.719 0.740 1 0.390 1.857 0.453 7.601 X BMI(3) -0.088 0.744 0.014 1 0.906 0.916 0.213 3.935 X EDU 2.592 2 0.274 X EDU(1) -0.498 0.473 1.107 1 0.293 0.608 0.240 1.537 X EDU(2) -0.504 0.332 2.310 1 0.129 0.604 0.315 1.157 X MAR 3.717 2 0.156 X MAR(1) 1.034 0.603 2.935 1 0.087 2.812 0.862 9.176 X MAR(2) 0.622 0.665 0.876 1 0.349 1.863 0.506 6.860 X INC 0.281 2 0.869 X INC(1) -0.350 0.724 0.234 1 0.629 0.705 0.170 2.914 X INC(2) -0.029 0.433 0.005 1 0.946 0.971 0.416 2.269 X Constant -6.211 1.370 20.550 1 0.000 0.002

The refitting of the multivariable model reveals the deviance (or the -2 log-likelihood value) is 450.253 with 5 degree of freedom (Table not shown). This model, at this stage, is referred to as the full model. The variables not selected for the original multivariable model are adding (one-by-one), refitting, and deleting until the model contains the most essential (and significant) variable and the best fit. This process is summarised in Table 7.13. Only one variable which was history of hypertension in parent(s) was added to the model according to their significance in both p-value and the Wald statistic (z). The preliminary main effects model of the second analysis is calculated (Table not shown). A deviance of this model equals 445.983 with 6 degree of freedom.

There is only one continuous variable (which is age) contained in the preliminary main effects model. Therefore, the linearity of age is checked. Firstly, the design variable

method is implemented. The quartile of age is calculated (which are Q1=36.25, Q2=48,

244 and Q3=60). The minimum and maximum values of age are 15 and 87. Thus, the quartile midpoints of age are 25.625, 42.125, 54, and 73.5, respectively. The age variable is then transformed into a categorical variable based on the quartile values. The preliminary main effects model replacing the continuous variable of age with those newly created categorical variables is refitted. The estimated coefficients of categorical age variable are plotted against the midpoint of each group presented in Figure 7.5.

Table 7.13: Adding, Refitting, and Deleting Process of the Second Analysis Deviance df p -value Wald statistic Full Model 450.253 5 SMK 428.263 7 0.000 0.056 PARSMK 428.934 6 0.000 0.829 PARHY 445.983 6 0.039 0.040 EXER 439.306 8 0.012 0.771 ANNOAIR 447.722 6 0.112 0.448 ANNOTRAF 446.729 6 0.060 0.280 NGI 449.595 6 0.417 0.532 INSUL 421.810 6 0.000 0.401

2.5

2.0

f 1.5

Coef 1.0

0.5

0.0 20 30 40 50 60 70 80 AGE

Figure 7.5: Plot of Estimated Coefficients of Age (AGE) vs. the Quartile Midpoints

Figure 7.5 shows a gradual increase followed by a sharp increase and then a return to a gradually increasing age coefficient. The results are ambiguous in terms of support for linearity in the logit for age. The fractional polynomial analysis was therefore applied and presented in Table 7.14. The table reveals that the best non-linear transformation of both J=1 and J=2 are not significantly different from the linear model. The decision to transform the continuous variable of age to another format was rejected. The preliminary main effects model then automatically becomes the main effects model as shown in Table 7.15.

245 Table 7.14: Fractional Polynomials for Age df Deviance G for Model vs. Linear p-value Powers Not in model 0 508.24 Linear 1 445.98 0.000* 1 J = 1 2 444.97 1.01 0.315** 0.5 J = 2 4 440.53 5.46 0.141*** 3, 3 * Compare the linear model to the model without AGE ** Compare the J = 1 model to the linear model *** Compare the J = 2 model to the linear model

Table 7.15: The Main Effects Model of the Second Analysis

Odds 95% CL Variable Coeff. Std. Err.Wald df Sig Ratio Lower Upper STR_7 1.008 0.290 12.124 1 0.000 2.741 1.554 4.835 GROUP -0.484 0.276 3.082 1 0.079 0.616 0.359 1.058 CHOL 1.350 0.262 26.562 1 0.000 3.858 2.309 6.446 AGE 0.067 0.009 51.618 1 0.000 1.070 1.050 1.089 PARHY 0.530 0.258 4.228 1 0.040 1.700 1.025 2.818 Constant -5.906 0.623 89.921 1 0.000 0.003

The first column of Table 7.16 lists all possible interactions from the main effects model. Each interaction was added into the model one at a time. The deviance, G-value, degree of freedom, p-value, and the Wald statistic (z) of each interaction were calculated as presented in the table. There is no any interaction that provides both significance in p-value and the Wald statistic (z). Therefore, the study decides not to include any interaction to the model. Consequently, the main effects model becomes the preliminary final model and ready for the goodness-of-fit test.

Table 7.16: Interaction Test for the Main Effects Model of the Second Analysis Interaction Deviance G df p-value Wald Main Effects Model 445.983 6 GROUP STR_7 445.745 0.238 7 0.626 0.627 GROUP CHOL 445.966 0.017 7 0.896 0.896 GROUP AGE 442.413 3.570 7 0.059 0.060 GROUP PARHY 445.855 0.128 7 0.721 0.721 STR_7 CHOL 445.787 0.196 7 0.658 0.658 STR_7 AGE 445.548 0.435 7 0.510 0.513 STR_7 PARHY 445.641 0.342 7 0.559 0.559 CHOL AGE 445.705 0.278 7 0.598 0.595 CHOL PARHY 444.727 1.256 7 0.262 0.263 AGE PARHY 445.951 0.032 7 0.858 0.859

The value of the Hosmer-Lemeshow goodness-of-fit statistic computed from the

^ contingency table in Table 7.17 is C = 5.082, and the p-value computed from the chi-

246 square distribution with 8 degree of freedom is 0.749. This indicates that the model is a good fit. Finally, from Table 7.15, the fit model of the logit of association between chronic noise stress and prevalence of hypertension is written as presented by equation (7.2):

^ g(x) = – 5.906 + (1.008)STR_7 + (-0.484)GROUP + (1.350)CHOL + (0.067)AGE +

(0.530)PARHY (7.2) where STR_7 is a chronic noise stress variable (1 = yes, 0 = no) GROUP is an aircraft noise exposure variable (1 = aircraft noise exposure group, 0 = control group) CHOL is a high cholesterol status variable (1 = yes, 0 = no) AGE is an age variable (range from 15 – 87) PARHY is a history of hypertension in parent(s) variable (1 = yes, 0 = no).

Table 7.17: Contingency Table of Hosmer and Lemeshow Test and Chi-Square Value of the Second Analysis

Decile of HY = no HY = yes Total Risk Observed Expected Observed Expected 1 64 63.10 0 0.90 64 2 63 62.29 1 1.72 64 3 63 61.41 1 2.59 64 4 60 59.13 3 3.87 63 5 56 58.69 8 5.31 64 6 56 56.72 8 7.28 64 7 52 54.15 12 9.85 64 8 48 49.50 16 14.50 64 9 44 42.14 20 21.86 64 10 28 26.89 36 37.11 64

Chi-square df Sig. 5.082 8 0.749

Even though the variable of aircraft noise exposure in the final model is not statistically significant (p-value = 0.079) (which implies that aircraft noise has no direct association with hypertension) this study still maintained it in the model for the purpose of controlling for the effects of aircraft noise exposure on an association between noise stress and hypertension.

247 7.2.2.2 Calculation of Odds Ratio and Confidence Interval

As there is no potential effect modifier (Wj) in the final model of the second analysis, the adjusted odds ratio when Wj=0 is simply a natural logarithm of the estimated coefficient of exposure variable (see Table 7.20) which can be written as:

^

Adjusted odds ratio = exp ( β1 ) = exp (1.008) = 2.74

^ ^ ^ The 95% confidence interval can be estimated as exp[ β i ± Z1−α / 2 SE(β i ) ] = exp[ 1.008 ± 1.96 (0.290)], which yields (1.554, 4.836).

7.2.2.3 Interpretation The binary logistic regression analysis revealed that there is some significant association between chronic noise stress and prevalence of hypertension in adults. The odds ratio of hypertension, after adjustment for some statistically significant covariates, is statistically significant because the 95% confidence interval does not range over one. Consequently, it can be interpreted that after controlling for aircraft noise exposure, cholesterol level, age, and history of hypertension in parent(s), subjects (aged 15 – 87) who have been suffering from chronic noise stress due to the long-term aircraft noise exposure have the odds of 2.74 of having hypertension compared with subjects without chronic noise stress. The study rejected the null hypothesis and then accepted the alternative hypothesis that chronic noise stress has an association with prevalence of hypertension.

7.3 SUMMARY, DISCUSSIONS AND CONCLUDING REMARKS

Previous studies, such as Evan et al (1998), Stansfeld et al (2000b), Haines et al (2001a, 2001b) and Morrel (2003), have paid much attention in studying blood pressure effects and stress in children exposed to aircraft noise level. Only two recent studies studied the association between aircraft noise and adult blood pressure. One research conducted by Rosenlund et al (2001) using history of hypertension data (“Have you been diagnosed

248 for hypertension by a physician during the past 5 years?”) of residents around Stockholm Airport, Sweden, and found significant association between aircraft noise and prevalence of hypertension. Another research conducted by Goto and Kaneko (2002) found no significant differences among residents in different aircraft noise levels based on previous blood pressure data from general health examination around Fukuoka Airport, Japan.

The objective of this chapter was to explore the second core research question (“Does long-term aircraft noise exposure associate with hypertension via noise stress as a mediating factor?”). Based on research assumption that has been set up by section 1.2.3, the study divided the main analysis into two parts. The first analysis was seeking to establish any association between long-term aircraft noise exposure and chronic noise stress. The null hypothesis of this analysis was set that there is no association between long-term aircraft noise exposure and chronic noise stress. On the other hand, the alternative hypothesis stated that there is an association between those two variables. The second analysis was concerned with any association between chronic noise stress and prevalence of hypertension. The null hypothesis of this analysis was that there is no association between chronic noise stress and prevalence of hypertension. The alternative hypothesis stated that there is an association between those two variables. The strategies to select variables involved in the model of both sub-section analyses were implement based on the guideline provided in Hosmer and Lemeshow (2000, pp.91-116) (see section 7.2.9).

From Appendix G, it recalls that the definition of odds is the possibility that event A will occur over the possibility that the event A will not occur. If odds is more than one, the possibility that event A will occur is higher than event A will not occur. If odds is less than one, the possibility that event A will occur is less than event A will not occur. Finally, if odds equals to one, it means the possibility that event A will occur is the same with event A will not occur. Odds ratio is the proportion of odds for group 1 over odds for group 2.

For the first analysis (see section 7.2.1), binary logistic regression analysis was performed to assess prediction of presence/absence of chronic noise stress based on: (1)

249 an exposure variable of aircraft noise exposure; and (2) four potential confounders of noise sensitivity score, traffic noise annoyance score, aircraft noise annoyance score higher or equal to seven points, and an interaction between traffic noise annoyance score and aircraft noise annoyance score higher or equal to seven points. Table 7.9 showed regression coefficients, Wald statistics, odds ratios and 95% confidence intervals for odds ratios for each of the exposure variables and the four potential confounders. According to the Wald criterion, aircraft noise exposure (z = 9.46, p-value = 0.002), noise sensitivity score (z = 11.35, p-value = 0.001), traffic noise annoyance score (z = 26.54, p-value < 0.001), aircraft noise annoyance score higher or equal to seven points (z = 43.80, p-value < 0.001), and interaction between traffic noise annoyance score and aircraft noise annoyance score higher or equal to seven points (z = 7.76, p-value = 0.005) reliably predicted chronic noise stress. The Hosmer-Lemeshow goodness-of-fit statistic revealed that this model is a good fit (see Table 7.10). The logit model was represented in Equation (7.1). The study rejected the null hypothesis and concluded that long-term aircraft noise exposure was significantly associated with chronic noise stress. After controlling for noise sensitivity, traffic noise annoyance, aircraft noise annoyance, and interaction between traffic noise annoyance and aircraft noise annoyance, subjects from aircraft noise exposure area have the odds of 2.61 of having chronic noise stress compared with subjects from the control area.

For the second analysis (see section 7.2.2), binary logistic regression analysis was performed to assess prediction of presence/absence of prevalence of hypertension based on: (1) an exposure variable of chronic noise stress; and (2) four potential confounders of high cholesterol status, age, history of hypertension in parent(s), aircraft noise exposure. Table 7.15 showed regression coefficients, Wald statistics, odds ratios and 95% confidence intervals for odds ratios for each of the exposure variable and the four potential confounders. According to the Wald criterion, chronic noise stress (z = 12.12, p-value < 0.001), high cholesterol status (z = 26.56, p-value < 0.001), age (z = 51.62, p- value < 0.001), and history of hypertension in parent(s) (z = 4.23, p-value < 0.04) reliably predicted prevalence of hypertension. Even though aircraft noise exposure (z = 3.08, p-value = 0.079) was insignificant to predict prevalence of hypertension implying that aircraft noise has no direct association with hypertension, it has been maintained to allow the model to control for the effect of aircraft noise exposure. The Hosmer-

250 Lemeshow goodness-of-fit statistic revealed that this model is a good fit (see Table 7.17). The logit model could be written as presented in Equation (7.2). The study rejected the null hypothesis and concluded that chronic noise stress was significantly associated with prevalence of hypertension in adult. After controlling for aircraft noise exposure, cholesterol level, age, and history of hypertension in parent(s), subjects who have chronic noise stress have the odds of 2.74 of having hypertension compared with subjects who do not have chronic noise stress.

Even though there was some form of association between NGI and aircraft noise annoyance score (see section 5.7.2), the NGI was insignificant in exploring the second core research question. For the first sub-section, the univariable logistic regression analysis demonstrated that NGI (G = 56.28, p-value < 0.001) was significant to predict chronic noise stress (STR_7) (see Table 7.7). However, when combining NGI with the others potential variables (such as noise sensitivity, aircraft noise annoyance, and traffic noise annoyance), the multivariable logistic regression analysis revealed that based on the Wald criterion NGI (z = 0.209, p-value = 0.647) was not longer sufficient to predict chronic noise stress (see Table 7.8). For the second sub-section, the univariable logistic regression analysis demonstrated that NGI (G = 0.863, p-value = 0.353) was unqualified to be a candidate for the multivariable model. Moreover, after the multivariable model was performed, the process of adding, refitting, and deleting to the full model revealed that based on the Wald criterion and the likelihood ratio test (see Table 7.18), NGI was insignificant to predict the prevalence of hypertension.

On the basis of what was summarised above, it was concluded that when controlling for noise sensitivity, traffic noise annoyance, aircraft noise annoyance, and interaction between traffic noise annoyance and aircraft noise annoyance, subjects from the aircraft noise exposure area have the possibility to have chronic noise stress 2.61 times that of subjects from the control group. In addition it was concluded that subjects, after controlling for high cholesterol status, age, history of hypertension in parent(s), and aircraft noise exposure, who have chronic noise stress have the possibility to have prevalence of hypertension 2.74 times of subjects who do not have chronic noise stress.

251 CHAPTER EIGHT

CONCLUSIONS AND RECOMMENDATIONS

This thesis aims to extend the existing knowledge of community health and well-being impacts by aircraft noise by using Sydney (Kingsford Smith) Airport as a case study. A significant of this research was to apply the concept of transdisciplinary framework in strengthening the study, which is currently rare in Australia, and overseas. The main problem addressed in this thesis is that even though aircraft noise is not loud enough to damage the auditory systems of residents in the vicinity of an airport, it deteriorates their quality of life by disturbing the daily activities (such as watching TV, listening radio, sleeping, conversation, or studying). Quality of life is one of the major components of health according to the definition declared by World Health Organisation (WHO). The contemporary health study has named the area of this knowledge as health related quality of life. Moreover, the everyday disturbance from aircraft noise could be a cause of emotional stress in groups of people who are susceptible to noise, or have noise sensitivity. Suffering from chronic stress could lead to future health problems, especially high blood pressure levels (called hypertension).

The knowledge of these research areas has been given little attention by previous researchers. The first objective of this research is to fill these gaps by exploring two core research questions. Firstly, “Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”; and, secondly, “Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”. The first core question compared health related quality of life of subjects from aircraft noise exposure group and the control group. The second core question attempted to develop an association between aircraft noise and hypertension. The second objective of this research was to improve the manner in which aircraft noise may be described and analysed by developing a ‘new’ easier-to-interpret aircraft noise index.

252 8.1 CONCLUSIONS

8.1.1 Exploring Core Research Questions

For the first core question (“Is health related quality of life worse in communities chronically exposed to aircraft noise than in communities not exposed?”), factorial analysis of covariance was employed to compare the mean scores of health measures of subjects from aircraft noise exposure area and the control group. After adjustment by significant covariate variables and potential confounding factors, the study revealed that Physical Functioning, General Health, Vitality, and Mental Health varied significantly with aircraft noise exposure (see Section 6.3). An answer for the first core research question would be: “Health related quality of life, in term of physical functioning, general health, vitality, and mental health, of community chronically exposed to high aircraft noise level were worse than the control area”.

Even though the present study found significantly associations between aircraft noise exposure and health related quality of life in term of Physical Functioning, General Health, Vitality, and Mental Health, these associations were weak. Only one to three percent of the variance in the adjusted mean scores of health related quality of life could be explained by aircraft noise exposure. This reflects a disadvantage of cross-sectional study which can only apply to a quasi-experimental study. Only levels of main effect independent variable, which is aircraft noise exposure (exposed, not exposed) can be manipulated by the study. The other independent variables, or secondary independent variables, have to be relied on data observed from the questionnaire. Groups of subject cannot be categorised equally into levels of secondary independent variables. An alternative method of analysis of covariance in a randomised complete block design, which will result in a stronger correlation between exposure variable and dependent variable, is prohibited in cross-sectional study.

For the second core question (“Does long-term aircraft noise exposure associate with adult high blood pressure level via noise stress as a mediating factor?”), binary logistic regression analysis was employed to develop an association among aircraft noise exposure, chronic noise stress, and prevalence of hypertension. After adjustment by

253 significant confounding factors, the study revealed that aircraft noise exposure reliably predicts chronic noise stress and chronic noise stress reliably predicts prevalence of hypertension (see Section 7.2). An answer for the second core research question would be: “Subjects (aged 15 – 87) who have been chronically exposed to high aircraft noise level have the odds of 2.61 of having chronic noise stress compared with the control group. In addition, subjects who suffered from chronic noise stress have the odds of 2.74 of having hypertension compared with those without chronic noise stress”.

A unique feature of the present study was to study association between aircraft noise and prevalence of hypertension via a mediating factor of noise stress. Nevertheless, it is important to note that this research did not intend to develop causality between aircraft noise and hypertension due to the limitations of cross-sectional study. The study appears to be an analytical study in examining the association between aircraft noise and hypertension. However, it still can not clearly explain whether long-term suffer from emotional stress from noise causes hypertension or hypertension makes people more easily to be stressed by noise. Nocturnal aircraft noise effects, such as induce cortisol increase, have not been considered by the present study due to the curfew policy at Sydney Airport.

The term “hypertension” in this thesis is self-assessed hypertension. As implied by its name, the individual’s blood pressure level has not been measured by any clinical tool. However, in fact, the self-assessed method is sometimes a better method than a clinical method because someone who has been told by their physician that they have hypertension will have it controlled. Therefore, the clinical method will find these people have no hypertension.

8.1.2 Developing a ‘New’ Easier-to-Interpret Aircraft Noise Index

The second objective of this research was to improve the manner in which aircraft noise may be described and analysed by developing a ‘new’ easier-to-interpret aircraft noise index. This ‘new’ index has been termed Noise Gap Index (NGI). The NGI incorporated the characteristics of background environmental noise into the N70. This index was established so it could distinguish between aircraft noise and background

254 environmental noise in a novel manner. An extensive noise monitoring program was implemented around Sydney Airport (see Section 5.2). Noise data were measured using a Bruel and Kjaer sound level meter Type 2236 (see Section 5.3). The measurement procedures were done according to the guidelines provided in Standard Australia Acoustics (SAA 1997).

Given that the comparisons between the background environmental noise and the N70 of aircraft noise could not be made directly because of the inherent differences between these two indices, this research transferred the N70 of aircraft noise into an index measured in dB(A) by employing the concepts of equivalent continuous sound pressure level (LAeq) and sound exposure level (LAE). In consequence, the relationship between long-term time average A-weighted sound pressure level of the aircraft noise level A ((L Aeq,(7am-6pm)) and N70 has been developed (see Section 5.5). By substituting both long-term time average A-weighted sound pressure level of the background B A environmental noise level (L Aeq,(7am-6pm)) and correlation between L Aeq,(7am-6pm) and N70 into the NGI definition, the formulae of NGI has been developed (see Section 5.6).

An advantage of the present study was an attempt to minimise the effects of background noise. Residents residing near major non-aviation noise sources, such as highway and train station, were excluded from the analysis. Effect from background environmental noise (as defined by research assumption in Section 1.2.3) was controlled by the NGI. Even though the NGI showed some form of correlation with aircraft noise annoyance, the final results revealed that this ‘new’ index was insignificant in either aircraft noise – noise stress correlation or noise stress – hypertension correlation. This may reflect an inaccuracy of N70 values obtained from a large scale N70 map (see Figure 5.11). The NGI can be improved by geo-coded N70 values of each dwelling household using both Geographic Information System and Transparent Noise Information Program. Note also that the NGI was derived from environmental noise data measured in local traffic area that free from major non-aviation noise sources. The major contribution of background environmental noise came from nearby streets or major roads which have been assessed by traffic noise annoyance scale in the questionnaire. This may result in analogous measurement between NGI and traffic noise annoyance scale. Moreover NGI also has correlation with aircraft noise annoyance. Therefore as both traffic noise annoyance and

255 aircraft noise annoyance were found to be significant in the analysis of aircraft noise – noise stress correlation, the NGI may became redundant and then insignificant in the analysis.

8.2 RECOMMENDATIONS

The contribution of this thesis is to establish robust hypothesises for future experimental study: (1) “long-term aircraft noise exposure has negative impacts to health related quality of life” and (2) “long-term aircraft noise exposure has indirect effects to hypertension via chronic noise stress as a mediating factor”. For the first proposed hypothesis, this thesis recommended further study on this area of knowledge by using standardised and validated tool (which allows comparison between studies), such as Short-Form 36, to measure health related quality of life. More stringent experimental study designs, such as a complete block design study, are recommended. The study enrols subjects from well-defined study populations (aircraft noise exposure area and the control area), equally assigns the study subjects of both study areas into levels (or blocks) of potential confounding factors, such as education level and smoking status, measures any necessary demographic characteristics, living behaviours, socio-economic status, and then statistically compares the mean value of health related quality of life between study groups.

For the second proposed hypothesis, this thesis recommended any future study attempting to study the effects of aircraft noise on hypertension to implement a prospective cohort study. Such a study enrols subjects from well-defined study populations (aircraft noise exposure area and the control area). To improve accuracy of blood pressure data, the study measures and records blood pressure of subjects by using automated blood pressure measuring equipment. A standard procedure to measure blood pressure and the threshold to define high blood pressure (either a systolic blood pressure, over 160 mmHg, or a diastolic blood pressure, over 90 mmHg), recommended by WHO should be followed. Subjects who have hypertension, history of hypertension, or undertaking hypertensive treatment should be excluded from the study at this initial stage. Information, such as demographic characteristics, noise sensitivity, noise stress,

256 noise annoyance, and potential confounding factors (for example, cholesterol level, history of hypertension in parent(s)) are measured and recorded. Study subjects are then followed through a suitable time intervals (for example, every two years). The rates of developing hypertension between aircraft noise exposure group and the control group are then statistically compared.

Moreover, this thesis recommended a future study to improve NGI by extending the environmental noise measurement to cover wider areas of aircraft noise exposure (20

Finally, this thesis recommended that the priority to protect health and well-being from aircraft noise exposure should be given first to community living in the vicinity of airports before the knowledge from the future experimental study emerges. By encouraging the policy maker(s) to interpret the meaning of ‘health’ in a broader way, the effects of aircraft noise on community health and well-being should be considered as a major issue in developing an aircraft noise management strategy.

257 REFERENCES

Abbas, K. A. (2004). “Traffic Safety Assessment and Development of Predictive Models for Accidents on Rural Roads in Egypt”, Accident Analysis & Prevention, Vol. 36, No. 2, pp. 149-163.

ABS. (1996). National Health Survey Australia, 1995, Users’ Guide, ABS catalogue no. 4363.0, Canberra: Australian Bureau of Statistics.

ABS. (1997). National Health Survey Australia, 1995: SF-36 Population Norms, ABS catalogue no. 4399.0, Canberra: Australian Bureau of Statistics.

AUSTRALIA. (2002). Aircraft Noise Levy Collection Act 1995, Act No. 98 of 1995, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2002). Air Navigation (Aircraft Noise) Regulations 1984, SR 1984 No. 188, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2003). Aircraft Noise Levy Act 1995, Act No. 97 of 1995, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2004). Air Services Act 1995, Act No. 81 of 1995, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2004). Aircraft Noise Levy Regulations, SR 1995 No. 279, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2004). Airports Act 1996, Act No. 42 of 1996, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2004). Airports (Environment Protection) Regulations 1997, SR 1997 No. 13, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

AUSTRALIA. (2004). Sydney Airport Curfew Act 1995, Act No. 134 of 1995, Canberra: Office of Legislative Drafting, Attorney – General’s Department. [Laws]

Agresti, A. (2002). Categorical Data Analysis, 2nd edition, New York: John Wiley & Sons, Inc.

Airservices Australia (1996). The Long Term Operating Plan for Sydney (Kingsford Smith) Airport and Associated Airspace, [Internet] Canberra: Airservices Australia. Available from: [Accessed December 2004]

Airservices Australia (1999). The Australian Noise Exposure Forecast System and Associated Land Use Compatibility Advice for Areas in the Vicinity of Airports. [Internet] Canberra: Airservices Australia. Available from: [Accessed November 2004]

258 Airservices Australia (2001). Guidelines for the Production of Noise Contours for Australian Airports, [Internet] Canberra: Airservices Australia. Available from: [Accessed December 2004]

Airservices Australia (2002). Environmental Principles and Procedures for Minimising the Impacts of Aircraft Noise, [Internet] Canberra: Airservices Australia. Available from: [Accessed December 2004]

Airservices Australia (2004a). Sydney Airport N436 – Australian Noise Exposure Index, 1 January – 31 December 2003, Canberra: Environmental Services Branch, Airservices Australia.

Airservices Australia (2004b). Sydney Airport Operational Statistics September 2004. [Internet] Canberra: Airservices Australia. Available from: [Accessed November 2004]

Airservices Australia (2004c). Noise and Flight Path Monitoring System, Sydney Quarterly Report, October – December 2003, SYQ4/03. [Internet] Canberra: Airservices Australia. Available from: [Accessed November 2004]

Airservices Australia (2004d). Noise and Flight Path Monitoring System, Sydney Quarterly Report, April – June 2004, SY04Q2. [Internet] Canberra: Airservices Australia. Available from: [Accessed November 2004]

Albrecht, G., Higginbotham, N., and Freeman, S. (2001). Transdisciplinary Thinking in Health Social Science Research: Definition, Rationale, and Procedures, pp 70-89 of Higginbotham, N., Albrecht, G., and Connor, L. (eds) Health Social Science: A Transdisciplinary and Complexity Perspective, Melbourne: Oxford University Press.

ANSI (1978). American National Standard Method for the Calculation of the Absorption of Sound by the Atmosphere. ANSI S1.26 – 1978, New York: American National Standards Institute.

Arana, M. and Garcia, A. (1998). “A Social Survey on the Effects of Environmental Noise on the Residents of Pamplona, Spain”, Applied Acoustics, Vol. 53, No. 4, pp. 245-253.

Babisch, W., Ising, H., Gallacher, J. E. J., Sharp, D. S., and Baker, I. A. (1993). “Traffic Noise and Cardiovascular Risk: The Speedwell Study, First Phase. Outdoor Noise Levels and Risk Factors”, Archives of Environmental Health, Vol. 48, No. 6, pp. 401- 405.

Babisch, W., Ising, H., Gallacher, J. E. J., Sweetnam, P. M., and Elwood, P. C. (1999). “Traffic Noise and Cardiovascular Risk: The Caerphilly and Speedwell Studies, Third Phase – 10 – Year Follow Up”, Archives of Environmental Health, Vol.54, No.3, pp. 210-216.

259 BACL (2004). 2003 Noise Management Strategies, [Internet] Hamilton Central, Queensland: Brisbane Airport Corporation Limited. Available from: , [Accessed December 2004]

Beranek, L. L., and Ver, I. L. (1992). Noise and Vibration Control Engineering (Principles and Applications), New York: John Wiley & Sons.

Berglund, B., and Lindvall, T. (eds). (1995). Community Noise, Document prepared for the World Health Organization (Centre for Sensory Research), Stockholm.

Berglund, B., Lindvall, T., and Schwela, D. H. (eds). (1999). Guidelines for Community Noise, Geneva: World Health Organization.

Bies, D. A., and Hansen, C. H. (1996). Engineering Noise Control, 2nd edition, London: E&FN Spon.

Black, J. A. (1997). “Airport Performance and Communication Strategies for Stakeholder Involvement”, Proceedings, 1st Air Transport Research Group, (ATRG) of the World Conference on Transport Research Society, Vol 2, No. 1, (Omaha, Nebraska: UNO Aviation Institute University of Nebraska), Section 4, pp. 1-20.

Blumrich, R., and Altmann, J. (1999). “Aircraft Sound Propagation Near to the Ground: Measurements and Calculations”, ACUSTICA – acta acustica, Vol. 85, No. 4, pp. 495- 504.

Bronzaft, A. L., Ahern, K. D., McGinn, R., O’Connor, J., and Savino, B. (1998). “Aircraft Noise: A Potential Health Hazard”, Environment and Behavior, Vol. 30, No. 1, pp. 101-113.

Bruel & Kjaer (1993). User Manual – Technical Documentation, Precision Integration Sound Level Meter Type 2236, Denmark: Bruel & Kjaer A/S.

CAGL (2004). Preliminary Version Draft Minor Variation Canberra International Airport Year 2020 Master Plan, [Internet] Canberra: Capital Airport Group Limited. Available from: , [Accessed December 2004]

Carlin, J. B., Stevenson, M. R., Roberts, I. B., and Catherine, M. (1997). “Walking to School and Traffic Exposure in Australian Children”, Australian and New Zealand Journal of Public Health, Vol. 21, No. 3, pp. 286-292.

Chatterjee, K., Hounsell, N. B., Firmin, P. E., and Bonsall, P. W. (2002). “Driver response to variable message sign information in London”, Transportation Research Part C: Emerging Technologies, Vol. 10, No. 2, pp. 149-169.

Chen, C.H., Lin, H.C., Kuo, H.S., Chang, M.S., and Chou, P. (1995). “Epidemiology of Hypertension in Kin-Hu, Kinmen”, The American Journal of Hypertension, Vol. 8, No. 4, pp. 395-403.

260 Chen, T. J., Chen, S. S. (1993). “Effects of Aircraft Noise on Hearing and Auditory Pathway Function of School-Age Children”, International Archives Occupational Environmental Health, Vol. 65, pp. 107-111.

Chen, T. J., Chen, S. S., Hsieh, P. Y., and Chiang, H. C. (1997). “Auditory Effects of Aircraft Noise on People Living near an Airport”, Archives of Environmental Health, Vol. 52, No. 1, pp. 45-50.

Cheuk, F. N. (2000). “Effects of Building Construction Noise on Residents: A Quasi- Experiment”, Journal of Environmental Psychology, Vol. 20, No. 4, pp. 375-385.

Cohen, S., Evans, G. W., Krantz, D. S., and Stokols, D. (1980). “Physiological, Motivational, and Cognitive Effects of Aircraft Noise on Children: Moving from the Laboratory to the Field”, American Psychologist, Vol. 35, No. 2, pp. 231 – 243.

Cohen, S., Evans, G. W., Krantz, D. S., Stokols, D., and Kelly, S. (1981). “Aircraft Noise and Children: Longitudinal and Cross-Sectional Evidence on Adaptation to Noise and the Effectiveness of Noise Abatement”, Journal of Personality and Social Psychology, Vol. 40, No. 2, pp. 331 – 345.

Cohen, S., Evans, G. W., Stokols, D., and Krantz, D. S. (1986). Behavior, Health, and Environmental Stress, New York: Plenum Press.

Crocker, M. J, (ed) (1996). Encyclopedia of Acoustics Vol.1, New York: Wiley- Interscience Publication.

Crook, M. A., and Langdon, F. J. (1974). “The Effects of Aircraft Noise in Schools around London Airport”, Journal of Sound and Vibration, Vol. 34, No. 2, pp. 221 – 232.

De Cesaris, R., Ranieri, G., Filitti, V., Bonfantino, M. V., Andriani, A. (1992). “Cardiovascular Effects of Cigarette Smoking”, Cardiology, Vol. 81, pp. 233-237.

De Vaus, D. A. (2002). Surveys in Social Research, St Leonards, NSW: Allen & Unwin.

Devore, J., and Farnum, N. (1999). Applied Statistics for Engineers and Scientists, New York: Duxbury Press.

DFT (2003). The Future of Air Transport – the White Paper presented to Parliament by the Secretary of State for Transport by Command of Her Majesty, [Internet] Great Britain: Department of Transport. Available from: , [Accessed December 2004]

Dillman, D. A. (2000). Mail and Internet Surveys, the Tailored Design Method, 2nd Edition, New York: John Wiley & Son, Inc.

261 Dobson, R. (1976). “The General Linear Model Analysis of Variance: Its Relevance to Transportation Planning and Research”, Socio-Economic Planning Science, Vol. 10, pp. 231-235.

DoTARS (2002). Discussion Paper: Expanding Ways to Describe and Assess Aircraft Noise, Canberra: Department of Transport and Regional Services, National Capital Printing Canberra.

DoTARS (2003a). Guidance Material for Selecting and Providing Aircraft Noise Information, [Internet] Canberra: Department of Transport and Regional Services. Available from: , [Accessed September 2004].

DoTARS (2003b). Transparent Noise Information Package: User’s Manual version 3.0, [Internet] Canberra: Department of Transport and Regional Services. Available from: , [Accessed September 2004].

DoTARS and DoEAH (2003). Guidance Material for Selecting and Providing Aircraft Noise Information, Canberra: Commonwealth Department of the Environment and Heritage and Commonwealth Department of Transport and Regional Services.

Edmonds, L. D., Layde, P. M., and Erickson, J. D. (1979). “Airport Noise and Teratogenesis”, Archives of Environmental Health, pp. 243-247.

Edwards, R. W. (2001). Statistical Geography Volume 1 Australian Standard Geographical Classification (ASGC), ABS catalogue no. 1216.0, ISSN 1441 1814, Canberra: Australian Bureau of Statistics.

EPA. (1990). The State of the Environment Report, Chatswood, NSW: Environment Protection Authority of NSW.

Evan, G. W., Bullinger, M., and Hygge, S. (1998). “Chronic Noise Exposure and Physiological Response: A Prospective Study of Children Living under Environmental Stress”, Psychological Science, Vol. 9, No. 1, pp. 75-77.

Fayers, P. M., and Machin, D. (2000). Quality of Life: Assessment, Analysis and Interpretation, New York: John Wiley & Sons Ltd.

FICAN (1997). Effects of Aviation Noise on Awakenings from Sleep, Federal Interagency Committee on Aviation Noise, United States.

Fidell, S., Barber, D. S., and Schultz, T. J. (1991). “Updating a dosage-effect relationship for the prevalence of annoyance due to general transportation noise”, Journal of Acoustics Society of America, Vol. 89, No. 1, pp. 221-233.

Fidell, S., Pearsons, K., Silvati, L., and Sneddon, M. (2002b). “Relationship between Low-Frequency Aircraft Noise and Annoyance due to Rattle and Vibration”, Journal of Acoustics Society of America, Vol. 111, No. 4, pp. 1743-1750.

262 Fidell, S., Pearsons, K., Tabachnick, B. G, and Howe, R. (2000). “Effects on Sleep Disturbance of Changes in Aircraft Noise near Three Airports”, Journal of Acoustics Society of America, Vol. 107, No. 5, pp. 2535-2547.

Fidell, S., Silvati, L., and Haboly, E. (2002a). “Social Survey of Community Response to a Step Change in Aircraft Noise Exposure”, Journal of Acoustics Society of America, Vol. 111, No. 1, pp. 200-209.

Fidell, S., Silvati, L., Pearsons, K., Lind, S., and Howe, R. (1999). “Field Study of the Annoyance of Low-Frequency Runway Sideline Noise”, Journal of Acoustics Society of America, Vol. 106, No. 3, pp. 1408-1415.

Fields, J. M., De Jong, R. G, Gjestland, T., Flindell, I. H., Job, R. F. S., Kurra, S., Lercher, P., Vallet, M., Yano, T., Guski, R., Felscher-Suhr, U., and Schumer, R. (2001). “Standardised General-Purpose Noise Reaction Questions for Community Noise Surveys: Research and A Recommendation”, Journal of Sound and Vibration, Vol. 242, No. 4, pp. 641-679.

Finegold, L. S., Harris, C. S., and Gierke, H. E. (1994). “Community Annoyance and Sleep Disturbance: Updated Criteria for Assessing the Impacts of General Transportation Noise on People”, Noise Control Engineering Journal, Vol. 42, No. 1, pp. 25-30.

Flindell, I. H., and Witter, I. J. (1999). “Non-Acoustical Factors in Noise Management at Heathrow Airport”, Noise and Health, Vol. 1, No. 3, pp. 27-44.

Foreman, J. E. K. (1990). Sound Analysis and Noise Control, New York: Van Nostrand Reinhold.

Franssen, E. A., van Wiechen, C. M., Nagelkerke, N. J., and Lebret, E. (2004). “Aircraft Noise around A Large International Airport and Its Impact on General Health and Medication Use”, Journal of Occupational and Environmental Medicine, Vol. 61, No. 5, pp. 405-413.

Garcia, A., Faus, L. J., and Garcia, A. M. (1993). “The Community Response to Aircraft Noise around Six Spanish Airports”, Journal of Sound and Vibration, Vol. 164, No. 1, pp. 45-52.

GCAL (2004). Draft Airport Environment Strategy 2004, [Internet] Queenland: Gold Coast Airport Limited. Available from: , [Accessed December 2004].

German, W. J., and Stanfield, C. L. (2002) Principles of Human Physiology, San Francisco: Benjamin Cummings.

Gjestland, T. (1994). “Assessment of Helicopter Noise Annoyance: A Comparison between Noise from Helicopters and from Jet Aircraft”, Journal of Sound and Vibration, Vol. 171, No. 4, pp. 453-458.

263 Goto, K., and Kaneko, T. (2002). “Distribution of Blood Pressure Data from People Living near an Airport”, Journal of Sound and Vibration, Vol. 250, No. 1, pp. 145-149.

Gregory, P. V., Eva A. K., Athanassios, T., Alexandros D. D., Konstantinos, A. A., Panagiota A. P., Dennis V. C., and Christodoulos I. S. (2005). “Effects of Dietary Salt Intake on Nocturnal Blood Pressure Fall in Essential Hypertension”, American Journal of Hypertension, Vol. 18, No. 5, Supplement 1, pp. A216-A217.

Gulding, Olmstead, Bryan, Mirsky, Fleming, D’Aprile, Gerbi, (1999). Integrated Noise Model (INM) Version 6.0 User’s Guide, Office of Environment and Energy, Final Report 1/97 – 9/99, Washington: Federal Aviation Administration.

Hardoy, M. C., Carta, M. G., Marci, A. R., Carbone, F., Cadeddu, M., Kovess, V., Dell’Osso, L., and Carpiniello, B. (2005). “Exposure to Aircraft Noise and Risk of Psychiatric Disorders: The Elmas Survey – Aircraft Noise and Psychiatric Disorders”, Social Psychiatry and Psychiatric Epidemiology, Vol. 40, No. 1, pp. 24-26.

Haines, M. M., Stansfeld, S. A., Job, R. E. S., Berglund, B., and Head, J. (2001a). “Chronic Aircraft Noise Exposure, Stress Response, Mental Health and Cognitive Performance in School Children”, Psychological Medicine, Vol. 31, pp. 265 – 277.

Haines, M. M., Stansfeld, S. A., Brentnall, S., Head, J., Berry, B., Jiggins, M., and Hygge, S. (2001b). “The West London Schools Study: The Effects of Chronic Aircraft Noise Exposure on Child Health”, Psychological Medicine, Vol. 31, pp. 1385-1396.

Haines, M. M., Breninall, S. L., Stansfeld, S. A., and Klineberg, E. (2003). “Qualitative Responses of Children to Environmental Noise”, Noise and Health, Vol. 5, No. 19, pp. 19-30.

Harris, C. M. (ed) (1991). Handbook of Acoustical Measurements and Noise Control, 3rd Edition, New York: McGraw-Hill, R. R. Donnelley & Sons Company.

Harten, N., and Olds, T. (2004). “Patterns of Active Transport in 11-12 year Old Australian Children”, Australian and New Zealand Journal of Public Health, Vol. 28, No. 2, pp. 167.

Hatfield, J., Job, R. F. S., Carter, N. L., Peploe, P., Taylor, R., and Morrell, S. (2001). “The Influence of Psychological Factors on Self-Reported Physiological Effects of Noise”, Noise and Health, Vol. 3, No. 10, pp. 1-13.

Hede, A. J., and Bullen, R. B. (1982). Aircraft Noise in Australia: A Survey of Community Reaction, National Acoustic Laboratories Report No. 88, Canberra: Australian Government Publishing Service.

Heinonen-Guzejev, M., Vuorinen, H. S., Kaprio, J., Heikkila, K., and Mussalo- Rauhamaa, H. (2000). “Self-report of Transportation Noise Exposure, Annoyance and Noise Sensitivity in Relation to Noise Map Information”, Journal of Sound and Vibration, Vol. 234, No. 2, pp. 191-206.

264 Hennekens, C. H., Buring, J. E., and Mayrent, S. L. (1987). Epidemiology in Medicine, Boston Massachusetts: Little, Brown, and Co.

Hoeger, R., Schreckenerg, D., Felscher-Suhr, U., and Griefahn, B. (2002). “Night-Time Noise Annoyance: State of the Art”, Noise and Health, Vol. 4, No. 15, pp. 19-25.

Hoeger, R. (2004). “Aircraft Noise and Times of Day: Possibilities of Redistributing and Influencing Noise Exposure”, Noise Health, Vol. 6, No. 22, pp. 55-58.

Holmen, B. A., and Niemeier, D. A. (1998). “Characterizing the Effects of Driver Variability on Real-World Vehicle Emissions”, Transportation Research Part D: Transport and Environment, Vol. 3, No. 2, pp. 117-128.

Horonjeff, R., and McKelvey, F. (1994). Planning & Design of Airports, 4th Edition, International Editions, New York: McGraw-Hill.

Hosmer, D.W. and Lemeshow, S. (2000). Applied Logistic Regression, 2nd edition, New York: John Wiley & Sons, Inc.

Hume, K., Terranova, D., and Thomas, C. (2002). “Complaint and Annoyance caused by Aircraft Operations: Temporal Patterns and Individual Bias”, Noise and Health, Vol. 4, No. 15, pp. 45-55.

Huynh, N. McLeod, I. and Kenna, L. (2004). Noise and Flight Path Monitoring System, Sydney Quarterly Report, April – June 2004, Canberra: Airservices Australia.

Hygge, S., Evans, G. W., and Bullinger, M. (2002). “A Prospective Study of Some Effects of Aircraft Noise on Cognitive Performance in Schoolchildren”, Psychological Science, Vol. 13, No. 5, pp. 469-474.

Hyndman, J. C. G., Holmana, D. J., and Pritchardb, D. A. (2003). “The Influence of Attractiveness Factors and Distance to General Practice Surgeries by Level of Social Disadvantage and Global Access in Perth, Western Australia”, Social Science & Medicine, Vol. 56, No. 2, pp. 387-403.

ICAO (1993), Environmental Protection, International Standards and Recommended Practices, Annex 16, Vol. 1, Aircraft Noise, 3rd Edition, Montreal: International Civil Aviation Organization.

Ignaccolo, M. (2000). “Environmental Capacity: Noise Pollution at Catania Fontanarossa International Airport”, Journal of Air Transport Management, Vol. 6, No. 4, pp. 191-199.

ISO 3891 (1978). Acoustics – Procedure for Describing Aircraft Noise Heard on The Ground, Geneva: International Organization of Standardization.

ISO 9613-1 (1993). Acoustics – Attenuation of Sound during Propagation Outdoors – Part 1: Calculation of the Absorption of Sound by the Atmosphere, Geneva: International Organization of Standardization.

265 Jatmika, H. E. (2001). “Economic, Social and Spatial Impacts of Major Commercial Airports and Their Management: A Case Study of Sydney (Kingsford Smith) Airport”, New South Wales, Australia, University of New South Wales. (unpublished PhD thesis)

Jenkins, L., Tarnopolsky, A., and Hand, D. (1981). “Psychiatric Admissions and Aircraft Noise from London Airport: Four-Year, Three-Hospitals Study”, Psychological Medicine, Vol. 11, pp. 765-782.

Jenkinson, C, Wright, L., and Coulter, A. (1994). “Criterion Validity and Reliability of the SF-36 in a Population Sample”, Quality of Life Research, Vol. 3, pp. 7-12.

Jenkinson, C. (ed) (1994). Measuring Health and Medical Outcomes, London: Biddles Ltd.

Jilda, C. G., Hyndman, A. C., D’Arcy, J. H., and Douglas, A. P. (2003). “The Influence of Attractiveness Factors and Distance to General Practice Surgeries by Level of Social Disadvantage and Global Access in Perth, Western Australia”, Social Science & Medicine, Vol. 56, No. 2, pp.387-403.

Job, R. F. S., Hatfield, J., Carter, N. L., Peploe, P., Taylor, R., and Morrell, S. (2001). “General Scales of Community Reaction to Noise (Dissatisfaction and Perceived Affectedness) Are More Reliable than Scales of Annoyance”, Journal of Acoustics Society of America, Vol. 110, No. 2, pp. 939-946.

Jones, D. M., and Chapman, A. J. (1984). Noise and Society, New York: John Wiley & Sons.

Karami, K. and Frost, S. (1999). “Nuisance caused by Aircraft Noise in the Vicinity of Tehran International Airport”, Environmental Management and Health, Vol. 10, No. 2, pp. 90-95.

Kennedy, B. P., Kawachi, I., Glass, R., Prothrow-Stith, D. (1998). “Income Distribution, Socioeconomic Status, and Self Rated Health in the United States: Multilevel Analysis”, British Medical Journal, Vol. 317, No. 7163, pp. 917-921.

Kikuo, A. (2004). “What is the most important in treating hypertension?”, International Congress Series, Vol. 1262, pp. 593-596.

Kleinbaum, D.G. and Klein, M. (2002). Logistic Regression: A Self-Learning Text, 2nd Edition, New York: Springer.

Ko, N. W. M. (1979). “Responses of Teachers to Aircraft Noise”, Journal of Sound and Vibration, Vol. 62, No. 2, pp. 277 – 292.

Kontogiannis, T., Kossiavelou, Z., and Marmaras, N. (2002). “Self-Reports of Aberrant Behaviour on the Roads: Errors and Violations in a Sample of Greek Drivers”, Accident Analysis & Prevention, Vol. 34, No. 3, pp. 381-399.

266 Kryter, K. D. (1990). “Aircraft Noise and Social Factors in Psychiatric Hospital Admission Rates: A Re-Examination of Some Data”, Psychological Medicine, Vol. 20, pp. 395 – 411.

Kryter, K. D. (1994). Handbook of Hearing and the Effects of Noise Physiology, Psychology, and Public Health, San Diego: Academic Press.

Lapham, S. C., Skipper, B. J., Chang, I., Barton, K., and Kennedy, R. (1998). “Factors Related to Miles Driven between Drinking and Arrest Locations among Convicted Drunk Drivers”, Accident Analysis & Prevention, Vol. 30, No. 2, pp. 201-206.

Levesque, T. J. (1994). “Modeling the Effects of Airport Noise on Residential Housing Markets”, Journal of Transport Economics and Policy, Vol. 28, pp. 199-210.

Li, J. (2001). “Explaining high-occupancy-toll lane use”, Transportation Research Part D: Transport and Environment, Vol. 6, No. 1, pp. 61-74.

Lucas, J. L. (2003). “Drivers’ Psychological and Physical Reactions after Motor Vehicle Accidents”, Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 6, No. 2, pp. 135-145.

Lukas, J. S. (1975). “Noise and Sleep: A Literature Review and A Proposed Criterion for Assessing Effect”, Journal of Acoustics Society of America, Vol. 58, No. 6, pp. 1232-1242.

Makoto, T., Hisashi, A., Yuji, H., Yoshihisa, F., and Tsutomu, I. (2000). “Association between Alcohol Intake and Development of Hypertension in Japanese Normotensive Men: 12-Year Follow-Up Study”, American Journal of Hypertension, Vol. 13, No. 5, pp. 482-487.

Martinez, K.L.H., and Porter, B.E. (2004). “The Likelihood of Becoming A Pedestrian Fatality and Drivers’ Knowledge of Pedestrian Rights and Responsibilities in The Commonwealth of Virginia”, Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 7, No. 1, pp. 43-58.

Mato, R. A. M. and Mufuruki, T. S. (1999). “Noise Pollution Associated with the Operation of The Dar es Salaam International Airport”, Transportation Research Part D: Transport and Environment, Vol. 4, No. 2, pp. 81 – 89.

Matsui, T., Stansfeld, S., Haines, M., and Head, J. (2004). “Children’s Cognition and Aircraft Noise Exposure at Home – The West London Schools Study”, Noise and Health, Vol. 7, No. 25, pp. 49-58.

Maxwell, L. E., and Evans, G. W. (2000). “The Effects of Noise on Pre-School Children’s Pre-Reading Skills”, Journal of Environmental Psychology, Vol. 20, pp. 91- 91.

Meister, E., and Donatelle, R. (2000). “The Impacts of Commercial Aircraft Noise on Human Health: A Neighbourhood Study in Metropolitan Minnesota”, Journal of Environmental Health, Vol. 63, No. 4, pp. 9-15.

267 Mickey R. M., Dunn, O. J., and Clark, V. A. (2004). Applied Statistics: Analysis of Variance and Regression, 3rd Edition, New Jersey, John Wiley & Sons.

Miedema, H. M. E., and Vos, H. (1998). “Dose-Response Relationships for Transportation Noise”, Journal of Acoustics Society of America, Vol. 104, No. 6, pp. 3432-3445.

Miedema, H. M. E., and Vos, H. (1999). “Demographic and Attitudinal Factors that Modify Annoyance from Transportation Noise”, Journal of Acoustics Society of America, Vol. 105, No. 6, pp. 3336-3344.

Milliken, G. A., and Johnson, D. E. (2002). Analysis of Messy Data, Volume III: Analysis of Covariance, New York, Chapman & Hall/Crc.

Miyakita, T., Matsui, T., Ito, A., Tokuyama, T., Hiramatsu, K., Osada, T., and Yamamoto, T. (2002). “Population-Based Questionnaire Survey on Health Effects of Aircraft Noise on Residents Living around U.S. Airfields in the Ryukyus – Part I: An Analysis of 12 Scale Scores”, Journal of Sound and Vibration, Vol. 250, No. 1, pp. 129- 137.

Morfey, C. L., and Howell, G. P. (1981). “Nonlinear Propagation of Aircraft Noise in the Atmosphere”, American Institute of Aeronautics and Astronautics Journal, Vol. 19, No. 8, pp. 986-992.

Morrell, S. (2003). “Aircraft Noise and Child Blood Pressure”, New South Wales, Australia, University of Sydney. (unpublished PhD thesis)

Morrell, S., Taylor, R., and Lyle, D. (1997). “A Review of Health Effects of Aircraft Noise”, Australian and New Zealand Journal of Public Health, Vol. 21, No. 2, pp. 221- 236.

NASA (1999). Making Future Commercial Aircraft Quieter: Glenn effort will reduce engine noise, FS-1999-07-003-GRC. [Internet] Ohio: National Aeronautics and Space Administration Glenn Research Center. Available from: [Accessed June 2005]

Nelson, P. M. (ed) (1987). Transportation Noise Reference, London: Butterworth.

NSW Department of Health (1997). The 1997 NSW Health Survey Questionnaire. [Internet] North Sydney, New South Wales: NSW Department of Health. Available from: , [Accessed November 2002]

OECD (1980). Noise Abatement Polices. Conference on Noise Abatement Policies 7-9 May 1980, Paris: Organisation for Economic Co-Operation and Development.

Ohrstrom, E., Bjorkman, M., and Rylander, R. (1988). “Noise Annoyance with Regard to Neurophysiological Sensitivity, Subjective Noise Sensitivity and Personality Variables”, Psychological Medicine, Vol. 18, pp. 605-613.

268 Patrick, D. L., and Erickson, P. (1993). Health Status and Health Policy, New York: Oxford University Press.

Pearsons, K., Barber, D., Tabachnick, B., and Fidell, S. (1995). “Analysis of the Predictability of Noise-Induced Sleep Disturbance”, Journal of Acoustics Society of America, Vol. 97, No. 1, pp. 331-338.

Pennington, G., Topham, N., and Ward, R. (1990). “Aircraft Noise and Residential Property Values Adjacent to Manchester International Airport”, Journal of Transport Economics and Policy, Vol. 24, pp. 49-59.

Raitanen, T., Törmäkangas, T., Mollenkopf, H., and Marcellini, F. (2003). “Why Do Older Drivers Reduce Driving? Findings from Three European Countries”, Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 6, No. 2, pp. 81-95.

Rebbeck, T. R., Turner, S. T., and Sing, C. F. (1996). “Probability of Having Hypertension: Effects of Sex, History of Hypertension in Parents, and Other Risk Factor”, Journal of Clinical Epidemiology, Vol. 49, No. 7, pp. 727-734.

Rice, T. M., Peek-Asa, C., Kraus, J. F. (2003). “Night time Driving, Passenger Transport, and Injury Crash Rates of Young Drivers”, Injury Prevention, Vol. 9, No. 3, p. 245.

Robert, S. A. (1998). “Community-Level Socioeconomic Status Effects on Adult Health”, Journal of Health and Social Behaviour, Vol. 39, No. 1, pp. 18-37.

Robertson, S., and Hall, L. A. (2001). “Impact of Road Conditions on Elderly Drivers”, Journal of Transportation Engineering, Vol. 127, No. 3, p. 244.

Rosenlund, M., Berglind, N., Pershagen, G., Jarup, L., and Bluhm, G. (2001). “Increased Prevalence of Hypertension in a Population Exposed to Aircraft Noise”, Journal of Occupational and Environmental Medicine, Vol. 58, No. 58, pp. 769-773.

Rutherford. A. (2001). Introducing ANOVA and ANCOVA a GLM Approach, London: SAGE Publications Ltd.

Rylander, R. and Bjorkman, M. (1997). “Annoyance by Aircraft Noise around Small Airports”, Journal of Sound and Vibration, Vol. 205, No. 4, pp. 533-537.

SAA (1997). Standard Australian, Acoustics – Description and measurement of environmental noise Part 1: General Procedures, SSA 1055.1 – 1997, 3rd Edition, New South Wales: Standards Australia.

SAA (2000). Standard Australian, Acoustics - Aircraft Noise Intrusion – Building Siting and Construction, SSA 2021 – 2000, New South Wales: Standards Australia.

269 SACL (2000). Noise Abatement Procedure – Aircraft Engine Ground Running Rules, [Internet] Mascot, New South Wales: Sydney Airport Corporation Limited. Available from: , [Accessed December 2004]

SACL (2004). Preliminary Draft Sydney Airport Environment Strategy 2005 - 2010, [Internet] Mascot, New South Wales: Sydney Airport Corporation Limited. Available from: , [Accessed December 2004]

Sadler, D. D., Perrine, M. W., and Peck, R. C. (1991). “The Long-Term Traffic Safety Impact of A Pilot Alcohol Abuse Treatment as An Alternative to License Suspensions”, Accident Analysis & Prevention, Vol. 23, No. 4, pp. 203-224.

Saenz, A. L., and Stephens, R. W. B. (1986). Noise Pollution Scope 24, New York: John Wiley & Sons.

Sarantakos, S. (1998). Social Research, 2nd Edition, South Melbourne, Victoria: Macmillan Publishers.

Schultz, T. J. (1978). “Synthesis of Social Surveys on Noise Annoyance”, Journal of Acoustics Society of America, Vol. 64, No. 2, pp. 377 – 405.

SSC (1995). Falling on Deaf Ears, Report of the Senate Select Committee on Aircraft Noise in Sydney, Canberra: Senate Select Committee, Department of the Senate.

Simon, J. G., van de Mheen, H., van der Meer, J. B. W., and Mackenbach, J. P. (2000). “Socioeconomic Differences in Self-Assessed Health in a Chronically Ill Population: The Role of Different Health Aspects”, Journal of Behavioural Medicine, Vol. 23, No. 5, pp. 399-420.

Singh, R. B., Beegom, R., Mehta, A. S., et al (1998). “Prevalence and Risk Factors of Hypertension and Age-Specific Blood Pressure in Five Cities: A Study of Indian Women”, International Journal of Cardiology, Vol. 63, pp. 165-173.

Smith, A. P., and Jones, D. M. (1992). Handbook of Human Performance Volume I: The Physical Environment, London: Academic Press.

Spreng, M. (2000). “Possible Health Effects of Noise Induced Cortisol Increase”, Noise and Health, Vol. 2, No. 7, pp. 59-64.

Spreng, M. (2004). “Noise Induced Nocturnal Cortisol Secretion and Tolerable Overhead Flights”, Noise and Health, Vol. 6, No. 22, pp. 35-47.

Stansfeld, S. A. (1992). “Noise, Noise Sensitivity and Psychiatric Disorder: Epidemiological and Psychophysiological Studies”, Psychological Medicine Monograph, Supplement 22, Cambridge: Cambridge University Press.

Stansfeld, S. A., Haines, M. M., Burr, M., Berry, B., and Lercher, P. (2000a). “A Review of Environmental Noise and Mental Health”, Noise and Health, Vol. 2, No. 8, pp. 1-8.

270 Stansfeld, S. A., Haines, M., and Brown, B. (2000b). “Noise and Health in the Urban Environment”, Noise and Health, Vol. 15, No. 1-2, pp. 43-82.

Stansfeld, S. A. and Matheson, M. P. (2003). “Noise Pollution: Non-Auditory Effects on Health”, British Medical Bulletin, Vol. 68, pp. 243-257.

Stansfeld, S. A., Sharp, D. S., Gallacher, J., and Babisch, W. (1993). “Road Traffic Noise, Noise Sensitivity and Psychological Disorder”, Psychological Medicine, Vol.23, pp. 977-985.

Streiner, D. L., and Norman, G. R. (1995). Health Measurement Scales: A Practical Guide to Their Development and Use, 2nd Edition, Oxford: Oxford University Press.

Sullman, M. J. M., and Baas, P. H. (2004). “Mobile Phone Use amongst New Zealand Drivers”, Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 7, No. 2, pp. 95-105.

Sumer, N. (2003). “Personality and Behavioral Predictors of Traffic Accidents: Testing a Contextual Mediated Model”, Accident Analysis & Prevention, Vol. 35, No. 6, pp. 949-964.

Tabachnick, B. G. and Fidell, L. S. (2001). Using Multivariate Statistics, 4th Edition, Boston: Allyn and Bacon.

Trewin, D. (2001). Information Paper Census of Population and Housing Socio- Economic Indexes for Areas, ABS catalogue no.2039.0, ISBN 0 642 47936 4, Canberra: Australian Bureau of Statistics.

Trewin, D. (2002). 2001 Census of Population and Housing, Selected Social and Housing characteristics, Australia, ABS catalogue no.2015.0, ISBN 0 642 47814 7, Canberra: Australian Bureau of Statistics.

Upham, P., Maughan, J., Raper D., and Thomas, C. (2003). Towards Sustainable Aviation, London: Earthscan Publication.

Uyeno, D., Hamilton, S. W., and Biggs, A. J. G. (1993). “Density of Residential Land Use and the Impact of Airport Noise”, Journal of Transport Economics and Policy, Vol. 27, pp. 49-59. van Kamp, I., Job, R. F., Hatfield, J., Haines, M., Stellato, R. K., and Stansfeld, S. A. (2004). “The Role of Noise Sensitivity in the Noise-Response Relation: A Comparison of Three International Airport Studies”, Journal of Acoustics Society of America, Vol. 116, No. 6, pp. 3471-3479.

Vandegrift, D., and Yoked, T. (2004). “Obesity Rates, Income, and Suburban Sprawl: An Analysis of US States”, Health & Place, Vol. 10, No. 3, pp. 221-229.

271 WACL (2004). Environment Strategy 2004, [Internet] Cloverdale, Western Australia: Westralia Airport Corporation Limited. Available from: , [Accessed December 2004]

Walton, D., Thomas, J. A., and Dravitzki, V. (2004). “Commuters’ Concern for the Environment and Knowledge of the Effects of Vehicle Emissions”, Transportation Research Part D: Transport and Environment, Vol. 9, No. 4, pp. 335-340.

Ware, J. E. (2000). “SF-36 Health Survey Update”, SPINE, Vol. 25, No. 24, pp. 3130- 3139.

Ware, J. E., and Sherbourne, S. D. (1992). “The MOS 36-Item Short-Form Health Survey (SF-36): I. Conceptual Framwork and Item Selection”, Medical Care, Vol. 30, pp. 473-483.

Ware, J.E., Snow, K.K., Kosinski, M. and Gandex, B. (1993) SF36 Health Survey Manual & Interpretation Guide, Boston, Massachusetts: The Health Institute, New England Medical Center.

Weathers, A., Minkovitz, C., O'Campo, P., and Diener-West, M. (2004). “Access to Care for Children of Migratory Agricultural Workers: Factors Associated with Unmet Need for Medical Care”, Pediatrics, Vol. 113, No. 4, p. E276.

Weinstein, N. D. (1978). “Individual Differences in Reactions to Noise: A Longitudinal Study in a College Dormitory”, Journal of Applied Psychology, Vol. 63, No. 4, pp. 458- 466.

Whitfield, A. (2003). “Assessment of Noise Annoyance in Three Distinct Communities Living in Close Proximity to a UK Regional Airport”, International Journal of Environmental Health Research, Vol. 13, No. 4, pp. 361-372.

WHO. (1980). Noise, Environmental Health Criteria 12, Geneva: World Health Organization.

Wijnen, R. A. A., and Visser, H. G. (2003). “Optimal Departure Trajectories with Respect to Sleep Disturbance”, Aerospace Science and Technology, Vol. 7, No. 1, pp. 81-91.

Wong, C. L., Chau, W., and Wong, L. W. (2002). “Environmental Noise and Community in Hong Kong”, Noise and Health, Vol. 4, No. 16, pp. 65-69.

Wu, T. N., Lai, J. S., Shen, C. Y., Yu, T. S., and Chang, P. Y. (1995). “Aircraft Noise, Hearing Ability, and Annoyance”, Archives of Environmental Health, Vol. 50, No. 6, pp. 452-456.

Xie, C., and Parker, D. (2002). “A Social Psychological Approach to Driving Violations in Two Chinese Cities”, Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 5, No. 4, pp. 293-308.

272 Yamamoto, K. J., and Donelson, M. J. (1993). “Long Distance Propagation Model and Its Application to Aircraft En Route Noise Prediction”, American Institute of Aeronautics and Astronautics Journal, Vol. 32, No. 1, pp. 207-209.

Young, T., Torner, J. C., Sihler, K. C., Hansen, A. R., Asa, C. P., and Zwerling, C. (2003). “Factors Associated with Mode of Transport to Acute Care Hospitals in Rural Communities”, Journal of Emergency Medicine, Vol. 24, No. 2, pp. 189-198.

Zimmer, K., and Ellermeier, W. (1999). “Psychometric Properties of Four Measures of Noise Sensitivity: A Comparison”, Journal of Environmental Psychology, Vol. 19, pp. 295-302.

273 APPENDIX A:

QUESTIONNAIRE AND CONTACT LETTERS

A-1

Figure A.1: Questionnaire

A-2

Figure A.1: Questionnaire (Continued)

A-3

Figure A.1: Questionnaire (Continued)

A-4

Figure A.1: Questionnaire (Continued)

A-5

Figure A.1: Questionnaire (Continued)

A-6

Figure A.1: Questionnaire (Continued)

A-7

Figure A.2: Cover Letter

A-8

Figure A.3: First Follow-Up Letter

A-9

Figure A.4: Second Follow-Up Letter

A-10

Figure A.5: Third Follow-Up Letter

A-11 APPENDIX B:

TRANSLATED QUESTIONNAIRE AND CONTACTS LETTERS

B-1

Figure B.1: Translated Questionnaire in Greek

B-2

Figure B.1: Translated Questionnaire in Greek (Continued)

B-3

Figure B.1: Translated Questionnaire in Greek (Continued)

B-4

Figure B.1: Translated Questionnaire in Greek (Continued)

B-5

Figure B.1: Translated Questionnaire in Greek (Continued)

B-6

Figure B.1: Translated Questionnaire in Greek (Continued)

B-7

Figure B.2: Translated Cover Letter in Greek

B-8

Figure B.3: Translated First Follow-Up Letter in Greek

B-9

Figure B.4: Translated Second Follow-Up Letter in Greek

B-10

Figure B.5: Translated Third Follow-Up Letter

B-11

Figure B.6: Translated Questionnaire in Arabic

B-12

Figure B.6: Translated Questionnaire in Arabic (Continued)

B-13

Figure B.6: Translated Questionnaire in Arabic (Continued)

B-14

Figure B.6: Translated Questionnaire in Arabic (Continued)

B-15

Figure B.6: Translated Questionnaire in Arabic (Continued)

B-16

Figure B.6: Translated Questionnaire in Arabic (Continued)

B-17

Figure B.7: Translated Cover Letter in Arabic

B-18

Figure B.8: Translated First Follow-Up Letter in Arabic

B-19

Figure B.9: Translated Second Follow-Up Letter in Arabic

B-20

Figure B.10: Translated Third Follow-Up Letter in Arabic

B-21 APPENDIX C: DESCRIPTION OF VARIABLES

Table C.1: Meaning and Coding of Variables

Variable Abbreviation Brief Description Value Meaning Hypertension in parent(s) PARHY A dichotomous variable assessing 1 Yes history of hypertension of parents 0 No High cholesterol level CHOL A dichotomous variable assessing 1 Yes high cholesterol level 0 No Aircraft noise exposure GROUP A dichotomous variable divide study 1 Aircraft noise exposure group sample into two groups 0 Matched control group Noise stress score STR Measure of noise sensitivity score range of 2 - 10 point Noise stress score STR_7 A dichotomous variable of STR 1 STR ≥7.0 higher or equal 7 using 7 as acutoff point 0 STR <7.0 Physical Functioning PF Measure of physical functioning range of 0 - 100 point General Health GH Measure of general health range of 0 - 100 point Vitality VT Measure of sense of vitality range of 0 - 100 point Mental Health MH Measure of mental health range of 0 - 100 point Employment status EMP A categorical variable assessing 1 White collar employment status 2 Blue collar 3Unemployed 4 Not in labour force Exercise level EXER A categorical variable assessing 1 High exercise activity level status 2 Moderate 3Low 4 Sedentary Parent smoking PARSMK A dichotomous variable assessing 1 Yes smoking status of the other person 0 No in the same house Smoking status SMK A categorical variable assessing 1 Current smoker smoking status 2 Ex-smoker 3 Never smoke Alcohol intake ALCO A categorical variable assessing 1 High alcohol consumption existing alcohol intake level 2 Low alcohol consumption 3 Never consume alcohol Nutrition NUTRI A dichotomous variable assessing 1 Yes whether or not often have salty food 0 No Noise sensitivity score SEN Measure of noise sensitivity score range of 0 - 45 point Aircraft noise annoyance scale ANNOAIRMeasure of aircraft noise annoyance scale range of 0 - 10 point Aircraft noise annoyance scale ANNOAIR_7 A dichotomous variable of STR 1 ANNOAIR ≥7.0 higher or equal 7 using 7 as acutoff point 0 ANNOAIR <7.0 Traffic noise annoyance scale ANNOTRAFMeasure of traffic noise annoyance scale range of 0 - 10 point Noise gap index NGI Measure of difference between aircraft noise and background noise Insulation INSUL A dichotomous variable assessing 1 Yes whether or not have insulated house 0 No Sex SEX A dichotomous variable assessing 1 Male gender 0 Female Age AGE Measure of age Body Mass Index BMI A categorical variable assessing 1 Obesity body mass index level 2 Overweight 3 Acceptable weight 4 Underweight Education level EDU A categorical variable assessing 1 Bachelor degree or higher education level 2 Certificate / Diploma 3 High school or lower Marrital status MAR A categorical variable assessing 1 Married/de facto marrital status 2 Widowed / divorded / separated 3 Never married Weekly income INC A categorical variable assessing 1 over AUD$2,000 household weekly income 2 AUD$401 - AUD$1,999 3 under AUD$400

C-1 APPENDIX D:

DATA SCREENING

SPSS_ANALYZE_DESCRIPTIVE_EXPLORE provides descriptive statistics, histograms, and block plots of continuous variables for each group. Table D.1 tabulates descriptive statistics of continuous variables by study groups. Figure D.1 and D.3 illustrate histogram of continuous variables and discrete variables by study groups, respectively. Figure D.2 displays block plots of continuous variables by study groups. To ensure the accuracy of input data, it was double-checked with the original data. As a result, from Table D.1, accuracy of input data is guaranteed by a plausible Mean and Std Deviation. The values of input data are in the range with reasonable Minimum and Maximum values.

The distributions of variables could be checked by statistical methods (such as skewness and kurtosis) and graphical methods (such as histogram, stem-and-leaf, and block plots). When a distribution is normal, the values of skewness and kurtosis are zero (Tabachnick and Fidell, 2001). A distribution will be considered as positive skewness if there is a pileup of cases to the left and the right tail is too long. On the other hand, a distribution with a pileup of cases to the right with too long a left tail will be considered as negative skewness. A positive kurtosis indicates a distribution that is too peaked with short, thick tails. Conversely, a negative kurtosis describes a distribution that is too flat with too many cases in the tails. The histogram is useful for quickly detecting the shape of distributions. The histogram of normal distribution has a symmetric bell-shape with mean and median at the same point in the centre.

From Table D.1 and Figure D.1, the age (AGE) and noise sensitivity (SEN) variables were considered normally distributed, especially in control group when the skewness values were nearly zero. As expected, distributions of noise stress score (STR) in noise exposure group has some form of negative kurtosis with more cases in the right tail, meaning that most cases were highly stressed by noise. Most of the cases expressed their noise stress level at noise stress score equals six. On the other hand, distributions of noise stress score in control group were slightly positive skewness. High numbers of

D-1 cases were in the left tail of distribution, implying that most cases in control group were not stressed by noise. The distribution of frequency was fluctuated from around thirty to eighty cases at STR ranges from two to six, but rapidly decreased to seven cases at noise stress score equals seven and reached the minimum at noise stress score equals nine.

According to the SF-36’s developer, when dealing with the large sample size the distribution of each SF-36 scale is considered normally distributed (Ware and Sherbourne, 1992). Table D.1 and Figure D.1 support this assumption by providing nicely normal distributions of General Health (GH), Vitality (VT), and Mental Health (MH). However, it was found that the distribution of Physical Functioning (PF) was relatively negative skewness. Further analysis of Physical Functioning should be done with caution.

D-2 Table D.1: Descriptive Statistics of Continuous Variables by Study Groups

STR PF GH VT MH ANNOTRAF ANNOAIR SEN AGE Noise Control Noise Control Noise Control Noise Control Noise Control Noise Control Noise Control Noise Control Noise Control Exposure group Exposure group Exposure group Exposure group Exposure group Exposure group Exposure group Exposure group Exposure group N of Valid Case 334 314 339 314 338 316 332 313 332 313 333 315 335 315 331 307 336 312 Mean 6.44 4.25 79.09 79.23 64.49 66.08 54.58 57.02 68.02 73.53 2.61 1.96 6.27 1.03 27.76 26.97 46.63 50.85 SE Mean 0.13 0.11 1.38 1.42 1.17 1.33 1.14 1.13 1.02 0.95 0.14 0.13 0.17 0.11 0.44 0.42 0.85 0.86 Median 6 4 90 90 67 72 56 56.25 70 75 2 1 7 0 28 27 44 52 Variance 5.328 3.718 642.27 633.42 463.58 558.453 431.507 398.068 344.522 282.714 6.615 5.349 9.213 4.037 62.696 54.452 242.551 231.719 Std Deviation 2.31 1.93 25.34 25.17 21.53 23.63 20.77 19.95 18.56 16.81 2.57 2.31 3.04 2.01 7.92 7.38 15.57 15.22 Minimum2200000010100000461516 Maximum 10 10 100 100 100 100 100 100 100 100 10 10 10 10 45 45 86 87 IQR 3 3303032302531.2525203.53 5 111102220 Q1 5 3 70 70 50 52 44 43.75 55 65 0.5 0 4 0 22 22 35 41 Q2 6 4 90 90 67 72 56 56.25 70 75 2 1 7 0 28 27 44 52 Q3 8 610010082826975.0080854 3 9 133325761 Skewness -0.013 0.805 -1.365 -1.453 -0.532 -0.759 -0.465 -0.504 -0.765 -0.888 0.965 1.249 -0.490 2.315 -0.263 0.043 0.444 -0.047 SE Skewness 0.133 0.138 0.132 0.138 0.133 0.137 0.134 0.138 0.134 0.138 0.134 0.137 0.133 0.137 0.134 0.139 0.133 0.138 Kurtosis -0.879 0.328 0.884 1.270 -0.160 -0.103 -0.107 -0.245 0.374 0.760 0.088 0.652 -0.934 4.890 -0.391 -0.122 -0.484 -0.313 SE Kurtosis 0.266 0.274 0.264 0.274 0.265 0.273 0.267 0.275 0.267 0.275 0.266 0.274 0.266 0.274 0.267 0.277 0.265 0.275

D-3 For GROUP= noise exposure For GROUP= control 80 100

80 60

60

40

40

20 20

Frequency 0 Frequency 0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

STR STR

For GROUP= control For GROUP= noise exposure 160 160

140 140

120 120

100 100

80 80

60 60

40 40

20 20

Frequency 0 Frequency 0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

PF PF

For GROUP= control For GROUP= noise exposure 80 70

60

60 50

40

40 30

20 20

10

Frequency 0

Frequency 0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

GH GH

For GROUP= control For GROUP= noise exposure 100 100

80 80

60 60

40 40

20 20

Frequency 0

Frequency 0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

VT VT

Figure D.1: Histogram of Continuous Variables

D-4 For GROUP= control For GROUP= noise exposure 100 80

80 60

60

40

40

20 20

0 0 Frequency Frequency 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

MH MH

For GROUP= noise exposure For GROUP= control 50 60

40 50

40 30

30

20

20

10 10

Frequency 0

Frequency 0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 7.5 12.5 17.5 22.5 27.5 32.5 37.5 42.5 7.5 12.5 17.5 22.5 27.5 32.5 37.5 42.5

SEN SEN

For GROUP= noise exposure For GROUP= control 60 60

50 50

40 40

30 30

20 20

10 10

0 0 Frequency Frequency 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 85.0

AGE AGE

For GROUP= noise exposure For GROUP= control 120 140

120 100

100 80

80

60 60

40 40

20 20

0 0 Frequency Frequency 0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0

ANNOTRAF ANNOTRAF

Figure D.1: Histogram of Continuous Variables (Continued)

D-5 For GROUP= noise exposure For GROUP= control 300 70

60

50 200

40

30 100

20

10

0 0 Frequency Frequency 0.0 2.0 4.0 6.0 8.0 10.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

ANNOAIR ANNOAIR Figure D.1: Histogram of Continuous Variables (Continued)

Since most of study samples were located in local traffic road (see Chapter 5) where traffic noise was considered low, skewness values of the traffic noise annoyance (ANNOTRAF) variable were relatively positive. Almost fifty percent (or Q2) of subjects from both groups were on traffic noise annoyance less than or equal two, implying that most of them were less annoyed by traffic noise. As expected, a distribution of aircraft noise annoyance (ANNOAIR) in control group was highly positive in both skewness and kurtosis. Since this area is free from aircraft noise, most of the cases were on the left tail, especially at aircraft noise annoyance equals zero (which means not at all annoyed by aircraft noise). On the other hand, a distribution of aircraft noise annoyance in aircraft noise exposure group was slightly negative in both skewness and kurtosis. The frequency of case was gradually increased at aircraft noise annoyance scale equals zero followed by slightly dropped at aircraft noise annoyance scale equals four and turn to steady until aircraft noise annoyance scale equals six, and then rapidly increased and reached the peak at aircraft noise annoyance scale equals ten. Further analysis of aircraft noise annoyance scale should be done with caution because the distribution of variable in both groups was relatively different.

The outliers could also be checked by statistical methods (such as Interquartile Range) and graphical methods (such as histogram, block plots). The interquartile range (IQR) is the difference between the third quartile and the first quartile (or Q3 – Q1). Devore and Farnum (2004) recommended that cases that are 1.5 times the IQR above the Q3 or below the Q1 are considered mild outliers, and cases that are 3 times IRQ from the nearest quartile are considered as extreme outliers. Whereas, Mickey et al (2004)

D-6 suggested that cases that are 2 – 3 times the IRQ above the Q3 or below the Q1 are outliers. SPSS provides block plot with procedure to detect the outliers (see Figure D.2). Basically, a block plot from SPSS consists of a block and its upper and lower intervals. The upper and lower edge lines of the block denote the Q3 and the Q1, respectively, while the middle line represents the median (or Q2). The upper and lower intervals denote the plausible largest and smallest cases. The outliers (1.5 times the IRQ) and the extreme outliers (3 times the IRQ) are represented by ‘ο’ and ‘∗’ symbol, respectively. While recognize the effects of distort statistics, this research considered a 1.5 times the IRQ as too conservative, so it was decided to exclude only extreme outliers from the analysis by treating them as the missing value.

D-7 12 120

100 10

80 8

60

6

40

4 21155265219184 577631 20 14420728525223174202 457413 284224 405404337 388 274 556622519387544 612 2 4039306 460573 0 1995686120 478645

STR 0 PF -20 N = 314 334 N = 314 339 control noise exposure control noise exposure

GROUP GROUP

120 120

100 100

80 80

60 60

40 40

20 20

361379556533428318 0 2352559 519 0 519465534652

GH -20

VT -20 N = 316 338 N = 313 332 control noise exposure control noise exposure

GROUP GROUP

120 50

100 40

80 30

60

20 40 107176 172129204 12584 10 20 94 199 360652 129 460 308 603556318 389

MH 0 SEN 0 N = 312 332 N = 307 331 control noise exposure control noise exposure

GROUP GROUP

100 12

10 300 581343415633405 80 240

8 269297922022812031330145

60 6

4 40

2

20 0

AGE 0

ANNOTRAF -2 N = 312 336 N = 315 333 control noise exposure control noise exposure GROUP GROUP Figure D.2: Block Plots of Continuous Variables

D-8 12

10 232

176240

8 14371301120159

30525820586202125

6 3326920894

121422918540100210271199

4 198297139151295265235

1339961207101041411151091325018055248182

2

0

ANNOAIR -2 N = 315 335 control noise exposure GROUP Figure D.2: Block Plots of Continuous Variables (Continued)

The histograms of discrete variables for each study group were generated by SPSS as shown in Figure D.3. The splits in discrete variables were about the same for both study groups, except in employment status (EMP), insulation (INSUL), body mass index (BMI), and highest education level (EDU). There were higher proportions of subjects who were “not in labour force”, “obesity”, and “low education” in the control group than the aircraft noise exposure group. As we would expect, very few cases in control group fall in right hand side (insulated house from noise) of distribution of insulation. On the other hand, there was almost one of three of subjects’ house in noise exposure group has been insulated from noise. This is a result of a noise management plan (called Sydney Airport Noise Insulation Project, SANIP) at Sydney Airport following the opening of the Third Runway in 1994.

D-9 For GROUP= noise exposure For GROUP= control 300 300

200 200

100 100

Std. Dev = .36 Std. Dev = .38 Mean = .15 Mean = .17 N = 334.00 N = 315.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

HY HY

For GROUP= noise exposure For GROUP= control 200 200

100 100

Std. Dev = .50 Std. Dev = .49 Mean = .46 Mean = .42 N = 335.00 N = 315.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

PARHY PARHY

For GROUP= noise exposure For GROUP= control 300 300

200 200

100 100

Std. Dev = .39 Std. Dev = .36 Mean = .19 Mean = .15 N = 333.00 N = 314.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

CHOL CHOL

For GROUP= noise exposure For GROUP= control 200 140

120

100

80 100

60

40

Std. Dev = 1.30 20 Std. Dev = 1.33 Mean = 2.1 Mean = 2.4 N = 330.00 N = 314.00

Frequency 0 Frequency 0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0 EMP EMP Figure D.3: Histogram of Discrete Variables

D-10 For GROUP= noise exposure For GROUP= control 160 140

140 120

120 100

100 80

80 60 60

40 40

Std. Dev = .98 Std. Dev = .98 20 20 Mean = 2.6 Mean = 2.8 N = 328.00 N = 310.00

Frequency 0 Frequency 0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0

EXER EXER

For GROUP= noise exposure For GROUP= control 300 300

200 200

100 100

Std. Dev = .41 Std. Dev = .40 Mean = .21 Mean = .20 N = 327.00 N = 303.00

Frequency 0

Frequency 0 0.00 .50 1.00 0.00 .50 1.00

P_SMK P_SMK

For GROUP= noise exposure For GROUP= control 160 200

140

120

100

80 100

60

40

Std. Dev = .80 Std. Dev = .72 20 Mean = 2.23 Mean = 2.36 N = 323.00 N = 305.00

Frequency 0 Frequency 0 1.00 1.50 2.00 2.50 3.00 1.00 1.50 2.00 2.50 3.00

SMK SMK

For GROUP= noise exposure For GROUP= control 200 200

100 100

Std. Dev = .63 Std. Dev = .61 Mean = 2.05 Mean = 2.09 N = 308.00 N = 294.00

Frequency 0 Frequency 0 1.00 1.50 2.00 2.50 3.00 1.00 1.50 2.00 2.50 3.00 ALC ALC Figure D.3: Histogram of Discrete Variables (Continued)

D-11 For GROUP= noise exposure For GROUP= control 300 300

200 200

100 100

Std. Dev = .40 Std. Dev = .37 Mean = .20 Mean = .17 N = 333.00 N = 309.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

NUTR NUTR

For GROUP= noise exposure For GROUP= control 300 400

300

200

200

100

100

Std. Dev = .49 Std. Dev = .17 Mean = .39 Mean = .03 N = 324.00 N = 295.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

INSULA INSULA

For GROUP= noise exposure For GROUP= control 200 300

200

100

100

Std. Dev = .50 Std. Dev = .47 Mean = .43 Mean = .33 N = 336.00 N = 314.00

Frequency 0 Frequency 0 0.00 .50 1.00 0.00 .50 1.00

SEX SEX

For GROUP= noise exposure For GROUP= control 140 100

120 80

100

60 80

60 40

40

20 20 Std. Dev = .87 Std. Dev = .92 Mean = 2.4 Mean = 2.2 N = 316.00 N = 280.00

Frequency 0

Frequency 0 1.0 2.0 3.0 4.0 1.0 2.0 3.0 4.0 BMI BMI Figure D.3: Histogram of Discrete Variables (Continued)

D-12 For GROUP= noise exposure For GROUP= control 140 160

120 140

120 100

100 80

80 60 60

40 40

Std. Dev = .83 Std. Dev = .67 20 20 Mean = 1.97 Mean = 2.30 0 N = 332.00 0 N = 312.00 Frequency Frequency 1.00 1.50 2.00 2.50 3.00 1.00 1.50 2.00 2.50 3.00

EDU EDU

For GROUP= noise exposure For GROUP= control 200 300

200

100

100

Std. Dev = .87 Std. Dev = .65 Mean = 1.72 Mean = 1.39 0 N = 334.00 0 N = 312.00 Frequency Frequency 1.00 1.50 2.00 2.50 3.00 1.00 1.50 2.00 2.50 3.00

MAR MAR

For GROUP= noise exposure For GROUP= control 300 300

200 200

100 100

Std. Dev = .55 Std. Dev = .52 Mean = 2.11 Mean = 2.15 0 N = 326.00 0 N = 302.00 Frequency Frequency 1.00 1.50 2.00 2.50 3.00 1.00 1.50 2.00 2.50 3.00 INC INC Figure D.3: Histogram of Discrete Variables (Continued)

D-13 APPENDIX E: ANALYSIS OF COVARIANCE

E.1 INTRODUCTION

Analysis of covariance (ANCOVA) was first introduced in 1930s by R.A. Fisher, and has become a widely used statistical method in many applications, especially in psychological research. Analysis of covariance is a combination of regression analysis and analysis of variance (ANOVA). Regression analysis attempts to explain a dependent variable by one or more independent variables and a residual component (or error), while analysis of variance tests the equality of two or more means at the same time. In fact, analysis of variance may be considered as an extension of the student t-test. Analysis of covariance assesses the equality of population means in which main effects and interactions of independent variables on the dependent variable are considered after the linear correlation (or covariance) between the covariate variables and the dependent variable was controlled.

In any experimental study, in which the covariates are expected to be equally distributed across the treatment groups, analysis of covariance is used to increase the sensitivity of the test of the main effects and interactions by reducing the residual error term that is used in making inferences. The residual variation can be reduced because part of it can be explained by the linear relationship between the dependent variable and the covariate variable(s). In a nonexperimental setting, in which there is a chance for covariates to be unequally distributed across the study groups, analysis of covariance is used as a statistical matching procedure to adjust the group means to what they would be if all subjects scored equally on the covariate variable(s). Differences between subjects on covariate variable(s) are removed so that, presumably, the remaining differences on dependent variable scores between groups are related to the effects of the grouping independent variable(s) (Tabachnick and Fidell, 2001). The following section briefly overviews the application of analysis of covariance in transportation research derived from published articles.

E-1 E.2 ANALYSIS OF COVARIANCE IN TRANSPORTATION RESEARCH

The first effort in introducing Analysis of covariance into transportation research appears to be by Dobson (1976). The proposed potential applications of analysis of covariance were on the analysis of trip generation, mode split, and market segmentation in transportation planning. Nevertheless, these proposals were constrained by the theoretical sophistication of analysis of covariance and the lack of appropriate computer programs available during that time period. Even though the computer is now much more advance than the past 30 years, published articles relating to the application of Analysis of covariance in transportation planning are limited.

Analysis of variance (including analysis of covariance) has become a functional tool in traffic safety and accident analysis. Kontogiannis et al (2002) conducted an aberrant- driving-behaviours survey of over 1400 drivers in Greece. The objective of this survey was to find the cause(s) of aberrant driver behaviours. Analysis of covariance (as part of the main statistical analysis) was used formally to determine the relationships between factor scales (which were violation factors and errors factors) and age and gender (as independent variables). A covariate variable was annual mileage. Analysis of covariance found that highway-code violations and aggressive violations were significantly different as a function of age. The younger drivers were more likely to report engaging in such violations.

Sumer (2003) applied analysis of covariance to distinguish the personality factors and aberrant driving behaviour factors in predicting traffic accident involvement in Turkey. The analysis was based on data (which were collected from 295 Turkish professional drivers) that included various measures of personality factors, driver behaviours, and accident history. The key finding was the aberrant driver behaviours have direct effect on accident involvement. Abbas (2004) used analysis of variance to compare the time series data of traffic and accidents, over a 10-year period on five selected rural roads in Egypt. These data were used further to calibrate the model for predicting the expected number of accidents, injuries, fatalities, and casualties on roads in Egypt.

Sadler et al (1999) used repeated measures analysis of covariance to compare alcohol- related accident rate between the driver license-suspension program and the pilot E-2 alcohol abuse treatment program. After the 4-year period follow-up, it was found that the use of license-suspension waiver as an incentive to participate in a drinking driver program had a negative impact on traffic safety. Lapham et al (1998) studied factors relating to the distance driven between drinking and arrest locations using data collected from 3107 offenders convicted of drink driving in United States. Analysis of covariance demonstrated that among those arrested outside the immediate vicinity of their drinking locations, persons who drank in a high or medium-high arrest intensity area, and those with blood alcohol concentrations of more than 200 mg/l, drove fewer miles compared to other offenders. Independent variables such as age, gender, and vehicle age were unrelated to how far drunk drivers travel before their arrests.

Lucas (2003) used multivariate analysis of variance to determine whether or not drivers involved in at least one motor vehicle accident within the past five years would report greater psychological and physical reactions than drivers not being in an accident in the past five years. The sample size was 124 drivers. As expected, it was found that drivers in motor vehicle accidents reported significantly greater fear for personal safety, worries about driving, exhaustion, and negative physical symptoms than did drivers not being in a motor vehicle accident.

Apart from the knowledge of traffic safety and accidents, some published articles have applied analysis of covariance and analysis of covariance in vehicle emission research. For example, Holmen and Niemeier (1998) conducted an experimental study to determine whether driving styles (i.e., intensity or duration of acceleration events) affect vehicle emission levels. Two emission substances including carbon monoxide (CO) and oxides of nitrogen (NOx) were measured from the twenty-four tested cars. Analysis of variance revealed that there were significant differences in emissions between driving styles.

Walton et al (2004) used analysis of covariance to study attitudes towards the environment and knowledge of the polluting effects of vehicle emissions using surveyed data from 566 commuters (which are train and bus commuters, private motor vehicle commuters and smoky vehicle commuters). No significant differences of both environmental attitudes and emission knowledge were found among groups.

E-3 Returning to the environmental noise pollution issue, Cheuk (2000) applied analysis of variance to study the effect of noise pollution, especially building construction noise, on the nearby residents. It was found that the structure of residence has significant effects on residents’ perceptions to noise. Finally, the last article, which is considered the most relevant to this thesis, was produced by Meister and Donatelle (2000). They applied multivariate analysis of covariance to study the effects of commercial aircraft noise on health related quality of life of residents living near a major commercial airport in Metropolitan Minnesota, USA. After controlling for potential confounding factors (such as marital status, income, education, smoking status, body mass index, gender, age, and sensitivity to noise), multivariate analysis of covariance revealed that the health related quality of life of subjects from aircraft noise exposure areas were significantly worse than the control areas.

E.3 ONE-WAY ANALYSIS OF VARIANCE WITH FIXED EFFECTS

E.3.1 General Model As the statistical operations of analysis of covariance are an extension of those for analysis of variance, it is considered helpful to firstly discuss the fundamental concepts of analysis of variance and then go through the analysis of covariance. The following paragraph overviews the fundamental concepts of the one-way analysis of variance model with fixed effects.

The one-way analysis of variance model with fixed effects is (Mickey et al, 2004, equation 2.4, p.38):

Yij = µ + α i + ε ij where th th Y ij is the dependent variable score of study sample j in the i population (i = 1, 2, …, a) (j = 1, 2, …, n) a is the number of groups n is the number of subjects per group and is the same for each group µ is the overall mean of dependent variable scores th α i is the effect of the i category (α i = µ i - µ )

E-4 th th ε ij is the error indicating how much the ij variable deviates from the i population

mean (µ i).

a Note that the ∑ i=1 α i = 0, and ε ij are independent and normally distributed with zero mean and variance σ 2.

2 The estimated parameters of the population parameters µ, µ i, σ , and ε ij are⎯Y.. ,⎯Yi . , 2 s , and (Y ij -⎯Yi . ) where⎯Yi . means average value of dependent variable scores of group i and⎯Y.. means overall mean of dependent variable scores of every groups.

Therefore, the terms in the model Yij = µ + α i + ε ij are replaced by their estimated parameters as:

Yij =⎯Y.. + (⎯Yi . -⎯Y..) + (Yij -⎯Yi . ) then

Yij -⎯Y.. = (⎯Yi . -⎯Y..) + (Yij -⎯Yi . ) (E.1)

E.3.2 Sum of Squares The equation (E.1) can be simply interpreted as (Mickey et al, 2004, p.41): “the total deviation of an individual observation from the overall mean is broken down into two parts: the deviation of the category mean from the overall mean, and the deviation of the individual observation from its category mean”. The last term represents the variation of the individual observations about their own sample means and quantifies the amount of variation that is attributable to sources other than the factor (or category). It may be referred to as “residual”, or “error”. Both sides of equation (E.1) are squared and summing over all observations (a × n) in the data set, which results as: (Mickey et al, 2004, equation 2.11, p.42):

a n a n a n 2 2 2 ∑∑()Yij −Y.. =∑∑(Yi . −Y..) + ∑∑(Yij −Yi .) (E.2) i==11j i==11j i == 11j

E-5 a n 2 The term ∑∑()Yij −Y.. is called “the total sum of squares (SSt)”. The term i==11j

a n 2 ∑∑()Yi . −Y.. is called “the between group sum of squares (SSb)”. The term i==11j

a n 2 ∑∑()Yij −Yi . is called “the within group sum of squares (SSw)”. Equation (E.2) can i==11j be re-expressed as SSt = SSb + SSw.

Note that for computational convenience, both SSb and SSw are presented in term of raw score equations rather than deviation equations as follows:

2 2 1 ⎡ a ⎛ n ⎞ ⎤ 1 ⎡ a n ⎤ SSb = ⎢ ⎜ Yij ⎟ ⎥ − ⎢ Yij ⎥ (E.3) n ∑∑⎜ ⎟ an ∑∑ ⎣⎢ i==1 ⎝ j 1 ⎠ ⎦⎥ ⎣ i==11j ⎦

2 a n 1 ⎡ a ⎛ n ⎞ ⎤ SS = Y 2 − ⎢ ⎜ Y ⎟ ⎥ (E.4) w ∑∑ ij n ∑∑⎜ ij ⎟ i==11j ⎣⎢ i==1 ⎝ j 1 ⎠ ⎦⎥

To obtain the mean squares (MS), the sum of squares is divided by a suitable degree of freedom (df). For one-way analysis of variance, the number of degrees of freedom associated with the SSt is equal to the total number of observations minus 1, or (a × n) – 1. The 1 df is lost when the overall mean is estimated. The number of degrees of freedom associated with the residual (or SSw) is equal to the total number of observations minus a, or (a × n) – a. The a df is lost when the means for each of the a groups are estimated. Finally, the number of degrees of freedom associated with the factor (or SSb) is equal to the total number of group minus 1, or a – 1.

E.3.3 Hypothesis Test

The proportion of MSb over MSw provides the variance for an F ratio to test the null hypothesis that µ 1 = µ 2 = … = µ a.

MS F = b df = (a – 1), (a × n) – a (E.5) MS w

E-6 If the calculated F from equation (E.5) exceeds the critical F (which is obtained from the standard F table), the null hypothesis is rejected. Thus, there is a difference among the means in the a groups. On the other hand, if the calculated F is less than the critical F, the researcher accepts the null hypothesis, and concludes that there is no difference among the means in the a groups.

E.4 ONE-WAY ANALYSIS OF COVARIANCE – COMPLETELY RANDOMISED DESIGN

This design of analysis of covariance consists of a number of groups in one-way treatment. Each of the a treatments is assigned by n samples randomly selected from a homogeneous study population. In the context of this research, a equals two: aircraft noise exposure group and control group, and a treatment is chronic aircraft noise exposure.

E.4.1 Assumptions There are seven main assumptions that need to be satisfied before applying analysis of covariance.

• Linearity – for each study group, the covariate should have a linear relationship with the dependent variable. The stronger the correlation, the more useful the covariate will be. Inspecting scatterplots between the covariate variable and the dependent variable for each group tests this assumption.

• Homogeneity of Variance – in analysis of covariance, it is assumed that the variability (or variance) of dependent variable scores is expected to be about the same across the study groups. Various statistical tools could test this assumption. However, the most popular method for analysis of covariance is called “Levene’s test”1. The null hypothesis is set as the error variance of the dependent variable is equal across groups. If the Levene’s test is significant (p<0.05), it would indicate a violation of homogeneity of variance. The null hypothesis is then rejected. This can

1 See more details in Milliken and Johnson (2002, pp.354-355) E-7 be solved by transforming the dependent variable scores, or using a more stringent α level (see Tabachnick and Fidell, 2001, pp.79-80, for more details).

• Independence – each of the study samples (or observations) must be independent. The study samples should be randomly selected from the defined study populations.

• Independence of Covariate and Independent Variable(s) – the covariate variable should be unrelated to the independent variables. The differences in the covariate variable between the study groups only occur by chance. For experimental study, this assumption could be achieved by measuring the covariate variable before the beginning of the experiment and also by randomly allocating subjects to the different levels of the independent variable.

• Normality – in analysis of covariance, it is assumed that the dependent variable have a normal distribution with the same score on the covariate variable and in the same group. Using statistical tools such as skewness test, kurtosis test, normal probability plots, and detrended normal plots could test this assumption. More details of the normality testing is found in Tabachnick and Fidell (2001, pp.72-77).

• Homogeneity of Regression – the linear relationship between the covariate variable and the dependent variable should have the same slope in each study group. The average of the slopes for all samples is used to adjust the dependent variable scores, therefore it is assumed that the slopes do not differ significantly either from one another or from a single estimate of the population value. A simple procedure to test this assumption is by inspecting scatterplots between the covariate variable and the dependent variable for each group. An alternative way is to use a suitable program (such as SPSS MANOVA) (see Tabachnick and Fidell, 2001, p.292, for more details). Tabachnick and Fidell (2001, pp.303-304) also recommend strategies for situations when this assumption is violated.

• Reliability of Covariates – in analysis of covariance, it is assumed that the covariate variable are measured without error. The instrument used to measure the covariate variables should be reliable. Unreliable covariate variables can cause either under-

E-8 or over-adjustment of the dependent variable scores. Tabachnick and Fidell (2001, p.283) suggest criteria for a reliable covariate variable in non-experimental study.

E.4.2 General Model In general, the one-way analysis of covariance model with fixed effects in completely randomised design is expressed as (Rutherford, 2001, equation 6.5, p.108):

Yij = µ + α i +β (Xij –⎯X..) + ε ij (E.6) where th th Y ij is the dependent variable score of study sample j in the i population (i = 1, 2, …, a) (j = 1, 2, …, n) th th X ij is the covariate variable score of study sample j in the i population (i = 1, 2, …, a) (j = 1, 2, …, n). a is the number of groups n is the number of subjects per group µ is the overall mean of the dependent variable scores (or⎯Y..) β is the regression coefficient X.. is the overall mean of the covariate variable scores th α i is the effect of the i category (α i = ⎯Yi. -⎯Y.. )

ε ij is the random variation due to any uncontrolled source (ε ij = Y ij -⎯Yi . ).

In analysis of covariance, it is assumed that the linear relationships between the covariate variable and the dependent variable have the same slope in each study group

(β1 = β2 = … = βa). The regression coefficient (β) can be calculated from Rutherford (2001, equation 6.2, p.107):

⎡ a n ⎤ ⎢∑∑()X ij − X i . ()Yij − Y i . ⎥ i==11j β = ⎣ ⎦ (E.7) a n ⎡ 2 ⎤ ⎢∑∑()X ij − X i . ⎥ ⎣ i==11j ⎦

From Equation (E.6), the adjusted dependent variable score (Y’ij) can be calculated as:

Y’ij = Yij - β (Xij –⎯X..) = µ + α i + ε ij E-9 The adjusted dependent variable mean (⎯Y’i.) can be calculated, when all subjects have obtained a covariate score equal to the overall mean of the covariate (⎯X..), as (Rutherford, 2001, equation 6.8, p.110):

⎯Y’i. = ⎯Yi. - β (⎯Xi. –⎯X..) (E.8)

The adjusted dependent variable means are useful for comparison of the mean response in the a different populations when the effects from covariate(s) are removed.

E.4.3 Sum of Squares and Cross Products

In addition to the analysis of variance’s model (SSt = SSb + SSw), analysis of covariance has two more partitions in the model. First, the differences between the covariate variable scores and their overall mean are partitioned into between- and within-groups sums of squares (Tabachnick and Fidell, 2001, equation 8.2, p.284):

SSt(x) = SSb(x) + SSw(x) (E.9) where

SSt(x) is the total sum of squares on the covariate variable scores

SSb(x) is the between group sum of squares on the covariate variable scores

SSw(x) is the within group sum of squares on the covariate variable scores.

Second, the linear relationship between the dependent variable and the covariate variable is partitioned into sums of products associated with covariance between groups and sums of products associated with covariance within groups (Tabachnick and Fidell, 2001, equation 8.3, p.284):

SPt = SPb + SPw (E.10) where

SPt is the total sum of products on the covariate variable scores

SPb is the between group sum of products on the covariate variable scores

SPw is the within group sum of products on the covariate variable scores.

Note that the term ‘sum of product’ involves taking two deviations of scores from their means from the same subject (for example, Yij -⎯Yi. and Xij -⎯Xi.), multiplying them

E-10 together (instead of squaring as the sum of square), and then summing the products over all subjects.

The mathematical formula of SSb(x), SSw(x), SPb, and SPw are presented as following (Tabachnick and Fidell, 2001, table 8.2, p.286):

2 2 1 ⎡ a ⎛ n ⎞ ⎤ 1 ⎡ a n ⎤ SSb( x) = ⎢ ⎜ X ij ⎟ ⎥ − ⎢ X ij ⎥ (E.11) n ∑∑⎜ ⎟ an ∑∑ ⎣⎢ i==1 ⎝ j 1 ⎠ ⎦⎥ ⎣ i==11j ⎦

2 a n 1 ⎡ a ⎛ n ⎞ ⎤ SS = X 2 − ⎢ ⎜ X ⎟ ⎥ (E.12) w( x) ∑∑ ij n ∑∑⎜ ij ⎟ i==11j ⎣⎢ i==1 ⎝ j 1 ⎠ ⎦⎥

1 ⎡ a ⎛ n ⎞⎛ n ⎞⎤ 1 ⎡ a n ⎤⎡ a n ⎤ SP = ⎜ Y ⎟⎜ X ⎟ − Y X (E.13) b ⎢∑∑⎜∑ ij ⎟⎜ ij ⎟⎥ ⎢∑∑ ij ⎥⎢∑∑ ij ⎥ n ⎣⎢ i==11⎝ j=1 ⎠⎝ j ⎠⎦⎥ an ⎣ i==11j ⎦⎣ i==11j ⎦

a n 1 ⎡ a ⎛ n ⎞⎛ n ⎞⎤ SP = Y X − ⎜ Y ⎟⎜ X ⎟ (E.14) w ∑∑ij ij ⎢ ∑⎜∑ ij ⎟⎜ ∑ ij ⎟⎥ i==11j n ⎣⎢ i = 1⎝ j=1 ⎠⎝ j = 1 ⎠⎦⎥

The sums of squares for the dependent variable scores are then adjusted by the partitions for the covariate variable (equation 6.9) and the partitions for the association between the covariate variable and the dependent variable (equation 6.10). The adjusted between group sum of squares (SS’b) is presented by following (Tabachnick and Fidell, 2001, equation 8.4, p.285):

⎡ 2 2 ⎤ (SPb + SPw ) (SPw ) SSb′ = SSb − ⎢ − ⎥ (E.15) ⎣⎢()SSb( x) + SS w( x) SS w( x) ⎦⎥

The adjusted within groups sum of squares (SS’w) is found by subtracting from the unadjusted within groups sum of squares (SSw) a term based on within groups sums of squares and products associated with the covariate variable and with the linear relationship between the dependent variable and the covariate variable as following (Tabachnick and Fidell, 2001, equation 8.5, p.285):

2 (SPw ) SS w′ = SS w − (E.16) SS w( x) E-11 To obtain the adjusted mean squares (MS’), the adjusted sum of squares is divided by a suitable degree of freedom (df). For one-way analysis of covariance, the number of degrees of freedom associated with the SS’b is equal to the total number of group minus

1, or a – 1. The number of degrees of freedom associated with the SS’w is equal to the total number of observations minus a and minus c, or (a × n) – a – c. The c df is lost according to the total number of covariate variables.

E.4.4 Hypothesis Test In analysis of covariance, there are two hypotheses that need to be tested. The first hypothesis is to test whether or not the regression lines are the same for the a different groups. Figure E.1a displays the null model of this hypothesis. For the null model, the regression lines are not only parallel, but they also overlap with each other. The Y intercepts are all the same (α 1 = α 2 = … = α a). This implies that the effect of the study group (or treatment group) is not significant. Type I sums of squares, which are adjusted for all previous but not following effects, are used to test this hypothesis (Tabachnick and Fidell, 2001).

Figure E.1: Population Regression Lines under Two Hypotheses:

(a) α 1 = α 2 = … = α a ; and (b) β1 , β2 , …, βa = 0 (source: reproduced from Mickey et al, 2004, figure 15.3, p.382).

The second hypothesis is to test whether the parallel regression lines have a slope of zero (β1 , β2 , …, βa = 0) (see Figure E.1b). If this null hypothesis is not rejected, it means that the knowledge of the covariate’s value contributes nothing to the prediction of Y. The objective of testing this second hypothesis is to verify that a reasonable E-12 covariate has been selected. Type III sums of squares, which are adjusted for all other effects, are used to test this hypothesis (Tabachnick and Fidell, 2001).

These two null hypotheses could be tested by hand-calculation, but this risks human error and is time consuming especially when dealing with a large amount of data. In general, these tasks are done by a suitable statistical program (such as SPSS). The analysis of covariance table is generated as shown in Table E.1.

Table E.1: Analysis of Covariance Table for Testing Effects of Factor and Covariate

Source of Adjusted Adjusted df F ratio Variation Sum of Squares Mean of Squares

Factor SS' b a - 1 MS' b MS' b / MS' w

Covariate SS' x cMS'x MS' x / MS' w

Residual SS' w (a x n) - a - c MS' w

Total SS' t (a x n) - 1 (source: reproduced from Mickey et al, 2004, table 15.3, p.383).

The proportion of MS’b over MS’w provides the variance for an F ratio to test the first null hypothesis. The second null hypothesis is tested by the proportion of MS’x over

MS’w. Similarly to the analysis of variance, if the calculated F for analysis of covariance exceeds the critical F, the null hypothesis is rejected.

If a statistically significant effect of a main effect or interaction of independent variables on the adjusted dependent variable scores is found (or the first null hypothesis is rejected), the strength of association for the effect is assessed using η2 (Tabachnick and Fidell, 2001, equation 8.7, p.288):

SS′ η 2 = b (E.17) ()SSb′ + SS w′

The value of η2 is in the range between zero to one when, for instance, 0.5 represents 50% of the variance in the adjusted dependent variable scores is associated with the main effect or interaction of independent variables.

E-13 E.4 AN EXAMPLE OF AANALYSIS OF COVARIANCE

Data used in this worked example was collected during the pilot test of the proposed health survey instrument using the suburb of Kurnell, Sydney as a case study (see section 4.4). The objective of this section is merely to describe and illustrate the process of analysis of covariance based on the analytical methods provided in section E.4. Note that the results of this example analysis should not be interpreted as a research outcome. This hand-calculating example is suitable for a small sample size of simple design analysis of covariance (for example, one-way analysis of covariance with a fixed effect). The intension of this section is to provide a better understanding of the underlying principle of analysis of covariance. However, only the first hypothesis (α 1 = α 2 = … =

α a) could be tested by this hand-calculating procedure. Both hypotheses from section E.4.4 could be easily tested by a suitable statistical program (such as SPSS). A computer program is also required when dealing with more complex designs involving unequal sample sizes, and multiple covariate variables and independent variables like in the main health survey analysis (see section 6.3).

For the purposes of illustration, the data set used in this example was assumed to obey all assumptions required for analysis of covariance. The example analysis was divided into two sections. Each section consists of two study groups (a = 2): high aircraft noise exposure (i = 1); and low aircraft noise exposure (i = 2). The first analysis compares the Physical Functioning score of a subject adjusted for age between high noise exposure group and low noise exposure group. It was hypothesised that age (as covariate variable) is a variable influencing the Physical Functioning score. The Physical Functioning score (as dependent variable) ranges from zero to one hundred when the greater score means the better physical functioning condition. The second analysis compares the Mental Health score of a subject adjusted for the noise sensitivity score, between the high aircraft noise exposure group and the low aircraft noise exposure group. It was hypothesised that noise sensitivity (as covariate variable) is a variable relating with the Mental Health score. The noise sensitivity score ranges from zero to forty-five. Subjects with higher noise sensitivity scores are more sensitive to noise than those with the lower scores. The Mental Health score (as dependent variable) ranges from zero to one hundred where the greater score means the better mental health condition.

E-14 The data of both examples are given in Table E.2. Note that from this table, without any control for covariate variables, it seems that the physical functioning of a subject from a high noise exposure area was better than for a low noise exposure area. On the other hand, it seems that the mental health condition of a subject from a high aircraft noise exposure area was worse than those from a low aircraft noise exposure area. Analysis of variance (see Table E.3) revealed that the difference of Physical Functioning mean score was statistically significant (F (1, 36) = 7.640, p-value = 0.009). Nevertheless, analysis of variance revealed that the difference of Mental Health mean score was not statistically significant (F(1, 36) = 2.574, p-value = 0.117). The calculations relating to sum of squares for dependent variable, sum of squares for covariate variable, and sum of products are presented in Table E.4.

Table E.2: Data on Physical Functioning (PF), Mental Health (MH), Age (AGE), and Noise Sensitivity (SEN) of Subjects in Two Different Aircraft Noise Exposure Levels for the Example Analysis of Covariance.

High Aircraft Noise Exposure Low Aircraft Noise Exposure PF MH AGE SEN PF MH AGE SEN 10075333175706830 10055453890806838 95 80 78 18 100 90 54 35 85 80 28 10 45 55 46 30 75 35 42 26 95 55 41 39 85 25 50 42 36 100 70 32 90 85 44 20 65 75 54 50 100 65 28 23 5 75 76 28 90 30 40 31 55 90 77 27 100 90 49 21 100 90 38 20 95 90 31 20 85 80 72 39 95 90 29 22 40 80 39 10 95 65 21 36 100 85 37 31 85 55 31 33 95 55 39 48 90 85 55 37 5 65 84 49 10075453975607535 85 50 39 20 95 95 53 22 95 75 43 32 100 100 37 4 95 100 43 23 100 90 35 10 sum 1755.00 1305.00 774.00 522.00 1360.71 1490.00 1063.00 577.29 mean 92.37 68.68 40.74 27.47 71.62 78.42 55.95 30.38

Table E.3: Analysis of Variance for the Example Analysis of Covariance Source of Variance Sum of Squares df Mean Square F Sig. PF Between Groups 4085.158 1 4085.158 7.64 0.009 Within Groups 19248.842 36 534.690 Total 23334.000 37 MH Between Groups 900.658 1 900.658 2.574 0.117 Within Groups 12594.737 36 349.854 Total 13495.395 37

E-15 Table E.4: Calculations Based on Section E.4 for the Example Analysis of Covariance

High Noise Exposure Low Noise Exposure PF2 MH2 AGE2 SEN2 PF x AGE MH x SEN PF2 MH2 AGE2 SEN2 PF x AGE MH x SEN 10000 5625 1089 961 3300 2325 5625 4900 4624 900 5100 2100 10000 3025 2025 1444 4500 2090 8100 6400 4624 1444 6120 3040 9025 6400 6084 324 7410 1440 10000 8100 2916 1225 5400 3150 7225 6400 784 100 2380 800 2025 3025 2116 900 2070 1650 5625 1225 1764 676 3150 910 9025 3025 1681 1521 3895 2145 7225 625 2500 1764 4250 1050 1276 10000 4900 1024 2500 3200 8100 7225 1936 400 3960 1700 4225 5625 2916 2500 3510 3750 10000 4225 784 529 2800 1495 25 5625 5776 784 380 2100 8100 900 1600 961 3600 930 3025 8100 5929 729 4235 2430 10000 8100 2401 441 4900 1890 10000 8100 1444 400 3800 1800 9025 8100 961 400 2945 1800 7225 6400 5184 1521 6120 3120 9025 8100 841 484 2755 1980 1600 6400 1521 100 1560 800 9025 4225 441 1296 1995 2340 10000 7225 1369 961 3700 2635 7225 3025 961 1089 2635 1815 9025 3025 1521 2304 3705 2640 8100 7225 3025 1369 4950 3145 25 4225 7056 2401 420 3185 10000 5625 2025 1521 4500 2925 5625 3600 5625 1225 5625 2100 7225 2500 1521 400 3315 1000 9025 9025 2809 484 5035 2090 9025 5625 1849 1024 4085 2400 10000 10000 1369 18 3700 429 9025 10000 1849 529 4085 2300 10000 8100 1225 100 3500 900 SUM 162975 98175 34440 15712 71515 34335 115851 120900 64605 20541 70375 43264

When applying data from both Table E.2 and Table E.4 for the six sums of squares and

products, the adjusted between groups sum of squares (SS’b), and the adjusted within

groups sum of squares (SS’w) are as follows:

Sums of squares and sums of products for the first example analysis:

()17552 +13612 ()1755 +1361 2 SS = − = 4091.085 b 19 ()()19 2

(17552 +13612 ) SS = ()162975 +115851 − = 19269.280 w 19

()7742 +10632 ()774 +1063 2 SS = − = 2197.921 b( x) 19 ()()19 2

(7742 +10632 ) SS = ()34440 + 64605 − = 8042.632 w( x) 19

()1755 (774)(+ 1361 )()1063 (1755 +1361)(774 +1063) SP = − = −2998.647 b 19 ()()19 2

E-16 ()1755 (774)+ (1361)(1063) SP = ()71515 + 70375 − = −5731.541 w 19

From these values, both SS’b and SS’w were calculated based on equation (E.15) and (E.16), respectively.

⎡()− 2998.647 − 5731.541 2 − 5731.5412 ⎤ SSb′ = 4091.085 − ⎢ − ⎥ = 733.054 ⎣ ()2197.921+ 8042.632 8042.632 ⎦

− 5731.5412 SS′ = 19269.280 − = 15184.726 w 8042.632

Sums of squares and sums of products for the second example analysis:

()13052 +14902 ()1305 +1490 2 SS = − = 900.658 b 19 ()()19 2

()13052 +14902 SS = ()98175 +120900 − = 12594.737 w 19

()5222 + 577 2 ()522 + 577 2 SS = − = 80.434 b( x) 19 ()()19 2

(5222 + 577 2 ) SS = ()15712 + 20541 − = 4372.168 w( x) 19

()1305 (522)(+ 1490 )()577 (1305 +1490)(522 + 577) SP = − = 269.154 b 19 ()()19 2

()1305 (522)+ (1490)(577) SP = ()34335 + 43264 − = −3525.940 w 19

From these values, both SS’b and SS’w were calculated based on equations (E.15) and (E.16), respectively. E-17 ⎡()269.154 − 3525.940 2 − 3525.9402 ⎤ SSb′ = 900.658 − ⎢ − ⎥ = 1362.032 ⎣ ()80.434 + 4372.168 4372.168 ⎦

− 3525.9402 SS′ = 12594.737 − = 9751.238 w 4372.168

Mean squares of the adjusted sums of squares were calculated by dividing sums of squares by appropriate degree of freedom. From section E.4.4, the number of degrees of freedom associated with the SS’b was equal to 2 – 1 = 1. The number of degree of freedom associated with the SS’w was equal to (2×19) – 2 – 1 = 35. The null hypotheses that there is no effect from aircraft noise exposure levels on the selected health measure scores (Physical Functioning and Mental Health) was tested as following:

Hypothesis testing for the first example analysis:

SS′ 733.054 MS′ = b = = 733.054 b ()a −1 ()2 −1

SS′ 15184.726 MS′ = w = = 433.849 w ()an − a − c ()()2 19 − 2 −1

MS′ 733.054 F = b = = 1.69 MS w′ 433.849

Note that, from Table E.3, the adjusted within group (or error) mean square (which is 433.849) was lower than the unadjusted within group mean square (which is 534.690).

From standard F-table at α = 0.05, F ratio = 1.69 with df1 = 1 and df2 = 35 has p-value = 0.202 which is higher than 0.05. Thus, the null hypothesis was not rejected. There was no significant difference in Physical Functioning mean score between high- and low noise exposure groups after controlling for a variable of age. Additionally, analysis of covariance (table not shown) performed by SPSS revealed that age was a significant covariate variable (F = 9.411, p-value = 0.004). The null hypothesis that there is no

E-18 effect from covariate variable (slope of the regression line (β) between the dependent variable and the covariate variable equals to zero) was rejected.

From equation (6.7), the slope of regression (β) can be calculated as follows (calculation table not shown):

()21.842 − 5753.4 β = = −0.713 ()2909.68 + 5132.95

From equation (6.8), the adjusted dependent variable mean (⎯Y’i.) can be calculated as: when i = 1,

⎯Y’1. =⎯Y1. - β (⎯X1. –⎯X..) = 92.37 – (-0.713)(40.74 – 48.34) = 86.95 when i = 2

⎯Y’2. =⎯Y2. - β (⎯X2. –⎯X..) = 71.62 – (-0.713)(55.95 – 48.34) = 77.05

After adjusting for the variable of age, the difference of the Physical Functioning mean score between both groups was decreased (from 92.37 – 71.62 = 20.75 to 86.95 – 77.05 = 9.90). It could be concluded that even though on average subjects from high noise exposure group have a higher adjusted Physical Functioning score than those in low noise exposure group, analysis of covariance revealed that this adjusted difference was not statistically significant.

Hypothesis testing for the second example analysis:

SS′ 1362.032 MS′ = b = = 1362.032 b ()a −1 ()2 −1

E-19 SS′ 9751.238 MS′ = w = = 278.607 w ()an − a − c ()()2 19 − 2 −1

MS′ 1362.032 F = b = = 4.889 MS w′ 278.607

Similar to the first example analysis, the adjusted within group (or error) mean square (which is 278.607) was lower than the unadjusted within group mean square (which is

349.854). From the standard F-table at α = 0.05, F ratio = 4.889 with df1 = 1 and df2 = 35 has p-value = 0.037 which is lower than 0.05. Thus, the null hypothesis was rejected. There were differences among groups in the mental health mean score after controlling for noise sensitivity. Since there was a statistically significant effect of a main effect on the adjusted dependent variable score, the strength of association for the main effects was calculated by equation (E.17) as follows:

SS′ 1362.032 η 2 = b = = 0.123 ()SSb′ + SS w′ ()1362.032 + 9751.238

It is concluded that 12.3% of the variance in the adjusted Mental Health score was associated with aircraft noise exposure level.

Additionally, analysis of covariance (table not shown) performed by SPSS revealed that Noise Sensitivity was a significant covariate variable (F = 10.207, p-value = 0.003). The null hypothesis that there is no effect from covariate variable (slope of the regression line (β) between the dependent variable and the covariate variable equals to zero) was, therefore, rejected.

Similar to the first sample analysis, the slope of regression (β) can be calculated by equation (E.7) which yields β = -0.807 (calculation table not shown). From equation

(E.8), the adjusted dependent variable mean (⎯Y’i.) can be calculated as:

E-20 when i = 1,

⎯Y’1. =⎯Y1. - β (⎯X1. –⎯X..) = 68.68 – (-0.807)(27.47 – 28.93) = 67.50 when i = 2

⎯Y’2. =⎯Y2. - β (⎯X2. –⎯X..) = 78.42 – (-0.807)(30.38 – 28.93) = 79.59

After adjusting for noise sensitivity, the difference of the Mental Health mean score between both groups was increased (from 78.42 – 68.68 = 9.72 to 79.59 – 67.50 = 12.09). It could be concluded that, on average, the subject’s mental health condition adjusted for noise sensitivity was significantly worse in a high noise exposure area than in a low noise exposure area.

This subsection has provided an example of a hand-calculation of analysis of covariance. This simple analysis of covariance design was presented for the purpose of illustration. The total sample size was only thirty-nine, and the number of each independent variable and covariate variable was limited to one. Analysis of variance revealed that the difference of the Physical Functioning mean score between groups was statistically significant, but the difference of the Mental Health mean score between groups was not statistically significant. However, after adjusting for significant covariate variable by analysis of covariance, it was found that the difference of the Physical Functioning mean score between groups turned out to be not statistically significant, but the difference of the Mental Health mean score between groups became statistically significant. Analysis of variance produced larger within group (or error) mean of squares than analysis of covariance, implying that use of the significant covariate variable has reduced the “noise” in the error term. The residual variation was reduced because part of its can be explained by the covariate variable.

E-21 APPENDIX F: LINEARITY TEST AND TEST OF HOMOGENEITY OF REGRESSION OF THE FIRST CORE RESEARCH QUESTION

100 100

80 80

60 60

40 40

GROUP GROUP

20 noise exposure 20 noise exposure Rsq = 0.2428 Rsq = 0.0018

control control

PF 0 Rsq = 0.2286 PF 0 Rsq = 0.0119 0 20 40 60 80 100 0 10 20 30 40 50

AGE SEN

100 100

80 80

60 60

40 40

GROUP GROUP

noise exposure 20 noise exposure 20 Rsq = 0.1103 Rsq = 0.0082

control control

GH 0 Rsq = 0.0207

GH 0 Rsq = 0.1089 0 20 40 60 80 100 0 10 20 30 40 50

AGE SEN

100 100

80 80

60 60

40 40

GROUP GROUP

noise exposure 20 noise exposure 20 Rsq = 0.0171 Rsq = 0.0492

control control

VT 0 Rsq = 0.0248

VT 0 Rsq = 0.0250 0 20 40 60 80 100 0 10 20 30 40 50 AGE SEN Figure F.1: Scatter Plots between dependent variables and covariate variables by Aircraft Noise Exposure

F-1 100 100

80 80

60 60

40 40

GROUP GROUP

noise exposure 20 noise exposure 20 Rsq = 0.0096 Rsq = 0.0383

control control 0 Rsq = 0.0403 0 Rsq = 0.0025 MH MH 0 20 40 60 80 100 0 10 20 30 40 50

AGE SEN

Figure F.1: Scatter Plots between dependent variables and covariate variables by Aircraft Noise Exposure (Continued)

F-2 APPENDIX G: LOGISTIC REGRESSION ANALYSIS

G.1 GENERAL APPLICATIONS

The term Logistic Regression was firstly introduced in 1944 by Joseph Berkson. It has been used in a variety of applications on biomedical research, social sciences, genetics, business and marketing (Agresti, 2002). Logistic regression is multivariate analysis. It is most often used for binary response variable where the response variable is categorical, for instance: effectiveness of new medicine (success, failure); diagnosis of angina (present, absent); and paying a credit card bill on time (yes, no). The obvious application of logistic regression analysis to transportation research would be to analyse the mode of travel (private car, public transport). The following paragraphs briefly overview application of logistic regression in transportation derived from the published articles.

The first example (Li, 2001) portrays a common practical application of logistic regression in transportation. It used logistic regression models to examine the determinants of high-occupancy-toll lane use using data surveyed on the State Route 91 Express Lanes in California. It found that controlling for other variables, household income, vehicle occupancy, commute trip, and age are important predictors of high- occupancy-toll lane use.

Two articles (Carlin et al, 1997 and Harten and Olds, 2004) used logistic regression to analyse a transport patterns of Australian children. Carlin et al (1997) found that 33% of children walk to school, while over 60% are driven to school by car, with very small proportions riding bicycles or taking public transportation. Harten and Olds (2004) found that distance was the strongest predictor of the likelihood that a trip would be active (walking, cycling). They also suggested improvement in neighbourhood safety and more accessibility for children to encourage usage of bicycles.

Most of articles used logistic regression to study a cause of traffic accident. Rice (2003) studied the association of night time driving and young drivers in California, USA. They found that the injury crash rate for young drivers (age 16 – 17) increases G-1 especially during night time hours and in the absence of adult supervision. Robertson and Hall (2001) examined the characteristics of crashes involving elderly driver (age 65 and older) in the state of Kentucky, USA on different road surface conditions (dry, wet, or snowy/slushy). The results revealed that all elderly drivers performed relatively better than the younger drivers as the road conditions worsened from dry to wet to snowy.

One study in New Zealand (Sullman and Baas, 2004) used logistic regression to study the relationship between crash involvement and mobile phone use whilst driving. They found that once all the contribution of the demographic and descriptive variables had been controlled the relationship between crash involvement and mobile phone use was no longer significant.

Martinez and Porter (2004) analysed pedestrian crashes in Virginia, USA from 1990 to 1999 and investigated variables believed to predict these crashes using logistic regression analysis. They found that such variables as location, sex, age, pedestrian drinking, driver drinking, driver violation, and time of day are significantly correlated with pedestrians involved in crashes. In China, Xie and Parker (2002) found a contribution of aggressive violations to traffic accident involvement when controlling for demographic variables by using the logistic regression model.

Some articles were dedicated to the study of accessibility of patients to health care. Young (2003) identified factors associated with the mode of transport to rural hospitals in Northwest Iowa, USA. Weather (2004) used logistic regression to assess the correlation of unmet need for medical care among migrant children in eastern North Carolina, USA. Finally, Jilda (2003) applied logistic regression to study the impact of attractiveness factors and distance to general practice surgeries of people by level of social disadvantage and global access in Perth, Western Australia.

Apart from the above areas of application of logistic regression, one article (Raitanen, 2003) studied the reasons which lead older drivers (Finnish, German, and Italian) to reduce their driving. Logistic regression found that retirement was associated with reduced driving in all locations. Older age, changes in leisure activities, and chronic conditions were significantly associated in at least one of the locations. Finally, one G-2 article (Chatterjee, 2002) applied logistic regression to the area of intelligent traffic engineering. They used logistic regression to determine the effectiveness of variable message signs (VMS) (which is used to notify motorists of planned events and current network problems) in London.

G.2 FUNDAMENTAL CONCEPTS

The logistic regression model has been proved as a multivariate statistic tool that best describes an epidemiological problem (which is exposure – disease relationship), especially, when the dependent variable (or disease variable) is binary or dichotomous (diseased or not diseased) (Hosmer and Lemeshow, 2000). The general form of logistic function is (Kleinbaum and Klein, 2002, p.8):

1 Y = 1+ e− X

The logistic function produces the S-shape curve which the value of Y approaches 1 as X approaches +∞, and Y approaches 0 when X approaches -∞. The variable X, in epidemiological research, is called an index of combined risk factors and has the form of linear summation as (Kleinbaum and Klein, 2002, p.8):

X = β0 + β1X1 + β2X2 + … + βkXk

= β0 + ∑βiXi

Studies use the quantity π(x) = E(Y⎪X) to represent the conditional mean of Y given X when the logistic distribution is used. The logistic regression model used throughout the study is: 1 π (x)= (G.1) (1+ e−(β0 +∑ βi Xi ) )

G-3 An alternative way to describe the logistic model is called the logit form of the model. A logit transformation is given by the natural log of the quantity π(x) divided by one minus π(x) as: ⎛ π (x) ⎞ logit π (x) = ln⎜ ⎟ ⎝1−π (x) ⎠ so the logit of π(x) (called g(x)) = ln [e(β0+ ∑βiXi)]

= β0 + ∑βiXi (G.2)

It is noted that g(x) has the same value as an index of combined risk factors which has many properties of a linear regression model. The logit can range from -∞ to +∞ depending on the range of X.

In the situation when some of the independent variables are nominal variables such as race, sex, occupation, and so forth, it is inappropriate to directly include them in the model. They have to be recoded into a series of design (dummy) variables. If a nominal variable has k categories, then k-1 design variables will be needed. For example, employment status variable (EMP) (employed, unemployed, and not in labour force) is coded as EMP(1) equals 1 if employment status is employed, and 0 if otherwise; EMP(2) equals 1 if employment status is unemployed, and 0 if otherwise. Therefore, the not in labour force is identified by codes of 0 on both design variables (EMP(1) = 0 and EMP(2) = 0).

G.3 THE E, V, W MODEL

An exposure – disease correlation can be distorted by some factors called confounding factors. A factor will be considered a confounder if it meets the three following criteria: 1) to be a known risk factor for the result of disease of interest; 2) to be a factor associated with exposure but not a result of exposure; and 3) to be a factor that is not an intermediate variable between exposure and disease. To prevent a distortion effect, all necessary confounding factors need to be carefully controlled. There may also be a

G-4 situation when some confounding factor(s) has/have interaction with an exposure variable called potential effect modifier variable(s).

Based on equation (7.2), the general logit formula for the E, V, W model is (Kleinbaum and Klein, 2002, p.66):

m p g(x) = β 0 + β1 E + ∑∑β iVi + E β jW j (G.3) i=+21j=m where E is an exposure variable

Vi is the potential confounders

Wj is the potential effect modifiers with E

β0 is a constant term

β1 is a coefficient of E

βi is the coefficient of Vi

βj is the coefficient of Wj

The confounding factor (C) can be in terms of a continuous variable (for example, age (AGE), noise sensitivity score (SEN)), ordinal variable (for example, body mass index category (BMI), exercise activity level (EXER)), dichotomous variable (for example, gender (SEX)), or nominal variable (for example, smoking status (SMK), employment status (EMP)). The potential confounders are the functions of C’s. It can have the form of a single C (such as AGE, SMK) and can also be a product of C’s such AGE×SEX, EXER×BMI, or SMK2.

The potential effect modifier variable(s) is/are a subset of the potential confounder(s). It is a product of terms with exposure variable (E) such as E×AGE×SEX, E×SMK.

G.4 FITTING THE LOGISTIC REGRESSION MODEL

To estimate the unknown parameters in the logistic regression model (or to fit the model), the Maximum Likelihood procedure has been introduced as the most preferable technique (Hosmer and Lemeshow, 2000, p.7-10). In order to describe the Maximum

G-5 Likelihood procedure, it is essential to construct the likelihood function, l(β) where β is the collection of unknown parameters being estimated in the model (βo, β1, β2, …, βp)

The l(β) is the joint probability of observing the data. It combines the contribution of all the subjects in the study (Hosmer and Lemeshow, 2000, p.7-10) either diseased (y =1) or not diseased (y =0).

n yi 1− yi l(β ) = ∏π (xi ) ()1−π (xi ) (G.4) i=1

Taking the natural log of equation (5.4) provides the log likelihood which is defined as:

n L(β ) = ln()l(β ) = ∑ (yi ln(π (xi )) + (1− yi )ln(1−π (xi )) ) (G.5) i=1

To determine the set of unknown parameters that maximise L(β), the L(β) is differentiated with respect to the total number of unknown parameters plus 1 (Hosmer and Lemeshow, 2000, p.33). The differentiated likelihood equations are expressed as follows:

n ∑()yi −π (xi ) = 0 i=1 and

n ∑ xij ()yi −π (xi ) = 0 i=1 for j = 1, 2, …, p

To solve the likelihood equations and obtain the unknown parameters, the suitable statistic software (such as SPSS) is required. The estimated parameter is represented by

^ ^ putting the symbol (^) over the coefficient symbol, for example, β 0 , β1 .

G-6 G.5 TESTING FOR THE SIGNIFICANCE OF THE MODEL

After the logistic regression model has been fitted, the obtained unknown parameters need to be tested for significance. The non-significant parameter(s) will be excluded from the fit to minimise the number of parameters in the model.

It has been proven that the difference between log likelihood statistics for two models, one with all parameters (p1) included and another with some parameters (from the full model) excluded (p2), has an approximate chi-square distribution with degree-of- freedom (p1 – p2) (Kleinbaum and Klein, 2002, pp.130-134). The G statistics for the log likelihood ratio test is presented as:

G = -2 [(log likelihood of smaller model) – (log likelihood of bigger model)] (G.6)

The log likelihood is calculated, normally by the statistic software, based on summing the probabilities associated with the predicted and actual outcomes for each case (Tabachnick and Fidell, 2001, pp.525-527). In accordance with equation (7.5) the log likelihood can be written as: n ⎛ ^ ^ ⎞ log likelihood = ∑⎜Yi ln(Yi )+ (1−Yi )ln(1−Yi )⎟ i=1 ⎝ ⎠ where n is the total number of cases

Yi is the actual outcome of case i

^ Yi is the predicted outcome of case i

The null hypothesis is set up, so there is no significant difference between the log likelihood of bigger and smaller model. If the null hypothesis is satisfied, it would say that the smaller model give an equivalent prediction of the outcome variable compared with the bigger model. Conversely, if the null hypothesis is rejected, it can be concluded that at least one and perhaps all of the parameters that are in the bigger model, but not in the smaller model, have a significant contribution to the model.

G-7 There is another statistical tool for the significance test of a parameter called the Wald test statistics. It is obtained by comparing an estimated coefficient of predictor

^ ^ ^ ^ parameter (such as β 0 , β1 ) to an estimate of its standard error (such as SE(β 0 ) ,

^ ^ SE(β1 ) ) (Hosmer and Lemeshow, 2000, p.16). It is noted that the previous formulae is usually applied in the univariate case. The univariate Wald test follows standard normal distribution (or Z distribution) and square of this Z statistic is approximately a chi- square statistic with one degree-of-freedom in a large sample size (Kleinbaum and Klein, 2002, pp.134-136).

G.6 CONFIDENCE INTERVAL ESTIMATION

In the univariate case, the general formula to calculate the confidence interval (CI) for the endpoint 100(1-α)% of the estimated coefficients (βi) are:

^ ^ ^ exp[ β i ± Z1−α / 2 SE(β i ) ] or

^ ^ ^ exp[ β i ± Z1−α / 2 Var(β i ) ]

^ ^ ^ where: Var(β i ) is the estimated variance of the estimated coefficient β i

G.7 CALCULATING ODDS RATIO

The probability that some event will occur over the probability that the same event will not occur is called an odds. Therefore, in the specific case of this study, the probability of getting hypertension over the probability of not getting hypertension can be written as: E(Y = 1 X ) odds = 1− E(Y = 1 X )

π (x) = 1−π (x)

G-8 ⎛ ^ ^ m ^ p ^ ⎞ ⎜ ⎟ = exp⎜ β 0 + β1 E + ∑ β i Vi + E ∑ β j W j ⎟ ⎝ i=2 j=m+1 ⎠

The ratio of two odds is called the odds ratio. If the exposure variable is coded as (0,1) variable, the general formula of odds ratio will be

odds ratio = (odds for group1) / (odds for group0)

⎛ ^ p ^ ⎞ ⎜ ⎟ = exp⎜ β1 + ∑ β j W j ⎟ (G.7) ⎝ j=m+1 ⎠

^ In a situation where there is no interaction ( β j in equation (7.7) = 0) in the logit model and the exposure variable is a dichotomous variable (a,b), then a 100(1-α)% confidence interval estimate for the odds ratio is obtained by

⎛ ^ ^ ^ ⎞ exp⎜(a − b) β ± Z a − b SE(β )⎟ ⎝ 1 1−α / 2 1 ⎠

^ If there is an interaction in the odds ratio (β j in equation (5.7) ≠ 0), the formula used to calculate a 100(1-α)% confidence interval of the odds ratio will be complicated. Let

^ p ^ ^ us rewrite the β1 + ∑ β j W j term of equation (7.7) as K . Thus, a 100(1-α)% j=m+1 confidence interval of the odds ratio will be (Kleinbaum and Klein, 2002, p.141):

⎛ ^ ^ ^ ⎞ exp⎜ K Z Var(K ) ⎟ (G.8) ⎜ ± 1−α / 2 ⎟ ⎝ ⎠ where

^ ^ ^ ^ p ^ ^ p ^ ^ ^ p p ^ ^ ^ 2 Var(K) = Var(β1 ) + ∑W j Var(β j ) + 2 ∑W j Cov(β1 , β j ) + 2 ∑∑W jWk Cov(β j , β k ) j=m+1 j=m+1 j=+m+1 k= j 1

The following subsection (7.3.8) briefly describes an example of the logistic regression analysis using short data collected from the pilot study.

G-9 G.8 AN EXAMPLE OF LOGISTIC REGRESSION ANALYSIS

Data for this example was obtained from the pilot study in the suburb of Kurnell, Sydney during June – July, 2003. It is noted that the purpose of the pilot study is to test the performance of the proposed health survey instruments for the main health survey. The sample size was only 100 subjects with a 62% response rate. After excluding some incomplete responses, fifty-six subjects were included into the analysis (see section 4.4 for more details).

The objective of this section is merely to describe the process of logistic regression analysis. The results of this example analysis should not be interpreted as a research outcome. The example analysis is run through SPSS version 12.0 using Binary Logistic Regression Function. The dependent variable is the self-reported hypertension status (HY) (0=Normotension, 1=Hypertension). The independent variables are age (AGE), obesity status (OBE) (0=No, 1=Yes), and highest education level (EDU) (0=Lower than high school, 1=Certificate – Diploma, 2=Higher than bachelor degree).

The example analysis assumed the obesity status as an exposure variable and set age and highest education level as the potential confounders. The analysis also assumed a product of obesity status and age (OBE×AGE) as the potential effect modifier variable.

Before getting started, the highest education level variable was recoded into a series of dichotomous (dummy) variables (see Table G.1). Then the logit of this example study was set as g(x) = β0+ β1OBE + β2AGE + β3EDU(1) + β4EDU(2) + β5OBE×AGE

Table G.1: Categorical Variables Coding of Example Analysis of Logistic Regression

Parameter coding Frequency (1) (2) Highest Education Lower than high 13 1.000 .000 Level school Certificate – 27 .000 1.000 Diploma Higher than 12 .000 .000 bachelor degree

G-10 The first iteration (Block 0: Beginning Block), which includes only constant terms, was carried on and the log likelihood value of the constant-only model was calculated as shown by Table G.2. The second iteration (Block 1: Enter Method), which includes all the independent variables, was proceeded. The log likelihood value and the coefficient parameters were estimated as presented in Table G.3. Therefore, the G statistics for the log likelihood ratio test between the full independent model and the constant-only model is calculated as

G = -2 [(log likelihood of smaller model) – (log likelihood of bigger model)] = 37.193 – 32.776 = 4.417

Table G.2: Iteration History (a,b,c) of Block 0: Beginning Block of Example Analysis of Logistic Regression

-2 Log Coefficients Iteration likelihood Constant Step 0 1 38.690 -1.538 2 37.225 -1.960 3 37.193 -2.035 4 37.193 -2.037 5 37.193 -2.037 a Constant is included in the model. b Initial -2 Log Likelihood: 37.193 c Estimation terminated at iteration number 5 because parameter estimates changed by less than .001.

Table G.3: Iteration History of Block 1: Enter Method of Example Analysis of Logistic Regression

-2 Log likelihood Coefficients Iteration Constant AGE OBE OBE x AGE EDU(1) EDU(2) Step 1 1 35.928 -1.337 -.004 -.326 .022 -.384 -.142 2 33.086 -1.502 -.011 -.672 .039 -.812 -.254 3 32.787 -1.320 -.019 -1.131 .052 -1.106 -.289 4 32.776 -1.225 -.022 -1.294 .057 -1.174 -.291 5 32.776 -1.219 -.022 -1.302 .057 -1.177 -.291 6 32.776 -1.219 -.022 -1.302 .057 -1.177 -.291

G-11 The degrees of freedom are five (the difference between degrees of freedom of the full model (df=6) and the constant-only model (df=1)). Thus the p-value of this test is P[χ2(5)>4.417] = 0.491 which is non-significant at the α = 0.05 level. It can be concluded that all the independent variables provide non-significant contribution to the model. As previously mentioned, the objective of this section is to explain the process of logistic regression analysis. The sample size of this example model is very small thus the non-significant contribution of independent variables to the model is expected.

For illustrative purposes, ignore the non-significance of the independent variables. The following information provided by Table G.4 is the fit of the logit model, written as:

^ g(x) = (-1.219) + (-1.302)OBE + (-0.022)AGE + (-1.177)EDU(1) + (-0.291)EDU(2) + (0.057)OBE×AGE

The Wald test of an individual coefficient and its p-value are presented by column four and five of Table G.4. At α = 0.05 level, it is obviously similar to the previous conclusion that the independent variables are not significant.

The odds and the confidence interval of individual coefficient were tabulated in columns six to eight. For example, the 95% confidence interval of the odds of age variable is 0.978 ± 1.96 (0.046) which yields (0.893, 1.071).

Table G.4: Variables in the Equation of Example Analysis of Logistic Regression

B S.E. Wald df Sig. Exp(B) 95.0% C.I.for EXP(B) Lower Upper Step Age -.022 .046 .230 1 .632 .978 .893 1.071 1(a) Obesity -1.302 4.561 .082 1 .775 .272 .000 2075.397 Age by .057 .081 .495 1 .482 1.059 .903 1.240 Obesity Edu .624 2 .732 Edu(1) -1.177 1.491 .623 1 .430 .308 .017 5.729 Edu(2) -.291 1.074 .073 1 .787 .748 .091 6.139 Constant -1.219 2.266 .289 1 .591 .296 a Variable(s) entered on step 1: Age, Obesity, Age * Obesity , Edu.

G-12 Since all the confidence interval values range over 1.00 at α = 0.05 level, the odds of individual coefficient are therefore not statistically significant. From equation (G.7), the odds ratio of this example model is

odds ratio = (odds for full model) / (odds for constant-only model)

^ ^ = exp [ β1 + β 5 AGE] = exp [-1.302 + (0.057)AGE]

For example, controlling for the highest education level and age and assuming that there is an interaction between obesity and age, the probability of hypertension development of people aged 55 with obesity is 6.25 times higher than people at the same age but without obesity.

odds ratio = exp [-1.302 + (0.057)(55)] = exp [1.833] = 6.25

The confidence interval of the odds ratio was checked through equation (G.8). The

^ ^ ^ ^ variance of β1 + β 5 AGE (called Var(K )) is defined as:

^ ^ ^ ^ ^ ^ ^ ^ ^ 2 Var(K) = Var(β 1 ) + AGE Var(β j ) + 2(AGE.Cov(β1 , β j )

The variance of the variable is equal to the square of the standard error of the variable.

^ ^ ^ ^ 2 2 Therefore, from Table G.4, Var(β1 ) = SE (β1 ) = (4.561) = 20.803 and

^ ^ ^ ^ ^ ^ 2 2 Var(β j ) = SE (β j ) = (0.081) = 0.00656. The covariance of β1 and β j was defined as

^ ^ correlation coefficient of β1 and β j multiplied by the standard error of both variables.

^ ^ ^ Therefore, from Table G.5, Cov(β1 , β j ) = (-0.967)(4.561)(0.081) = 0.3572. At AGE

^ ^ equals 55, the Var(K ) equals 20.803 + (552)(0.00656) + 2(55)(0.3572) = 79.939.

G-13 At age 55, a 95% confidence interval of the odds ratio will be exp [-1.302 + (0.057)(55) ± 1.96 79.939 ] = (1.52E-07, 2.55E08). It is obvious that the 95% confidence interval of the odds ratio is not statistically significant.

Table G.5: Correlation Matrix of Example Analysis of Logistic Regression

Constant AGE OBESITY AGE by OBESITY EDU(1) EDU(2) Constant 1.000 -0.907 -0.411 0.493 -0.195 -0.452 AGE -0.907 1.000 0.404 -0.543 0.014 0.152 OBESITY -0.411 0.404 1.000 -0.967 0.311 0.019 AGE by OBESITY 0.493 -0.543 -0.967 1.000 -0.316 -0.010 EDU(1) -0.195 0.014 0.311 -0.316 1.000 0.383 EDU(2) -0.452 0.152 0.019 -0.010 0.383 1.000

G.9 VARIABLE SELECTION STRATEGY

The previous section has presented an example calculation of logistic regression analysis. It involved a small number of independent variables and all of them were put into the fit model. However, for the main health and well-being survey where there are many independent variables involved, it is very important to recognise that it is not a good idea to include a large number of independent variables in the model. The more variables included, the greater the estimated standard errors become (Hosmer and Lemeshow, 2000, p.92). The variable selection strategy, therefore, should be performed carefully. The following sections demonstrate variable selection strategies and methods for modelling the logistic regression based on the recommendation provided by Hosmer and Lemeshow (2000, pp.91-116).

The first step of variable selection strategy should begin with a careful univariable analysis of each variable. The log likelihood ratio test with one degree of freedom (the difference between degree of freedom of the univariable model (df =2) and the constant-only model (df =1)) should be calculated with its associated p-value. Any variable whose univariable test has a p-value ≤ 0.25 (Hosmer and Lemeshow, 2000, p.95) is a candidate for the multivariable model along with all variables of known clinical importance (such as age and gender) and the set up exposure variable. Once the variables have been identified, the multivariate analysis is performed.

G-14 The second step of variable selection strategy is to examine the Wald statistic (z) and its p-value for each variable resulting from the multivariate analysis. Variables that have non-significance of either p-value or Wald statistic (z) should be discarded. The variable that has not been selected for the original multivariable model may be added back into the model at this step. The addition variable(s) should be checked by the Wald statistic (z) and the likelihood ratio test. The process of deleting, refitting, and verifying continues until the analyst confidently believes that all of the important variables are included in the model and those excluded are clinically and/or statistically unimportant (Hosmer and Lemeshow, 2000, p.97). The model at the end of second step is called the preliminary main effects model.

The next step of variable selection strategy is to test the linearity of the continuous variable in the logit. Note that this step can be ignored if there is no continuous variable included in the model. The linear relationship between the logit and the selected continuous independent variable is necessary since the rationale for development of the logit is based on a linear model. Continuous independent variables that do not satisfy the linearity test should be transformed to the other form of variables such as dichotomous.

This research accepts two methods to test the linearity of a continuous variable: (1) design variables and (2) fractional polynomials (Hosmer and Lemeshow, 2000, p.99). The design variable method (called the graphical method) is performed by obtaining the quartiles of the distribution of the variable. Then create a categorical variable with 4 levels using three cut-off points based on the quartiles. Fit the multivariate model replacing the continuous variables with the created categorical variables. It is noted that, as k = 4, three design variables (k-1) must be used with the lowest quartile serving as the reference group.

The estimated coefficients are plotted against the midpoint of each group. The estimated coefficient of the reference group is zero. The analyst visually inspects the shape of the plotted curve and decides the most logical parametric shape(s) (i.e., linear, quadratic) for the scale of the variable. The model is refitted by using the selected parametric forms suggested by the plot. The form that is most significantly different from the linear model and makes clinical sense will be chosen. G-15 The fractional polynomials method (called the analytical method) has a general formula as:

J g(x, β ) = β 0 + ∑ F j (x)β j j=1 where

Fj(x) is a particular type of power function.

The function of Fj(x) is defined as:

p j Fj(x) = x when pj ≠ pj-1 or

= Fj-1(x) ln(x) when pj = pj-1 for j = 2, …, J

It is noted that pj can be any value of number, however Hosmer and Lemeshow (2000,

p1 p.100) suggest a set of {-2, -1, -0.5, 0, 0.5, 1, 2, 3}, where p1 = 0 the x equal ln(x). -1 For example, if J =2 with p1 = -1 and p2 = -1, then the logit is g(x, β) = β0 + β1 x + -1 2 3 β2 x ln(x). If J =2 with p1 = 2 and p2 = 3, then the logit is g(x, β) = β0 + β1 x + β2 x . Hosmer and Lemeshow (2000) recommended J equal 1 or 2. For J =1, eight models are required, and for J =2, thirty-six models are required. The best model is the one with the largest log likelihood. The degree of freedom for linear model, J = 1 and 2 are 1, 2 and 4, respectively. The key issue is whether either of the two best transformed models is significantly better than the linear model. The likelihood ratio test is performed and if the p-value is less than 0.05, the selected continuous dependent variable should be transformed. We refer to the model at the end of this step as the main effects model.

The final step of variable selection strategy is to check for interactions among the variables in the model. Any interaction will be added to a model if it is statistically significant based on likelihood ratio test (the main effects model vs. the main effects model plus an interaction). The final consideration should be made that any interaction term in the model must make sense from a clinical perspective. At the end of this final step, the model is called the preliminary final model. Before using any model for inferences, the fit of the model must be assessed.

G-16 G.10 GOODNESS-OF-FIT

The technique adopted by this thesis for evaluating the goodness-of-fit is based on Hosmer-Lemeshow Tests. The subjects are divided into 10 groups (g = 10) according to their estimated probability; those with estimated probability less than 0.1 will be allocated in the lowest group, and so on, up to those with estimated probability more than 0.9 which will be placed in the highest group. Subjects in each group are divided into two groups based on the outcome variable (for instance, hypertension and normotensive). The observed frequencies and the expected frequencies (calculated by the preliminary final model) are compared using the Hosmer-Lemeshow goodness-of-fit statistic. The distribution of the Hosmer-Lemeshow goodness-of-fit statistic is well approximated by the chi-square distribution with g-2 degree of freedom. The good model produces a nonsignificant chi-square (p-value > 0.05). The mathematical details of Hosmer-Lemeshow goodness-of-fit statistic is beyond the scope of this thesis but can be found in Hosmer and Lemeshow (2000, pp.147-156). Statistical software (such as SPSS) provides a function to calculate a Hosmer-Lemeshow goodness-of-fit statistic. At the end of this section, the model is referred to as the final model.

G-17