The Perth Photochemical Smog Study

A joint project of

Western Power Corporation

and the

Department of Environmental Protection

with contributions from

Environment Protection Authority of VictoriaCSIRO Division of Atmospheric Research

CSIRO Division of Coal and Energy TechnologyFlinders Institute for Atmospheric and Marine Sciences

Western Australian Department of TransportUniversity of Adelaide Department of Community Medicine

Murdoch University

May 1996

Western Power Corporation

Report CS20/96363 Wellington Street

Perth, Western Australia 6000

Department of EnvironmentalProtectionReport 16

141 St Georges TerracePerth, Western Australia 6000

PREFACE

This report on the Perth Photochemical Smog Study was prepared for submission to the EnvironmentalProtection Authority (EPA). Completion of the study to the satisfaction of the EPA was a condition of approval,imposed by the Minister for the Environment in August 1991, on a proposal by SECWA (now Western PowerCorporation) to expand the Pinjar Gas Turbine Power Station.

The Perth Photochemical Smog Study, while being the responsibility of Western Power Corporation, wasundertaken as a joint project with the Department of Environmental Protection (DEP) in view of the widerbenefits of understanding photochemical smog development in the Perth region. (Before 11 January 1994, theDEP and the EPA were a single organisation known as the Environmental Protection Authority. For simplicityin this report, the name DEP is used as if it existed throughout the whole study, thereby avoiding confusion withthe EPA, which is an independent statutory authority.)

As is apparent from the context of the study, the scope of work was limited to scientific investigation anddevelopment. Development of photochemical smog control strategies was not part of the scope of work for thisstudy, but may proceed on the basis of study findings.

ACKNOWLEDGEMENTS

The following persons and organisations are acknowledged for their contribution to the nominated aspects of thePerth Photochemical Smog Study.

Scoping and conceptual design of the study: Roman Mandyczewsky(1) and Ken Rayner(2).

Joint management of the study: Pel Weir(1) and Ken Rayner.

Preparation and editing of the final report: Peter Rye(2), Ken Rayner and Pel Weir.

Contributors to the study report text: Charles Guest(3) and Robert Hughes(4) . Dr Guest prepared those sectionsrelating to health effects.

Installation and operation of the monitoring network and data management systems: Robert Kleinfelder(2), StevePrice(2), Arthur Grieco(2), Steve Lang(2), Peter Mountford(2), Bob Crowe(2), Henry Berko(2), Peter Rye, DavidGriffiths(2), Pel Weir and Iain Cameron(2).

Atmospheric profiling: Peter Mountford, Gene Tulimowski(10), Bureau of Meteorology staff.

Preparation of the emissions inventory: Pel Weir, Bruce James(5), Anthony Stuart(6), Frank Carnovale(6), MartinCope(6), Peter Rye, Otto Muriale(2) and Brendon Hewson(2).

Intensive field measurement program:

management: Ken Rayner, Pel Weir, Peter Rye and Graham Johnson(7).

aircraft and truck measurements: Jorg Hacker(8), Neville Clark(8) and staff, John Carras(7), David Williams(7),Peter Nelson(7), Mark Drummond(7) and Tony Lange(7).

measurements of reactive organic compounds: Ian Galbally(9), Simon Bentley(9), Ian Weeks(9), Rose Ye(9) andMalcolm Elsworth(9), Pel Weir and Henry Berko.

radiosonde program: David Griffiths, Peter Mountford, Michael Bell(2), Adrian Blockley(10), Ian Yull(2), ChrisTushingham(1), Neal Masters(1), Tyne Archer(2), staff of the Bureau of Meteorology, students from MurdochUniversity and UWA and others.

Modelling and data analysis: Peter Rye, Martin Cope, Josef Ischtwan(6), Julie Noonan(9), Peter Manins(9), PeterHurley(9), Graham Johnson, Merched Azzi(7), Adrian Blockley.

Associate Professor Rod Simpson of Griffith University and Dr Neville Bofinger of the Queensland Universityof Technology reviewed the study report.

The study was funded by Western Power Corporation, with the Department of Environmental Protectionproviding facilities, scientific expertise and resources.

This report was produced using Microsoft Word for Windows 2.0, with graphics also by CorelDRAW 4 andlocally-created software.

(1) Western Power Corporation.(2) Department of Environmental Protection.(3) University of Adelaide Department of Community Medicine.(4) Visiting Environmental Officer from the Department of the Environment, New Brunswick, Canada.(5) WA Department of Transport.(6) Environment Protection Authority of Victoria.(7) CSIRO Division of Coal and Energy Technology.(8) Flinders Institute for Atmospheric and Marine Sciences.(9) CSIRO Division of Atmospheric Research.(10) Murdoch University.

CONTENTS

Preface

Acknowledgements

Executive Summary ................................................................................................................................ I

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

1.1. HISTORY 1

1.2. OBJECTIVES 2

2. The Nature of Photochemical Smog.......................................................................................... 3

2.1. SMOG CONSTITUENTS 3

2.2. THE CHEMISTRY OF SMOG FORMATION 4

2.2.1. Natural Ozone Cycle 4

2.2.2. Ozone Cycle In Polluted Air 4

2.3. SOURCES 5

2.3.1. Fuel Combustion 5

2.3.2. Petroleum Products 6

2.3.3. Natural and Miscellaneous Sources 6

2.4. HEALTH EFFECTS OF PHOTOCHEMICAL SMOG 6

2.4.1. Air Pollution and Health 6

2.4.2. Methods for the Investigation of Health Effects 6

2.4.3. Overview of Smog Health Effects 7

2.4.4. Summary 9

2.5. OTHER EFFECTS 9

2.5.1. Effects on Vegetation 9

2.5.2. Protective Standards for Vegetation 10

2.5.3. Effects on Materials 10

2.6. PHOTOCHEMICAL SMOG AROUND THE WORLD 11

2.7. AIR QUALITY STANDARDS 13

2.8. CONTROL MEASURES 15

2.8.1. Industrial Control Measures 15

2.8.2. Motor Vehicle Controls 16

2.8.3. Other Sectors 17

2.8.4. Episode Control 18

2.8.5. Economic Instruments 18

3. Perth’s Photochemical Smog Climate ..................................................................................... 19

3.1. METEOROLOGY AND ATMOSPHERIC DISPERSION 19

3.2. THE SEA BREEZE 19

3.3. SMOG CONDITIONS 20

4. Overview of Activities ............................................................................................................... 22

4.1. CONTINUOUS AIR QUALITY AND METEOROLOGICAL SURFACEMEASUREMENTS 22

4.1.1. Monitoring Site Locations 22

4.1.2. Overview of the Air Quality Monitoring Stations 23

4.1.3. Daily Telemetry 26

4.1.4. Instrument Calibrations 26

4.1.5. Data Processing 27

4.2. BUREAU OF METEOROLOGY SURFACE MEASUREMENTS 27

4.3. VERTICAL MEASUREMENTS OF WIND AND TEMPERATURE 27

4.3.1. Sodar Sites 27

4.3.2. Radar Profiler 29

4.3.3. Perth Airport Radiosonde and Wind Data 30

4.4. EMISSIONS INVENTORY 30

4.4.1. Motor Vehicle Emissions 31

4.4.2. Industrial Emissions 31

4.4.3. Area-Based Emissions 32

4.4.4. Biogenic Emissions 32

4.5. FIELD EXPERIMENTS 33

4.5.1. Aircraft Survey 33

4.5.2. Truck Survey 34

4.5.3. ROC Speciation Sampling 35

4.5.4. Radiosonde and Pibal Releases 36

4.6. DATA STORAGE AND AVAILABILITY 37

4.7. COMPUTER MODELLING 38

5. Summary and Analysis of Smog Measurements.................................................................... 39

5.1. AIR QUALITY DATA 39

5.1.1. Ozone Concentrations 39

5.1.2. Nitrogen Dioxide Concentrations 44

5.1.3. Fine Particle Concentrations 45

5.2. ASSESSING THE IMPACT OF SMOG LEVELS IN PERTH 46

5.2.1. Assessing Health Risks of Smog Levels 46

5.2.2. Assessing the Potential for Vegetation Impacts 47

5.3. METEOROLOGY OF SMOG EVENTS 48

5.3.1. Inland Events 48

5.3.2. Kwinana Events 49

5.3.3. Coastal Events 50

5.3.4. Bushfire Smoke Events 50

5.3.5. Light Westerly Wind Events 51

5.4. ANALYSIS USING THE INTEGRATED EMPIRICAL RATE METHOD 52

5.4.1. Description of the Integrated Empirical Rate Model 53

5.4.2. Application of the IER Model 56

6. Emissions Analysis ................................................................................................................... 58

6.1. MOTOR VEHICLE EMISSIONS 58

6.1.1. Methodology 58

6.1.2. Spatial and Temporal Distribution 59

6.1.3. Temperature Effects 59

6.1.4. Effects of Emission Controls 60

6.1.5. Verification of Motor Vehicle Emission Estimates 61

6.2. INDUSTRIAL EMISSIONS 63

6.2.1. Distribution Across Industry 63

6.2.2. Seasonal Variation in Industrial Emissions 65

6.2.3. Aircraft Validation of Industrial Emission Estimates 65

6.2.4. Evaluation of the Reactivity of Emissions from Kwinana 66

6.3. AREA-BASED EMISSIONS 67

6.3.1. Contributions to Area-based Emissions 67

6.3.2. Spatial and Temporal Variations 68

6.3.3. Verification of Area-based Emissions 69

6.3.4. Trends for Area-based Emissions 69

6.4. BIOGENIC EMISSIONS 69

6.5. SUMMARY 70

7. Modelling .................................................................................................................................... 71

7.1. METEOROLOGICAL MODELLING 71

7.1.1. Modelling Using 3DSB 73

7.1.2. Modelling Using LADM 75

7.2. PHOTOCHEMICAL MODELLING 76

7.2.1. The CIT Model 76

7.2.2. The Urban Airshed Model 79

7.3. METHODS FOR ASSESSING TRENDS AND CONTROLS 81

7.3.1. Sensitivity to Possible Trends 81

7.3.2. Effects of Expected Trends 82

7.4. SUMMARY OF MODELLING 84

8. Conclusions and Recommendations....................................................................................... 85

8.1. SUMMARY OF FINDINGS 85

8.2. RECOMMENDATIONS FOR FURTHER WORK 86

8.3. CONCLUSIONS 86

9. References.................................................................................................................................. 88

Appendix Glossary of Chemical and Mathematical Symbols......................................................... 93

I

OVERVIEW OF PHOTOCHEMICALSMOG

Photochemical smog is characterised by highconcentrations of ozone at ground level. It formswhen urban air pollutants, principally nitrogen oxides(NOx) and reactive organic compounds (ROC) frommotor vehicles and many other sources, react togetherfor a few hours under the influence of sunlight andhigh temperatures.

Ozone at high concentrations can reduce lungfunction and exacerbate asthma. It is also harmful tovegetation. In severe cases, photochemical smog istypified by a whitish haze and causes eye irritation.Photochemical smog is the type of air pollution oftenassociated with Los Angeles, but is now known tooccur around many of the world’s large cities.

Photochemical smog requires strong sunlight andrelatively high air temperature to reach significantconcentrations. Hence, in the case of Perth, it islimited to the period late spring through to earlyautumn. By way of contrast, highly visible smokehaze occurs more frequently in the colder months, ondays when smoke from domestic fires and othersources is trapped near the ground under atemperature inversion.

BACKGROUND TO THE PERTHPHOTOCHEMICAL SMOG STUDY

The Department of Environmental Protection (DEP)began measurements of photochemical smog inPerth’s air in 1989, at a single site in the suburb ofCaversham, 15 kilometres NE of the city centre.Despite the common perception that Perth is a windycity and therefore not prone to air pollution, the firstsummer of measurements revealed that the city wassometimes subjected to smog levels whichapproached or exceeded the guidelines recommendedby the National Health and Medical Research Councilof Australia (NHMRC).

In 1991 the State Energy Commission of WesternAustralia (SECWA, now Western Power Corporation)sought to extend the capacity of the gas turbine powerstation it operated at Pinjar, some 40 kilometres northof the Perth central business district. In view of theCaversham data, the Environmental ProtectionAuthority expressed concern that increasing the NOxemissions at Pinjar could contribute to Perth’s

emerging photochemical smog problem which, at thatstage, was poorly defined.

A consequent condition on the development at Pinjarwas that SECWA undertake a study of the formationand distribution of photochemical smog in Perth, aparticular outcome of which would be to determinethe effect of the Pinjar power station’s emissions onsmog in the region.

Given the DEP’s concerns and responsibilit y inrelation to urban air quality, the Perth PhotochemicalSmog Study (PPSS) was developed as a jointlyoperated and managed project, funded by SECWAand with DEP contributing faciliti es and scientificexpertise.

Study activities started in early 1992.

PHOTOCHEMICAL SMOG LEVELS INTHE PERTH REGION

The primary objective of the Perth PhotochemicalSmog Study was to measure, for the first time, themagnitude and distribution of photochemical smogconcentrations experienced in the Perth region and toassess these against Australian and internationalstandards, with consideration given to health andother environmental effects.

Summary of Measurements

The study’s monitoring and data analysis programwas very successful in defining the distribution ofPerth’s smog.

The Perth region experiences photochemical smogduring the warmer months of each year. On average,during the three year period July 1992 to June 1995,there have been 10 days per year on which the peakhourly ozone concentration has exceeded 80 parts perbilli on (ppb) somewhere over the Perth region (Figure1).

Canada and the World Health Organisation have setgoals equal to, or close to, 80 ppb as the level beyondwhich ozone concentrations become unacceptable. InAustralia, the NHMRC has set a 1-hour average of100 ppb and a 4-hour average of 80 ppb as goalswhich should not be exceeded (NHMRC 1995). Thesegoals are both exceeded about two times per yearsomewhere over Perth. It is worth noting that theprofessional review commissioned by the NHMRC

Executive Summary

II

before revising the goals (Woodward et al. 1993)recommended a 1-hour goal of 80 ppb on the basis ofrecent health effects research and the magnitude ofpotential exposures in Australia. The previousNHMRC goal of 120 ppb for one hour has beenexceeded only twice at Caversham since monitoringbegan in 1989.

Ozone events occur across a large region surroundingthe city. Figure 2 shows that monitoring sitescovering an area extending from Rottnest to RollingGreen (near Toodyay) and Two Rocks toRockingham recorded exceedances of 80 ppb duringthe study.

Nevertheless, the air over Perth on most summer daysis relatively clean, due to the windy climate andisolation from other cities. Ozone events, except those

associated with bushfire smoke, tend to be of shortduration (an hour or two) following the arrival of thesea breeze. This contrasts with regions of the USAand elsewhere which experience elevated ozone levelsover many hours or days, caused by the transport ofpollutants from distant cities and industrial areas.

Based on a comparison of measurement statistics, it isapparent that Perth experiences photochemical smogto an extent similar to Brisbane, greater than Adelaideand somewhat less than Sydney and Melbourne(Figure 3). Given the differences in population andvehicle numbers between Perth, Melbourne andSydney, it may be surmised that Perth has a potentialfor smog problems similar to that of the larger cities.

Year

0

50

100

150

200

250

300

79 81 83 85 87 89 91 93

Ozone

(ppb)

MelbourneSydneyBrisbaneAdelaidePerth

Figure 3. Peak 1-hour ozone concentrations in Australian capitalcities. Averages are for clock hours, except for Queensland values,

which are based on rolling half-hour averages.

Concentrations of nitrogen dioxide approached theNHMRC goal in central Perth on some occasions butwere well below the goal at other monitoring sites.Concentrations of fine particulate matter were lowduring high ozone events, except where theseconcentrations were caused by bushfire smoke.

Assessment of Health andEnvironmental Effects

In his assessment of the study data, Dr Charles Guestof the University of Adelaide Department ofCommunity Medicine concluded that current healtheffects of photochemical smog in Perth, includingreductions in lung function and increased risk ofasthma attack, were probably mild. He noted howeverthat the number of times that the new NHMRC goalswould be exceeded was set to grow rapidly ifemissions and the consequent smog concentrationswere to increase by modest amounts. He thereforerecommended, from the perspective of public health,that control of photochemical smog should beconsidered now.

0

1

2

3

4

5

Year/Month1992 1993 1994 1995

Days

Figure 1. Number of days per month on which peak 1-hour ozoneconcentrations exceeded 80 ppb somewhere in the Perth region.

0 5 10 15 20

Scale (km)

Quinns Rocks

Two Rocks

Caversham

Kenwick

Rolling Green

Cullacabardee

Maximum Hourly Average Ozone

1 October 1992 to 31 March 1995

KEY

Shaded box representshighest measurement

Bold box represents80 ppb WHO goal

Swanbourne

Rottnest

Rockingham

Figure 2. Maximum hourly average ozone concentrations duringthe period of PPSS. The Rockingham site only operated from 28

January to 28 February 1994.

III

Vegetation impacts from ozone in the Perth regionwere not measured as part of this study but are likelyto be low in light of the relatively low frequency ofsmog events and the low concentrations of ozoneexperienced most of the time. However, an 8-houraverage of 50 ppb, set by the Victorian EPA as theacceptable limit for protection of vegetation, isexceeded several times per year in the Perth region. Itis possible therefore that vegetation in the Perthregion (including horticultural crops) experiencestransient stress.

UNDERSTANDING SMOGDEVELOPMENT OVER PERTH

The secondary set of objectives for the study relatedto assembling the information and tools needed tounderstand and be able to predict photochemicalsmog development over Perth.

In addition to desktop studies and computer analyses,a major field experiment was conducted to intensivelygather information on emissions and smogdevelopment. The field experiment centred around theuse of a research aircraft operated by the FlindersInstitute for Atmospheric and Marine Sciences andthe CSIRO, supported by specialised ground-basedmeasurements.

A summary of f indings, including those provided bycomputer models, is given below.

Meteorology of Smog Events

There is a dominant, well -defined weather patternrelated to Perth’s smog events. The highest smogconcentrations were found to occur on those daysduring spring to autumn when a weak low pressuretrough is situated very close to the coast (as seen in

Figure 4) and subsequently crosses the coast in theafternoon.

Under these conditions, emissions from morning peakhour traff ic are blown off-shore by north easterlywinds into the light wind region of the trough (Figure5), where smog reactions proceed rapidly under thetypically high temperatures. A strong temperatureinversion, also typical of these conditions, keeps thesmog plume concentrated near the ocean surface.

The study has revealed two well -defined classes oftrough-related smog events, namely:

• inland events, which show a pattern of peakconcentrations in the eastern metropolitan area.These correspond to days when the trough movesinland, and recirculates the smog plume backacross the metropolitan area where it receives aboost from fresh afternoon emissions; and

• coastal events, in which high concentrations ofsmog form in a warm, stable air mass offshore,are returned across the coast by the sea breeze,but are then dispersed rapidly in unstableconditions over the land.

It also became apparent in the course of the study thatsome smog plumes may return on-shore well south ofthe monitoring network (e.g., near Mandurah).

The behaviour of the trough on any particular day isdiff icult to forecast with the accuracy needed for

Kwinanaindustrial area

Perth centralbusiness district

Figure 5. The path of air masses offshore, then onshore, on atypical Perth smog day (21 March 1994). The black line

shows the path of air which passed the Perth central businessdistrict in the morning, and the shaded line shows that for

the air which passed the Kwinana industrial area.

1016

1018

1020

1012

30

25

35

40105 110 115 120 125

H

1014

1016

1024

1026

1018

L

1020

1022

Figure 4. Pressure chart for the morning of 16 March 1994, when atypical example of a coastal low pressure trough was present.

IV

smog event predictions. Nevertheless, there is someprospect of generating reliable smog forecasts shouldthat become necessary in the future.

Significant smog concentrations were observed inother conditions, particularly when bushfire smoke,which contains high concentrations of ROC, wasblown across the metropolitan area during daytime.

Compilation of an EmissionsInventory

The major sources of emissions which lead to theformation of photochemical smog were quantified infour categories, as follows:

• Motor vehicles: Vehicle movement data, from aMain Roads Western Australia computer modeland traff ic counts, were combined with emissionsdata for the Perth vehicle fleet, to estimate hourby hour contributions to emissions across theregion.

• Industrial emissions: Individual industries wereidentified from several sources of information,including Australian Bureau of Statistics andGovernment agency data bases. Emissionestimates for each significant industry werederived from questionnaire responses and/orengineering calculations.

• Area sources: This category includes a broadrange of sources such as service station vapourlosses, surface coatings and thinners, natural gasleakage, bitumen, lawn mowing, domesticsolvents and off-road engines. Emissions wereestimated from Australian Bureau of Statisticsinformation and other statistics and apportioned

across the metropolitan area according topopulation density.

• Biogenic emissions: This category representsemissions of reactive organic species fromvegetation, which are significant in thephotochemical smog process. The approach takenin estimating biogenic emissions was to relatethem to vegetation density using satellit ephotographs. The vegetation species distributionfor the Perth region was derived using naturalvegetation maps. Published canopy emissionrates were then used to derive biogenic emissionestimates for the region.

Emissions estimates of ROC, NOx, carbon monoxide,sulphur dioxide and particulates from all sourcegroups were combined into a composite data base,giving hour-by-hour emissions for three day types(average winter day, average summer day and hotsummer day) for each cell of a 3 by 3 km grid acrossthe region.

Figure 6 shows, as an example, the relativecontributions from the first three source categories(all due to human activity) to annual ROC and NOxemissions. Biogenic emissions have been excludeddue to the uncertainty attached to their estimation.

On an annual average basis, motor vehicles accountfor 51% of the total NOx and 44% of the total ROCemissions caused by human activity in the region.Furthermore, motor vehicle ROC emissions arerelatively more reactive than other urban sources,which account for a further 37% of the ROC.

Computer Modelling ofPhotochemical Smog

Several characteristic smog events were simulated bycomputer models which represented the dynamics ofregional meteorology, emissions and photochemicalprocesses. The validity of model predictions wasdemonstrated with some success, by comparison withmeasurements.

The inventory of smog precursor emissions, coupledwith modelli ng results, confirmed that motor vehiclesare the dominant cause of Perth’s smog. Tests of thesensitivity of model results to variations in emissionswere carried out, as ill ustrated in Figure 7 for one ofthe smog events. These tests showed that reduction ofmotor vehicle ROC emissions (with or withoutvehicle NOx reductions) would yield the greatestbeneficial reduction of ozone, whereas reduction ofNOx from vehicles in the absence of ROC reductionscould paradoxically lead to an increase in peak ozone(although not for the 18 March 1994 as ill ustrated). Insummary, modelli ng results point to control of motorvehicle kilometres travelled and ROC emissions as

MotorVehicles

51%

Industrial44%

Area5%

NOx

ROC

Industrial19%

Area37% Motor

Vehicles44%

Figure 6. Relative contributions to NOx and ROC emissions,excluding biogenic sources, for the PPSS region.

V

the most beneficial options to reduce regional ozonelevels.

However it needs to be noted that the accuracy of thechemistry models has not been demonstrated (inAustralia or elsewhere) for the low ratios of ROC toNOx found in Perth’s air. Under such conditions, themodels predict that ozone production will besuppressed when more NOx is added. Furtherassessment of the models is required.

The Kwinana industrial area was identified in theemissions inventory as a major source of NOx and alesser source of ROC. The resultant effect seen inmodelli ng predictions was a significant suppression ofozone across those portions of the metropolitan areaimpacted by the Kwinana NOx plume, for the reasonexplained above. Two issues in relation to Kwinanaemissions require further investigation as a matter ofpriority:

• whether, as indicated by independentmeasurements, the ROC emissions from Kwinanaare higher, by a factor of two to four, than thosecalculated in the emissions inventory;

• whether the models correctly simulated thebehaviour of the bulk of Kwinana NOx which ispresent in elevated narrow plumes from tallchimneys.

The potential significance of biogenic emissions ofROC from natural vegetation was recognised.Estimation of the magnitude and reactivity of theseemissions was necessarily coarse. Modelli ng

confirmed that this source of emissions may be asignificant contributor to the magnitude and extent ofhigh ozone concentrations. Further work is needed toimprove the accuracy of biogenic emissionsestimates.

The above-mentioned computer models whichsimulate the three-dimensional meteorology andatmospheric smog chemistry are extremely complex.While the meteorological and smog dispersionmodelli ng was based on the research developments ofthe Environment Protection Authority of Victoria(EPAV), CSIRO and DEP, no attempts were made toimprove upon the atmospheric chemistry mechanismsemployed within the smog models (developed in theUSA). In total, six one-day smog events weremodelled by the EPAV, CSIRO and DEP.Configuring and testing of these models to achievevarying degrees of agreement with observations tookseveral months of intense effort.

The model simulations and associated data setsprovide a useful initial basis for formulating andtesting smog management strategies. However, it isclear that improvements and additions will need to bemade to this collection of smog event simulations toprovide a reliable and representative basis forpredicting trends and testing smog managementstrategies. Continued monitoring of regional smoglevels and meteorological parameters is essential, bothto obtain data to characterise individual smog eventsfor modelli ng purposes and to maintain a record ofsmog levels over time. This record is needed toestablish trends and to determine the effectiveness ofmanagement strategies.

THE NEED TO MANAGE PERTH’SSMOG

Perth experiences photochemical smog levels whichexceed goals set by the NHMRC and other bodies.The potential exists for the problem to grow.

Control of photochemical smog is complex. It is notamenable to simple and uniform solutions. A controlmeasure designed to improve air quality in oneportion of the metropolitan region might have anegative effect elsewhere. The consequences ofregulating NOx emissions are far from obvious.Given the potentially enormous costs to thecommunity, industry and Government of variouscontrol options, it is critically important to thoroughlyanalyse the cost-effectiveness of alternatives beforedecisions are made. Adequate analyses will t ake timeand resources.

Fortunately, Perth is not yet experiencing acute smogproblems. The planning, transport and environment

BC MV x 0

MVROCx 1.5

MVROCx 0.5

MVNOX x 1.5

MVNOXx 0.5

IND x 0

INDROCx 0

INDNOX x 0

DASx 0

BIO x 0

100

0

20

40

60

80

120

140

Figure 7. Estimated maximum ozone concentrations (ppb) in the

Perth airshed for the meteorology of 18 March 1994 and various

emission scenarios. Key to graph labels:- BC : base case inventory

used in model simulation; MV x 0 : motor vehicle emissions

scaled by zero (i.e., deleted); IND x 0 : industrial emissions scaled

by zero; DAS x 0 : domestic area source emissions scaled by zero;

BIO x 0 : biogenic emissions scaled by zero; similar notation for

ROC or NOx emissions scaled by 0.5 and 1.5. (Cope and Ischtwan

1995).

VI

agencies therefore have time to learn from theexperience of other cities and, in consultation withaffected parties, develop sound, cost-effective smogmanagement strategies.

The understanding of Perth’s smog and the modelli ngcapabilit y developed in this study have beenimmediately applied by Western Power Corporationto assess the impact of the Pinjar Gas Turbine PowerStation. This assessment will be the subject of aseparate report to the EPA; the DEP has had noinvolvement in its preparation.

Western Power Corporation does not have a centralrole in developing smog management strategies forthe Perth region. The development of strategies forconsideration by government will be progressed bythe DEP in consultation with relevant agencies andaffected parties. The Perth Photochemical SmogStudy has provided a sound foundation on which tobuild.

1

1.1. HISTORY

In large cities, emissions from everyday activities canhave an impact on air quality. Concern hashistorically centred on pollution which had obvioussources, such as sulphurous smoke emissions fromindustrial and domestic chimney stacks. Perth hasbeen relatively free of such pollution, although smokefrom domestic fires (the subject of a separate study) isof growing concern.

Since the 1940s, a class of pollution events with lessobvious origins has arisen around large cities andfurther afield. This type of pollution, named“photochemical smog” , was first observed in LosAngeles, Cali fornia. Analysis of a sequence of thesephotochemical smog events gave the first indicationsthat emissions from motor vehicles were a majorcontributor.

Photochemical smog, characterised by highconcentrations of ozone at ground level, forms whenurban air pollutants (principally nitrogen oxides andreactive organic compounds from motor vehicles andmany other sources) react together for a few hoursunder the influence of sunlight and elevatedtemperatures. Ozone at high concentrations mayreduce lung function and may adversely affectasthmatics. In severe cases, photochemical smog istypified by a whitish haze and eye irritation.Nevertheless, significant photochemical smogconcentrations may be present without a noticeablereduction of visibilit y.

Since photochemical smog requires strong sunlightand relatively high air temperature to reach significantconcentrations, it is limited to times of the year whenthese conditions prevail (e.g., late spring through toearly autumn in Mediterranean climates). By way ofcontrast, highly visible smoke haze occurs morefrequently in the colder months, on days when smokefrom domestic fires and other sources is trapped nearthe ground under a temperature inversion.

Measurements of photochemical smog in Perth’s airbegan in 1989. These were undertaken by theDepartment of Environmental Protection (DEP) at asingle site in the suburb of Caversham, 15 km NE ofthe city centre (Figure 1.1). Despite the commonperception that Perth is a windy city and therefore notprone to air pollution, the first summer ofmeasurements revealed that the city was sometimessubjected to smog levels which approached orexceeded the limits recommended by the National

Health and Medical Research Council of Australia(NHMRC).

In 1991 the State Energy Commission of WesternAustralia (SECWA, now Western Power Corporation)sought to extend the capacity of the gas turbine powerstation it operated at Pinjar, some 40 km north of thePerth central business district. Gas turbines, like allcombustion processes, produce nitrogen oxides(NOx) which are one of the precursors tophotochemical smog. In view of the Caversham data,the Environmental Protection Authority expressedconcern that increasing the NOx emissions at Pinjarcould contribute to Perth’s emerging photochemicalsmog issue which, at that stage, was poorlyunderstood.

SECWA’s efforts to address this issue in itsenvironmental impact assessment of the proposedextension were limited by the lack of suff icientinformation on the meteorological and air chemistrycharacteristics of the region. A consequent conditionon the development at Pinjar was that SECWAundertake a study of the formation and distribution ofphotochemical smog in Perth, a particular outcome ofwhich would be to determine the effect of the Pinjar

1. Introduction

Rottnest

Caversham

Pinjar

0 5 10 15 20

Scale (km)

Perth CBD

Kwinana

Fremantle

Figure 1.1 Location map of the Perth region.

2

power station’s emissions on smog in the region.

The Perth Photochemical Smog Study was thusoriginated. While it was a SECWA initiative with aspecific objective, it was clear from the outset that theresults of the study would be of great benefit in amuch broader context, namely providing a foundationfor strategic management of Perth’s air quality. Giventhe DEP’s concerns and responsibilit y in this area, thestudy was developed as a jointly operated andmanaged project, funded by SECWA and with DEPcontributing faciliti es and scientific expertise.

The study began in early 1992.

1.2. OBJECTIVES

The primary objective of the Perth PhotochemicalSmog Study was to measure, for the first time, themagnitude and distribution of photochemical smogconcentrations experienced in the Perth region and toassess these against Australian and internationalstandards. In this respect the study was different fromconcurrent studies and ongoing work in otherAustralian capitals (Sydney, Melbourne and Brisbane)where photochemical smog monitoring had beenundertaken for about two decades. A large part of theeffort expended in this study was directed towards theestablishment and maintenance of a monitoringnetwork.

The secondary set of objectives for the study relatedto assembling the information and tools needed tounderstand photochemical smog development overPerth. Specific objectives were:

• to identify and quantify all sources of smogprecursor emissions across the Perth region;

• to develop the capabilit y of predicting themagnitude and distribution of photochemicalsmog across the region, employing “state ofknowledge” computer models of the meteorologyand atmospheric chemistry;

• to develop an understanding, with the aid of thesecomputer models, of the meteorological regimeswhich give rise to significant photochemical smogevents; and

• to assess, with the aid of the computer models, thestatus of smog development in the Perth regionand the relative significance of various emissionsource groups.

The understanding of Perth’s smog and the modelli ngcapabilit y developed in this study have beenimmediately applied by Western Power Corporationto assess the impact of the Pinjar Gas Turbine PowerStation. This assessment will be the subject of a

separate report to the EPA; the DEP has had noinvolvement in its preparation.

It is the intention of the DEP to continue, after thestudy is complete, to develop and apply further thecomputer models and associated information toinvestigate likely trends in smog levels associatedwith the projected growth of Perth, and to assess themerits of possible smog control strategies. This typeof work is not within the charter of Western PowerCorporation and has not been undertaken as part ofthe study.

3

Urban photochemical smog was first identified in LosAngeles, USA during the 1940s, where it wascharacterised by an atmospheric haze andaccompanying irritation of the eyes and lungs. Thiskind of smog occurred under light wind conditions onwarm sunny days, whereas the term “smog” wasoriginally coined to describe a mixture of smoke andfog, which is more common in winter. The term maytherefore be misleading, since neither smoke nor fogare key components of photochemical air pollution.The term “photochemical” reflects the fact that suchsmog occurs due to the influence of strong sunlightand high temperatures on anthropogenic airpollutants; it is therefore a summertime phenomenon.Another more accurate term often used is“photochemical oxidants” .

In time, photochemical smog was being experiencedin many other cities of the world, particularly thosewhich were industrialised, had large motor vehiclefleets and were in latitudes where frequent warmsunny days were experienced. Australia was noexception, with Sydney and Melbourne sufferingsignificant smog events in the 1970s. In WesternAustralia, the city of Perth underwent rapid growth toa population of more than one milli on during the1980s, and associated anthropogenic emissionscombined with the city’s warm, sunny summers havecreated the potential for smog formation.

2.1. SMOG CONSTITUENTS

The constituents of photochemical smog are notemitted directly in significant amounts. They are thesecondary reaction products which result whennitrogen oxides (NOx) and reactive organic

compounds (ROC), which have numerous emissionsources, react together in the presence of sunlight.The principal component of photochemical smog isozone. Consequently, ozone concentration is used todefine smog levels. Some common terms used todescribe the groups of compounds associated withphotochemical smog formation are given in the boxbelow.

Normally a colourless gas, ozone occurs naturally inthe atmosphere. Concentrations in unpolluted, remoteregions at moderate elevations are currently 15 to 35parts per billi on (ppb) (Lefohn, Krupa, andWinstanley 1990). Ozone is a powerful oxidant whichreduces pulmonary function and can damagevegetation and susceptible materials such as rubberand cloth. Measurable effects (discussed later) onplants and animals are observed at levels of 40 to 80ppb and above. Such levels are normally only attainedas a result of ozone production from emissionsgenerated by human activity or from natural pollutionevents such as bushfires.

It is important to make a distinction between theozone in photochemical pollution at ground level,where high concentrations create serious pollution,and ozone in the upper atmosphere, where itspresence in the “ozone layer” is critical to protect theearth from excessive ultraviolet radiation.

Other typical constituents of photochemical smoginclude nitrogen dioxide, peroxyacetyl nitrate (PAN),nitric acid, aldehydes and organic and inorganicnitrates in particle form. PAN, nitric acid andaldehydes are all strong irritants of the eyes, nose andthroat.

Photochemical smog consists mainly of gases which

2. The Nature of Photochemical Smog

Common Abbreviations for Photochemical Smog Precursors

ROC: Reactive organic compound. An organic compound which can take part in chemical reactions in theatmosphere.

ROG: Reactive organic gas. Synonymous with ROC.

VOC: Volatile organic compound. A compound which has a high vapour pressure, such that it evaporatesreadily at ambient temperatures. Includes compounds which do not take part in photochemical smog reactions(for example chlorofluorocarbons, or CFCs).

NMHC: Non-methane hydrocarbons. A group of hydrocarbons which excludes methane, carbonyl compounds,alcohols, and halocarbons.

NMOC: Non-methane organic compound. All organic compounds other than methane.

NOx: Compounds of nitrogen and oxygen, usually taken to mean nitric oxide (NO) and nitrogen dioxide (NO2).

4

are generally invisible at low concentrations.However, at higher concentrations, condensed smogparticulates remain suspended in the air, causing haze.Although it is common to associate visual haze withphotochemical smog, it is possible to have significantphotochemical smog pollution without any obvioushaze and equally, to have atmospheric haze that is notthe result of photochemical smog. For example, hazeis also caused by dust and smoke emissions, andmarine salts.

2.2. THE CHEMISTRY OF SMOGFORMATION

Ozone, nitrogen oxides and hydrocarbons may takepart in numerous reactions in the atmosphere. The fullsuite of photochemical reactions which are thought tooccur is highly complex and no attempt will be madeto deal comprehensively with them in this report: onlythe basic reaction pathways will be described.

2.2.1. Natural Ozone Cycle

Ozone is generated naturally by the combination ofmolecular oxygen (O2) with atomic oxygen (O):

O + O2 + M → O3 + M (2.1)

Both monatomic oxygen and the ozone moleculeformed are unstable. For the reaction to proceed,excess energy must be absorbed by another moleculewhich does not take part in the reaction itself. Thismolecule, which could be nitrogen (N2) or anotheroxygen molecule (O2) is indicated by “M” in theequation. The free, unstable oxygen atom in equation2.1 results from the splitti ng of nitrogen dioxide bylight with a wavelength of 440 nanometres or less,which includes light in the far blue and violet parts ofthe visible spectrum:

NO2 + light → NO + O (2.2)

The oxygen atom split off is in an unstable, excited

state which promotes its combination with molecularoxygen in equation 2.1. Another important reaction isthat of ozone with nitric oxide:

O3 + NO → NO2 + O2 (2.3)

Considering just these three reactions, equations 2.1and 2.2 lead to ozone formation, whereas equation 2.3results in ozone removal. The final or equili briumozone concentration depends on the rates of eachreaction and the relative concentrations of NO2 andNO. If all the monatomic oxygen in equation 2.1leads to ozone formation, the rate of formation will bedetermined by the rate of dissociation of NO2 inequation 2.2, whereas the rate of destruction willdepend on how much NO is present. These reactionsrepresent one of the probable pathways by whichozone levels at remote locations are determined. Eventhough NO and NO2 are present at very lowconcentrations in unpolluted air (less than 0.5 ppb),their relative abundance is believed to controlbackground ozone levels in the troposphere (Seinfeld1986, Finlayson Pitts and Pitts 1986).

Other ozone sources in the absence of significantanthropogenic emissions include occasional direct“ injection” from the stratosphere (where it is formedin reactions involving the photodissociation ofmolecular oxygen) and another reaction seriesinvolving the oxidation of methane (CH4) and carbonmonoxide (CO) in the presence of hydroxyl radicals(OH⋅� ). The significance of ozone formation via theCH4/CO pathway is debatable.

2.2.2. Ozone Cycle In Polluted AirEmissions of NO are substantial in urban regions.High concentrations of ozone are prevented fromdeveloping in the presence of significant NO by thetitration reaction in equation 2.3. For high ozonelevels to occur, other reactions are required whichwill remove NO (by oxidation to NO2) at a ratesimilar to or faster than the rate at which the NO

Units of Measurement for Smog Gases

Conventionally, the concentration of any gas in the air is expressed as a density, for example, micrograms percubic metre (µg/m3), or a fraction by volume, such as parts per million (ppm). For pollutants occurring in lowerconcentrations, the fraction by volume is often expressed in parts per billion (ppb), where 1 ppb is1/1,000,000,000. Ozone concentrations during smog episodes in Australia often exceed 80-100 ppb, and mayreach 200 ppb.

Since the density of air itself varies with changes of pressure and temperature, a concentration expressed as adensity does not convert to a unique fraction. So, when a density is used, it is usually converted to the valuewhich would have occurred if the air had been at a standard pressure and temperature. The standards normallyused are 0oC and 1013.25 hPa (hectopascal).

With this adjustment, the conversion between density and fractional measures is a constant, depending only onthe molecular weight of the gas. For the 0oC reference, 1 ppb of ozone equates to 2.145 µg/m3. Some countriesuse a 20oC reference, giving a conversion factor close to 2.

5

reaction removes ozone, while not at the same timeleading to ozone removal. This perturbation of thenatural ozone cycle is provided by a series ofreactions involving ROC, most of which are emittedin the form of hydrocarbons. A key fast oxidationreaction for NO is the reaction with peroxy radicals(RO2⋅� ). The peroxy radicals are formed by thereaction of an organic, for example a hydrocarbon(RH) with the hydroxyl radical (OH⋅� ), followed byreaction with molecular oxygen:

RH + OH⋅� → R⋅� + H2O (2.4)

R⋅� + O2 → RO2⋅� (2.5)

The peroxy radicals so formed can then oxidise NO toNO2:

RO2⋅� + NO → NO2 + RO⋅� (2.6)

This is the important path by which NO is oxidisedwithout involving ozone. The hydrocarbon inequation (2.4) could be one of a large varietycommonly found in urban areas, for examplepropane, hexane, benzene, toluene, xylene orhydrocarbons of natural origin such as isoprene.During the oxidation reactions of these hydrocarbons,the hydroxyl radical is usually regenerated,whereupon it can take part in further oxidationreactions. Other products may be formed which arecharacteristic of photochemical smog. For examplethe oxidation of ethane and propene results in thegeneration of acetaldehyde, which can undergofurther reactions with the hydroxyl radical, oxygenand nitrogen dioxide to form PAN. Non-gaseouscompounds also form as fine particles, giving rise tothe haze which is observed during severe smog

episodes in the USA and elsewhere.

In general terms, the amount of NOx available affectsthe ultimate quantity of smog that can be producedand the amount of ROC, sunlight and temperatureaffect the rate at which it forms. Nevertheless, therelationship between peak smog levels and NOx andROC emissions defies simple analysis, as will be seenin later chapters.

2.3. SOURCES

As noted previously, the main constituents ofphotochemical smog are not emitted in significantquantities directly by human or natural processes.Rather, they are secondary pollutants arising fromprimary emission species (or precursors). Typicalemission sources of the precursors are summarised inTable 2.1.

2.3.1. Fuel Combustion

The release of a large proportion of anthropogenicNOx and ROC is intimately tied to the cycle ofrefining, distributing and use of a range of petroleum-based products, notably fuels.

Internal combustion engines are responsible for amajor proportion of anthropogenic NOx and ROC.Combustion in these engines occurs at hightemperatures and pressures, favouring the productionof NOx. Approximately 60% of ROC emissions fromvehicles result from exhaust gases containingincompletely burnt fuel, and 40% from evaporativelosses during operation and refilling. Sourcecategories include light and heavy duty petrol anddiesel-fuelled vehicles, ships, rail and air traffic, andagricultural equipment.

Small engines such as those used to power lawnmowers and other garden equipment, chain saws,outboard motors, and other off-road engines are alsoremarkably large emitters of NOx and ROC. Two-stroke engines tend to be higher emitters than four-stroke designs.

On a global basis, approximately 30-50% of NOxemissions are thought to be due to fossil fuelcombustion (Seinfeld 1986), with much of theremainder due to biomass burning. Many industrialactivities produce NOx emissions, includingelectricity generation, the use of industrial boilers,high temperature incineration, smelting operations,kilns and other processes, especially those involvinghigh temperatures (e.g., glass manufacture and blastfurnaces). Other fixed industrial sources includecombustion and process emissions from metalsmelting, pulp and paper making, petroleum refiningand the production of industrial chemicals. Lesseramounts are released from home heating systems.

Table 2.1. Sources of ozone precursors

NOx ROC

Natural sources

Bushfires Vegetation

Lightning Bushfires

Bacterial action

anthropogenic sources

Power generation Motor vehicles

Motor vehicles Painting, solvent use

Industrial boilers Fuel evaporation

Incineration Printing industry

Kilns Wood processing

Jet engines Off-road engines

6

2.3.2. Petroleum Products

Petroleum fuels are moderately to highly volatile andconstitute a major source of ROC. Evaporative releaseto the air occurs during the refining process andsubsequent storage and handling. Evaporative lossesare increased at higher temperatures. Substantialhydrocarbon emissions occur during the manufactureand application of asphalt, and the use of hot bitumen.Hydrocarbon gases such as propane and butane areused as propellants in some aerosol cans, and their useresults in release of the contents which adds to theatmospheric burden of ROC.

The evaporation of organic solvents also constitutes amajor source of ROC release to the air. Thesesolvents are found in many commercial andhousehold products including degreasing agents,adhesives, paints, varnish, polish, and other surfacecoatings. Printing and dry cleaning are otherprocesses using volatile organic solvents.

2.3.3. Natural and MiscellaneousSources

Natural sources are relatively more importantcontributors to ROC than to NOx. Whereas naturalsources of NOx are usually ignored in emissionsinventories, natural ROC emissions (commonlycategorised as “ forest” or “biogenic” emissions) mustbe included, and can constitute a significantproportion of the total ROC inventory. Thecharacteristic smell of coniferous and eucalypt forestsis due to the release of these hydrocarbons, especiallyalkenes such as pinene (conifers), cymene andlimonene (eucalyptus). Isoprene is emitted frombroadleaved trees. These emissions are thought to beresponsible for the characteristic haze which affectsthe Blue Mountains of New South Wales and theSmoky Mountains in the eastern United States. Thefull complement of ROC species emitted byvegetation is more than 350 (e.g., Finlayson-Pitts andPitts 1986). The role of many of these compounds insmog formation is unknown, although isoprene andpinene are in a moderate to high reactivity class, andare thought to contribute significantly.

Bushfires can feature significantly in photochemicalsmog incidents, although their sporadic occurrencemeans they are not usually accounted for in emissioninventories. Research in Western Australia (Evans etal. 1977) and the Northern Territory (Hurst, Griff ithand Cook 1994) has shown that large quantities ofboth NOx and ROC are released by bushfires. Thishas significance for Perth where severalphotochemical smog events, caused by a combinationof urban emissions and smoke from either naturalbushfires or hazard reduction burning, have beenrecorded.

Based on emission factors derived by Evans et al.(1977), and using typical fuel loads, it is possible toestimate total smog precursor emissions frombushfires. For a prescribed burn of the size oftenconducted in the south west of Western Australia(about 5000 ha), the total non-methane hydrocarbonmass released would be of a similar order to a fullday’s anthropogenic emissions from the Perth region.The emission of nitrogen oxides would beapproximately a third of that emitted by humanactivity during a Perth day. These relative magnitudesare generally consistent with observations of theeffects of bushfire smoke in Perth.

Fermentation processes can release ROC, as can theburning of refuse or agricultural wastes. Bakeries,meat cooking and charcoal barbecuing (especiallywhen organic starter fluid is used) have also beenidentified as significant ROC sources, and controlmeasures for these sources have been undertaken insome jurisdictions in the USA.

Some minor additional NOx sources includebiological processes in the ocean and oxidation ofammonia. Chemical processes such as nitric acidproduction can also be the source of significant NOxemissions (Bretschneider and Kurfurst 1987).

2.4. HEALTH EFFECTS OFPHOTOCHEMICAL SMOG

2.4.1. Air Pollution and Health

Air pollution can harm human health by directinhalation and by other routes of exposure. Many airpollutants affect the respiratory and cardiovascularsystems.

In Australian cities, air pollution levels sometimescause acute health effects, including asthma, and maycontribute to chronic disease (pulmonary fibrosis andemphysema) (Woodward et al. 1995). Worldwide,respiratory disease is a major cause of lost work timeand disabilit y. Photochemical smog plays a part,although its impact on the total burden of respiratorydisease is small compared with, for example, theeffects of smoking cigarettes.

2.4.2. Methods for the Investigation ofHealth Effects

To establish any health effects caused by pollutants,experimental research and epidemiological studies areneeded. Experiments in this case typically include theuse of air chambers. This method allows theinvestigator to specify the amount and duration of theexposure to air pollutants that the subjects experience.

7

The subjects chosen for these experiments are oftenyoung adults in good health who exercise at specifiedlevels inside the chamber, under close observation forchanges in rates and volumes of breathing. However,other sub-populations, such as the very old or the veryyoung, or those with respiratory disease, could bemore vulnerable to photochemical smog. For ethicalreasons, these other groups are usually excluded fromexperimental research. The health effects from anytype of pollution vary not only according to theintensity and the duration of exposure but also withthe age and health status of the population exposed.To obtain information about the health effects ofpollutants on a wider spectrum of the population,epidemiology is needed.

Epidemiological studies involve the detailedassessment of possible causes of disease, sometimescalled exposures, and the health outcomes (includingrecognised diseases and other loss of normal bodyfunction) in real settings. Epidemiology complementsexperimental research to establish the nature of healthproblems affecting parts of the population. As well asanswering “what disease?”, the combined researcheffort aims to describe “when” , “where” and, mostdiff icult, “why” health effects occur.

Surveys are undertaken on population groups who areexposed to high levels of air pollutants as part of theirusual li ving conditions, focusing on the health statusof the group in relation to these “normal” conditions.After careful examination of all the evidence,epidemiologists, together with other scientists,consider the likelihood of cause and effect in therelationship between exposure and disease.

It is necessary to rule out other potential causes beforeassuming that air pollution is causing the respiratorydisease. Other possible causes, to which many of usare exposed, include living in houses with unflued gasheaters and environmental tobacco smoke (passivesmoking). The combined effects of several pollutants,termed synergistic effects, must also be considered,before attributing respiratory disease to one possiblecause.

2.4.3. Overview of Smog HealthEffects

Photochemical smog is comprised of gases and veryfine particles (or aerosols). The health implications ofthe main gaseous components of photochemical smoghave been well known for many years. The healthimpact of f ine particles from a wide range of sources(not just photochemical smog) has emerged, withinthe period of this study, as a major issue (Dockery etal. 1993). A concurrent study, the Perth Haze Study,will examine the concentration of f ine particles

present in Perth’s air and the possible impact onhealth and visibilit y.

As will be demonstrated in Chapter 5, theconcentrations of f ine particles measured during smogevents over Perth show only marginal variations fromnormal levels. Smog events induced or enhanced bybushfires are the exception; bushfire smoke yieldsrelatively high particle concentrations. In the absenceof bushfires, particle concentrations associated withphotochemical smog are insignificant compared tothose which occur during winter (e.g. due to theaccumulation of smoke and vehicle emissions undertemperature inversions). For this reason, the healtheffects of f ine particles are not considered within thecontext of this study.

Symptoms associated with ozone exposure haveincluded cough, dryness of the throat, wheeze, chesttightness and lassitude. Thus, the main adverse humanhealth effects of photochemical smog occur in theairways. While experimental studies (e.g., Drechsler-Parks 1987) suggest that, at the usually encounteredconcentrations, ozone is the most important cause ofadverse health effects, other components ofphotochemical smog are noted briefly.

Some acute symptoms, including eye irritation, maybe due to other oxidant gases (such as peroxyacetylnitrate, PAN), aldehydes and hydrocarbons in thegaseous phase (Woodward et al. 1995). Studies ofpulmonary function have not demonstrated effects ofPAN at 250 ppb, a higher concentration thanencountered in the ambient air in most cities.

Oxides of nitrogen (NOx) are precursors of ozone, buthave health effects in their own right. Nitric oxide andnitrogen dioxide (NO2) are considered the mostimportant of the NOx in the air close to the earth’ssurface, as they are often found in significantconcentrations in polluted air masses that undergoshort- or long-range transport. Of the two, NO2 ismore toxic and irritating. At ambient levels in apolluted environment, it may have acute or chronichealth effects. The type of exposure experiencedoutdoors is very different from that used in chamberstudies, and possible interactions with other pollutantsneed to be considered (Tattersfield 1993).

Like ozone, NO2 is relatively insoluble in water andtherefore can reach far into the respiratory tract.Unlike ozone, however, NO2 does not show a clearexposure-response relationship (i.e., there is nosimple proportionate increase in health effects for agiven increase in dose of NO2). For this reason, theeffects of ozone are the primary focus in this report.At concentrations of NO2 less than 300 ppb, it is veryunlikely that health effects would be noted by any

8

individual (ibid.). Australian air quality goals for NO2are set at more stringent levels than this.

Exposure to low levels of ozone causes tissue injuriesdeep in the lung, adjacent to the areas where gasexchange occurs. Acute tissue reactions afterexposures of several days include inflammation. Suchchanges have been observed in several species,including humans. For example, inflammatorychanges may follow single or repeated exposures to100 ppb ozone, a level found frequently withmoderate photochemical air pollution. Human tissueinjury may be progressive with prolonged exposure,but this is not well defined (Crapo et al. 1992).

In many experimental studies on humans, significantreductions of pulmonary function have been reported.One measure of respiratory function commonly testedis the forced expiratory volume (the volume of air thata person can blow out in a fixed time, usuallyspecified as one second, FEV1). Reductions in forcedexpiratory volume are usually accompanied bychanges in breathing pattern that are noticeable, andsometimes unpleasant. The severity of thesesymptoms parallels the impairment of lung functionwhich occurs in persons exposed to ozoneconcentrations of 80 ppb for several hours (Table2.2). Exposures of this order occur during theafternoons of the warmer months, in and aroundmany cities.

Epidemiological studies have shown that reducedrespiratory function frequently occurs at exposurelevels lower than 120 ppb (averaged over one hour)(Table 2.2). The range of individual susceptibility iswide (Woodward et al. 1995). Children may be athigher risk for detrimental effects of ozone thanadults because they spend more time outdoors duringsummer months when ozone levels are higher andbecause their lungs are still developing (White et al.1995). Studies have not demonstrated an exposure

threshold below which respiratory function remainsintact, so it is not possible to set a level at whichsafety from the adverse health effects of ozone can beguaranteed.

For a daily maximum 1-hour average ozoneconcentration of between 50 and 100 ppb, the WorldHealth Organisation (1992) classifies the overalleffect on human health as mild. In this range of ozoneexposure, eye, nose and throat irritation may occur ina sensitive minority, with slight reductions in FEV1averaged over the whole population, possiblyincreasing to a 10% loss in the most sensitive 10%. Aminority of adults may experience some chesttightness and cough. Some athletes could note a slightreduction in peak performances at this level ofexposure to ozone.

The occurrence of asthma, a common clinicalcondition characterised by intermittent respiratorysymptoms with variable obstruction of the airways,remains a controversial subject in relationship toozone levels (Woodward et al. 1995). While theevidence that photochemical smog itself causesasthma in the first place is not strong, ozone mayincrease the sensitivity of persons with asthma toother agents that cause wheezing (a synergisticeffect).

Other chronic effects of ozone remain uncertain, acritical gap for public health considerations. Theseeffects could include very subtle changes in the lung,eventually resulting in increased rates of respiratorydisease, with significant implications for health carecosts. Chronic ozone exposures probably contribute todeaths in and around cities, but the impact on totalmortality rates is probably very small (Lippmann1993). Although increased risks of cancer in humansdue to ozone exposures have not been proven, thispossibility has recently re-surfaced in the media

Table 2.2. Photochemical smog and health events in humans

Ozone concentration(ppb)

Exposure duration Health event

Experimental studies (controlled exposure in chambers)

> 80 Hours Reduced lung functionAirway inflammation

Epidemiological studies

< 120(daily 1-hour maximum)

Days to weeks Reduced lung functionIncreased severity of asthmaRespiratory symptoms

9

(Edwards 1995).

2.4.4. SummaryPhotochemical smog may cause irritation of the eyes,nose and throat. At current levels in many cities,smog may produce, over long periods of time, mildinflammation of the airways. In addition, people withasthma may experience narrowing of their airwayswith attacks of wheezing.

It appears that the risks associated with ozone andother pollutants are especially increased for childrenand adults with asthma. However, for children withno underlying lung disease there is also a risk ofadverse health effects associated with these pollutants.

There is no threshold for the effects of ozone, henceno safe level can be identified. The increasedstringency of the air quality goal for ozone (recentlyrevised by the NHMRC to the equivalent of 100 ppb,1-hour average) reflects this point. It is possible thathealth effects may occur for some people at lowerexposures than this. The difficulties of detecting andattributing health effects at these levels of exposure,however, are very considerable. Epidemiological andother studies are needed to identify interactive effectsbetween pollution and the respiratory system inAustralia and elsewhere. Effects are often small, solarge populations need to be studied. Cumulativemeasures of exposure are needed.

2.5. OTHER EFFECTS

2.5.1. Effects on Vegetation

All plants have some tolerance to ozone, as it isnaturally present in the atmosphere. However, atelevated concentrations, ozone is potentially injuriousto horticultural crops, ornamentals, commercial andnon-commercial forest trees, native shrubs and annualplants. The ozone dose which may cause adverseeffects varies between species and is influenced byplant nutrition, water availability, weather and otherfactors. Although still the subject of active research, itis apparent that harmful effects are felt in manyspecies under the ozone exposure regimes which aretypical in many industrialised parts of the world. Theresults of a large number of controlled exposureexperiments indicate that ozone has no beneficialeffects on plants. This contrasts with the effects ofSO2 and NOx, which under some circumstances canenhance plant growth by providing nutritionalbenefits (e.g., Pye 1988).

The toxicity of ozone to commercial crops and otherplants was noted as long ago as the 1950s (e.g.,Richards et al. 1958, Heggestad and Middleton 1959).Tropospheric ozone is now believed to cause moredamage to crops than all other air pollutantscombined, resulting in a 5-10% loss in crop yields inthe United States at an annual cost of three billiondollars, one billion of which is in California (Elsom1992). Other pollutants such as PAN and NO2 whichtypically occur in polluted atmospheres containingozone, can also be significantly toxic to plants (e.g.,

Effects of Ozone on Terrestrial Vegetation

PROPERTY EFFECT

Whole plantPhotosynthesis DecreasesLeaf conductance DecreasesWater use efficiency DecreasesLeaf area ReducedCrop maturation rate DecreasedFlowering Reduced abundance, lower and delayed fruit setDry matter production and yield DecreasedDrought tolerance ReducedNutrient stress More prone to ozone injurySusceptibility to pathogens May increase or decrease, depending on species

Plant community/populationCommunity structure Change favouring ozone tolerant speciesReproduction/life cycle Genetic change in progeny, reduced seed set and viability

Adapted from Runeckles and Krupa 1994. This table is necessarily generalised and indicates the most commonresponse as determined from experiments on a wide range of species. In each case, plant varieties, orindividuals, exhibit a range of responses, which may include no effect. The variability in response betweenspecies is large. Effects on reproduction and community structure are subject to large uncertainties.

10

Cabrera et al. 1988), although unlike ozone,concentrations are seldom in the phytotoxic rangeover large areas.

Many commercial crops may be adversely affected byexposure to ozone, including wheat, maize, oats,barley, alfalfa, soybeans, rice, citrus, lettuce, andonions (Olszyk, Cabrera and Thompson 1988,Olszyk, Thompson and Roe 1988, Kromroy, Tseng,Olson, and French 1988). Grapes, tobacco, tomatoand potato are especially sensitive. Direct effects maybe manifest by reductions in photosynthesis andhence growth rate, resulting in reduced yield, orvisible foliar injury and premature leaf drop. Theeffects of reduced quality may be more economicallysignificant for some crops than reductions in yield.

Ozone may also have indirect effects by altering plantsusceptibilit y to pathogens such as fungi, insects orviruses (e.g., Dohmen 1988) or by reducing coldhardiness (Barnes et al. 1988). More subtle, yetimportant effects may be caused by interference withplant reproductive processes, which may causeprogressive unpredictable changes in the geneticstructure of plant progeny (Wolters and Martens1987, Cox et al. 1989, Cox and Malcolm 1994).

Ozone is thought to be a factor contributing to thedeteriorating condition of some forests in Europe andNorth America. Relatively littl e work has been doneso far in Australia to characterise the response ofnative forest trees to ozone. O’Connor et al. (1975)examined the sensitivity of more than 100 nativeAustralian shrubs and trees to acute ozone injury, andconcluded that most species were resistant. Theexposure regime used (one hour exposures toconcentrations of 100-800 ppb) was, however,somewhat unrealistic, and the response measure(visible injury) was at best a crude indicator of ozoneimpact.

Monk (1994) carried out an experiment in whicheight species of eucalyptus were exposed tocontrolled doses of ozone in open-topped chambers.Seven of the eight species exhibited a significantreduction in growth (20-30%). Tuart, which occurs inthe Perth area, showed littl e leaf injury, but had thegreatest reduction in growth, demonstrating thatvisible symptoms are not reliable indicators of anozone effect. More research to assess the potentialimpact of ozone episodes in Perth on native species ofvegetation is needed.

2.5.2. Protective Standards forVegetation

Determining an ambient ozone standard of a typewhich will provide adequate protection to vegetation

has been the subject of considerable study.Cumulative exposure indices, such as the W126exposure parameter developed by Lefohn andRuneckles (1987), seem to be most appropriate. Theform of most current ambient ozone standards doesnot explicitly take into account the importance ofcumulative dose. However, it is possible under somepollutant climatologies that a 1-hour standard may bea surrogate for longer term averages. For example,Pearson, Linzon and Donnan (1988) concluded, in adetailed study of the effects of ozone on crops inOntario, that attainment of the 80 ppb 1-hour Ontariostandard would essentially eliminate foliar damageand crop productivity losses due to ozone in thatprovince.

The US Forest Service has determined a “red line”value of 100 ppb for one hour as unacceptable fordesignated wilderness areas in the United States(Adams et al. 1991). The need for a longer termstandard is being carefully considered during currentreassessments of ambient ozone standards in NorthAmerica.

As plants have some abilit y to repair ozone damage,the duration of episodes and the time betweenepisodes is important. The greatest impact is likely tooccur when an ozone episode coincides with otherstresses, or at significant development stages in theplant’s li fe cycle.

2.5.3. Effects on Materials

The oxidative nature of ozone and other smogconstituents results in direct chemical attack onsusceptible materials by photochemical smog.Materials affected include metals, paintwork, plasticsand rubber, concrete, masonry and stone, works ofart, textiles and dyes (Yocom and Upham 1977,Grosjean, Whitmore and Cass 1988).

Ozone causes erosion of paint coatings including oil -based house paint, latex paint, automotive finishesand industrial coatings. Oil -based house paint isaffected the most. This has been verified in bothcontrolled environment tests and field exposures (e.g.,Reiss et al. 1995).

The accelerated deterioration and cracking of rubberby ozone has been recognised for years, and was usedin Los Angeles in the past as a means of assessing theambient ozone concentration. The smog problem inLos Angeles was partially detected by the observationthat rubber products in the city area were prone toexcessive cracking. This form of deterioration isespecially critical for vehicle tyres and rubber-insulated electrical wires. Rubber cracking is ofparticular concern when rubber products such as

11

tyres, conveyor belts, seals and hoses, are beingstored for a significant time before sale or use.

Gaseous and particulate nitrates, which are present inphotochemical smog, were found to be responsiblefor the failure of electrical and telecommunicationsswitchgear in parts of the United States, due tocorrosion of the contacts. This required replacementwith more expensive metals which could resist thecorrosive attack (Yocom and Upham 1977).

2.6. PHOTOCHEMICAL SMOGAROUND THE WORLD

Photochemical smog can occur in any region, givenadequate concentrations of the required precursorspecies, and enough sunlight. The crucial role ofsunlight means that smog is most common in lowerlatitudes, but long summer days at higher latitudescan compensate. Also, smog and its precursors can betransported considerable distances by air massmovement, to affect areas hundreds of kilometresfrom major pollutant source regions. This isespecially pronounced where the transport path isover water, as in this case thermal mixing and ozoneremoval processes are reduced. Long-rangetransported ozone affects areas in the north east of theUnited States and Canada, and parts of north-westernEurope including the United Kingdom andScandinavia.

The first detailed measurements were made in citiesof the United States. On the west coast, from SanFrancisco to Los Angeles, ozone has been recordedsince the late 1950s. In the period 1958-1968, peakhourly average concentrations in the Los Angelesregion were in the range 400-600 ppb (Elsom 1992).Progressive tightening of emissions controls has sincereduced these values to the 200-300 ppb range (seeFigure 2.1).

Routine ozone measurements have been maderegularly in Europe since about 1970. Ozone

concentrations exceed 80 ppb each summer over awide area. Peak concentrations are in the range 150-220 ppb during major episodes. The European ozonemonitoring network is less extensive than that inNorth America and the full geographic extent ofEuropean ozone episodes is still poorly defined.During three major events analysed (in 1980 and1982) the highest concentrations across Europe were150-180 ppb, with most values under 120 ppb (OECD1990). These values appear lower than those oftenexperienced in the United States during smogepisodes.

Records from Melbourne clearly ill ustrate theinfluence of year-to-year fluctuations of meteorologyin controlli ng the occurrence of smog events (Figure2.2). The number of days exceeding 80 ppb hasvaried interannually by a factor of more than 10 overthe 15-year period, against a general downward trendin frequency. Large interannual variations in ozonelevels, driven by variations in summer weather type,are also experienced in the eastern United States,Canada (e.g., Fuentes and Dann 1993) and Europe(Elsom 1992).

For a broader perspective, Figure 2.3 presents data onhighest 1-hour ozone concentrations in several citiesin Australia, the USA, Canada, Europe and Asia.Values in Australian cities are seen to be lower thanLos Angeles (still t he worst affected city in the UnitedStates in terms of ozone pollution), similar to NewYork, and higher than Zurich or Helsinki. Note thatthere are no data for Perth before 1990, and that therewas only one ozone monitoring station in Perth in1991, hence the peak annual values for these yearsmay be underestimated. During 1992, one siteoperated all year and a further four sites for part of theyear.

Detailed examination of monitoring results revealsthat each city has a smog “climate” which is to somedegree unique. This can make comparisons based onsimple indicators such as peak annual levels

Year

0

100

200

300

400

500

600

60 64 68 72 76 80 84 88

(ppb)

Ozone

Figure 2.1. Annual maximum hourly average ozoneconcentration in Los Angeles, for the period 1958 to

1989 (Elsom 1992).

Year

0

10

20

30

40

80 82 84 86 88 90 92 94

No.

ofdays

Figure 2.2. Number of days when hourly ozoneconcentrations exceeded 80 ppb in Melbourne, 1980-

1994 (Ahmet 1995).

12

misleading. For instance, two sites may experiencesimilar annual ozone peaks, but one site may recordmany more hours per year above some intermediateconcentration value, and have a much higher annualmean concentration. This is important whenconsidering ozone effects in cases where cumulativedose is significant, for example vegetation impacts.Additionally, different locations within the sameurban area may record substantially different ozone

concentrations due to the local effects of trafficemissions and meteorology. The number ofmonitoring sites and their placement also has animportant effect on the data recorded and whetherpeak values are accurately represented. These factorscomplicate comparisons between different urbancentres.

All Australian state capital cities except Hobart have

0

50

100

150

200

250

300

Los AngelesNew York

BostonVancouver

TorontoMelbourne

SydneyPerth

Hong KongZurich

Helsinki

1988 1989 1990 1991 1992

Peakozone

conc.

(ppb)

Figure 2.3. Peak annual 1-hour average ozone values in selected world cities, 1988-1992. (Source:Kiely et al. 1995).

Year

0

50

100

150

200

250

300

79 81 83 85 87 89 91 93

Ozone

(ppb)

MelbourneSydneyBrisbaneAdelaidePerth

Figure 2.4. Trends of peak hourly average ozone concentrations in Australian capital cities. (Source:annual reports of Australian state environment agencies, personal communications)

13

experienced significant photochemical smog events.Figure 2.4 shows that, with the exception of Adelaide,annual peak ozone concentrations are normally nearor more than 120 ppb. While there has been somedownward trend of peak values in Sydney andMelbourne since the early 1980s, there are indicationsthat much of this trend was due to meteorologicalfactors, or limitations of the ozone monitoringnetworks. Peak ozone readings in Perth since 1990have been in the 100-130 ppb range, higher than thosein Adelaide, but lower than Sydney and Melbourne.Peak values have declined slightly since 1991,although other statistics have suggested a smallupward trend. Greater detail on Perth’s ozone levelsand comparative data from other Australian capitalsare given in Chapter 5.

2.7. AIR QUALITY STANDARDSAir quality criteria may be mandated nation-wide aslegal standards by central government agencies (as forexample in the United States by the US EPA) orrecommended by central agencies as objectives orgoals. The latter approach is followed in Australia andCanada, and requires individual states or provinces toembody the objectives in their respective statutes, orotherwise reference them in guidelines, approvals orlicences, if they are to have legal status. In Australia,air quality goals for major air pollutants such as ozoneare recommended by the NHMRC, with input fromexperts in fields other than health, as required.Individual states have, to date, generally notdeveloped and legislated their own air qualitystandards. The exception is Victoria, which in its

Table 2.3. 1-hour average air quality standards for ozone in various countries and states.

Country/State Value(ppb)

Criterion

Australia, NHMRC (1995) 100 not to be exceeded more than once a year

250 public warning to be given if levels are expected to exceed 250

Australia, EPA Victoria 120 acceptable level, not more than one day per year

(1981) 150 detrimental level, not to occur

Canada (1989) 50 maximum desirable, long-term goal for protection of unpolluted areas

80 maximum acceptable, for protection against adverse effects on soil ,water, flora, fauna, materials, personal comfort and well -being

150 maximum tolerable, giving serious deterioration to air quality such asto pose substantial risk to public health, requires urgent abatement

European EconomicCommunity

60

120

guide, not more than one day per year

limit, not more than five days per year

The Netherlands (1986) 60 guide value

120 limit value

Japan (1990) 60 limit value

120 alert level, public warnings issued

Mexico (1984) 100 standard

Brazil , Sao Paulo 81 standard, not more than one day per year

United States (1990) 120 standard, not more than one day per year

Cali fornia 90 standard

World Health Organisation(1987)

76-100 guideline to protect public health

Sources: IUAPPA 1988, Japan Environment Agency 1985, Streeton 1990, Cochran et al. 1992, Government ofVictoria 1981

14

State Environment Protection Policy (Government ofVictoria 1981) established standard concentrations fornumerous air contaminants, including ozone. Detailsare given in Tables 2.3 and 2.4.

As indicated in Section 2.4, ozone has a range ofeffects on human health. Recognition of the adverseeffects of oxidants resulted in the establishment ofambient air quality standards for ozone in manynations, especially during the 1980s. At least 17nations or national agencies have now issued ozonestandards (Cochran, Pielke and Kovacs 1992). Theexistence of an effects threshold in the range 60 to100 ppb may not be supported by recent research, butprimary standards adopted by most countries of theworld and states of Australia, which are intended toprotect public health, are roughly in this range. Manyexisting standards, especially those below 100 ppb 1-hour average, were based significantly onextrapolation of effects on vegetation. Since theshortest averaging time used in most air qualitydatabases is one hour, this period is the one usuallynominated in ozone standards. However, the need forstandards covering longer averaging times is beingexamined as part of current re-assessments of ozonestandards in North America and elsewhere.

In air quality management, it is common practice todefine several concentration values as objectives orstandards, which cover the range from desired cleanair to a general public health alert. Table 2.3 showsthat there is some consistency at the lower end of thescale, where the words “guide” and “desirable” areused. Values in the range 50-60 ppb match the lowerend of the range where physiological effects havebeen detected.

Above this range is a set of “acceptable” levels, forwhich the term “standard” is closely equivalent.These are generally taken as the concentration whichthe population will accept only occasionally. Valuesrange from 80 to 120 ppb, with lower values usuallybeing of more recent origin. For example, theAustralian air quality goal for ozone was recentlyrevised (June 1995) from 120 ppb to 100 ppb (1-houraverage) with a recommendation that it be reviewedagain within five years. A 4-hour goal of 80 ppb wasalso recommended by the NHMRC (NHMRC 1995,Guest et al. 1994). In the initial submission to theNHMRC (Woodward et al. 1993), the suggested newvalues for adoption were 80 ppb (1-hour) and 60 ppb(4-hour).

Finally, there is a range of levels which are judged to

Table 2.4. Long term air quality standards for ozone in various countries and states.

Country/state Averagingperiod

Value(ppb)

Criterion

Australia, Victoria (1981) 8 h

8 h

50

80

acceptable level, to protect vegetation, not tobe exceeded more than three times per year

detrimental level, to protect vegetation, not tobe equalled or exceeded

Australia, NHMRC (1995) 4 h 80 not to be exceeded more than once per year

WHO (1987) 8 h 50-60 to protect public health

24 h 32.5 to protect vegetation

8 h 30 to protect vegetation, average 09-17h daily,May-Sep.

Canada (1989) 24 h 25 maximum acceptable

24 h 15 maximum desirable

1 yr 15 maximum acceptable

Austria (1987) 8 h 50

Israel (1992) 8 h 80

New Zealand (1986) 8 h 30

South Africa (1965) 24 h 50

Sources: Government of Victoria 1981, NHMRC 1995, Elsom 1992, IUAPPA 1988, Cochran etal. 1992, OECD 1990.

15

be suff iciently damaging to a portion of thepopulation that they should ideally not occur at all ,and if they do, public health advisory notices may beissued and special industrial control measures put intoeffect. These range from 120 to 250 ppb.

Standards applicable at longer averaging times existin some nations. A summary is given in Table 2.4. Itseems likely that these long-term standards willreceive increased attention in the future.

The amount and quality of information available toexpert panels for the setting of standards hasincreased dramatically since the mid 1980s, whenmost existing ozone standards were set. Accordingly,reviews have recently been carried out or are inprogress in several nations including Australia,Canada and the USA.

2.8. CONTROL MEASURES

The following discussion of smog control measures isbased on documented experience around the worldand is presented here for information. Assessment ofcontrol measures applicable to Perth’s smog has notbeen undertaken within the context of this study.

As ozone is not emitted directly, but formed in the airfrom other air pollutants (principally NOx and ROC)control measures must focus on reducing theconcentration of these precursors. In most developednations a variety of controls are being applied in anattempt to bring the photochemical smog problemunder control. However, despite more than 15 yearsof effort, ozone continues to reach undesirable levelsin Europe, many parts of North America, andAustralia. There are several probable reasons for thislimited success. The understanding of real-worldemissions continues to be deficient, hence the scaleand nature of reductions which are needed isimperfectly defined. Perhaps more significantly,emissions from the mobile source sector have notbeen reduced in line with improvements in emissionscontrol technology, as vehicle numbers and triplengths have continued to increase.

A consensus view in the USA is that controls on bothNOx and ROC emissions are needed if ozone is to bebrought under control. However, the effect of NOxand ROC controls on ozone levels is variable andcomplex, depending on many factors including localclimatology and atmospheric transport, and therelative magnitudes of NOx and ROC emissions in agiven airshed. An ironic feature is that in some largeurban areas, ozone levels may actually rise inresponse to a lowering of NOx emissions, althoughregions downwind may benefit. Whereas NOxcontrols can potentially bring benefits over asubstantial area, the effects of ROC controls areusually felt closer to the source of emissions.

In many areas where there is extensive forest cover,such as the United States and Canada, emissions ofreactive biogenic ROC are very large and essentiallyuncontrollable. As a result, reductions inanthropogenic ROC emissions may have relativelylittl e impact on ozone formation regionally. Theemission of reactive biogenic ROC is proportional totemperature, boosting photochemical activity on hotdays. Biogenic ROC emissions are also significant inEurope (OECD 1990) and Australia (Carnovale et al.1991).

Despite these obstacles, it is felt that the controlmeasures which have been applied have at leasthelped prevent the situation getting worse. Somenotable successes have occurred, such as thedesignation early in 1995 of San Francisco as an“ozone attainment area”, despite a doubling of themotor vehicle population in the past 25 years, and thehuman population increasing by 50% to reach sixmilli on. In 1969, San Francisco endured 65 days peryear over the 120 ppb limit. This fell to once per yearby 1994. The success of the Bay Area Air QualityManagement District is ascribed to an active programof legislation beginning with automobile emissionstandards in 1966, fuel vapour recovery (1974),regional oxidant modelli ng (1975), vehicle emissionstesting (1984), transport management methods tominimise single-passenger automobile use, andprogressively stricter ROC controls in 1989 and 1992.Another major component was public education andoutreach, emphasised in the 1991 Bay Area Clean AirPlan (Air Quality Week 1995).

2.8.1. Industrial Control Measures

As noted in Section 2.3.1, the bulk of anthropogenicNOx is due to combustion. Control options for NOxuse technology which reduces the production ofnitrogen oxides by changes in the design ofcombustion chambers or burners. NOx formation isgreatly affected by the way in which fuel is deliveredand mixes with air. The entrainment of atmosphericair (mainly nitrogen, a potential source of NOx) maybe restricted or even prevented completely, with thesubstitution of a pure oxygen source. NOx formedfrom entrained air (thermal NOx) constitutes themajority of the NOx formed by high-temperaturecombustion of gas or light oil . Some NOx is alsoformed from the nitrogen in the fuel itself (fuel NOx).

NOx formation increases with increasing combustiontemperature and pressure. Therefore, engineeringcontrol options focus on reducing peak flametemperatures in boilers and kilns, and reducing excessoxygen levels. Off-stoichiometric combustion (OSC)regulates the oxygen content in a furnace, reducingboth fuel and thermal NOx by about 30%. Flue gas

16

recirculation can also be used to reduce excessoxygen. Low-NOx burners, which may be fitted orretrofitted to steam boilers, can reduce NOx emissionsby 30-50% (OECD 1988).

Changes in burner design, oxygen or fuelmanagement cannot reduce NOx emissions muchbeyond 50% in thermal power plants. Other methodssuch as catalytic scrubbing of f lue gases arepotentially more effective, and can reduce emissionsby up to 90%. For example, NOx can be reduced tonitrogen and water by use of ammonia and a catalyst(selective catalytic reduction, or SCR). Some otheradvanced combustion technologies, used primarily forenergy eff iciency benefits and to control sulphurdioxide emissions, also offer benefits of reduced NOxemissions, for example integrated gasificationcombined cycle (IGCC) and atmospheric fluidisedbed combustion (AFBC). NOx output is reduced byabout 40% with these methods, applicable to coalplants. NOx emissions from gas turbines can beabated by diluent injection, in which water or steam isintroduced with the fuel, or by SCR. Recentlyintroduced “dry low NOx” technology for gasturbines employs premixing of fuel and air at a verylean ratio to achieve low emissions (Schorr 1992).

Some power utiliti es in the USA have opted to reducesmog-forming emissions indirectly by offering to buyand scrap older model cars, thereby obtaining anemissions credit. These so-called scrappage or“clunker” programs have been accepted in some casesby the US EPA, but not without criticism. Oldermodel vehicles are able to run cleanly if adequatelymaintained, and it is not certain that those peoplescrapping an old vehicle will replace it with a cleanerrunning model, or even if the vehicle scrapped wasbeing used much (e.g., Beaton et al. 1995).

Industrial approaches to ROC control vary accordingto process. Major emission reductions have beenachieved using paints, varnishes, adhesives andfinishing compounds with reduced organic solventcontent, or which use water as the carrier mediuminstead of organic solvents. Industries using thesecompounds include manufacturers of furniture andwood panelli ng, the automotive manufacture andrepair industries, and appliance manufacturers.

The printing industry is reducing ROC emissions withthe use of low-ROC inks and ROC recovery. Vapourrecovery is also used in dry cleaning and metaldegreasing operations.

For petroleum refineries, emissions are reduced byroutine inspection for, and repair of, leaks fromvalves, gaskets and compressors, and the use offloating roofs in storage tanks. Recent practice is toinstall double seals on these tanks. The covering of

still wells, used for monitoring liquid levels in storagetanks, offers the potential to reduce residual ROCemissions from these tanks by about 50%.Wastewater treatment systems (e.g., oil separators)are covered to reduce fugitive emissions.

2.8.2. Motor Vehicle Controls

Between 40 and 60% of the NOx present in the air inindustrialised regions is emitted from motor vehicles.This proportion may be higher in large, low-densityurban regions with relatively littl e heavy industry. InAustralian cities, motor vehicles have been estimatedto generate 50-60% of the hydrocarbons and as muchas 80% of the NOx (Carnovale et al. 1991 − see alsoChapter 6). Despite major efforts, adequate control ofemissions from this sector has proven diff icult.Although the emission rate per kilometre has fallenconsiderably for many makes of vehicle in the past 15years, the total number of vehicles in use and vehiclekilometres travelled (VKT) has increased inexorably,cancelli ng out a proportion of potential gains in airquality.

In addition to the emission control devices which maybe fitted, motor vehicle emissions depend on manyfactors, including the total kilometres driven, enginecondition, weather, fuel quality, and driver behaviour.Emission control devices deteriorate and are alsotampered with by owners. With the exception of fuelquality, these factors are diff icult or impossible tocontrol. This may be an important reason for the slowprogress in bringing ozone under control in parts ofthe United States and elsewhere where major controlinitiatives have been undertaken.

Catalytic converters were adopted in the United Statesin the mid 1970s to control tailpipe emissions when itbecame apparent that greatly improved motor vehicleemission performance was needed to bring the seriousozone problems in Cali fornia under control. Thenecessary reductions in CO, NOx and hydrocarbonemissions could not all be achieved via tuning orair/fuel ratio modifications, since the lean mixtureswhich reduce CO and hydrocarbons maximise theproduction of NOx. The introduction of unleadedpetrol in Australia in 1985 enabled the use of catalyticconverters on 1986 and later model vehicles. James(1995) estimated that 85% of the post-1985 light-dutyvehicles in use in Perth in 1994 were fitted with three-way catalytic converters, the remaining 15% havingtwo-way catalysts.

A revised Australian Design Rule (37/01) whichgoverns emissions from light duty vehicles wasgazetted in March 1995, and will affect vehicles soldin 1997. The new standard will be equivalent to USstandards for the 1981-1993 period. However, new

17

vehicles for the Australian market may have loweremissions than the 37/01 requirements, as controldevices required for more stringent standardselsewhere are likely to be fitted (ACVEN 1994).

Vehicle inspection and maintenance programs (oftenabbreviated to I&M) have become one of the majorinitiatives aimed at reducing the output of automotivesmog precursors in North America and elsewhere.These programs are intended to ensure that theemission characteristics of new vehicles are as far aspossible retained during use. Vehicles not passing thetests are typically denied re-registration until repairsare made to bring emissions within acceptable limits.

In Australia, New South Wales is developing a motorvehicle maintenance program, which currentlyemploys a visual check of the catalytic converter aspart of the annual vehicle safety check. Emissionstesting options are being evaluated to help design aneffective I&M program (RTA 1995). In Victoria andNSW, police and EPA off icials operate a “spotting”program for vehicles visibly emitting excessivepollution. Offenders are required to have theirvehicles repaired and to provide proof of repair. Asimilar program is being piloted in Perth.

Despite the heavy emphasis placed on I&M programsin the United States and elsewhere, the I&M approachis not without critics. The emissions test is onlyapplied once per year at most, and the results may notapply to the vehicle when in use the rest of the time.Additionally, the I&M test does not consider howmuch a vehicle is used. A relatively high emitter maybe a smaller contributor over a year than a cleanervehicle which is driven much more.

Reformulated motor fuels are in use in the USA andare being evaluated in Europe, Canada and elsewhere(Burke 1995, BC MOE 1995, US EPA 1994). Incomparison to normal fuel, reformulated blendsusually have a lower Reid vapour pressure, sulphur,aromatic and olefin content, and a higher level ofoxygenates. The impact of reformulated blends maybe diff icult to evaluate with confidence. There is noexpectation that the use of these new blends will onits own solve urban ozone problems. Whereasreformulated petrol is expected to reduce hydrocarbonemissions by 15%, the use of engines optimised formethanol or ethanol offers reductions of 80-90% (USEPA 1993). However, emissions of air toxics such asformaldehyde may rise, and the small marketpenetration of special fuel vehicles means potentialbenefits are small compared to fleet-wide use ofreformulated fuels.

Cali fornia is requiring the phasing in of vehicles withprogressively cleaner engines starting in 1994.Legislation for the Low Emission Vehicle (LEV)

program mandates that the new vehicle fleetemissions average must meet increasingly demandingstandards each year through to 2003. This can bedone using any combination of four types of specifiedvehicle (Table 2.5).

Two per cent of all vehicles sold in Cali fornia in 1998must be Zero Emission Vehicles (ZEVs), rising to10% in 2003. The only vehicles able to meet the ZEVdesignation at present are electric vehicles.

At the date of writing, five states in the east of theUSA have adopted the Cali fornia LEV program, andfour more are proposing to do the same, althoughthere has been significant politi cal opposition frommajor vehicle manufacturers, in particular concerningthe ZEV requirement. In Canada, the LEV programhas received active consideration (Schwartzel 1994),and new vehicles will be required to meet the LEVstandard from 2001.

Environmental agencies are recognising the ultimatelimitations of technology alone to combat the urbanvehicle-related smog problem, in the face ofconstantly expanding low-density suburbs, inhabitedby residents who are culturally and often logisticallydependent on cars. Reducing the need to travel, bychanging work practices, redesigning urbancommunity layout and making public transit moreeff icient and attractive are measures which will becritical in the long term if vehicle emissions are to beadequately controlled. These approaches will stillneed to be coupled with increasingly stringentemissions standards (BC MOE 1995, Oge 1995).

2.8.3. Other Sectors

In addition to vehicle testing programs, regionssuffering from particularly severe ozone pollutionhave adopted technology to prevent evaporativelosses of petrol and other organic fuels duringshipment and handling, when storage tanks are fill ed,and at the petrol pump. Other controls on evaporativelosses of petroleum fuel include absorptive canisters

Table 2.5. Controls on vehicle emissions inCali fornia.

Vehicle Category Characteristics

Transitional lowemission vehicle

50% less ROC than1994 models

Low emission vehicle 75% less ROC, 50%less NOx

Ultra-low emissionvehicle

84% less ROC, 50%less NOx and CO

Zero emission vehicle(ZEV)

No exhaust orevaporative emissions

18

on vehicles to capture vapours from the fuel tank, andreducing the vapour pressure of the fuel itself (Reidvapour pressure).

In 1995, Cali fornia began regulating pollutantemissions from off-road engines, such as those fromlawn mowers, outboard motors, chain saws andtractors. Many of these engines, even when new, emitan amount of ROC and NOx in one hour equivalent todriving a typical passenger car 300-500 km. Off-roadengines are estimated to produce 15% of the totalNOx and 20% of the non-biogenic ROC in the UnitedStates (US EPA 1991). The US EPA is expected toadopt the Cali fornia off-road engine standardsnationally (Schwartzel 1994).

The open burning of agricultural, forestry orbiomedical wastes, domestic or municipal refuse,natural bushfires or prescribed fuel reduction burns allhave the potential to produce both NOx and ROC.Some of these activities are controllable, for exampleNOx controls may be fitted to biomedical ormunicipal waste incinerators. Natural bushfires andcontrolled burns are managed to minimise risk toproperty and to preserve commercial forest stands,but considerable emissions may still result, whichmay occasionally have a significant influence onozone concentrations (e.g., Rye 1996a).

2.8.4. Episode Control

In some parts of the world, smog may occur inepisodes of more than one day. Episode control is amanagement approach employed to prevent smogreaching hazardous levels. For example, publicadvisory notices may be issued on days of high smog-forming potential, determined by analysis of theweather and current smog levels. The public may beadvised to avoid the use of solvents and oil -basedpaints, lawn mowers and outboard motors, to avoidunnecessary vehicle trips, to car pool i f possible, andto refuel vehicles after dark to minimise ROCemissions during periods of peak temperatures andsmog-forming potential. The public are also advisedto ensure that their cars are well maintained, to driveat reduced speeds, make sure tyres are properlyinflated, and to reduce electricity consumption in thehome. Smog advisory notices may also contain amessage warning against heavy outdoor exercise.Open burning is usually banned during the ozoneseason, but a reminder may be included that suchburning (including the use of barbecues) makes smogepisodes worse. The effectiveness of these notices inreducing smog is uncertain.

In Australia, there are as yet no smog advisoryprograms containing prescriptive messages requiringthe public to modify their behaviour. In Sydney andMelbourne, information is added to routine air qualitybulletins warning the public of expected breaches of

the NHMRC ozone guidelines on the next day. At thetime of writing, New South Wales was using 120 ppbas the guide value, whereas Victoria was employing90 ppb, which corresponds to their “ light” pollutiondescriptor. Values more than 120 ppb are considered“significant” and a “breach of policy” . In bothjurisdictions, those with respiratory complaints areadvised to use their medication and avoid strenuousactivity.

2.8.5. Economic Instruments

A variety of economic instruments are in use whichare designed to control air pollutant emissions. Bymaking use of f iscal incentives, economic instrumentsare intended to provide rapid response and greaterflexibilit y and cost-effectiveness in comparison withthe more traditional “command and control” approachto pollution management.

Emission charges are fees assessed for each tonne ofpollutant released, and are intended to inducecompanies to reduce pollution as much as possible.The use of emission permits, which may be bought,sold and traded in a regulated market is anothermarket-based approach which has been used in airpollution control. Companies must collectively meetoverall emissions caps, and individual firms can dothis by reducing their emissions as required or, if theyconsider it more cost-effective, by purchasing anemissions allowance, which only exists if anothercompany within the trading zone can achieve areduction beyond their allowance. The companywhich has made emission cuts benefits by selli ng thesurplus allowances.

The experience with trading schemes so far seemsmixed. There may be regional inequities if companiesin a given region buy permits rather than cleaning up.Some companies are not suff iciently aware of thecontrol options available, therefore their decisions arenot made as the “theoretical” market would predict. Itis unclear whether even in the United States, themarkets are large enough to function effectively aspredicted by economists (A’Hearn, Wilson andHardiman 1994). There is also an increased need toinvest in monitoring equipment, verify the actualemissions produced by each source, and enforcesanctions effectively, if the emissions tradingapproach is to succeed.

Some control of motor vehicle emissions has alsobeen attempted in the USA, Canada and Sweden bythe use of rebates and “ feebates” applied to newvehicles when sold. A tax is applied to the sale priceof the most polluting cars, and rebates are paid topurchasers of cleaner models. Critics suggest that thefeebate/rebate values typically applied are insuff icientto have a major influence on buyer behaviour.

19

3.1. METEOROLOGY ANDATMOSPHERIC DISPERSION

Atmospheric pollutants will only reach highconcentrations when weather conditions are suitable.The main determining factors are wind speed, and theintensity and depth of atmospheric turbulence (Figure3.1).

Wind is probably the most obvious influence. When itis windy, emitted pollutants are spread over a broadregion, and concentrations are reduced. Light windsallow emissions to remain close to the source, andaccumulate in higher concentrations. However, evenwhen wind speeds are greater, a change of winddirection can result in the return of pollutants to thearea where they were emitted. Such “ recirculation”events cause problems in many of the world’s cities.

Atmospheric turbulence reduces the concentrations ofpollutants by mixing them with clean air. The flow ofwind over rough surfaces, and convection over heatedland, both generate atmospheric turbulence. Bycontrast, water surfaces remain aerodynamicallysmooth and cool. The layer of undispersed haze oftenvisible offshore near coastal cities is evidence of thelimited turbulent mixing over water bodies.

The layer of the atmosphere in which turbulenceoccurs is termed the “boundary layer” . The depth ofthe turbulence (the “mixing depth” in Figure 3.1) setsan ultimate limit to the benefits of turbulentdispersion. Over land during the day time, this depthgrows to be large, typically in the range of a kilometreor more. But at night and over water, mixing depthmay remain only a few hundred metres, or less.

3.2. THE SEA BREEZE

The Perth summer season is characterised byprevaili ng offshore winds, with long periods of f ineand warm weather. The resulting temperaturedifference between sea and land regions creates anonshore pressure difference, which leads to regularsea breezes. Due to the Coriolis effect and surfacefriction, the sea breezes are directed clockwise aroundthe low pressure region (the land) and onshore, so aresouth westerly at Perth.

Photochemical smog events in the Perth region areclosely linked to this daily wind pattern. In theDepartment of Environmental Protection’smonitoring records, high smog concentrations haveonly ever occurred when a sea breeze was present.

3. Perth’s Photoche mical SmogClimate

Figure 3.1. Factors affecting the dispersion of pollutants in the atmosphere.

20

The sea breeze is one example of a recirculation event(Figure 3.2), since the onshore flow normallydevelops after an offshore easterly wind. Particularlywhen morning winds are north easterly, the morning’spollutant emissions can be returned across the city.

The influence of the sea breeze also arises through itseffect on inland mixing depths. It brings cool aironshore, replacing a layer of air already warmed byconvection. Due to the temperature difference, the seabreeze is denser than the inland air mass, so that airwithin it can not mix with air above. This means thatmixing depths when the sea breeze occurs arecommonly reduced to only a few hundred metres.

The importance of this effect has previously beenidentified in Western Australia. Its consequences werefirst studied in detail during the Kwinana AirModelli ng Study (Department of Conservation andEnvironment 1982).

3.3. SMOG CONDITIONS

For photochemical smog to form over Perth, twoweather conditions other than the presence of a seabreeze must be met:

• suff icient heat and sunlight must be present todrive the chemical reactions which form smog;and

• the air mass containing Perth’s emissions mustremain in the Perth region, generally offshore andlargely undispersed as the smog develops.

The temperature and sunlight of a Perth summer dayare generally more than suff icient, so the secondcondition is usually the controlli ng factor.

There are several points in the sea breeze cycle wheredispersion is reduced and smog development isenhanced. In the early morning, low mixing depthsresult from persistence of the stable night-timeconditions, before convection becomes active. Thewind flow takes the city’s concentrated pollutantemissions offshore, where the absence of convectionand the smooth water surface ensure that lowturbulence and low mixing depths persist.

During this period, the chemistry and meteorology ofthe smog-forming processes interact. Chemicalreactions proceed faster when concentrations arehigher. This means that, while the emissions remainoffshore and poorly diluted, smog reactions progressmore rapidly.

When the sea breeze subsequently moves onshore, thereduction of mixing depth described in Section 3.2takes effect. Both the morning’s emissions, and themid-day contribution added as the air moves inland,remain poorly dispersed.

The Perth Photochemical Smog Study has shown thatthere are several quali fications and variations to theseprinciples:

• Normally, the sea breeze circulation begins somedistance offshore. Winds seaward and landwardof its formation zone retain an offshore directionfor longer. If this zone is close inshore, or windspeeds are more than a few metres per second,

Kwinanaindustrial area

Perth centralbusiness district

Figure 3.2. The path of air masses offshore, thenonshore, on a typical Perth smog day (21 March

1994). The black line shows the path of air whichpassed across the Perth business district in the

morning and the shaded line shows that for the airwhich passed across the Kwinana industrial area.

8am

12 noon

Sea Breeze

Figure 3.3. When offshore flow is strong, or thesea breeze forms close to the shoreline, the city’smorning emissions may be lost from the region.

21

the air mass containing morning emissions willreach the seaward side of the initial sea breezezone, and so be lost from the region (Figure 3.3).

Typically, the sea breeze forms close to the shorewhen morning wind directions are betweensoutherly and easterly. On days when the initialwind direction is in this range, Perth’s emissionsare carried generally northward, and do not returnto the city. These two effects combine to ensurethat, on such days, significant smog events do notoccur in the Perth region.

• When morning winds are from north of east, thesea breeze forms further offshore. Morningemissions may still be on the land side of the seabreeze when it develops. In such cases, theemissions are drawn into the growing sea breezeflow. It is in these conditions that significantconcentrations of photochemical smog are mostoften seen in Perth (Figure 3.4).

• If a low pressure trough is located off the southwest coast in the morning (as in Figure 3.5), awind direction change from offshore to onshoreexists already, at the trough axis. Perth’semissions can therefore not escape past thetrough. A sea breeze generally develops, and mayreturn some of the city’s emissions. The wind

change at the trough also propagates landwardduring the day, in the same manner as the leadingedge of the sea breeze. This brings furtheremissions onshore, possibly after the passage ofthe sea breeze.

In these conditions, the winds near Perth tend tobe lighter than normal, and typically from a northeasterly direction. These conditions have thehighest potential for smog formation in Perth.

8am

12 noon

Sea Breeze

Figure 3.4. When the morning easterly is light, or thesea breeze forms well offshore, the forming smog

may be trapped in the sea breeze, and returned to thecity.

30

1016

1018

1016

1020

1012

25

35

40105 110 115 120 125

H

1014

1016

1022

1024 1026

1018

L

Figure 3.5. The coastal low pressure trough on the morning of 16 March 1994, a typical example of this feature.The trough axis is shown by the broken line. Pressures shown are in hPa (hectopascal).

22

The greatest focus of effort during the PerthPhotochemical Smog Study was directed at measuringthe city’s photochemical smog events and theiraccompanying meteorology. At the same time,estimates were assembled of gaseous emissions whichwere known to contribute to photochemical smogformation.

These data collection activities had two primaryobjectives:

The first objective was to provide detailedinformation, never previously obtained, on theconcentrations of photochemical smog occurring overthe Perth region. This information on Perth’s airquality is of primary importance in its own right andrepresents a major outcome of the study.

The second objective was to provide acomprehensive data base for use in analysing how,when and where photochemical smog forms in the

Perth region. This analysis was done using a varietyof computer models to represent smog events, asdescribed in the following chapters.

4.1. CONTINUOUS AIR QUALITY ANDMETEOROLOGICAL SURFACEMEASUREMENTS

4.1.1. Monitoring Site Locations

Before the study, the DEP monitoring station atCaversham was the only station in the Perth region atwhich the key photochemical smog parameters, ozoneand oxides of nitrogen, were measured together. As apart of the study, seven new stations were initiallyestablished, recording these smog parameters andmeteorological parameters, plus five stationsrecording meteorology only.

Monitoring sites were chosen on the basis of the need

4. Overview of Activities

Rottnest Island

Swanbourne

Quinns Rocks

Two Rocks

Queen's Buildings

Caversham

Kenwick

Rolling Green

Hope Valley

Cullacabardee

North Rockingham

Belmont

Canning Vale

Gidgegannup

Middle Swan

Red Hill

Jandakot

Pinjar Tiwest Chandala

Ocean Reef

Mount Lawley

0 5 10 15 20Scale (km)

Gingin

KEYAQMS Sites

Meteorology only

Bureau of Met.

Radar

Sodar

Figure 4.1. Locations of measurement sites used for the PerthPhotochemical Smog Study. The abbreviation “AQMS” represents “Air

Quality Monitoring Station” (see Section 4.1.2)

23

to define the areas of smog influence and tounderstand how regional meteorological patternsinfluence the distribution and concentration of smog.To ensure that site choices were as informed aspossible, the CSIRO Division of AtmosphericResearch was contracted before the start of themonitoring program to examine the wind and smogdata available from the existing site at Caversham.This yielded preliminary analyses of air parceltrajectories and photochemical activity across theregion. Sites for monitoring stations recommended byCSIRO were consistent with those identified by thestudy team, based on an understanding of the localmeteorology.

Final site choices are shown in Figure 4.1. These werea compromise between the preferred locations andsites that were physically and logistically viable.

In brief, the logic for site choices was as follows:

• Rolling Green was chosen as representative of theregion likely to receive Perth’s smog in the lateafternoon. The study team was aware of f indingsfrom Sydney and Melbourne that peak smog levelscould occur many tens of kilometres from thesource of emissions. Rolli ng Green was theeasternmost of a line of stations, namely Rottnest −Swanbourne − Caversham − Rolli ng Green, whichcould be expected to measure the inlandprogression of smog during on-shore winds.

• Quinns Rocks and Two Rocks were selected todetect recirculation over northern residential areasof the previous day’s smog after it had been blowninland, and to measure the inflow of smog when thesea breeze returned it to the northern beaches. Noevents of the first kind were detected during thestudy, possibly due to the non-occurrence of therelevant weather conditions, but these sites revealedmany of the second kind of event.

• Swanbourne was chosen as a measurement site fortwo reasons. The first was the need to obtain aregular sample of the offshore flow of the city’smorning emissions, before photochemical smog hadstarted to form. The location was also expected tobe the site which often received the first inflow ofsmog with the afternoon sea breeze.

• Cullacabardee was chosen on the basis of CSIROmodelli ng, which showed a strong possibilit y of thesmog mass passing north west of Caversham ondays when the sea breeze did not penetrate wellinland. Monitoring confirmed that this did happen.The location, well removed from suburbanresidences, also proved ideal for siting noisy sodarand radar instruments (see Section 4.3).

• Kenwick, located to the south east of the city, wasnot expected to experience smog levels as high as

Caversham. It was selected to ensure that the effectsof smog carried by occasional north westerly seabreezes were measured, and that there was a sitewhich would measure smog originating fromKwinana industry during on-shore winds.

• Rottnest Island was the only location where it waspossible to measure smog development offshore. Itwas selected on the basis of a strong impression(proven correct) that the data from this site wouldbe of great value in developing an understanding ofPerth’s smog events. It proved to be a valuablesource of on-line data during field experiments,since it usually received the sea breeze, and anyassociated photochemical smog, first.

The sites listed above were the original seven. Aftermonitoring across two summer periods at Two Rocks,the study team was satisfied that a clear decline insmog concentrations occurred north of Quinns Rocks,so that continued monitoring at Two Rocks was not ahigh priority. Other stations were established forshorter periods in the course of the study toinvestigate specific issues, as follows:

•� Gingin. The Two Rocks station was moved to theGingin airfield in September 1994 to measure theeffect of the Pinjar gas turbine power station.

• Jandakot. This station was established in March1994 to measure the plume from Kwinana, close tothe source.

•� North Rockingham. This existing DEP station wasequipped with monitors for ozone and nitrogenoxides for a portion of the summer of 1993-1994, tosee whether smog recirculation southward of themajor emissions sources could be detected.

Figure 4.1 also shows the DEP stations at QueensBuildings in the Perth central business district andHope Valley near Kwinana. Both stations are in theimmediate vicinity of large emissions of nitrogenoxides. This parameter is measured at the stations butozone is not, for reasons explained in Chapter 5.

4.1.2. Overview of the Air QualityMonitoring Stations

Normally, each Air Quality Monitoring Station(termed an “AQMS”) consisted of a transportableshed surrounded by a 12 metre square and 1.8 metrehigh galvanised chain-linked security compound. Theclearest exception was the Queens Buildings site,located in a building in the central business district.Figure 4.2 shows a typical AQMS, using aprefabricated, skid mounted shed measuringapproximately 4.8 x 2.4 x 2.4 metres. All sheds wereair conditioned, due to the temperature requirementsof the sensitive electronic instruments inside.

24

With the exceptions previously noted, monitoringequipment at each station included instrumentsmeasuring ozone, nitric oxide and nitrogen dioxide.

Carbon monoxide, sulphur dioxide and hydrocarbonswere also measured at some stations. Theseinstruments, plus an IBM compatible personalcomputer and chart recorder (see Figure 4.3), werearranged along benches in each shed. At somestations a nephelometer, used to measure visibilityreduction due to airborne particulate matter, was fixedto an internal wall. Where possible, theinstrumentation at each site was standardised toenable equipment interchange.

Airtrak 2000 instruments were installed at Cavershamand Swanbourne. These instruments were designed tomeasure the constituents and the reactivity ofphotochemical smog. The instruments did not operatereliably throughout the study, except for brief periodswith intense maintenance.

Most sites also included a 10-metre tower supportingwind speed, wind direction and temperature sensors.The DEP stations at Caversham and Hope Valleyhave higher towers to allow two-level measurementsof temperature well above the top of the shed. Thesetwo stations also include instruments for measuringradiation (global, net and ultra violet) and othermeteorological parameters.

Table 4.1 summarises the variables measured at eachsite and the site commissioning date.

Figure 4.2. A typical Air Quality Monitoring Station (AQMS) shed, at the Caversham site.

Figure 4.3. AQMS shed interior. The personalcomputer, logger and chart recorder used by the

telemetry system are shown.

25

Table 4.1. Summary of data recorded at all sites during the Perth Photochemical Smog Study

Site Data Parameters Measuredstart O3 NOx SO2 CO HC TSP VISI PM25 PM10 WS10 WD10 AT10 RH10 DELT NR SR UV RAIN PRES ATRKSODR RASS

Caversham 10/89 • • • • • • • • • • • • • • • • •Swanbourne 08/92 • • • • • • • • • • •Kenwick 09/92 • • • • • • • • •Quinns Rocks 10/92 • • • • •Cullacabardee 12/92 • • • • • • • • • •Rolling Green 12/92 • • • • •Rottnest Island 12/92 • • • • • •Jandakot 03/94 • • • • • •Two Rocks 11/92 • • • • •Gingin 08/94 • • • • •QueensBuildings

07/89 • • • • • •

Hope Valley 01/89 • • • • • • • • • • • •Rockingham 10/91 • • •Gidgegannup 02/93 • • • •Middle Swan 09/92 • • •Pinjar 07/93 • • •Belmont 08/92 • • •Canning Vale 12/92 • • •Red Hill 12/92 • • •

Explanation of symbols:

Parameter Parameter DescriptionModel Manufacturer Method

O3 ozone TEI49 Thermo Environmental Instr Ultraviolet absorption

NOx nitrogen oxides TEI42 Thermo Environmental Instr Chemiluminescence

SO2 sulphur dioxide 8850S Monitor Labs Ultraviolet fluorescence

CO carbon monoxide TEI48 Thermo Environmental Instr Gas filter correlation

HC hydrocarbons APHA-350E Horiba Ltd. Flame ionisation

TSP total suspended particulates Mk3 Cairns Instrument Services High volume sampler

VISI visibility 1591 / 1550 Belfort Instrument Co. Integrating nephelometer

PM25 sub 2.5 micron particles 1400a Rupprecht & Patashnick Tapered element oscillatingmicrobalance (TEOM)

PM10 sub 10 micron particles Mk3 Cairns Instrument Services High volume sampler with sizeselective head

WS10 wind speed at 10 m. WMIII Climatronics Three cup anemometer

WD10 wind direction at 10 m. WMIII Climatronics Wind vane (single potentiometer)

AT10 air temperature at 10 m 6507 Unidata Thermistor, non aspirated

RH10 relative humidity at 10 m MP100 Rotronics Capacitative hygrometer

DELT temperature change 6-18 m - DEP Electronic subtraction

NR net radiation CN1 Middleton Net pyrradiometer

SR solar radiation 8-48 Eppley Black and white pyranometer

UV ultra violet radiation TUVR Eppley Selenium PE cell

RAIN rain 491001 Rimco Tipping bucket

PRES atmospheric pressure PTA427 Vaisala Barocap silicon capacitive gauge

ATRK Rsmog Airtrak 2000 Mineral ControlInstrumentation

Gas-phase titration

SODR wind profile Mk4 Australian Defence ForceAcademy

Doppler acoustic sounder

RASS wind and temperature profile Lap3000 Radian 915 MHz doppler radar and RASS

26

4.1.3. Daily Telemetry

Logging of data at each site was performed using aUnidata STARLOG Macro data logger. This deviceuses a microprocessor to control the conversion ofmeasured electrical signals to numerical values.

In operation, the logger scans all the air quality andmeteorological channels at the site at one secondintervals. The logger’s microprocessor then generatesa 10-minute average for each parameter, which issubsequently stored in the logger’s memory.

The stored data can be accessed by a suitablyprogrammed personal computer. The DEP’s system,used for the PPSS, employs a customised programwhich can be controlled using either remotecommands transmitted via a modem, or localcommands from the station computer’s keyboard. Theprogram allows many site software maintenance tasksto be performed remotely (from DEP head off ice),and allows transmission of logged data at any time(Rye 1991).

Retrieval of data by telemetry can be performedmanually or automatically, using a companionprogram on the central computer system located at theDEP head off ice (Figure 4.4).

Normally at 7am and 3pm each day, all sites areautomatically contacted by the central computer. Anynew data gathered by each logger since the last dial-up are transferred.

Data from each site are returned in compressedhexadecimal format. As part of the telemetry retrievalprocess, the measurements recorded in the data filesare converted firstly to the voltages measured by thelogger. These are then transformed, using the mostcurrent calibration (see Section 4.1.4) to measurementunits, such as metres per second for wind speed, andplaced into the data base for that site.

Each morning, time series plots of all 10-minuteaveraged data for the previous day are examined todetermine whether any faults or breakdowns haveoccurred.

On days when a smog event was expected, this dataretrieval was often performed hourly. Any site canalso be contacted manually at any time throughout theday, and the data viewed or retrieved.

4.1.4. Instrument Calibrations

Calibrations were performed routinely on allinstruments. During the summer months (Novemberto March), all i nstruments were calibrated twicemonthly. This process involved measuring outputvoltages from the monitor at conditions of zero input,and at three other levels covering the instrument’s

sensitivity range. At all other times, the monitors werecalibrated monthly.

The calibrations were performed on-site usingcertified span gases (i.e., of accurately knownconcentration), or a calibrated ozone source, andcalibrated gas blending equipment. The monitor’svoltage outputs, typically at 90%, 60%, 30% and 0%of its full scale input range were recorded in the sitelog book.

A linear calibration expression of the type:

Concentration = Slope x Voltage - Offset

was determined using the method of least squares.Generally, a correlation coeff icient better than 0.9999was achieved.

The sole deviation from this process was for thenephelometers, for which two-point (rather than four-point) calibrations were done twice monthly usingsuitable span and zero gases.

The gas blending equipment was calibrated at three-monthly intervals at the DEP’s workshop.

Calibrations of the meteorological equipment wereperformed at longer intervals (namely six to 12months) as these instruments tend to be more stablethan the gas monitoring equipment.

Further detail on quality control and preventativemaintenance is provided by Grieco, Mountford andKleinfelder (1996).

Computer

Site PC

Chart

RecorderMonitor

Data

Logger

Modem

Controlling

Figure 4.4 Monitoring site data flow.

27

4.1.5. Data Processing

Once a month, all the data that had entered each site’sdata base for the previous month were reprocessed toreplace any unwanted or incorrect data with an errorcode. The replaced data included calibrations, powerfailures, monitor repairs, monitor baseline drifts andother spurious occurrences that incorrectly reflectedthe ambient conditions prevaili ng at the time. The sitelog-book contained the dates and times ofmaintenance, such as calibrations and monitor repairs.The remainder, such as power failures and instrumentfailures, were found by cross-checking a computerplot of the monitor’s voltage output and themonitoring station chart recorder trace.

When reprocessing the data, any drift in the monitor’soutput during the period between any two calibrationswas assumed to be linear. A linear interpolation wasconsequently performed over time, from onecalibration equation to the next. An individualcalibration equation was generated from theinterpolation for every 10-minute average. The newlycalculated value was then re-entered into the database, overwriting the original value.

The initial stages of processing of data as describedabove are equally valid for the data obtained from themeteorological equipment. Data processing for thetwo types of equipment is different to the extent thatthere is no need for the secondary processing ofmeteorological data, the calibrations ofmeteorological equipment being essentially morestable. Meteorological data in the database wereedited monthly for errors that occurred due to powerfailures, instrumental problems, and occasionalunexplained spikes which were clearly incorrect data.

Grieco, Mountford and Kleinfelder (1996) providegreater detail on the daily and monthly dataprocessing procedures.

4.2. BUREAU OF METEOROLOGYSURFACE MEASUREMENTS

At the start of the study, the Bureau of Meteorologywas operating automatic weather stations at RottnestIsland, Swanbourne and Ocean Reef (see Figure 4.1).Later, sites at Mount Lawley and Gooseberry Hillwere added. Wind measurements from these siteswere available on the same 10-minute averaginginterval as used for the study’s data logging.

The Bureau provided data for the Rottnest Island,Swanbourne and Ocean Reef stations, initially as ablock of data commencing 1 January 1991, followedby monthly updates. Commencing in January 1994,the Mount Lawley data were also provided.

All measurements were provided on diskette, so thatprocessing was a simple procedure. A computerprogram was used to perform automatic format anddata validity checks. When passed, the wind datawere added to a data base compatible with the rest ofthe study data.

Of the four sites, two were near air quality monitoringstations, operated for the study, at which winds werealso measured. The Bureau of Meteorology’s data forthese locations formed a backup data set. In the caseof Rottnest Island, the Bureau of Meteorology’s siteproved slightly more representative of offshore winddirections, and was used in preference to the AQMSwind data when available.

4.3. VERTICAL MEASUREMENTS OFWIND AND TEMPERATURE

While photochemical smog is forming, the air masscarrying it can pass through a number of windregimes. Where these winds converge, verticalmotions develop and the forming smog can be carriedto higher levels. Later, the smog may descend in aregion where the winds diverge. Such complicationsmean that, to follow the movement of smog, details ofwinds at levels up to a kilometre or more are required.

In addition, to estimate the rate of dispersion of smog,the stabilit y of the atmosphere must be determined.To estimate this, air temperatures up to severalhundred metres are also required.

4.3.1. Sodar Sites

The term “sodar” is a derivative of “ radar” , and standsfor “sound detection and ranging” . Like a radar, asodar transmits pulses of waves, but uses sound ratherthan radio signals. Where small temperaturevariations in the atmosphere occur, the speed of soundchanges, causing some of the sound to be reflected.

Sodars incorporate three large enclosures, eachhousing loudspeakers mounted at the focus of aparabolic dish. One enclosure points verticallyupwards, while the other two point at an angle fromvertical (see Figure 4.5) and are mounted at rightangles to each other.

The three units emit, in sequence, a brief frequencytone of approximately 2kHz. The emitting speaker isthen switched to become a microphone for the nextfew seconds, and records the frequency of thereturned signal. The frequency is altered due to themovement of air towards or away from the receiver(the “doppler shift” ), so that radial velocities can becalculated. The three sets of velocity components are

28

transformed to generate horizontal and verticalvelocities for each measurement level.

During the period of the study, two sodar instrumentswere operated, as described below.

A sodar designed and constructed by the AustralianDefence Force Academy was placed at theCullacabardee monitoring station where it operatedfrom September 1992 onwards. It measured 30-minute average winds at 30-metre height intervals,with a quoted maximum range of 900 metres. Inpractice, 500 metres proved to be the effective upperlimit for reliable data (considered to be good

performance). The data recovery rate was fairly goodthrough 1993, at more than 85%, and better in 1994,at about 95%.

Data were initially recorded on a 10-minute averaginginterval. This period was chosen for compatibilit ywith the rest of the network data logging, but wasextended to 30 minutes on 18 February 1993, in theinterests of improving the range of the instrument.The benefit of the change was seen as an increase inrange by 50-100 metres (see Figure 4.6).

By way of example, Figure 4.7 presents the windprofile measured by the Cullacabardee sodar at 12noon on 4 February 1994, at the time of a significantozone event during the major field experiment. Itshows westerly winds changing through southerly atabout 500 metres (the top of the coastal inflow layerat the time).

The second sodar was operated on behalf of the studyby Murdoch University’s Institute of EnvironmentalScience. This was a commercial model, manufacturedby Remtech. It was located at the Gidgegannupmeteorological site from January 1993 to November1994, then moved to Rottnest Island, where itoperated from January to March 1995.

This instrument also measured winds at 30-metreheight intervals, to a quoted maximum of 620 metres.Data were again logged on a 30-minute averaginginterval. The range where 50% of measurements werevalid was 400 metres. This was less than that of theCullacabardee sodar, but still more than adequate forits intended task.

Figure 4.5. Arrangement of sodar transmitter / receivers used at Cullacabardee. One antennapointed east, one north, and one vertically.

0

20

40

60

80

100

0 100 200 300 400 500 600

HEIGHT (M)

Percent Frequencyof Maximum Range

30-minute averages

10-minute averages

Figure 4.6. Frequency of valid wind velocitymeasurements made by the Cullacabardee sodar, forboth 10-minute and 30-minute averaging intervals.

29

Some reliabilit y problems were experienced, with adata recovery rate of only 71%, but the main lossesoccurred in the winter seasons, when smog eventswere not to be expected.

Both sodar units provided a near-continuous record ofthe vertical profile of wind velocities in the lowestfew hundred metres of the atmosphere. Thisinformation provided valuable detail i n thereconstruction of wind fields.

The Murdoch sodar was shifted to Rottnest for thefinal smog season (1994-1995 summer) to captureany “multi -day” smog events which occurred.Experience had shown the predictabilit y of theseevents to be low. The sodar proved a less expensiveand more reliable monitoring method than thealternative, namely sending a radiosonde team toRottnest Island on several occasions with a stronglikelihood of false alarms and missed opportunities.

4.3.2. Radar Profiler

A radar system for measuring wind and temperatureprofiles was operated at Cullacabardee from January1994 to the end of the study. This site was chosenbecause is was expected to be representative of themeteorology of the coastal plain.

The system employs a 915 MHz doppler radar tomeasure vertical profiles of winds and a “radioacoustic sounding system” (abbreviated RASS) tomeasure temperature profiles.

Figure 4.8 shows the instrument, a Radian LAP3000model, sited at Cullacabardee. The central feature isthe electromagnetic antenna that can configure itselfelectronically to point in one of three differentdirections (unlike a conventional sodar system, forwhich three acoustic antennae are needed). Directioncontrol is achieved using electronic phasing of thesignal, to simulate an angled antenna.

To measure the wind speed and direction profiles, theelectromagnetic antenna operates in its threecomponent mode to measure radial velocities via thedoppler shift (the same principle as a sodar).Horizontal and vertical velocities at incrementalaltitudes are calculated by the radar software.

To measure the temperature profile, theelectromagnetic radar switches to a vertical beam-only mode, and the acoustic speakers are switched onto provide a continuous sound. Using sophisticatedsignal processing, the radar software is able todetermine the local speed of sound at successiveheights. Since the local speed of sound is related tothe local virtual temperature, the latter can then becalculated.

The RASS radar system operated with high reliabilit y.It remained in operation for 97.7% of the installedperiod, with valid temperature returns 93.8% of theperiod.

The wind profili ng radar operated in two continuouslyalternating modes. In its high-resolution mode, itmeasured winds from 110 metres to a maximumheight of 2.5 km, at 60-metre intervals. Itsimultaneously measured winds at lower verticalresolution (105 metres) to a maximum height of 4.2km. Temperature measurements commenced at 120metres, with a 60-metre resolution to a maximumheight of 1.5 km. Data from all profiles were storedevery 30 minutes, the winds being averaged over 25

200

400

600

1 2 3 4

Figure 4.7. Cullacabardee sodar wind vector plot for12 noon on 4 February 1994. The view is an obliqueprojection, with the elli pses corresponding to windspeeds at 1 m/s intervals. The west-to-east direction

is left-to-right. Vertical axis labels are heights inmetres.

Figure 4.8. The RASS radar system operated atCullacabardee. The white dome in the centre covered

the radar antennae, while the concrete cylindershoused the acoustic transmitters.

30

minutes and the temperature values being averagedover the five minutes in each half hour during whichthe RASS was switched on. The RASS radar systemproduced enormous quantities of data − about5,000,000 separate wind speed, wind direction, andtemperature measurements over the period of itsoperation.

The detail available is typified by Figure 4.9. Thisshows wind vectors overlaid on a potentialtemperature (defined on page 50) contour plot. Onthis day, a weak sea breeze developed about noon, buta major trough passage about 3pm brought asignificant smog event to the region. The figureshows wind direction changes as expected, but theslow cooling of the near-surface layers (with fewcontours in evidence) identified this day as a “coastalevent” . See page 50 for further explanation.

4.3.3. Perth Airport Radiosonde andWind Data

Until July 1995, the Bureau of Meteorology releasedtwo balloon-borne radiosondes every day from PerthAirport, one at 0600 and one at 1800 WST during thehotter months. Since July 1995, only a singleradiosonde has been released each day at 0600.

The radiosondes, which measure temperature andhumidity, are tracked by radar to determine locationand wind velocity. In practice the radar takes a shortperiod to lock on to the radiosonde, causing loss ofwind data over the lowest few hundred metres of the

atmosphere, which is the region of most interest tothis study. Nevertheless the data are a valuablesupplement to other information on upper level windsand temperature.

Radiosonde data have been obtained from the Bureauon a routine basis throughout the study and stored in adata base at the DEP.

4.4. EMISSIONS INVENTORY

One of the key elements in developing anunderstanding of the formation of photochemicalsmog is a detailed knowledge of the nature andamount of precursor emissions from all sources in theregion. In common with other studies of a similarnature in Australia and the USA (e.g., Carnovale et al.1991, Dickson et al. 1991) the approach taken in thePPSS was to consider emissions in four basic sourcegroups:

• motor vehicles;

• major industrial sites, commonly referred to aspoint sources;

• area-based sources - emissions from relativelysmall but widely distributed sources; and

• biogenic (natural) sources.

The approaches taken to investigate emissions fromthese source groups are briefly described in thissection. Discussion of the results and the processes

0 6 12 18 24Time (WST)

0

500

1000

1500

2000

2628

3030

3232

32

34

34

343434

36

36

Height(m)

Figure 4.9. RASS radar time-height plot for 19 February 1994. The arrows represent wind velocityvectors, a southerly (northward) direction being up the page. The heavy lines are contours of potential

temperature.

31

employed to check the validity of the estimates ispresented in Chapter 6.

As described in Chapter 2, the emitted species that areimportant in the development of photochemical smogare nitrogen oxides and reactive organic compounds.These were estimated from all sources in preparingthe emissions inventory for Perth. Other speciesquantified were carbon monoxide, carbon dioxide,sulphur dioxide, particulate matter and lead. Althoughthese are not all directly involved in thephotochemical smog process, knowledge of theirorigins assisted the interpretation of emissionsdistribution.

For airshed modelli ng purposes, emissions wereresolved in both time and space. For spatial resolutionthe study area was overlaid by a grid of cells, eachthree kilometres square. Individual source emissionswere allocated to grid cells on the basis of theirlocation. Account was also taken of the elevation atwhich emissions occurred with, for instance,emissions from tall stacks (over 45 metres) andaircraft being inserted into higher levels of the modelgrid. Hourly variations of source emission rates werealso recorded, giving a diurnal profile of emissionfluxes for each cell .

The year 1992 was taken as the base year for theinventory as it was the most recent year for whichsome crucial elements of the required informationwere available. A final step in preparing theinventory, therefore, was to develop forwardestimates of emissions for the years beyond 1992, toenable realistic air pollution predictions for thoseyears. The forward estimates were based on predictedpopulation and economic growth.

4.4.1. Motor Vehicle Emissions

Estimates of emissions from the Perth motor vehiclefleet were prepared for the PPSS by the Departmentof Transport (James 1995) utili sing information fromseveral sources. Spatial and temporal distribution ofvehicle movements, drawn from a traff ic modeldeveloped by Main Roads Western Australia and roadtraff ic counts, were combined with vehicle fleetemission factors derived from dynamometer testing ofa representative range of vehicles by authorities inVictoria and NSW (see Section 6.1 for more detail ).Resultant motor vehicle emission estimates wereassigned to the inventory grid cells according to theirlocation and the time of day.

Only on-road vehicles were considered in thissegment of the inventory. Other mobile sources suchas off-road vehicles, rail transport, boating andshipping were collected in the area-based segment.

4.4.2. Industrial Emissions

In preparing emissions inventories, industrial pointsources are identified and evaluated individuallybecause a relatively small number of sites usuallyaccount for a large proportion of the industrialemissions. They were estimated by the PPSS teamusing a methodology based on that developed for aMelbourne inventory (Carnovale et al. 1991). Thisentailed identifying industrial sites having significantemissions of photochemical smog precursors anddetermining their emissions via a questionnaireprocess.

Information sources utili sed in identifying sitesincluded:

• the Planning Land Use System (PLUS) databasedeveloped by the Ministry for Planning whichdetails more than 50,000 sites in a range ofindustrial and commercial land use categories inPerth;

• the Australian Bureau of Statistics (ABS)Business Counts for Western Australia;

• the Classified Plant database operated by theDepartment of Occupational Health Safety andWelfare, which identifies boiler plant in WesternAustralia;

• the Telecom “Yellow Pages” ;

• gas and electricity consumption data from theState Energy Commission; and

• the licensing database operated by theDepartment of Environmental Protection.

An initial li st of more than 8000 potential sites wasidentified from these sources. Application of furtherquali fying criteria such as the nature of processes onsite, fuel consumption and the type and probableamount of emissions (a lower cutoff of 10 tonnes peryear was applied), reduced the final li st of sites to330.

The operators of each of these sites were surveyed byquestionnaire to detail operations and emissions.Information sought included emission rates, daily andseasonal operational profiles (to enable temporalresolution of emissions), fuel and raw materialsconsumption (to characterise the process if emissionrates were unknown and had to be calculated), stackdetails (to characterise elevated emissions) and detailsof organic compound storage and handling (tocharacterise fugitive ROC emissions).

Questionnaires were reviewed to reconcile energy andraw materials consumption with site processes, annualproduction and emission estimates. Where noemissions information was available for a site,

32

estimates were made using published emission factors(US EPA 1985) for the process, sometimes modifiedto account for known local effects.

Point source emissions were located on the inventorygrid cell network according to their grid reference (to0.1 km). They were estimated in two groups:

• surface emissions which included fugitives andother emissions occurring from stacks less than45 metres; and

• elevated emissions occurring above 45 metres,with stack dimensions, exhaust velocities andtemperatures detailed for the modelli ng process.

Included in the latter group and afforded specialconsideration were aircraft emissions. These wereestimated for each of Perth’s three airports, takingaccount of aircraft types and their emissioncharacteristics, flight schedules and movementpatterns from the ground to 1000 metres altitude.

4.4.3. Area-Based Emissions

While atmospheric pollution is commonly associatedwith industrial emissions and motor vehicles, thereare many other human activities which give rise tophotochemical smog precursors. These sources tendto be individually small (for example domesticactivities and light industry) but large in number andwidely distributed.

Evaluation of these area-based emissions for the PPSSwas undertaken by the EPAV (Stuart and Carnovale1994). They analysed a range of area sourceclassifications:

• surface coatings and thinners − evaporation ofROC from decorative or protective painting ofbuildings, manufactured goods and other objects;

• service stations and refuelli ng − losses of ROCdue to vapour displacement during tank filli ng,tank breathing and vehicle refuelli ng;

• domestic and commercial use of aerosol products − hydrocarbon propellants and active ingredients;

• natural gas leakage − mainly due to imperfectionsin gas pipes, low in Perth because of therelatively new gas distribution system;

• cutback bitumen − ROC emissions from roadsurfacing operations;

• lawn mowing − a relatively minor activity withdisproportionately large emissions, partly due tothe use of two-stroke engines;

• dry cleaning − ROC emitted as a result ofevaporative loss of dry cleaning solvents;

• natural gas combustion − NOx and CO producedin the flame and some ROC from imperfectcombustion;

• domestic solid and liquid fuel combustion −NOx, ROC, CO and particulates from oil ,kerosene and wood combustion;

• domestic and commercial solvents - evaporativeloss of ROC from products such as cleaningfluids, toiletries, adhesives and polishes;

• railways − emissions from diesel locomotives;

• marine craft − engine emissions from leisure andcommercial vessels; and

• off-road vehicles − primarily from vehicles suchas tractors and construction vehicles.

In estimating the emissions from these sources, Stuartand Carnovale used a similar approach to thatpreviously taken for an emissions inventory inMelbourne (Carnovale et al. 1991). This involvedevaluation of each of the above classifications basedon gross consumption/activity data for Perth, derivedfrom sources such as the ABS, local utiliti es andmanufacturers’ and traders’ associations.

From this information, emission estimates were madeusing published US EPA emission factors (US EPA1985) for particular activities, or factors establishedfrom Australian experience where they wereavailable.

4.4.4. Biogenic Emissions

Biogenic sources of ROC are receiving increasingattention in regional oxidant studies (e.g., Carnovaleet al. 1991, Tanner and Zielinska 1994). Thepredominant source of these is vegetation.

Australian vegetation is a known source of ROC. Forthe Perth region, which is subject to morning windsfrom inland forest and pasture areas, there is potentialfor biogenic emissions to be significant in thephotochemical smog process.

The approach taken in estimating biogenic emissionswas to relate them to vegetation density. Satellit ephotographs of the Perth region were overlaid withthe inventory cell grid, and vegetation density wasestimated on a scale of 0 to 5 for each grid cell . Thevegetation species distribution for the Perth regionwas derived using natural vegetation maps (Beard1979a, 1979b). These provided the dominantvegetation class for each grid cell . From theserecords, measured canopy densities (Hingston,Dimmock and Turton 1980) and published canopyemission rates (e.g., Lamb, Gay and Westberg 1993)were used to derive biogenic emission estimates for

33

the region. More details are provided by Cope andIschtwan (1995).

4.5. FIELD EXPERIMENTS

4.5.1. Aircraft Survey

During the intensive study period of the 1993-1994summer, an instrumented aircraft was operated overthe PPSS region, with four objectives:

• to provide aircraft measurements of the totalregional emissions of nitrogen oxides andhydrocarbons, from both the Perth urban area andKwinana industry;

• to map out the extent of the photochemical smoggenerated on suitable days;

• to measure and track the Pinjar power stationplume, and monitor its interaction with the Perthurban plume if possible; and

• to provide micrometeorological data forconditions associated with the first threeobjectives.

The aircraft, a Cessna 340A, was owned and operatedby the Flinders Institute for Atmospheric and MarineSciences. The field exercise was jointly undertaken byscientists from the Institute and the CSIRO Divisionof Coal and Energy Technology. A detaileddescription of the aircraft and the experimental workis provided by Carras et al. (1995).

The aircraft was large enough to carry a crew of threeand a load of several chemical and meteorologicalsensors (Figures 4.10, 4.11). With an operating

endurance of six to seven hours, and a cruise speed of165 knots, it was well suited to the tasks involved insurveying the full extent of Perth’s smog plume.

In summary, the aircraft recorded the following:

Navigation: position, altitude above mean sea level,accelerations, time, altitude above ground.

Meteorology: wind velocity, ground surfacetemperature, air temperature, humidity, static anddynamic pressure, global radiation, vegetation index.

Chemistry: nitrogen oxides, ozone, methane, totalhydrocarbons, carbon monoxide, carbon dioxide.

Measurements were logged at rates between 1 and 50per second, being stored and preprocessed using anon-board computer. Some of the data could be

Figure 4.10. The instrumented nose of theCessna 340A used for the summer 1994

fieldwork. Struts support temperature, humidity,velocity and pressure sensors. The nose cone

encloses weather radar.

Figure 4.11. General view of the Flinders Institute for Atmospheric and Marine Sciences’ Cessna 340A.

34

displayed in real time in the aircraft, to provide acheck on instrument function, and to aid in-flightdecisions.

In addition to the electronic record gained this way,air samples were also taken in Tedlar® plastic bagsand in glass flasks. After the flight, these wereanalysed for chemical composition using gaschromatography.

The days and purposes of the flights completed areshown in Table 4.2.

The outcome of the aircraft survey is typified by theresults presented in Figure 4.12. This shows estimatesof the total mass flux of nitrogen oxides from thePerth metropolitan area, based on aircraftmeasurements of concentrations and wind velocitiesin the region immediately offshore, on the four dayslisted as “emissions measurement” in Table 4.2.

The graph shows significant scatter which ischaracteristic of measurements along the flight pathof an aircraft. However a general peak-hour flux ofclose to 1.2 kg/s is evident, as well as a trend of lesscertain magnitude from 7:30am to 9:30am. Thesemeasurements have been used in validation of thePerth emissions inventory, a task which otherwisewould not have been economically feasible. Furtherdetail on emissions estimates from the aircraftmeasurements is given in Chapter 6.

While there were no days during the experimentalperiod on which smog reached high levels, a well -defined smog event occurred on 4 February 1994. Onthis day, the aircraft tracked the emissions plume overthe ocean in the morning and then tracked thedeveloping smog plume inland to York and beyond inthe afternoon. Figure 4.13 shows a composite pictureof the path followed by the smog plume between 1pmand 3pm, based on the latter measurements. It clearly

identifies the separate contributions to the smogplume of emissions from the urban area and Kwinanaindustries.

4.5.2. Truck Survey

During the intensive study period of the 1993-1994summer, an instrumented truck was operated by theCSIRO Division of Coal and Energy Technology.

On-board monitoring equipment measuredconcentrations of nitrogen oxides, ozone, carbonmonoxide/dioxide, hydrocarbons and fine particles,and was capable of being operated while the vehiclewas in motion (Carras et al. 1994). The primaryfunction of the vehicle during the intensive studyperiod was to operate as a mobile measurementplatform or as a rapidly deployable monitoringstation. It worked in conjunction with the aircraft,taking ground level measurements of specific plumesthat were being identified aloft by the aircraft. Onother occasions it moved to strategic locations toprovide extra measurement data which characterisedthe inland progress of the sea breeze front and itssmog characteristics.

A notable observation from this aspect of theintensive field program was that no evidence of thePinjar NOx plume could be detected at ground levelby the truck when it was located under the plumeconcurrently being detected aloft by the aircraft.

Table 4.2. Aircraft operations during the PPSSintensive study period

Date Duration Purpose

27 January 1815-1915 Test flight

28 January 0752-1109 Emissions measurement

31 January 0755-1126 1526-1759

Emissions measurement Smog tracking

3 February 0759-1115 Emissions measurement and Pinjar plume

4 February 0714-1123 1257-1628

Sea breeze structure Smog tracking

7 February 0740-0915 Background air measurements

8 February 0657-1054 1235-1502

Emissions measurement and Pinjar plume

9 February 0800-1100 Kwinana emissions measurement

10 February 1108-1513 Sea breeze structure

0.0

0.5

1.0

1.5

2.0

7 8 9 10Time (hour WST)

NOx

(kg/s

as NO)

Figure 4.12. Estimates based on aircraftmeasurements, of the flux of nitrogen oxides from the

Perth metropolitan area over four mornings.

35

4.5.3. ROC Speciation Sampling

Most instruments used for continuous monitoring oforganic compounds in ambient air measure these inbulk, expressing the result as a “hydrocarbon”concentration. However, some knowledge of theoccurrence of individual organic species is importantfor a number of reasons:

• some organic species are far morephotochemically reactive than others and relativeconcentrations must be taken into account formeaningful photochemical modelli ng;

• the relative amounts of particular species canprovide information on the photochemical age ofthe reacting air mass;

• specific ROC sources often emit a characteristicmix of compounds − for example, acetylene is acharacteristic tracer for motor vehicle emissions(Galbally et al. 1995) − and speciation of ROC inair masses can often be used to identify theirorigin; and

• a knowledge of the amount and type of organiccompounds detected at specific locations can beused to verify estimates made in the emissionsinventory.

Consequently, an important component of theintensive field studies was a program to collect airsamples at selected locations for detailed ROCspeciation analysis and subsequent interpretation.

This work was undertaken by the CSIRO Division ofAtmospheric Research (Galbally et al. 1995).Sampling regimes were developed specifically tocharacterise:

• motor vehicle fleet emissions;

• the photochemical smog plume;

• the Kwinana industrial plume;

• rural (forest) air; and

• Perth urban air.

The approach taken in the analysis of air affected bymotor vehicle emissions was twofold. “ In-traff ic”sampling was achieved using a standard Commodoresedan equipped with a purpose-built pumping system.This fill ed a Tedlar® bag from a roof-mountedsnorkel with its intake approximately one metre abovethe roof of the car. Such sampling was mainly usedfor freeway and arterial road traff ic conditions. Tosample motor vehicle emissions more representativeof congested, stop/start driving conditions, a staticsampler was mounted about four metres above a CBDstreet at the Queens Buildings monitoring station.

From analysis of these samples and fuel consumptiondata, it was possible to estimate average fleetemission factors (in g/km) for CO2, NOx,hydrocarbons and CO under the different traff icconditions. These provided an independent estimateto compare with fleet emission factors derived fromthe motor vehicle emissions inventory. Additionally,the ratios of species measured in the sampled aircould be compared with ratios suggested by theemissions inventory. The advantage of the ratio

Quinns Rocks

Mullalloo

ObservationCity

Pearce

Perth

Jandakot

Northam

O3 ppb7060504030

Figure 4.13. The path of the smog plume measured between 1pm and 3pm on 4 February 1994,showing the separate contributions from urban and Kwinana emissions.

36

technique is that it is largely immune to dilutioneffects and the number of vehicles effectivelysampled.

Use of these techniques to verify the motor vehicleemissions inventory is described in more detail i nChapter 6. Given the potential for errors in thesampling and estimating processes, agreementbetween the measured and derived values wasfavourable. This work also supported the allowancesmade for the effects of deterioration of exhaustcatalysts in deriving the motor vehicle emissionsinventory.

For all other sampling regimes (for smog plumes,urban air, source plumes and forest air) a pumpedsystem was employed to draw air samples into eitherspecially prepared Tedlar® bags or stainless steelcanisters. This was done in such a way that the airwas drawn directly into the sample container, soavoiding the possibilit y of any contamination fromthe pump.

Based on the xylenes/toluene ratio observed in Perthsmog samples, Galbally et al. (1995) estimated thatthe smog is not significantly aged. This implies thatunder the meteorological conditions applying duringthe sampling campaign, all smog measured developedfrom ROC emitted on the same day.

4.5.4. Radiosonde and Pibal ReleasesBalloon-borne radiosondes were used to measurevertical wind and temperature profiles during periodswhen smog events were probable, and to provide datain support of other measurements during intensivefield experiments.

The radiosondes transmitted a signal containingpressure, temperature and humidity measurements.Using theodolite measurements of the elevation andazimuth of the sonde from the release site, and thepressure measurement to give sonde altitude, theposition of the sonde, and so wind velocity, could beestimated.

For the summers of 1992-1993 and 1993-1994, AIRradiosonde systems were hired from Queensland andVictorian electricity authorities. In addition, anOmega sonde system was operated under contract bythe Bureau of Meteorology during the 1993-1994summer. The DEP’s equipment, enhanced with theaid of an electronic data logger, was used for the1994-1995 summer.

Balloons without attached sondes (termed “pibals” )were also used at two sites over the 1993-1994summer. These were weighted to give apredetermined rate of rise. Elevation and azimuthmeasurements were again used to give position andwind velocity.

Release sites were at Rottnest Island, Rolli ng Green,Gidgegannup and Swanbourne. The first two of thesewere AQMS sites. During the 1992-1993 summer theSwanbourne radiosondes were released at the ovalbehind Swanbourne beach, the next year’s at theSwanbourne AQMS site, and in 1994-1995 within theSwanbourne army barracks area.

During the 1992-1993 summer, only the Swanbourneand Gidgegannup sites were used. The pattern ofoperation involved a nominal 6am radiosonde releaseat Gidgegannup, followed by others typically at noonand 3pm at Swanbourne. Dates were 5, 6, 7, 19, 20,21 and 22 January, and 4, 5, 12 and 16 February.

Figure 4.14 ill ustrates the reason for the use of theinland site. Winds measured at Gidgegannup by theradiosonde and sodar were very similar, up to thelimit of the sodar’s range. However, measurements atPerth Airport were unavailable below 300 metres’height, due to the rapid sonde ascent rate and timerequired for the tracking radar to lock on to theradiosonde. Winds at Cullacabardee weresignificantly lighter, probably due to the effects of theDarling Scarp. Without the Gidgegannup sonde,knowledge of the low-level winds inland would havebeen significantly lacking.

The greatest number of radiosonde releases was madeduring the 1993-1994 summer, coinciding with themost intensive phase of the PPSS measurementprogram, including the two-week field experiment.Radiosondes were released at Rottnest Island,Swanbourne and Rolli ng Green.

Normal operations comprised two teams, one atRottnest and the other at Swanbourne, both releasingradiosondes three-hourly from 6am to 3pm, withpibals at 90-minute intervals between. Pibal releasescontinued through the night in crucial periods,employing disposable torches tied under the balloon.The Rottnest team was generally staffed and managedby Murdoch University’s Institute of EnvironmentalScience, and the Swanbourne group by a teamcomprising DEP and contracted staff , although someinterchange occurred due to the demands of trainingand staff ing flexibilit y. The Rolli ng Green radiosondesystem was operated by the Bureau of Meteorology,but only on a subset of the operational days.

Operational days during 1993-1994 were 16 and 17December, 12, 13, 19, 20, 21, 30 and 31 January and3, 4, 8, 9, 10, 11, 12, 18 and 19 February. The Rolli ngGreen site was used on 31 December, 20, 21 and 31January and 12, 13 and 19 February.

The main purpose for the maintenance of radiosondemeasurements during the 1994-1995 summer was toensure no multi -day smog events were missed. Withsodars operating at Cullacabardee and Rottnest, and

37

radar temperature measurements inland atCullacabardee, the greatest value of radiosondemeasurements was to record the stabilit y of theinflowing marine air mass. Swanbourne was thereforethe only site used.

Operational days over the 1994-1995 summer were30 November, 2, 13, 15, 16, 22, 28 and 30 December,4, 5, 9, 10, 11, 16, 17, 18, 19, 26, 27, 30 and 31January, 1, 3, 4, 5, 6, 9, 10, 13, 14, 15, 17, 21, 22, 23,28 February and 7, 8, 10, 13, 14, 15, 16, 21, 22 and23 March. Most of these days involved only oneradiosonde release, as a precaution or in support ofother work (most often about 3pm). However, on 27January, 10 February, 22 February and 22 Marchseveral radiosondes covering the period from 6am to3pm were tracked.

4.6. DATA STORAGE ANDAVAILABILITY

The Department of Environmental Protection has anextensive suite of data processing software, developedoriginally during the Kwinana Air Modelli ng Study,and continuously enhanced since then. The system isaimed at eff icient storage and retrieval of timesequence measurements, and is readily interfaced to

the output of the Department’s data loggers, and tovarious data formats provided by sources such as theBureau of Meteorology and consultants.

The DEP’s data bases can accept continuous orbroken time sequences of any number of parallelchannels, at time intervals from one second to 24hours. This flexibilit y permitted the following datasets to be stored in the same structures:

• continuous surface measurements of air qualityand meteorology;

• Perth Airport radiosonde and balloon windprofiles;

• Cullacabardee, Gidgegannup and Rottnest Sodarwind profiles;

• Cullacabardee radar wind and temperatureprofiles; and

• air quality and meteorological measurementsmade by the Flinders University Cessna 340Aaircraft.

The only major data set gathered during the studywhich is not in this format is the collection of verticaltemperature and wind measurements made using

Airport sonde Time 0543

Gidgegannup sodar Time 0600

Cullacabardee sodar Time 0600

0 5 10 15 20

wind speed

0

200

400

600

800

1000

Gidgegannup sonde

0 5 10 15 20

wind speed

0

200

400

600

800

1000

0 5 10 15 20

wind speed

0

200

400

600

800

1000

0 5 10 15 20

wind speed

0

200

400

600

800

1000

Time 0600

Figure 4.14. Comparison of wind speed profiles (in m/s) measured using radiosondes at Gidgegannup and PerthAirport, and by the two sodar instruments, on the morning of 5 February 1993. The vertical axis is height above

ground level, in metres.

38

radiosondes. These measurements have been stored astext files, and in personal computer database formats.

With the primary exception of data from the Bureauof Meteorology, which may be obtained from thatorganisation, all of these measurements are readilyavailable. (Depending on the application, and thequantity of data requested, a charge may be applied.)

4.7. COMPUTER MODELLING

The three years of measurements undertaken duringthe study revealed the pattern of photochemical smogevents across much of the Perth region. By analysingthe regional winds, it was also possible to estimate thesources and destinations of the smog.

However, to understand in more detail which sourceswere the dominant contributors, and how changes inthese contributions would affect Perth’s air quality infuture, more sophisticated analysis was required. Thiswas because:

• meteorological measurements cannot representcompletely and accurately all the processesoccurring across the whole region; and

• the complexity of the photochemical reactionsrendered any simple analysis of sourcecontributions pointless.

To achieve a more sophisticated analysis, it wasnecessary to use computer models to represent theprocesses (both meteorological and chemical)contributing to smog events in the region.

The modelli ng procedures are described in detail i nChapter 7. Particular emphasis was placed on theapplication of a range of models, to the dual tasks ofsimulating the meteorology and the chemistry ofsmog events. In this manner, weaknesses inherent inparticular models were revealed, and the models mostapplicable for future air quality planning wereidentified.

39

5.1. AIR QUALITY DATA

The air quality and meteorological data collectedduring the Perth Photochemical Smog Study arestored in data bases which are accessed by computermodels and other computer-based systems asrequired. A comprehensive summary of air qualitydata has been prepared (Grieco 1996).

Presentation of data in this report is limited to a broadoverview of the measurements of ozone, which is thekey indicator of photochemical smog, together with asummary of nitrogen dioxide measurements in theCentral Business District. The latter, as well as beinga significant contributor to ozone formation, is also arecognised pollutant.

5.1.1. Ozone Concentrations

Table 5.1 provides a condensed set of statisticsshowing the number of occasions, over the indicatedtotal number of “smog seasons” (October to March,inclusive), on which the various nominated criteriawere exceeded. The number of “smog seasons” isdifferent for different sites, therefore the table shouldnot be used for site to site comparison unless theexceedance numbers are divided by the associated

number of “smog seasons” (i.e., converting to a “persmog season” basis).

Table 5.2 (page 40), which is drawn from Chapter 2,Tables 2.3 and 2.4 (pages 13 and 14), provides anexplanation of the source of the criteria tested inTable 5.1. An assessment of the health implications isgiven in Section 5.2.

The highest 1-hour average ozone measurements atthe various monitoring stations, given in the last rowof Table 5.1, are displayed as bar graphs in Figure 5.1(page 40), where the horizontal marker on each graphis the 80 ppb level. The figure ill ustrates thatphotochemical smog is a regional, not a local, issue.

Table 5.3 (page 41) takes another row of Table 5.1,namely “Number of hours > 80 ppb” and expands thetime dimension to show the dates on which theseoccurred. Also shown are the number of stationswhich recorded an event on each day and the numberof separate hours during which 80 ppb was exceeded.

Usually the smog events are of short duration, withonly one hour per day exceeding the 80 ppb level.Figure 5.2 shows ozone trends for a representativesmog event at Caversham, with a sharp rise associatedwith sea breeze arrival, then a decrease shortly

5. Summary and Analysis of SmogMeasurements

Table 5.1. Summary of ozone measurements over the period 1989 to 1995.

Monitoring Results CA CU GG JA KE QR RG RI SW TR

Number of smog seasons (October to March) 6 3 1 1 3 3 3 3 3 2

Number of hours > 120 ppb (NHMRC, Vic EPA) 2 0 0 0 0 0 0 0 0 0

Number of hours > 100 ppb (NHMRC pending) 9 1 0 0 0 0 2 1 2 0

Number of hours > 80 ppb (Canada, WHO) 35 10 2 0 8 8 19 8 12 1

Number of 4-hours > 80 ppb (NHMRC pending) 5 0 0 0 0 0 2 1 1 0

Number of 8-hours > 50 ppb (Vic EPA, WHO) 23 9 4 1 11 14 16 8 14 2

Number of days with one hour > 80 ppb 20 7 1 0 6 5 10 5 8 1

Number of days with one hour > 100 ppb 6 1 0 0 0 0 1 1 1 0

Number of days with 4-hour periods > 80 ppb 5 0 0 0 0 0 2 1 1 0

Number of days with 8-hour periods > 50 ppb 23 9 4 1 11 14 16 8 14 2

Highest one hour concentration (ppb) 133 101 91 73 99 91 112 103 107 94

Site codes: CA - Caversham, CU - Cullacabardee, GG - Gingin, JA - Jandakot, KE - Kenwick,QR - Quinns Rocks, RG - Rolli ng Green, RI - Rottnest Island, SW - Swanbourne, TR - Two Rocks

40

afterward. Measurements have shown that this patternis typical of all i nland sites.

However, Table 5.3 shows that at some sites therehave been smog events stretching over three hours atleast. Generally these occur under differentmeteorological conditions - for example, when smogoccurs due to the effect of bushfire smoke, with windssouth westerly all day.

The short duration of smog events is a commonAustralian phenomenon. Events tend to be of muchlonger duration in many parts of the United States andEurope.

Figure 5.3 extracts from Table 5.3 the number of daysduring each of the smog seasons in the study onwhich the 1-hour average ozone concentrationexceeded 80 ppb at one or more of the monitoringstations, and presents this information as a bar graph.Essentially the figure indicates “ the number of smogdays per year” . This information is valuable forcomparison with other cities, since it is not directlydependent on the number of monitoring stations. Overthe three smog seasons of the study, there have been9, 9 and 12 such smog days respectively, giving anaverage of 10 per year.

Figure 5.3 also gives a clear indication thatphotochemical smog is a seasonal phenomenon.Figure 5.4 has been included to confirm this point,showing highest 1-hour ozone concentrations permonth at four selected monitoring sites. The sitesstretch roughly in a line from Rottnest (20 kmoffshore) to Rolli ng Green (60 km inland). Each plotshows a characteristic annual cycle peaking in latesummer, with winter values close to the naturalbackground ozone concentration.

0 5 10 15 20

Scale (km)

Quinns Rocks

Two Rocks

Caversham

Kenwick

Rolling Green

Cullacabardee

Maximum Hourly Average Ozone

1 October 1992 to 31 March 1995

KEYShaded box representshighest measurementBold box represents80 ppb WHO goal

Swanbourne

Rottnest

Rockingham

Figure 5.1. Maximum hourly average ozoneconcentrations during the period of PPSS. The

Rockingham site only operated from 28 January to 28February 1994.

Table 5.2. Sources of ozone concentration criteria

Ozone concentration guideline Source of guideline

1-hour average of 120 ppb(NHMRC, Vic EPA)

National Health and Medical Research Council and Australian and NewZealand Environment Council (NHMRC 1990).Government of Victoria (1981).

1-hour average of 100 ppb(NHMRC)

National Health and Medical Research Council (NHMRC) (1995).

1-hour average of 80 ppb(Canada, WHO)

Canadian Federal-Provincial Committee on Air Pollution (IUAPPA 1988).World Health Organisation (Europe) (1987).

4-hour average of 80 ppb(NHMRC)

National Health and Medical Research Council (1995).

8-hour average of 50 ppb(Vic EPA, WHO)

World Health Organisation (Europe) (1987).Government of Victoria (1981).

0

30

60

90

120

150

Ozone

(ppb)

0 6 12 18 24Time (hours WST)

Figure 5.2. Ozone concentrations at Caversham on atypical photochemical smog day, 18 January 1991.

41

Table 5.3. Number of separate hours per day on which the hourly average ozone concentrationexceeded 80 ppb.

Summer CA CU GG JA KE QR RG RI SW TR

89-90 28 Feb 90 1

18 Jan 91 2

90-91 20 Jan 91 1

22 Feb 91 3

02 Mar 91 1

91-92 29 Jan 92 2

07 Mar 92 1

22 Oct 92 3 2 - 2 -

05 Dec 92 - - - 1 -

08 Jan 93 2 1 2 2 - - 1 -

12 Jan 93 1 - - - - - - -

92-93 13 Jan 93 2 - - - 3 - - -

29 Jan 93 1 1 - - 2 - - -

30 Jan 93 1 - - - 3 - - -

12 Feb 93 - - - 1 - 1 - -

17 Feb 93 1 - - - 1 - - -

19 Dec 93 - - 1 - - - - -

22 Jan 94 3 1 - - 1 - - -

23 Jan 94 3 - 1 - - - - -

19 Feb 94 - - - - 1 1 1

93-94 16 Mar 94 - - - - - 1 - - -

17 Mar 94 1 - - - - - - - -

18 Mar 94 - 2 - - - - - - -

21 Mar 94 - 1 - - - - - - -

14 Apr 94 - - - - - - - 1

17 Dec 94 - 2 2 - - - 3 - -

16 Jan 95 2 2 - - - - 3 - -

17 Jan 95 - - - - - - 1 - -

18 Jan 95 2 - - - - - - - -

31 Jan 95 - - - - - - 1 - -

94-95 10 Feb 95 - - - - - 2 - 2 -

12 Feb 95 2 - - - 1 - - - -

17 Feb 95 - - - - - - - - 1

20 Feb 95 - - - - - 2 - 3 3

21 Feb 95 - - - - - 1 - - 2

22 Mar 95 - - - - 1 - - - -

23 Mar 95 - - - - - - - 1 -

Site codes: CA - Caversham, CU - Cullacabardee, GG - Gingin, JA - Jandakot, KE - Kenwick,QR - Quinns Rocks, RG - Rolli ng Green, RI - Rottnest Island, SW - Swanbourne, TR - Two Rocks.

“ -” means the station was operating but no exceedance was recorded. A blank means that the station was notoperating at the time.

42

0

1

2

3

4

5

O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A

Number

Month

1992 1993 1994 1995

of days

Figure 5.3. Number of days per month on which peak 1-hour ozone concentrations exceeded 80 ppbsomewhere in the Perth region.

CAVERSHAM

0

40

80

120

(ppb)

Ozone

SWANBOURNE

0

40

80

120

(ppb)

Ozone

ROTTNEST ISLAND

0

40

80

120

(ppb)

Ozone

ROLLING GREEN

0

80

120

O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A

(ppb)

Ozone

40

Month

1992 1993 1994 1995

Figure 5.4. Peak hourly average ozone concentration for each month, from October 1992 to April 1995.

43

Statistics of ozone concentrations in other states ofAustralia are presented in Table 5.4. Beforecomparing the figures within the table and withpreceding tables or figures, an explanation ofdiffering data processing methods is required.

The Department of Environmental Protection storesall air quality data as 10-minute averages. When dataare processed to produce, for example, the statistics inTable 5.1, processing is done on a rolli ng 10-minutebasis. The highest hourly average at Caversham, 133

ppb (see Table 5.1), was found via data processing tohave occurred between 1430 and 1530 on the 22February 1991. In other words, processing of hourlyor other averages is not restricted to “clock hours”such as 1400 to 1500 or 1500 to 1600.

If calculated hourly averages are constrained to startand finish on a clock hour, it is inevitable that bothmaximum 1-hour averages and the number ofexceedances of standards will be underestimated.

Table 5.4. Ozone statistics for selected Australian cities, 1993-1994.

Criterion Year Perth(a)

Perth(b)

Sydney(b)

Melbourne(b)

Adelaide(b)

Brisbane(c)

Number of sites in theurban network

1993

1994

8

10

8

10

12

14

11

12

2

2

4

7

Highest 1-hour ozoneconcentration for the yearin the network

1993

1994

112

104

112

103

155

142

172

133

83

68

123

138

Second highest 1-hourozone concentration forthe year in the network

1993

1994

112

103

110

103

150

130

170

110

81

65

107

105

Number of days when the1-hour ozoneconcentration exceeded80 ppb somewhere in thenetwork

1993

1994

8

9

5

7

17

31

14

19

1

0

3

6

Number of days when the1-hour ozoneconcentration exceeded100 ppb somewhere inthe network

1993

1994

2

2

2

2

6

15

8

4

0

0

1

2

Number of days when the1-hour ozoneconcentration exceeded120 ppb somewhere inthe network

1993

1994

0

0

0

0

3

1

3

1

0

0

1

1

Number of days when the4-hour mean ozoneconcentration exceeded80 ppb somewhere in thenetwork

1993

1994

1

3

1

3

9

7

0

0

1

3

Number of days when the8-hour mean ozoneconcentration exceeded50 ppb somewhere in thenetwork

1993

1994

13

11

13

11

20

31

0

0

6

18

Source: Environmental Protection Agencies of NSW, Vic., SA, Qld and WA.(a) running averages with 10-minute resolution.(b) clock hours.(c) running averages with 30-minute resolution.

44

This is because sharp ozone peaks, such as Perthexperiences, often straddle the boundary of clockhours. Health symptoms occur independently of thetime within an hour when an ozone event occurs, sothe results of data processing should similarly beindependent of when events occur.

Nevertheless it is common practice among the otherstates of Australia and overseas to process data inclock hours. To accommodate this mismatch, Table5.4 contains two columns for Perth data. The firstcolumn contains figures based on rolli ng 10-minutecalculations and is therefore consistent with thepreceding figures and tables. The second columncontains figures based on clock hours. As can be seenthere are significant differences for 1-hourexceedance frequencies but not for 4-hour or 8-houraverages.

In summary, Table 5.4 shows that ozone levels inSydney and Melbourne are similar, as are those inPerth and Brisbane, the former pair experiencinglevels somewhat higher than the latter pair.

It is worth noting that the number of times a particularozone goal (e.g., 80 ppb 1-hour average) is exceeded

is a very sensitive indicator. A modest increase inPerth’s ozone levels may cause a rapid rise in thefrequency with which ozone goals are exceeded.

This is ill ustrated in Table 5.5, which has beenproduced by taking Caversham ozone measurements,scaling these measurements up by 10% and 20% andcalculating the number of times (based on rolli ngaverages) that the various ozone goals would beexceeded in each case. The table indicates how theresults might change if modest increases in ozonelevels, associated with the growth of Perth, were tooccur.

5.1.2. Nitrogen DioxideConcentrations

Figure 5.5 presents nitrogen dioxide measurements atQueens Buildings in the Central Business District(CBD) of Perth. The data are presented as annualcumulative frequency curves, which show the numberof 1-hour periods for a year that any particularconcentration was reached or exceeded. Such graphsprovide a simple way to view a full year’smeasurements.

Figure 5.5 shows that the highest nitrogen dioxideconcentration recorded in five years was 150 ppb in1991. This can be compared to Australian 1-houraverage ambient guidelines (Table 5.6).

It is apparent that the nitrogen dioxide concentrationsin the CBD are not far below the guidelines. It is to beexpected that nitrogen dioxide concentrations in theCBD will be significantly higher than at otherlocations in the metropolitan area because:

• emissions of fresh nitrogen dioxide from vehicleexhausts are most concentrated in the CBD; and

• titration of ozone by nitric oxide to form nitrogendioxide will be most vigorous in the CBD, againbecause of the greater quantity of exhaustemissions containing nitric oxide.

Table 5.5. Projected increases in the number of times that ozone goals will be exceeded at Caversham if ozonelevels increase by 10% or 20%.

Current Data + 10% + 20%

CAVERSHAM 91-92 92-93 93-94 94-95 91-92 92-93 93-94 94-95 91-92 92-93 93-94 94-95

Number of hours > 80 ppb 3 11 7 6 10 19 14 12 23 28 15 27

Number of hours > 100 ppb 1 3 2 0 1 4 4 3 2 10 6 5

Number of hours > 120 ppb 1 0 0 0 1 1 0 0 1 3 2 0

Number of 4 hours > 80 ppb 0 1 2 0 1 1 2 2 2 3 3 4

Table 5.6. Ambient hourly nitrogen dioxideconcentration guidelines employed in Australia.

Source ofGuideline

Concentration(ppb)

Maximumpermissibleexceedances

Victorian EPA

acceptable level

detrimental level

150

250

3 days peryear

none

NHMRC 160 once permonth

45

Inspection of nitrogen dioxide measurements fromother monitoring stations showed that concentrationsat Caversham reached about half of those recorded inthe CBD, with other stations showing lower levels.

5.1.3. Fine Particle Concentrations

The monitoring sites at which fine particle mass orvisibilit y reduction (associated with fine particles)were measured during the study were indicated inTable 4.1.

The instruments measuring fine particle mass(PM2.5) were installed in mid 1994 as part of aseparate study, so there was only a partial overlapwith ozone measurements.

A nephelometer, measuring visibilit y reduction, wasinstalled at the Caversham site in late 1989 andoperated throughout the period of the smog study.This instrument provides a measure of the presence ofairborne particles which scatter visible light. Some ofthe particles produced in photochemical smog fallwithin this size range, as evidenced by the visiblehaze they produce when present in suff icientconcentration. Similarly, particles from bushfiresmoke, domestic fires and other combustion processesare detected by the instrument.

To determine whether there is evidence for theformation of significant amounts of f ine particles

during smog events, the highest hourly average ozoneconcentration for each of the Caversham smog eventsidentified in Table 5.3 (starting from 1991) wasplotted against the coincident hourly averagenephelometer reading (Bsp, a measure of lightscattering). The plot in Figure 5.6 includes four points(marked) which have been identified from separaterecords as being bushfire events, with associated highparticle concentrations from smoke. The remaining

NO (ppb)

1

10

100

1000

10000

0 20 40 60 80 100 120 140

01/01/90 - 31/12/9001/01/91 - 31/12/9101/01/92 - 31/12/9201/01/93 - 31/12/9301/01/94 - 31/12/94

2

Hours

Figure 5.5. Number of hours that measured nitrogen dioxide concentrations exceeded levels ranging from 0 to150 ppb at Queen’s Buildings, central Perth, for all completed years since 1990.

0.0

0.5

1.0

1.5

80 100 120 140Ozone (ppb)

January-February 1991-1995,1200-1500 averagesp

B

Identified bushfire smoke events

Figure 5.6. Relationship between visibilit y reduction(Bsp) and ozone concentration for ozone events at

Caversham with an hourly average concentration over80 ppb, from 1991 to 1995.

46

points lie (on average) a littl e above the average valueof Bsp for the early afternoon on summer days (seethe horizontal li ne). This is likely to be explained inpart by the generally lighter winds on smog eventdays which result in a higher concentration ofparticles from all sources. Nevertheless the values arelow, indicating good visibilit y. It is significant thatthere is no apparent trend of visibilit y with increasingozone, suggesting a very weak dependence, if any, atCaversham.

A similar plot has been produced for the smaller set offine particle mass data, yielding the same conclusions.

During the winter months of 1994 and 1995, therewere several occasions on which the hourly averagenephelometer readings at a suburban site exceededBsp = 8 during the night and were still above Bsp = 3well after sunrise. These events were clearly causedby smoke from domestic fires over a broad suburbanarea. The particle loading from photochemical smog(during summer) is negligible by comparison.

Accordingly, fine particle formation withinphotochemical smog is not included within the healthassessment or modelli ng components of this study.

5.2. ASSESSING THE IMPACT OFSMOG LEVELS IN PERTH

5.2.1. Assessing Health Risks ofSmog Levels

A wide range of information is needed on which tobase a comprehensive assessment of health risksassociated with photochemical smog in Perth. Indetermining the links between the level of exposureand the risk of a particular health effect, bothexperimental and epidemiological research isconsidered to establish exposure-response (also called“dose-response’) relationships. While we have someinformation on the relationship between ozone levelsand reductions in lung function, just how much of anincrease in ozone concentrations is suff icient to causean asthma attack or the development of respiratorydisease is poorly defined.

Measurement of exposure is also needed for riskassessment. It is diff icult to estimate precisely thehealth risks of ozone in the general populationbecause the profile of population exposure has not yetbeen established accurately. Further, the measuredlevels of ozone may not directly indicate the dosereceived by individual people out in the community.The dose each person’s lungs receive is related tobreathing rates and activity patterns (such as timespent outdoors).

In Perth and other Australian cities, people generallyspend more time indoors than outdoors. The groupmost exposed to photochemical smog may thereforeinclude outdoor workers, school children and peoplewho are not in regular daytime work indoors, becauseozone is most prevalent during the daylight hours ofthe warmer months. Exposure levels from earlieryears may also be relevant to the development ofchronic conditions, in contrast with acute episodes ofshortness of breath or asthma. Without thisinformation, estimates of exposure of the populationof Perth remain imprecise.

In a review of the health impacts of ozone inAustralia, potentially adverse health episodes weredefined as including temporary, but clinicallysignificant, reduction in forced expiratory volume.With incomplete information to quantify the risks,only a very approximate estimate of adverse healtheffects was made (Woodward 1993). Exposure of thepopulation to ozone levels greater than 80 ppb (1-hour average) was considered, taking account of themeasurements of photochemical oxidants madeduring the 1980s, and the geographical distribution ofthe population of the five largest cities in Australia. Ifozone concentrations were kept at levels notexceeding a 1-hour average of 80 ppb, possiblyseveral milli on brief episodes of respiratorysymptoms could be prevented each year. This averagefigure applies to the national population of 17 milli on(Guest et al. 1994), but the effects would mostprobably be experienced by a small number of peoplein the most exposed parts of the worst affected cities.

Because asthma is so common (about 10% of allAustralians have asthma), a small i ncrease in the riskof an episode of asthma for an individual could addsubstantially to the national burden of ill ness, if largenumbers of people are exposed to ozone. Theindividual risk of an asthma attack could rise by asmuch as 25% following an increase in ozoneconcentration of 40 ppb (Holguin and Buff ler 1985).Again, if ozone concentrations were kept at levels notexceeding a 1-hour average of 80 ppb, many episodesof asthma might be averted each year in Australia.Avoidance of the chronic, mild asthma tendencycould be even more important. Many lives might besomewhat improved if fewer episodes of asthma wereexperienced by that proportion of the populationwhich is susceptible.

Considering the levels of ozone monitored in Perth(Table 5.1), only two hours in six seasons have beenrecorded in which the old NHMRC goal of 120 ppbwas exceeded (both at Caversham). It is thus unlikelythat acute, severe respiratory diff iculty has so fardeveloped in this city because of outdoor exposures toozone. However, several monitoring sites that reflect

47

ozone concentrations over more densely populatedareas have identified periods of up to three hoursexceeding 80 ppb. The discussion above referring to aslight additive risk to the development of asthmaattacks or respiratory symptoms is relevant here.

Ozone concentrations exceeding the level of 80 ppboccur on average on about 10 days per year (“highozone” days) somewhere in the Perth region. Toproceed further with a risk assessment requiresestimation of the “dose” of ozone actually inhaled bythe population (the measurement of ambient levelsshould not be equated with the dose humans receive),and their sensitivity to it. For example, assume that10% of the residential area of the city is affected, andthat 10% of the population is outside for more thanone hour. Thus about 1% of Perth’s population, orabout 10,000 people, might be affected on each “highozone” day. Perhaps 1000 of these people couldexperience some respiratory symptoms, although thenumber is likely to be much less, given thatepidemiological studies have shown that middle-agedand older people are probably less susceptible to therespiratory effects of ozone than young adults andchildren. The number of affected asthmatics on eachoccasion is also likely to be well below 1000 in Perth,but this estimate must be considered highly tentative.The number of asthmatics affected severely by thelevels of ozone so far recorded in Perth is probablyvery small , and could be zero.

The limitations of the health risk assessmentsassociated with exposures to ozone applied to thenational population have been noted elsewhere (Guestet al. 1994). Those national estimates provide,however, a guide to the interpretation of the Perthozone monitoring data. It is reasonable to infer thatozone concentrations exceeding 80 ppb, sustained forthree hours, may cause symptoms in a small minorityof the exposed population, and exacerbate asthma in afew individuals in Perth on each occasion. It isunlikely that nitrogen dioxide levels in Perthcontribute to respiratory disease.

Attribution of shortness of breath or the occurrence ofan asthma attack in any one individual to levels of airpollution would go beyond our current understandingof the effects of photochemical oxidants. Many partsof Perth now experience periods of at least eighthours when ozone concentrations exceed 50 ppb. It ispossible, but unproven, that exposures of this kind,over months and years, are associated with thedevelopment of chronic respiratory disease in avulnerable proportion of the population (Woodwardet al. 1995).

In summary, the current health effects ofphotochemical oxidants in Perth are considered mild.

Exceedances of the new NHMRC air quality goals forozone do occur, however, so it should be assumedthat potentially preventable and possibly cumulativeeffects are occurring each year during the warmermonths. Table 5.5 indicates that the number of timesthe NHMRC and other goals are exceeded aroundPerth is set to grow rapidly if emissions andconsequent smog concentrations increase by modestamounts. In other words, Perth appears likely toexceed the ozone goals regularly. From theperspective of public health, therefore, the control ofphotochemical smog in Perth should be considerednow.

5.2.2. Assessing the Potential forVegetation Impacts

Consideration of cumulative exposure indices andannual averages for ozone, as noted in Section 2.5, isuseful for the assessment of ozone impacts onvegetation. Lefohn, Krupa, and Winstanley (1990)presented information on ozone exposure regimes atso called “clean” locations around the world. Most ofthese sites are remote from significant humansettlement. Some basic statistics on cumulativeexposure at some of these remote sites, and selectedsites in Perth, are given in Table 5.7 for comparison.

Although Perth occasionally experiences episodes ofhigh ozone concentrations, it is clear that in terms ofcumulative exposure and annual means, ozone levelsin Perth are comparable to or lower than those at mostof the sites selected by Lefohn et al. (1990) as beingrepresentative of clean, background sites on a globalscale. The SUM0 parameter, which is the sum of allhourly concentrations above zero, is sensitive to datacapture and baseline, but the sites in the table may becompared with a high degree of confidence, as datacapture was high, and the variabilit y of ozone levels islow at remote sites. Within Australia, another cleansite for comparison is the baseline monitoring site atCape Grim, Tasmania, which had annual averageozone levels of 23-25 ppb (1976-1980), derived fromobservations with winds greater than 20 km/h andtrajectories from the southern ocean (Galbally et al.1986). Once again, Perth values are comparable orlower than those even at this remote location.

Lower annual mean values are often reported fromsites influenced by large urban centres, where there issignificant scavenging of ozone by nitric oxide. Theremay be some evidence of NO scavenging atCaversham, which has lower annual averages thanRottnest Island. However, values at Two Rocks,Rottnest and Rolli ng Green, where such scavenging isexpected to be minimal, are still l ow in global terms.

48

Background ozone concentrations are also influencedby the elevation of the monitoring site.

Despite these low annual means, the data presented inTable 5.1 show that exceedances of the VictorianEPA 8-hour 50 ppb standard, which was establishedfor vegetation protection, occur several times eachyear in Perth. Exceedances of some of the other long-term standards, such as the WHO 24-hour and 8-hourstandards (Table 2.4), also occur at times. Theseepisodes probably cause some transient stress tosensitive crops and vegetation, the impact of which ishard to assess reliably in the absence ofcomprehensive field studies of ozone impacts underAustralian conditions.

The low cumulative ozone exposure indices for Perthand relatively infrequent episodes of highconcentrations provide some confidence thatvegetation impact is not likely to be substantial.Despite this, the situation warrants close monitoring.As noted previously with regard to human healtheffects, a modest increase in emissions could result ina disproportionate increase in the frequency of smogevents exceeding ambient standards. The effects oncrops and other vegetation may then become a moreimmediate cause for concern.

5.3. METEOROLOGY OF SMOGEVENTS

Before the Perth Photochemical Smog Study, smogmeasurements were available from only theCaversham site. One purpose of the study was toprovide a broader perspective of smog events in thePerth region. The monitoring program has shown thatthe conditions causing increased ozone concentrationsat Caversham are only part of a range which can leadto smog events. A comprehensive assessment of thesmog events observed during the study is provided byRye (1996b). The classes of smog events identified inthis assessment are described below.

5.3.1. Inland EventsThe inland event class is the one primarily responsiblefor the smog events observed at the Caversham site.

The initial conditions identifying these events aremorning wind direction between north east and east,and wind speed at inland sites of 3 m/s or less.Usually (but not always) there is a low pressuretrough with its axis just offshore. In thesecircumstances, the day’s maximum temperature isusually more than 30oC, and the sea breeze arrives atthe coast around noon, or slightly earlier.

Because of the northerly component of the morning

Table 5.7. Ozone concentrations at “clean” sites around the world, compared to some PPSS sites.

Site Data capture (%)Annual Mean

Concentration (ppb) SUM0 (ppm hr)

WORLDWIDE

South Pole 93.8 29.4 238

Barrow, Alaska 94.0 26.6 208

Mauna Loa, Hawaii 96.1 38.6 321

American Samoa 92.4 13.6 105

PERTH

Caversham 1993 98.0 14.4 124

Caversham 1994 99.0 14.5 126

Rolli ng Green 1993 95.0 19.1 160

Rolli ng Green 1994 99.1 19.4 168

Two Rocks 1993 99.1 18.5 161

Rottnest Island 1993 86.5 21.4 162

Rottnest Island 1994 99.4 21.0 183

Periods: South Pole and Mauna Loa, 1980-1986; Barrow and Samoa, 1980-1987.

49

winds, the sea breeze has a more westerly directionthan usual. This ensures that the morning’s emissionstend to return across the city, rather than being carriedto the northern suburbs.

For normal rates of progress inland of the sea breezefront, peak ozone concentrations develop at inlandsites between 2pm and 3pm.

The mass of air which receives morning peak-houremissions is initially carried offshore by the lighteasterly winds. When the sea breeze arrives, thedirection of the wind carrying the air mass reverses.

The air then returns across the main urban region. Theresulting double dose of the received emissions is themajor factor in the generation of high ozone levelsinland (see Figure 5.7(a)).

Several other characteristics of these conditions arealso important in the process. Mixing depths arelimited by a strong temperature inversion capping theinflow (Figure 5.8). Experience also shows that whenwinds are stronger than about 3 m/s, morning peak-hour emissions may be carried too far offshore, andbe lost from the region before the sea breeze forms.

The coastal low pressure trough is also a major factorin these events. Due to the decrease of the offshorewind velocity to zero near the trough axis, morning

emissions remain closer to the coast. In addition, theair to the west of the trough axis is much cooler. Asthe sea breeze forms, this air is brought onshore,restricting the depth of convection over land andkeeping mixing depths low. The developing smogmass is therefore less dispersed as it moves inland.

5.3.2. Kwinana EventsThis set of smog days is generally similar in nature tothe “inland event” set, except for more southerlymorning wind directions. Urban emissions return overthe northern suburbs, and Kwinana emissions returnover the main urban region (see Figure 5.7(b)). Theozone which forms in the metropolitan area is theproduct of emissions from both the Kwinana andPerth regions. Ozone levels may be increased atCaversham, but the more northern site atCullacabardee tends to receive the highestconcentrations.

Although air trajectories passed through the Kwinanaregion on these days, the contribution of emissionsfrom Kwinana industry to photochemical smog levelsin Perth was unclear. One major role of analysesconducted during the study was to evaluate theimportance of Kwinana emissions on such days (seeSection 7.3).

12

11

10

9 8 7

8 9

10 11

13

12

(a)

11

12

13

14

1516

17

18

19

(b)

10

11

12

13

14Inlandevent

eventKwinana

Caversham

Cullacabardee

10

Figure 5.7. Air trajectories based on estimated winds at 100 metres’ height for days representative of types ofozone events: (a) an inland event (4 February 1994), when urban emissions returned across the urban region,

while emissions from the Kwinana region remained isolated; (b) a Kwinana event (18 March 1994, right), whenKwinana emissions returned across the urban area. Numbers on the trajectories are hours WST.

50

5.3.3. Coastal Events

This class represents a major additional regimerevealed by the Perth Photochemical Smog Study. Itis characterised by high concentrations of smogmeasured at coastal sites, but generally quitemoderate levels inland. Figure 5.9(a) shows that forsuch days, the appearance of the air mass trajectorycan be much the same as inland events.

However, in comparison to inland events, there aresignificant differences in the meteorology of thisclass. Although general air mass trajectories can besimilar, morning wind speeds are generally greater.The coastal low pressure trough is initially furtheroffshore, so the cold air mass to its west is not drawninto the sea breeze. The upper levels of the sea breezeinflow are therefore warmer than on days of inlandevents (see Figure 5.8).

This has two effects. The presence of warm air

offshore, overlying the cool sea surface, maintainsconditions of high atmospheric stabilit y. The city’smorning emissions are therefore held in highconcentrations, a condition which also enhances therates of the chemical reactions forming ozone. Then,when the smog mass reaches the coast, the warmerinflow temperature allows mixing depths to growmore rapidly inland. The ozone in the inflow istherefore largely dispersed before it reaches inlandmonitoring sites.

5.3.4. Bushfire Smoke Events

A significant fraction of photochemical smog eventsin Perth is due to chemical reactions betweencompounds in bushfire smoke, and Perth’s normalurban emissions of organic compounds and nitrogenoxides.

This smoke can contribute to ozone levels in theconditions already described, but can also produce

Potential Temperature (C)

0

500

1000

1500

2000

20 24 28 32 36 28 32 36 40Potential Temperature (C)

Swanbourne12 noon, 4 February 1994

Swanbourne2 p.m., 21 February 1995

Height(m)

Figure 5.8. Potential temperatures (see below for definition) at Swanbourne at 12 noon on 4 February1994 (left), and at 2pm on 21 February 1995 (right). The strong cooling of the layer below 500 metres

in the first case restricted vertical dispersion inland, a typical condition on “ inland event” days. Theshallower temperature inversion near the surface in the second case was characteristic of conditions

expected on “coastal event” days.

Potential Temperature

When a mass of air rises, the reducing pressure means that the air expands and cools. This change oftemperature is the reference against which atmospheric stability is measured. Because of the cooling, in a plot ofair temperature against height, the reference line marking the transition between unstable (convective) andstable conditions has a slope. Potential temperature is a concept which corrects for this effect.

It is defined mathematically by θ = T (1000/P)2/7, where T is temperature on the Kelvin scale, and P is pressurein hPa (hectopascal). For a rising air mass, θ is constant. Graphs of θ against height slope to the right in stableconditions, to the left in unstable conditions.

51

high concentrations when smog would otherwise notoccur. This appears to be due to the highconcentrations, and high reactivity, of organiccompounds in the smoke.

The trajectory shown in Figure 5.9(b) is for an event

when winds were south westerly all day. A lowpressure trough had passed inland on the previousday, and the effects of the trough on atmosphericstability were still evident. Figure 5.10 shows that astrong, low level temperature inversion was present inthe morning. The maximum surface air temperatureon the day was 26oC, and the evening temperatureprofile was similar to Figure 5.10, indicating that lowmixing depths persisted throughout the day.

Normally, in these conditions, no significant ozoneconcentrations would be expected. However,increased levels of ozone were detectable by the timethe inflow had reached Cullacabardee, andconcentrations at Caversham and Rolling Green, werehigher still (Figure 5.11).

The absence of air mass recirculation is a commonindicator of these events. The relatively gradualincrease, then decrease, of ozone concentration alsoidentifies them. Bushfire-related smog peaks reachmagnitudes similar to those due to urban emissions,but their longer duration makes their potential healtheffects greater.

5.3.5. Light Westerly Wind Events

These events are relatively uncommon, and arecharacterised by moderate levels of ozone − butrelatively high levels of nitrogen oxides − in the Perthmetropolitan area. High levels of ozone can form inthe rural region to the east. The typical pattern is of a

121110

9 8 7

(a)

14

13

12

11

10

Cullacabardee

Caversham

Rolling Green

(b)

Coastalevent

Bushfiresmokeevent

Figure 5.9. Air trajectories representative of types of ozone events: (a) a coastal event (21 February 1995) −the trajectory shown is not significantly different from that in Figure 5.7(a), (b) a bushfire smoke event (13January 1993) − high levels of ozone were detected at Cullacabardee, Caversham and Rolling Green about

2pm. Numbers on the trajectories are times, in hours WST.

18 24 30 36Potential Temperature (C)

0

500

1000

1500

2000

Height(m)

Figure 5.10. Potential temperatures at Perth Airportat 6am on 13 January 1993.

52

southerly-to-westerly wind all day, of relatively lowspeed. Emissions from the urban area and possiblyKwinana (as in Figure 5.12) accumulate in light wind,low mixing depth conditions in the morning, and theair mass is carried inland as the onshore flowstrengthens. The Rolli ng Green measurement siteshows increased ozone concentrations in the earlyafternoon.

5.4. ANALYSIS USING THEINTEGRATED EMPIRICAL RATEMETHOD

The Integrated Empirical Rate (IER) model ofphotochemical smog formation was developed byGraham Johnson and colleagues at the CSIRODivision of Coal and Energy Technology (Johnson,1983). Its main innovation is that it provides a way tounderstand the key aspects of photochemical smogprocesses from ambient monitoring data alone,without requiring a detailed emissions inventory.

The IER theory focuses on the essential features ofphotochemical smog development. It employs theexperimental results of smog chamber studies toderive a relatively simple set of algebraic expressions.The theory has also been recast into a set ofdifferential equations, called the Generic Reaction Set(GRS), which may be substituted for the complexequation sets of most smog models.

13

12

11

10

9

8

Rolling Green

Figure 5.11. Trajectory ending at Rolli ng Green on 31January 1995 at 1:30pm, the start of an ozone episodeat that site. The path crossed major emissions sources

during the morning, and is representative of thoseleading to a “light westerly wind” smog event.

00 06 12 18 24 0 20 40 60 80100120

Roll ing Green A.Q.M.S.

00 06 12 18 24 0 20 40 60 80100120

Cullacabardee A.Q.M.S.

O

(ppb)

00 06 12 18 24 0 20 40 60 80100120

Caversham A.Q.M.S.

3

Figure 5.11. Ozone concentrations (ppb) at three inland sites on 13 January 1993. The smog onthis day was formed by a mixture of bushfire smoke and urban emissions.

53

5.4.1. Description of the IntegratedEmpirical Rate Model

Johnson (1983) provided a derivation of the IERmodel. Later additions have clarified somerelationships, giving a version as summarised below:

An important starting point is the recognition that theconcentration of ozone at any particular location isnot necessarily a complete indicator of photochemicalsmog production. This is particularly true close to thecentral business district (CBD) of a city like Perth,where the nitric oxide emitted by motor vehiclesreacts with ozone in the air to produce nitrogendioxide:

NO + O3 → NO2 + O2 (5.1)

An ozone monitor sited just downwind of the CBDwould therefore measure low values. However, theboost to photochemical potential provided by theemissions of nitrogen oxides could mean that amonitor some kilometres downwind would measurelevels higher than those upwind of the CBD.

To allow for the ozone − nitrogen dioxide exchange,the IER theory defines a parameter called SMOG:

{ SMOG} t = { NOy} t - { NO} t + { O3} t (5.2)

where { } t denotes number of moles of the particularspecies present in the air mass at time t.

The term NOy represents the full range of gaseousoxygenated nitrogen species

{ NOy} = { NO} +{ NO2} +{ HNO3} (5.3)

+{ PAN} +{ gaseous organic nitrates} + ...

The symbol NOy is used as an extension of apreceding convention, in which the combination ofNO and NO2 only is denoted as NOx.

In addition to the gaseous species, there is also arange of stable non-gaseous nitrogen compoundsformed in photochemical smog. These are denoted bythe symbol SNGN.

The value of { SMOG} , which is approximately equalto { O3} + { NO2} , is a much more stable measure ofphotochemical activity than { O3} alone. It remainsconstant when NO is added to an air parcel,producing NO2 at the expense of O3.

The most commonly used instrument which measuresoxides of nitrogen records { NOy} directly. Thismeans that { SMOG} is directly measurable, andprovides a robust measure of smog development overspace and time.

Developing the theory, Johnson defined a secondparameter, SP, which stands for “smog produced” .This combines ozone and nitrogen compounds with

the same rationale as SMOG. However, it representsonly the products of smog reactions, both gaseous andnon-gaseous. It therefore excludes those componentsof the smog mixture which were present in theatmosphere initially (mainly background ozone) andthose which were directly emitted into the air (mainlythe NO2 emitted from various sources such as motorvehicles).

This definition means that SP is the sum of the totalNO consumed (which must become smog products ofsome sort) plus the total ozone produced:

{ SP} t = { NO} 0,t - { NO} t +

{ O3} t - { O3} 0 (5.4)

where { } 0 denotes the number of moles present at astart time (i.e., background) and { } 0,t denotes thenumber of moles of the species initially present in theair plus the moles added via emissions up to time t.

The IER model relies on a set of empiricalrelationships derived from smog chamberexperiments (Johnson 1983). The central finding ofthese experiments was that the production of smog,described by the SP parameter, follows a well -behaved pattern, ill ustrated by Figure 5.12.

The horizontal axis may be thought of as cumulativesunlight exposure, since the parameter kNO2, which isthe rate coeff icient for the photolysis of NO2, isdirectly determined from the ultraviolet light intensity(Johnson 1984). The function f(T) accounts for theeffect of air temperature, T, on the photolysis rate.

As shown in Figure 5.12, smog production proceedsinitially as a linear function of cumulative sunlight

0

200

400

600

SP

0 100 200 300

(ppb)

NO -limited regimex

Lightlimitedregime

∫k f(T) dNO

2

τ

Figure 5.12. Concentrations of Smog Produced as afunction of sunlight exposure, measured in a CSIROsmog chamber experiment (after Azzi, Johnson and

Cope 1992).

54

exposure. As ozone is generated, NO and NO2 aredepleted, by reactions forming other nitrogencompounds which do not participate further in ozoneformation. Then smog production ceases, with thetotal moles of SP (comprised largely of ozone)remaining constant.

The IER model separates the two phases of smogproduction. The first phase is called the “light-limitedregime” because the rate of smog production dependson sunlight intensity. The rate also depends on theamount and reactivity of ROC species in the air andthe air temperature. The second phase is called the“NOx-limited regime”, since it only occurs whenNOx has been consumed.

In this second phase, reactions which produce SNGNwill continue even though O3 production has ceased.

To understand the options available to controlphotochemical smog in a city like Perth, it is mostimportant to consider the implications of Figure 5.12.Perth experiences strong sunlight and hightemperatures in summer so that, given a mix of ROCand NOx emissions typical of an urban centre, we canexpect smog reactions to occur relatively quickly.

Reducing ROC emissions without reducing NOxemissions will slow the smog reactions, but may havelittl e effect on the ultimate amount of ozone produced − although the extra time taken could be beneficial, inallowing more dispersion.

Corresponding reductions of ROC emissions in a citywith less sunlight and lower temperatures might be farmore beneficial. The potentially highest smogconcentration occurs at the onset of the NOx-limitedregime, and this might not be reached beforenightfall .

The IER model is valuable in enabling an assessmentof the status of smog development given the datafrom a network of monitoring stations. The modelequations which permit this assessment are derived asfollows.

Empirical Relationships fromExperimental Observations

The Light-Limited Regime: Here, the smog chamberobservations represented in Figure 5.12 yield therelationship

SP = Rsmog ∫0

tkNO2

f(T) dτ (5.5)

where Rsmog is the photolytic rate coeff icient fororganic species, T is air temperature and τ is timethrough the day.

Rsmog can be calculated from the product of theconcentration of an organic species [ROC] with itsphotochemical reactivity coeff icient, aROC, summedover all such species, that is,

Rsmog = a [ROC(i )]ROC( i )i 1

n

=∑

= a [ROC]ROC total (5.6)

Johnson (1984) indicated that a good estimate ofaROC, the effective overall photochemical reactivity

coeff icient, in an urban atmosphere, is 0.0067.

Assessing the Mix of Emitted Nitrogen Oxides: Theemissions of NOy occur as NO and NO2 only (and soare given the combined name NOx). In general,across the various urban sources of NOx, NO is alarge fraction of the total emission, that is,

{ NO} 0,t = F{ NOx} 0,t (5.7)

where F has a nominal value of 0.9. This relationshipassumes that there is no significant initialconcentration of NO2 in the air mass.

Significant concentrations of NO2 might be found inan urban air mass at the start of a day if the previousday’s smog plume has been recirculated back to theurban area (i.e., a multi -day smog event). In this casethe model needs to be applied to the whole event,starting at the first day when the assumptions inequation (5.7) are correct.

CSIRO terminology for { NOx} 0,t is { NOx} em − thatis, the emissions of NOx. This term is used below.

Removal to Stable Non-Gaseous NitrogenCompounds: In the light-limited regime, the rate offormation of SNGN is proportional to the rate of totalsmog production:

{ SNGN} t = P{ SP} t (5.8)

where the value of P has been determined to be about0.125.

Attainment of the NOx-Limited Regime: When allnitrogen oxides have been consumed, no more smogmay be produced. The maximum quantity of smogproduced is found to be proportional to the initialsupply of nitrogen oxides − that is,

SPmax = β { NOx} em (5.9)

where β has been measured to have a value ofapproximately 4.

55

Model Equations for the Light-limitedRegime

Conservation of nitrogen species requires that

{ NOx} em = { NOy} t + { SNGN} t (5.10)

Using this relation, and the preceding ones, we findthat the amount of NOx emitted into the air, and theamount of smog produced, can be estimated asfollows:

{ NOx} em= {NO } P({O } {O } ) {NO} )

1 FPy

t3

t3

0 t+ − −−

(5.11)

{ SP} t = {O } O } NO} F{NO }

1 FP3

t3

0 ty

t− − +−

{ {

(5.12)

Having evaluated { NOx} em we can evaluate SPmaxfrom equation (5.9). The extent to which the smogreactions have progressed towards the onset of NOx-limited conditions can then be expressed as:

E = { SP} t/{ SP} max (5.13)

where E will always lie in the range zero to one. E is acrucial measure of the smog formation process, andits use is discussed in Section 5.4.2.

Model Equations for the NOx-limitedRegime

We can also derive equations equivalent to (5.11)and (5.12) for the NOx-limited regime:

{ NOx} em ={O } {O }

F3

t3

0−−β

(5.14)

{ SP} t = SPmax=β

β({O } {O } )

F3

t3

0−−

(5.15)

Determining the Regime

Before the above equations can be employed it is firstnecessary to know which regime is current at thelocation and time of interest. This determination canbe made by calculating, using equations (5.11) and(5.14), the term

G = {NO }

{NO } x em

x em

(equation 5.14)

(equation 5.11)(5.16)

If the value of G is less than one then the regime islight-limited. If G is greater than or equal to one thenthe regime is NOx-limited.

The derivation of this method for determining theregime is based on the observation that equation(5.14) will underestimate { NOx} em in light-limitedconditions, just as (5.11) will underestimate{ NOx} em in NOx-limited conditions.

Effect Of Dispersion: Converting toConcentration Units

In all of the above equations, the species areexpressed as moles present or emitted. This can bethought of as the numbers of molecules within thevolume of an experimental smog chamber.

In the fixed volume of a smog chamber, conversion tounits of concentration is simple. In the realatmosphere, the initial air mass (the urban plume)containing all of the initial emissions grows overtime, due to the mixing of additional air into theplume. Mixing occurs due to both lateral dispersion,and vertical dispersion as the turbulent layer abovethe earth's surface grows deeper during daytime.

Consequently, while the number of moles of somespecies increase and others decrease via chemicalreactions, the effect of dilution due to atmosphericmixing is superimposed on the concentrations of all .

The single exception is ozone, for which abackground concentration is present which may varyacross the region. Neglecting the variabilit y, it isapparent that only the excess ozone, { O3} t-{ O3} 0,will be diluted at the same rate as other species. In theforegoing equations, ozone is expressed as excessozone in every instance.

Since the dilution effect is linear and uniform acrossall species including excess ozone, the concentrationsof the various species measured at a monitoringstation are in the same ratio as the number of moles inthe diluted “air mass” near the station. Therefore, theequations for SMOG, SP, G and E at time t may berewritten in terms of measured concentrations ratherthan moles. However, the term { NOx} em makes littl esense expressed as a concentration.

Input Data Requirements

In summary, the IER model allows an assessment ofthe status of smog formation from the results ofambient monitoring. Equations (5.7) to (5.16) allowthe status of smog development to be determinedfrom routine monitoring data (NO, NOy and O3), butonly at the time of measurement and at the location of

56

the monitoring station. If analyses forward orbackward in space and or time are to be made, thenthe following additional data are required:

• a calculation of Rsmog from equation (5.6), or adirect measurement of this parameter (e.g. by anAirtrak instrument);

• solar radiation and temperature records orprojections to allow equation (5.5) to beevaluated; and

• model calculations or wind records for forwardor back trajectory modelli ng.

5.4.2. Application of the IER Model

Extent Calculation from AQMS Data

The parameter E in equation (5.13) was defined as the“Extent” to which the smog reactions have proceededtowards NOx-limited conditions. It is one of the set ofparameters which can be determined from routineNO, NOy and O3 monitoring data at a monitoringstation. Given the data from a network of monitoringstations, the Extent parameter is particularly useful inproviding an insight into smog processes over aregion.

In the light-limited regime (LLR), where E is lessthan unity, smog production is determined by theavailabilit y of ROC, the air temperature and theintensity of sunlight. In the NOx-limited regime(NLR), E is equal to unity and smog production islimited by the availabilit y of NOx, since the smogplume has utili sed all the NOx available and hasproduced the maximum possible amount of smog(SPmax). Smog production will cease even in thepresence of abundant ROC and sunlight. Only

injection of more NOx would result in an increase insmog concentrations.

Strategies to control smog concentrations in aparticular locality need to take proper account of thestage of smog development indicated by typicalvalues of E during smog events. If the air is in thelight-limited regime (E < 1), regional ROC reductionswould be effective but NOx control could be counter-productive. If the air is clearly in the NOx-limitedregime (E = 1), regional reduction of NOx emissionswould reduce smog levels, while the benefit of ROCreductions is less clear. However, as explained byBlanchard, Roth and Jeffr ies (1993), the IER modeltends to over-estimate E, thereby predictingpremature onset of NOx-limited conditions.Accordingly, IER predictions should not be used inisolation when considering the merits of NOxcontrols.

Furthermore, smog control strategies need a region-wide focus, to ensure that a “fix” for one locality doesnot worsen the situation elsewhere.

Calculated values of E during smog events across thestudy period, as presented below, provide a basis forassessing possible regional smog control strategies.

Figure 5.13 is a graph of the cumulative frequency ofoccurrence of E values for smog events (averagedover one hour) at representative monitoring stationsnominated in each plot (see Figure 5.14 for locations).Only events which had a 1-hour average [SP] (fromequation 5.4) greater than 50 ppb were considered assmog events. A background ozone concentration of15 ppb was used in the calculation of [SP]. The datawere compiled as 1-hour averages over the periodOctober 1992 to April 1995.

This method of presentation is designed to extractvalues of E for every clearly identifiable smog eventat the various monitoring stations and to plot thesevalues in a way which allows a comparison of thedifferent characteristics of smog at each site (i.e.,tendency to NOx or light-limited regimes), neglectingthe other details of the smog events (i.e., how many,what concentrations).

The graphs show, for each station, the percentage of1-hour events for which the calculated E was equal toor greater than the E value on the horizontal axis. Forexample, the graph for Caversham indicates that 80%of the identified smog events for that station had Eequal to or greater than 0.6, whereas 100% of theevents at Rolli ng Green (further inland) had E greaterthan 0.6 (i.e., further towards being NOx-limited thanCaversham).

In summary, stations which show graphs staying highand close to horizontal are in a region which is

0.2 0.4 0.6 0.8 1.0 Extent of smog formation

20

40

60

80

100

Swanbourne

Rolling Green

Rottnest

Jandakot

Caversham

Cullacabardee

Kenwick

%

(E)

Figure 5.13. Cumulative frequency with which IERExtent values were exceeded at several PPSS sites.

57

predominantly NOx-limited. Graphs which dropsignificantly from left to right indicate that the stationis in a region which is frequently light-limited.

The point at which each graph ends (E = 1) indicatesthe probability of NOx-limited conditions occurringat that station. In reality the transition to NOx-limitedconditions begins before E fully reaches unity, as canbe seen in Figure 5.13. The vertical dashed line on thegraphs at E = 0.95 indicates the possible start of thistransition zone, within which both ROC and NOxemission controls may be effective (Chang and Suzio1995). As seen from the graph for Caversham, 35% ofthe smog events have E equal to or greater than 0.95,i.e., a probability of 0.35 that smog events (asdefined) will be NOx-limited. The correspondingprobability for Swanbourne is 0.8.

In Figure 5.14, the probability of smog events beingNOx-limited is plotted as contours across the Perthregion, based on the method explained above for thevarious monitoring stations.

Interpretation of Extent Calculations

The cumulative frequency plots and the contour plotstogether provide a basis for comparing andinterpreting the smog measurements at each station.

It must firstly be stressed that the data extracted foranalysis relate only to clearly identified smog eventsat monitoring stations. By definition these events willonly occur after there has been adequate time forsmog to form. For example, the many morning hourson which fresh urban emissions pass out to sea overSwanbourne and Rottnest (with significant potentialto react) will not be extracted in this analysis as thereactions will be in the early light-limited stages, withminimal ozone having been formed.

The assessment of smog events in Section 5.3confirmed that significant smog events in the Perthregion involve urban and/or industrial emissionsbeing blown out to sea, reacting under sunlight, andthen returning back over the land. Hence if amonitoring station indicates a relatively low value ofE coincident with a significant smog concentration,the most likely explanation is that the air mass beingmeasured has been out to sea and, while returningacross the land, has picked up fresh emissions ofNOx.

This hypothesis is successful in explaining the broadfeatures of the contour plot, as follows:

• Onshore winds reaching Rottnest, Swanbourne,Quinns Rocks and Two Rocks are unlikely tohave trajectories which have recently passed overfresh NOx sources, hence the probability ofNOx-limited conditions is high (Swanbourne is

sometimes affected by fresh Kwinana emissions,which distorts the overall result).

• Rolling Green is well downwind of significantfresh NOx emissions and therefore almost alwaysNOx-limited. The same appears to be true atGingin also, indicating that NOx from the Pinjargas turbines does not strongly influence smogconcentrations downstream in the sea breeze.

• Caversham, during onshore winds, is downwindof a major source of fresh motor vehicleemissions from the CBD (NOx and ROC) andtherefore has a high probability of light-limitedsmog events. The same is true to a lesser extent atCullacabardee which is downwind of densesuburban areas.

• Both Jandakot and Kenwick are downstream,during south westerly winds, of the Kwinanaindustrial area, a major source of NOx and ROC.

360 380 400 4206440

6460

6480

6500

6520

8070

60

60

50

50

4070

80

90

Kenwick

Jandakot

Caversham

Rottnest

Swanbourne

Quinns Rocks

Two Rocks Gingin

RollingGreen

Cullacabardee

90

40 30

Pinjar

Figure 5.14. Probability across the Perth region ofsmog events being NOx-limited. Contours are omitted

where there are insufficient ozone measurements.Axis labels are AMG coordinates, in kilometres.

Reduced probabilities occur in areas receiving freshurban or Kwinana emissions.

58

In Chapter 4 the procedures used to develop the PPSSemissions inventory were outlined. In this chapter theanalyses and results are described in detail , togetherwith practical methods used to verify emissionestimates where possible.

Development of the Perth inventory followed in manyrespects the methodology used to develop aninventory for Melbourne (Carnovale et al. 1991).

Overall , the relative contributions to Perth’s smogprecursors from the various source groups showedsome similarities to Melbourne and also somedifferences. Figure 6.1 shows the relative contributionto annual emissions of the three main precursorspecies from the anthropogenic source groups (i.e.,excluding biogenic emissions).

NOx emissions are dominated by motor vehicles,being about half the total. However, of the remainder,the fraction attributable to industry is significantly

larger than in the Melbourne inventory. This reflectsthe existence of large NOx sources in the Kwinanaindustrial area, which is within the PPSS region.

Source contributions to ROC emissions are similar inratio to those of Melbourne, with motor vehicles andarea sources dominating industrial sources.

Carbon monoxide emissions are predominantly frommotor vehicles, as is normal for most cities. Industrialemissions of CO are almost negligible.

6.1. MOTOR VEHICLE EMISSIONS

6.1.1. Methodology

An overview of the methods employed by theDepartment of Transport (James 1995) to assemblethe vehicle emissions inventory is described in thissection, with specific issues being discussed in greaterdepth in subsequent sections.

The motor vehicle emissions inventory utili sed datafrom several sources:

• Vehicular movement data were obtained from acomputer-based traff ic model developed by MainRoads Western Australia (MRWA). This modelrepresents all metropolitan roads by a node andlink network. Modelled traff ic flows arecalibrated against actual vehicle counts made at arange of sites over the metropolitan area. Themodel classifies roads according to their capacityand type of construction, both of which influencevehicle behaviour. Model output, together withvehicle count data for hourly andweekday/weekend variations (see Section 6.1.2)allowed daily and hourly vehicle kilometrestravelled (VKT) and traff ic flow conditions oneach link to be estimated.

• For each major emitted species, emission factors(grams emitted per kilometre travelled) forspecific makes and models of vehicles werederived. These were based on vehicle test dataproduced by the Environment ProtectionAuthorities of Victoria and New South Wales(EPA NSW 1994). The emission factors tookaccount of the fact that vehicle emissions varyaccording to the mode of travel (e.g., freewayversus congested). The effect of ambienttemperature was also included, as discussed inSection 6.1.3. Fleet average emission factors alsoreflected the effects of regulatory controls onemission characteristics of new vehicles

6. Emissions Analysis

MotorVehicles

51%

Industrial44%

Area5%

ROC

CO

NOx

Industrial19%

Area37% Motor

Vehicles44%

Industrial2%

Area18%

MotorVehicles

80%

Figure 6.1. Source group contributions to annualphotochemical smog precursor emissions, Perth 1992.

59

introduced from 1986 onwards under AustralianDesign Rule (ADR) specifications, as discussedin Section 6.1.4.

• The state Register of Motor Vehicles (1991) wasused to provide the information on the makes andmodels which comprise the Perth motor vehiclefleet (James 1993). The distance travelled as afunction of vehicle age was found from the 1991Survey of Motor Vehicle Usage (AustralianBureau of Statistics 1991).

Combination of these databases enabled vehicle fleetemissions to be estimated on an hourly basis for eachlink in the MRWA model. Resultant link data werethen assigned to the emissions inventory grid cell i ntowhich the link fell , providing an estimate of the motorvehicle fleet emissions resolved in time and space.Gross annual emissions from motor vehicles in thePerth region in 1992 are shown in Figure 6.2.

Forward projections of estimated motor vehicleemissions were made based on traff ic volume changespredicted by the MRWA model, plus other expectedeffects such as the attrition of older (uncontrolledemission) vehicles from the fleet. Expected trends inmotor vehicle emissions are discussed further inSection 6.1.4.

6.1.2. Spatial and TemporalDistribution

The spatial distribution of emissions from motorvehicles is naturally dependent on traff ic density. Asill ustrated in Figure 6.3, emissions are mostconcentrated in the immediate neighbourhood of theCBD and main traff ic arteries through themetropolitan area. This pattern is typical of resultsfound in other cities (e.g., Carnovale et al. 1991).

It is important to define correctly the timing ofmorning peak hour emissions in relation to Perth’ssummer wind cycle (morning easterly followed by anafternoon sea breeze), so that computer models cancorrectly estimate the trajectory and ultimate impactof these emissions.

The temporal variation in emissions from motorvehicles was taken to be dependent on the number ofvehicles on each link for a given hour of the day. Thiswas determined by analysis of traff ic count datacollected during 1992-1993 by the MRWA at 44 sitesaround the metropolitan area. Variation of the diurnalprofile of vehicular activity was notably small fromone site to another. A representative temporal profilefor the region was therefore derived by adding thehourly vehicle counts for each site, as shown inFigure 6.4 (James 1995).

Hourly vehicle counts were then expressed as apercentage of the daily total so that the profile couldbe applied to individual li nks in the traff ic model. Theresult was a temporal distribution of VKT andtherefore emissions for each link.

6.1.3. Temperature Effects

While the spatial distribution of vehicular emissions isconsistent with traff ic distribution, the actual quantityof emissions varies from day to day according to theambient temperature. This is especially so for ROCemissions, as a significant portion of these arise fromevaporative losses and are therefore temperaturedependent.

0

5

10

15

20

25

30

NOx ROC CO/10 SO Lead Partic.2

Figure 6.2. Annual emissions (thousands of tonnesper year) from motor vehicles, Perth, 1992. The CO

value has been divided by 10 to fit on the graph.

2

5

10

20

50

100

200

Figure 6.3. Estimated 1994 vehicle emissions of ROCon an ozone event day, in kg/km2.

60

A common approach in estimating motor vehicleROC emissions is to consider them in two categories − tailpipe, or exhaust, emissions and evaporatives(also known as running losses). The latter has beentaken to include:

• evaporation from the fuel injection/carburationsystems;

• crankcase losses (in pre-1969 vehicles);

• hot soak losses (evaporation from hot enginesafter they have been turned off); and

• diurnal losses (loss of fuel vapour from parkedvehicles as a result of the daily ambienttemperature change).

Estimates used for these evaporative losses areexpressed as an emission factor (measured in g/km) tobe added to the exhaust emission of ROC. In deriving

these for the Sydney Metropolitan Air Quality Study,(EPA NSW 1994) they were categorised by day typeas either winter (temperature less than 20oC), summer(temperature more than 20oC but below 35oC) orhigh oxidant day (temperature more than 35oC).

Figure 6.5 shows the resulting daily motor vehicleROC emissions for Perth, according to these daytypes. The low variation of exhaust emissions and thelarge variation of evaporative emissions are bothclearly evident.

6.1.4. Effects of Emission Controls

A major factor in estimating emissions from themotor vehicle fleet is the effect of the variousemission control measures introduced underAustralian Design Rules. ADR 37 in 1986 wasprimarily aimed at reduction of hydrocarbon and COemissions. Vehicle manufacturers met therequirements by fitting evaporative-trapping canistersto fuel systems and catalytic converters to exhaustsystems. A side benefit of the latter was the need forsuch vehicles to use unleaded petrol.

Figure 6.6(a) shows the effects of these changes onsummer day fleet emissions, modelled for the period1986 to 2011 with no consideration given to catalystdeterioration. The forward projections are based onthe assumption that current emission standards fornew vehicles will apply unchanged to 2011. Theyindicate expected trends as ageing non-ADR emissionregulated vehicles drop from the fleet.

The combined effect of ADR 37 and the introductionof unleaded fuel is evident in the dramatic fall in leademissions and a general decline in ROC emissions toabout half their 1986 levels by 2011, despiteincreasing VKT associated with population growth.Nitrogen oxides are not so well controlled by currentvehicle emission standards and their general increaseover the period is predominantly due to increasingVKT.

However, estimates of fleet emissions, and inparticular forward projections, are strongly influencedby the variable performance of exhaust catalysts.From the limited tracking that has been undertaken ofemissions performance of Australian vehicles inservice (EPA NSW 1994) it is evident that catalystperformance falls away as the kilometre age of thevehicle increases. Furthermore this effect is quitevariable in vehicles of the same make and model.

This process is modelled by estimating catalystdeterioration factors. For the Perth inventory thedeterioration factors used were provided by the EPANSW, having been derived from data obtained fromin-service vehicle tests conducted by that authority

0

20

40

60

80

Time (hours WST)

0 6 12 18 24

Weekday

Saturday

Sunday

Figure 6.4. Hourly distribution of total daily motorvehicle traffic counts, Perth, 1992-1993. The verticalaxis is total traffic count, at 44 measurement sites, in

thousands.

0

20

40

60

80

100

120

Winter SummerHigh

Oxidant

EvaporativeExhaust

Figure 6.5 Exhaust and evaporative ROC emissions(tonnes) from motor vehicles on different day

categories, Perth weekday, 1994.

61

and the EPAV. Performance decline for each species(nitrogen oxides, ROC and carbon monoxide) wasexpressed as a function of kilometres travelled for allvehicle categories tested and average deteriorationfactors determined by linear regression. The resultingcorrections were then applied to the fleet emissionsaccording to the kilometre age mix of vehicles in thefleet, determined from the Register of MotorVehicles.

Figure 6.6(b) shows motor vehicle emission estimatesincluding the effect of catalyst deterioration, for aPerth summer day. Lead emissions again show adramatic decline from 1986, and ROC initially showsa downward trend. However, from 2001 the effect ofincreasing VKT overtakes ROC reductions and grossfleet ROC emissions begin to increase again. NOxshows similar trends to the undeteriorated catalystcase but overall NOx levels are significantly higherand by 2011 are nearly double the value for theundeteriorated catalyst case. Clearly, the issue ofcatalyst performance is critical to evaluation of motorvehicle fleet emissions and their effects.

In computer modelling, the estimates of catalystdeterioration discussed above were included in theemissions inventory, being the more realistic option.This decision was supported by results from samplingprogrammes to measure motor vehicle emissionsduring the intensive field studies carried out in thesummer of 1993-1994, as described below.

6.1.5. Verification of Motor VehicleEmission Estimates

The validity of the PPSS motor vehicle emissionsinventory was tested using several differentapproaches:

• comparison of annual values (e.g., VKT, tonnes ofemission species and average fleet emissionfactors) derived from the detailed PPSS inventorywith those from a simpler inventory developedfrom independent data;

• comparison of average fleet emission factors fromthe PPSS inventory with those derived fromanalysis of roadway samples; and

• comparison of mass fluxes of precursors predictedfrom the inventory with those measured by theaircraft during the intensive field programme in1994 (giving consideration to the extent of themotor vehicle contribution to general urbanemissions at the time of measurement).

Comparison of Annual Estimates

The simple, independent motor vehicle emissionsinventory was developed by the CSIRO Division ofAtmospheric Research (Galbally et al. 1995) usingvehicle emission factors from the NationalGreenhouse Gas Inventory Committee Workbook forTransport, and fuel consumption and VKT estimatesfrom the Australian Bureau of Statistics Survey ofMotor Vehicle Use, 1991. Comparison of thoseestimates with the PPSS inventory (1991) values areshown in Figure 6.7.

Agreement between the two approaches is within 6%for estimates of VKT, CO and CO2. For NOx andROC the PPSS estimates are about 14% higher andlower, respectively, than the CSIRO estimates.

A possible explanation for the poorer NOx and ROCagreement lies in the fact that the CSIRO estimatesare based on a single average emission factor for eachspecies. This was derived from dynamometer testing

0

20

40

60

80

100

120

140

1986 1991 1996 2001 2011

NO , tonnesROC, tonnesLead, 10s kgVKT, millions km

x(a) (b)

1986 1991 1996 2001 2011

Figure 6.6. Summer day motor vehicle emissions, Perth, 1986 to 2011, (a) ignoring catalyst deterioration and (b)taking exhaust catalyst deterioration into account.

62

of vehicles over a set drive pattern and applied to totalVKT.

The more detailed PPSS inventory categorised theVKT according to mode of travel and applied mode-specific emission factors. As traff ic conditions inPerth are relatively uncongested compared to thedynamometer drive pattern, a higher proportion ofVKT is spent in the freeway mode.

Figure 6.8 shows the PPSS inventory “mode oftravel” emission factors for the three main emissionspecies, compared to the single factor used in theCSIRO inventory.

NOx emissions are higher and total ROC emissionsare lower for freeway mode, leading to a relativedifference in the estimates consistent withobservations.

Comparison of Fleet Emission Factors

One of the objectives of the ROC speciation samplingprogram in the PPSS (see Section 4.5.3) was tomeasure real, on-road concentrations of motor vehicleemissions under particular traff ic conditions. Thisenabled determination of average fleet emissionfactors for comparison with those used to develop themotor vehicle emissions inventory.

The CSIRO Division of Atmospheric Researchsampled motor vehicle emissions using two platforms − a car within traff ic, for freeway/arterial roadconditions, and a static sampler mounted over thestreet, for congested CBD conditions (Galbally et al.1995). Analysis of these for CO, CO2, NOx and ROCprovided a “fingerprint” of motor vehicle fleetemission concentrations.

Emission factors for smog precursors were thenestimated, based on an emission factor for CO2. Thelatter was determined from the fleet VKT and fuelconsumption, obtained from the Survey of MotorVehicle Usage (Australian Bureau of Statistics 1991),and the known amount of CO2 produced per unit offuel consumed. The emission factor (g/km) for eachspecies was determined as the product of the CO2emission factor and the species-to-CO2 concentrationratio.

Fleet emission factors quantified in this way were inreasonable agreement with those used in the PPSSinventory, as shown Figure 6.9.

The field measurements therefore verified the vehicleemission factors used in the inventory. This impliesthat catalyst deterioration effects included in theinventory were valid.

Comparison of Gross Emission Fluxes

Estimates of the gross emissions flux in the Perthurban plume were made from airborne measurementsduring the summer of 1993-1994. The urban plume isa combination of motor vehicle and area sourceemissions. Discussion of the aircraft results will bedeferred until Section 6.3.3 (after area sources havebeen described). However, it may be noted here thatthe aircraft results provided good support forinventory estimates of total urban emissions. Morningpeak hour vehicle emissions were dominant at thetime of the aircraft flights.

Further corroboration of inventory estimates of motorvehicle emissions was provided by detailed analysisof the organic species detected in smog plume,industrial plume and urban air samples, collected in

0

5000

10000

15000

20000

25000

30000

VKTmillions

CO2tonnes

COtonnes/10

NOxtonnes

ROCtonnes

PPSS

CSIRO

/1000

Figure 6.7. Comparison of annual estimates of VKT,CO, CO2, NOx and ROC, Perth, 1991, from the PPSS

and CSIRO motor vehicle emissions inventories.

0

1

2

3

4

g /

km

ROCsummer

ROCwinter

NO CO/10x

FreewayArterialCongestedResidentialCSIRO

Figure 6.8. Comparison of the PPSS mode-specificfleet emission factors with the fleet average factor

used in the CSIRO inventory.

63

field programmes over the summers of 1992-1993and 1993-1994. Galbally et al. (1995) quantified thecontribution of motor vehicle emissions to pollutantsmeasured in the samples by comparing individualROC species concentrations with those of acetylene, akey low reactivity tracer for motor vehicle emissions.Species associated only with motor vehicle emissionsshow a strong linear concentration relationship withacetylene, while those from other sources do not. Thisprinciple was used to estimate the proportion of ROCspecies measured at given locations that hadoriginated from motor vehicles. It was calculated thatabout:

• 62% of the ROC in urban air;

• 53% of the ROC in photochemical smog plumes;and

• 4% of the ROC in Kwinana industrial air

was attributable to motor vehicle emissions.

The indicated 62% contribution from motor vehicleemissions to ROC observed in urban air comparesfavourably with the 58% contribution to gross urbanROC estimated from the emissions inventory (wheregross urban emissions are defined as total inventorysummer weekday ROC emissions minus Kwinanaindustrial ROC emissions).

6.2. INDUSTRIAL EMISSIONS

6.2.1. Distribution Across Industry

As outlined in Section 4.4.2, estimation of emissionsfrom industrial point sources in the Perth regioninvolved a survey of 330 industrial sites byquestionnaire. Persistence on the part of the inventoryteam achieved a very high survey response rate ofmore than 90%. Of the surveyed sites, 115 were

significant enough to warrant more detailedevaluation by follow-up interviews or site visits.

All industrial emission estimates were based on 1992data. No forward projections have been made forindustrial point source emissions as they are by naturemore constant than motor vehicle or area emissions.Changes are usually by discrete amounts associatedwith new projects or enhancements of existinginstallations, and are best incorporated into theemissions inventory as they occur.

The surveyed sources were grouped into industrycategories generally aligned with the AustralianStandard Industry Classification. The categories usedfor the inventory are listed in Table 6.1.

The relative contribution of the different industrial

0

0.5

1

1.5

2

2.5

3

3.5Inventory(Congested)Inventory(Residential)Measured, CBD

CO/10 ROC NOx

(a)

CO/10 ROC NOx

(b)Inventory(Freeway)Inventory(Arterial)Measured, Road

Figure 6.9. Comparison of PPSS inventory fleet emission factors (g/km) with those derived from measurementsin the CBD (a) and on freeway/arterial roads (b).

Table 6.1. PPSS Emissions Inventory, IndustrialSource Categories.

Code Description

10 Food and beverage, and animal by- products

20 Non-metallic mineral processing

30 Metals and metallic mineral processing

40 Chemical Products and services

50 Paper, paper products and printing

60 Utilities - electric power generation

70 Community services

80 Building and construction materials

90 Petroleum refining and fuel storage

100 Manufacturing, fabrication

110 Aviation

120 Paint & polymer manufacturing

64

activities to smog precursor emissions is shown inFigure 6.10.

In common with experience in other cities, it wasfound that a relatively small number of sourcesaccount for the bulk of industrial emissions. Forexample the 10 largest NOx emitters in the regionaccount for 92% of Perth’s total industrial NOxemission of 20,327 tonnes/annum (1992). Similarly,the 10 largest ROC emitters account for 91% of the11,706 tonnes of industrial ROC emissions. Thepattern was repeated for other emission species.

The industrial emissions inventory also highlights theconcentration of the Perth region’s heavy industry inthe Kwinana area. For the purposes of this study theKwinana area was defined by Australian Map GridCoordinates 6424.0 to 6443.0 km N and 375.0 to395.0 km E. Industrial point sources located in thisarea accounted for the following fractions of the Perthregion’s emissions:

• 88% of industrial NOx;

• 76% of industrial ROC;

• 56% of industrial CO; and

• 94% of industrial SO2.

The main observations from these data are:

• power generation and mineral processing are theprincipal industrial sources of NOx;

• petroleum refining and fuel storage, the food andbeverage industry and paper products andprinting are the principal industrial sources ofROC;

• the bulk of industrial NOx emissions (82%) areemitted from stacks taller than 45 metres. (Thispattern also applies to sulphur dioxide andparticulate emissions from industry − not shownin Figure 6.10); and

• the bulk of industrial ROC emissions (96%) areemitted at the surface.

The latter two points are potentially significant to thedynamics of smog formation as, under somemeteorological conditions, the NOx and ROC couldbe emitted into different air streams and thereforemight not mix effectively.

ROC Tonnes / year

0 100 200 300 400 500 600 700 800 900 1000

Petroleum refining and storage(/10)

Food and beverage

Paper products and printing

Aviation

Chemical products and services

Paint & polymer manufacture

Manufacturing, fabrication

Power generation

Non-metallic mineral processing

Building and construction materials

Metallic mineral processing

Community services

0 2000 4000 6000 8000 10000 12000

Power generation

Metallic mineral processing

Non-metallic mineral processing

Chemical products and services

Petroleum refining and storage

Building and construction materials

Aviation

Food and beverage

Community services

Paper products and printing

Manufacturing, fabrication

Paint and polymer manufacture

NOx Tonnes / year

ElevatedSurface

Note: divided by 10

ElevatedSurface

ROC

NOx

Figure 6.10. Contributions of different categories of industry to smog precursor emissions of NOx (lower) andROC (upper), Perth, 1992. Note that ROC emissions for petroleum refining/storage are divided by 10 for

presentation purposes. “Elevated” refers to emissions emanating from above 45 metres.

65

6.2.2. Seasonal Variation in IndustrialEmissions

Activity on industry sites was expected to be uniformthroughout the year, and responses to the inventoryquestionnaire confirmed this. The variation fromsummer to winter in hourly emission rates of mostspecies was very small , with the exception of ROC(Figure 6.11).

Industrial ROC emissions include a significantcomponent of fugitive emissions due to evaporativelosses from storage tanks and plant components suchas valves, flanges and seals. These losses increasewith increasing ambient temperature. The relativesummer and winter emission rates shown in Figure6.11 reflect this effect. Fugitive losses of ROC wereestimated for particular industries using theprocedures of API Bulletins 2517, 2518 and 2519(American Petroleum Institute 1989, 1991, 1993), orUS EPA publication AP-42 (US EPA 1985).

6.2.3. Aircraft Validation of IndustrialEmission Estimates

As the bulk of industrial emissions in Perth originatefrom the Kwinana area, airborne measurement ofpollutant flux in the Kwinana industrial plume was aneffective way to assess the accuracy of inventoryestimates of industrial emissions. This part of theemissions inventory verification was a component ofthe aircraft survey described previously in Section4.5.1.

Special purpose inventory flights by the FIAMSaircraft intersected the Kwinana plume close to thecoast in the early morning, generally under easterly(offshore) wind conditions (Carras et al. 1995). Thisexercise was conducted on four different days (Table

4.2, page 34). Instruments on the aircraft measuredpollutant levels when the plume was intersected andallowed the activation of equipment to collect airsamples for subsequent detailed analysis. Figure 6.12shows the flight path and typical instrument responsereturned from such an exercise.

The main features evident are the broad urban plumereflected in both the hydrocarbon and NOx traces,followed by relatively sharp peaks of NOx andhydrocarbons caused by individual industrial plumesin the Kwinana area.

From the concentration measurements, estimates ofthe flux of NOx and ROC were made using the windspeed and atmospheric mixing depth applying at thetime. The meteorological parameters were determinedeither from observations made by the aircraft’sinstrumentation, or from independent radiosondeflights at times close to the time of sampling. Theideal conditions for measurement of emission fluxesin this manner are low wind speeds and shallowmixing heights (conditions most likely to occur in theearly morning). Potential sources of uncertainty influx determination by this method include lowpollutant concentrations, slow instrument responsetimes, estimation of effective mixing height and

0

500

1000

1500

2000

2500

3000Summer dayWinter day

NO ROC CO SOx 2

Figure 6.11. Seasonal variation of industrial emissionsestimates for daytime hours (kg/hr). Only ROCemissions change significantly from winter to

summer.

NOx

ROC

Aircraftflightpath

QuinnsRocks

Fremantle

Kwinana

Figure 6.12. Aircraft flight path and correspondinginstrument response (plotted in arbitrary units)

traversing the urban and Kwinana plumes, on themorning of 28 January 1994. The broad central peakwas due to urban (mainly vehicle) emissions, while

the sharper southern peaks came from Kwinanaindustry.

66

variation of the instantaneous emission rates fromindividual emission sources.

Using this approach, Willi ams et al. (1995) providedestimates of the flux of hydrocarbons, NOx and CO2from Kwinana. In Table 6.2 these are compared toflux estimates from the emissions inventory, derivedfrom indicated emission rates and mixing depths atthe time of sampling.

Taking into account the above-mentioned areas ofuncertainty in the estimates, there is reasonableagreement between the measured and inventory-derived fluxes of CO2 and NOx. There was, however,a significant discrepancy in the ROC flux estimates,with the airborne measurements suggesting a ROCflux from the Kwinana area of about three to fourtimes that indicated by the inventory. This is asignificant finding because, if correct, Kwinana ROCemissions would be quantitatively the most significantin Perth, exceeding those of the motor vehicle fleetand therefore potentially playing a key role in Perth’ssmog events.

6.2.4. Evaluation of the Reactivity ofEmissions from Kwinana

Since aircraft measurements indicated that actualemissions of ROC appeared to be significantly greaterthan inventory estimates, it was decided to conductfurther investigations.

The Swanbourne Airtrak monitor was moved to theJandakot site (see Figure 4.1) on 21 February 1994 tomeasure the parameter Rsmog, which provides ameasure of the photochemical reactivity of all ROC,not just hydrocarbons. As described in Section 5.4,

Rsmog = a [ ]ROC(i )

ROC(i )i 1

n

=∑ (6.1)

where aROC(i) and [ROC(i)] are the photochemicalreactivity coeff icient and the concentration of theindividual ROC species respectively.

Examination of the emissions inventory data baserevealed one dominant source of ROC emissions inthe Kwinana industrial area, namely the fugitiveemissions from the oil refinery. This source accountedfor about 95% of the total (see Figure 6.10). It shouldbe noted that this estimate was derived fromengineering calculations rather than actualmeasurements.

To investigate the refinery ROC emissions estimates,it was decided to use the calculated refinery fugitiveemissions alone to predict, via a computer model, thevalues of Rsmog that would be expected at Jandakot,downwind of the industrial area in sea breezeconditions. The predicted values could be comparedwith Airtrak measurements at Jandakot. The methodis summarised below.

The practical diff iculty which arose in calculatingRsmog was that there was not a complete set of aROCvalues for all of the expected species of ROC in theemissions. It was possible, however, to derive themfrom aROC values calculated for Carbon Bond Four(CB-IV) surrogate compounds by Cope and Johnson(1994).

Given these ROC estimates, the steps involved inmodelli ng Rsmog at Jandakot were:

• convert emission rates for each ROC species tomolar units, of “moles carbon per second” ;

• for each of the species, multiply the emission rateby the derived aROC for the species, then sumacross all species to give the “total reactivity-weighted molar emission” with units “ reactivemoles carbon per second” ;

• use this emission value in a Gaussian plumedispersion model to calculate, at Jandakot, theconcentration in units of “ reactive moles carbonper cubic metre of air” , corrected to STP; and

• next use the fact that one mole of air occupies avolume of 0.0224 m3 at STP, to determine thenumber of “ reactive moles carbon per mole ofair” . Finally, multiply by 109 to obtain units ofppb (parts-per-billi on by volume).

The result is the calculated value of [Rsmog] in unitsof ppb, as defined in equation (6.1).

The Gaussian plume dispersion model DISPMOD(Rayner 1987) was used for this exercise. The modelhas been used extensively for modelli ng elevatedplumes in the Kwinana area. While recognising thatsuch models are not well suited to predicting pollutantconcentrations at particular points in time and space,they are nevertheless quite capable of predicting the

Table 6.2. Comparison of measured and inventory-estimated emission fluxes from the Kwinana area,

1994 (from Willi ams et al. 1995).

Species Measured(kg/s)

Inventory(kg/s)

CO2 57-120 140-180

NOx 0.2-0.65 0.38

ROC (as C) 1.2-2.5 0.45

67

statistics of high values and longer term averages.Consequently, if the model shows a consistent biasabove or below ambient measurements, that is astrong indication of errors in the emissions estimatesor the ambient measurements.

Comparison of Rsmog measurements and modelpredictions were made for selected events within theperiods February to April 1994 and December 1994to February 1995. An event was defined on the basisof favourable wind direction and availabilit y ofreliable data from the Airtrak instrument. Examples ofthese selected events are shown in Figure 6.13. Boththe modelled and measured concentrations are halfhour averages.

The modelled Rsmog values were generally less thanthe Airtrak Rsmog values. For example, on 16December 1994 the Airtrak measured a peak Rsmogvalue of 1.5 ppb, while the peak modelled Rsmogvalue for the same event was 0.85 ppb. The offset inthe time of these peak values is within reasonablebounds, given the uncertainties in meteorologicalvariables used for modelli ng and the nature ofatmospheric turbulence. For each identifiable event inthe data, the ratio of the Airtrak measured peakRsmog to the modelled peak for the same event wascalculated, yielding the following statistics:

Average ratio 1.75

Standard deviation 0.65

Median ratio 1.61

Number of events 56

This result supports the indication of the airbornestudy that ROC emissions from the Kwinana area areunderestimated in the PPSS emissions inventory.However, the degree of underestimation is notconfirmed and uncertainty about Kwinana ROCemissions remains a variable to be considered inairshed modelli ng interpretations.

6.3. AREA-BASED EMISSIONS

6.3.1. Contributions to Area-basedEmissions

The various classifications of sources considered tobe the most significant contributors to area-basedemissions were identified in Section 4.4.3. Therelative contributions of these to Perth’s estimated1992 (inventory base year) annual area source totalsof 22,600 tonnes ROC, 2,300 tonnes NOx and 47,000tonnes CO, are shown in Figure 6.14. This indicatesthat, on an annual basis, the most significant area-based contributors of smog precursors are domestic

16/03/94

0.0

0.5

1.0

1.5

2.0

13 14 15 16 17 18 19

08/01/95

0.0

0.5

1.0

1.5

2.0

9 12 15 18 21

16/12/94

0.0

0.5

1.0

1.5

2.0

12 13 14 15 16 17 18

Rsmog

(ppb)

Time (hours WST)

Figure 6.13. Reactivity of the Kwinana IndustrialArea plume on 16 March 1994, 16 December 1994and 8 January 1995: broad shaded lines − Rsmogfrom Airtrak measurements, solid lines − Rsmog

modelled from emissions inventory.

0 10 20 30 40 50 60 70

Surface

Service station,refueling

Aerosol

Nat gas leakage

Cutback bitumen

Lawnmowing

Dry cleaning

Natural gascombustion

Solid and liquidfuel combustion

Domestic wastecombustion

Railways

Marine craft

Off-roadvehicles

products

coatings

CO

ROCNOx

Solvents, domesticand commercial

Figure 6.14. Percentage contribution of varioussource classifications to annual area-based emissionsof NOx, ROC and CO, Perth, 1992 (from Stuart and

Carnovale 1994).

68

solid and liquid fuel combustion, off-road vehicles,marine craft, service stations and vehicle refuelli ng,and use of surface coatings. On a given summertimesmog event day, however, the relative importance ofthese sources changes significantly as their emissionsare not uniformly distributed over the year. In theiranalysis of Perth’s area emissions, Stuart andCarnovale (1994) found that, in summer, surfacecoatings/thinners, service station and fuelli ng losses,domestic/commercial solvent use, lawn mowing, off-road vehicles and marine craft are the main sources ofarea emissions.

6.3.2. Spatial and Temporal Variations

Area source emissions were assumed to be spatiallydistributed according to the pattern of populationdensity in the PPSS region, based on the premise thatactivities giving rise to them are population-based.Gross emissions were assigned to the inventorynetwork of grid cells in proportion to the grid cellpopulations. The latter were determined from theAustralian Bureau of Statistics 1991 Census data, andare shown graphically in Figure 6.15. The extensionof shading across the coast is an artefact of themethod used to smooth the data for graphicalpresentation.

The temporal distribution of area-based emissions isstrongly subject to seasonal, day-of-week and time-

of-day influences. The seasonal and day-of-weekinfluences are evident in Figure 6.16 which showsestimates of emissions from area-based sources onsummer and winter weekdays and weekend days.

The most significant observations are that:

• the area source contribution of ROC and COgreatly exceeds that of NOx;

• area emissions are always higher on weekend daysthan weekdays (because of the influence ofdomestic activities on area emissions); and

• CO and particulate emissions are much higher inthe winter (Note the different scales). This ispredominantly due to domestic solid and liquidfuel burning for heating so that, although thisactivity is a significant emitter of smog precursorson an annual basis, the emissions occur in thewinter when smog activity is suppressed by lowtemperatures and reduced sunlight. Therefore thissource has no significant impact on Perth’sphotochemical smog.

200

600

1000

1400

1800

2200

Figure 6.15. Estimated population density in each3km grid square for the Perth region in 1992, per

square kilometre.

0

100

200

300

400Winter weekend dayWinter weekday

0

20

40

60

80

100

120

ROC COx 10

Part.SOx 10

2NOx

Summer weekend daySummer weekday

Figure 6.16. Emissions from area-based sources,Perth, 1992, showing day-of-week and seasonalvariations in summer (lower) and winter (upper)

emissions, in tonnes per day. Vertical scalesdiffer between the two graphs (From Stuart and

Carnovale 1994).

69

To determine the temporal distribution of theemissions, the gross estimates were divided accordingto the days of the year on which the activities weremost likely to occur, taking into account the seasonaland weekday/weekend influences. For instanceemissions from home heating are almost entirelyconfined to the winter and while there is reducedactivity at petrol stations on weekends, lawn mowerusage is increased.

Resultant daily estimates were further dividedaccording to the likely diurnal profile to provide anhourly emission for each grid cell . For example,natural gas leakage was taken to occur around theclock, seven days per week whereas domestic andcommercial solvent use or aerosol product use wasassigned a more “9 to 5” profile.

6.3.3. Verification of Area-basedEmissions

The aircraft operations described in Sections 4.5.1and 6.2.3 also provided measurements of the flux ofemissions from the Perth urban region. The urbanplume, centred on the CBD, is almost entirely due toemissions from area-based sources and the motorvehicle fleet, since the region’s industrial emissionsare concentrated in the Kwinana area further to thesouth.

Willi ams et al. (1995), compared the measured fluxeswith those derived from the emissions inventory(vehicle plus area sources) for the time of day of thesampling, generally between 6.30am and 8.30am.Their results, reproduced in Table 6.3, indicate goodagreement between the measured and inventory-derived flux estimates.

Under the conditions at the time of sampling,uncertainties in the flux estimates were estimated tobe ± 0.05 kg/s for NOx and ± 0.1 kgC/s for ROC.

This agreement suggests that the combination of themotor vehicle and area source emission estimates inthe inventory are reasonable. Given that the motor

vehicle emission estimates have been corroborated byother approaches (Section 6.1.5), the inventoryestimates for area-based sources also appear to bereasonable.

6.3.4. Trends for Area-basedEmissions

In estimating emissions from area sources in Perth,Stuart and Carnovale also made forward projectionsto the year 2002. As area emissions are most stronglya function of population, forward estimates werebased on population growth projections. For Perth,these were derived from Local Government Areagrowth projections developed by the WA Ministry forPlanning, using data from the 1991 Census. Resultantestimates of annual emissions of NOx, ROC and COfrom area-based sources in 1992 and 2002 are shownin Figure 6.17.

0

10

20

30

40

50

60

NO ROC CO

1992

2002

x

Figure 6.17. Estimates of area-based emissions ofNOx, ROC and CO in Perth for the years 1992 and

2002 (kilotonne/year).

6.4. BIOGENIC EMISSIONS

Biogenic emissions are generated predominantly inthe daytime and are strongly temperature dependent.Estimates of daily emissions of biogenic ROC in thePPSS region for 16 March 1994, when the maximumtemperature was about 37oC, are shown in Figure6.18.

Total biogenic ROC emissions for the region shownwere calculated to be 60 tonnes per day. In massterms this was of the same order of magnitude asother ROC source group emissions in the region.However, as can be seen from Figure 6.18, theemissions were broadly spread.

The occurrence of maximum emission rates in theinland part of the PPSS region is clearly evident,particularly to the south. The dominant source in thisarea is the Jarrah forest.

Table 6.3. Comparison of measured and inventory-derived emission fluxes in the Perth urban plume,

1994 (from Willi ams et al. 1995).

Species Measured(kg/s)

Inventory(kg/s)

CO2 90 − 210 85 − 190

NOx 0.7 − 1.6 0.7 − 1.6

ROC(as C) 0.4 − 2.5 1.2 − 2.4

70

There remains some doubt as to the quantification ofbiogenic ROC emissions by the techniques used, but amore comprehensive investigation of this source wasoutside the resources and timing limits of the currentwork. The relative importance of these emissions tophotochemical smog development in Perth isassessable using photochemical smog models, andthis is discussed further in Chapter 7.

6.5. SUMMARY

Sources of photochemical smog precursors in thePerth region have been evaluated and resultsassembled into databases (motor vehicle, industrial,area and biogenic) suitable for use in computermodelli ng of regional air pollution. These databasesare all similarly constructed, based on a standard gridof 3 km-square cells. A total mass emission of eachspecies is assigned to each grid cell for each hour ofthe day, derived by summing all species emissionsfrom individual sources located in that cell . Speciesare “lumped” into chemical groups recognised by thephotochemical models.

Generally, the contributions to Perth’s emissions fromthe sources evaluated show similarities to those foundelsewhere (e.g., Carnovale et al. 1991), with motorvehicles being the dominant source of all the mainphotochemical smog precursors. However, thepresence in the region of a major industrial complexat Kwinana is reflected in the relatively higher

contribution from industrial point sources to overallemissions of NOx in Perth.

The spatial distribution of emissions in Perth is suchthat there are two general “plumes” identifiable − awidely spread one resulting from motor vehicle, areaand biogenic sources, and a more intense one arisingfrom the concentration of industry in the Kwinanaarea. While the latter may be more locally intense, theformer represents a larger mass of emissions and hasthe greater potential for photochemical smogproduction.

Several approaches were used to verify the emissionestimates in the PPSS inventory. Generally, theseindicated that estimates of NOx from motor vehicles,area sources and industrial point sources are realistic.Similarly, inventory estimates of ROC from motorvehicle and area sources appear to be valid. There isevidence, however, that ROC emissions from theindustrial sources (essentially in the Kwinana area)have been underestimated by a factor of two to four.Biogenic emissions of ROC also remain an area ofuncertainty.

A summary of 1992 annual emissions in the Perthregion for the three categories of human activity isgiven in Table 6.4.

Table 6.4. Annual emissions (tonnes) due to humanactivity, Perth, 1992.

NOx

(as NO2)ROC CO

MotorVehicles

23,100 27,000 213,900

Industry 20,300 11,700 5,100

Area-based

2,300 22,600 47,000

TotalEstimate

45,700 61,300 266,000

Scale

(kg/km 2

/day)

0.1

0.2

0.5

1

2

5

10

20

Figure 6.18. Distribution of daily total biogenic ROCemissions (in kg/km2/day), calculated for 16 March

1994.

71

In its current state of development, modelli ng ofphotochemical smog events is very different frommodelli ng the dispersion of primary air pollutantssuch as sulphur dioxide around an industrial area. Inthe latter case, modelli ng may extend across a fullyear or more, requiring relatively routinemeteorological data. Photochemical smog modelli ng,by virtue of its complexity and the computerresources required to undertake it, is currently limitedto modelli ng one or at most a few days.

Consequently the best current approach to assessalternative smog control strategies is to firstly model alimited number of smog events which span the rangeof likely event types. One or more of these events arethen re-modelled, with emissions varied to reflectalternative control options (or anticipated trends),comparing the results to the initial (“base case”)model results.

The context of model development and application inthis study is again stressed:

• model development was a secondary objective,the primary objective of this study beingmeasurement and understanding of Perth’sphotochemical smog for the first time;

• application of models in the current study waslimited to a demonstration of their capabiliti esand potential uses (noting that use of models toassess the impact of Pinjar Power Station, asdescribed in Chapter 1, was a parallel butindependent investigation being undertakenwithout DEP participation); and

• development of regional smog control strategiesand (possibly) regulatory controls is not withinthe mandate of Western Power, hence applicationof models for this purpose was not undertaken aspart of the study.

Modelli ng of photochemical smog has two linkedcomponents:

• modelli ng the three dimensional meteorology ofthe region, including important features like seabreezes; and

• modelli ng the transport, dispersion and chemicalreaction of the smog-forming emissions.

To maximise the likely success of modeldevelopment, two initiatives were simultaneouslypursued:

• a consulting assignment undertaken by theCSIRO Division of Atmospheric Research andthe EPAV, scientists from which are at theforefront (at least within Australia) ofmeteorological modelli ng and smog chemistrymodelli ng respectively; and

• development of a meteorological model andapplication of a US EPA smog model by theDEP.

These efforts and their outcomes are described in thischapter.

7.1. METEOROLOGICALMODELLING

To produce reliable estimates of the place and time ofpeak photochemical smog concentrations, the three-dimensional meteorology of the studied region mustbe accurately represented. This includes horizontaland vertical wind velocity, intensity of turbulentmixing (by which smog precursors and reactionproducts are dispersed) and air temperature, humidityand radiation (which affect the rates of manychemical reactions).

It would be ideal to have a monitoring network whichprovided continuous meteorological data at manypoints on the ground and, at a few key locations,continuous vertical profiles of meteorologicalparameters. The full field of winds and temperaturescould then be generated by interpolation betweenmeasurement points, and careful extrapolation to theboundaries of the region under investigation.However, to provide suff icient accuracy across thewhole region, a large number of measurement pointswould be necessary, and the associated costs wouldbe very high.

During intensive field investigations, it is possible tomaintain a high density of measurements usinglabour-intensive methods. Such was the case on 4February 1994, when a smog event occurred withinthe main PPSS intensive field study. But to maintainsuch a measurement program continuously is beyondthe reach of most study budgets. As a result, theapproach normally taken is to model the meteorologyof the day being investigated, and compare themodelled fields of wind, mixing depth andtemperature with available measurements. When thereis close agreement between modelled fields andmeasurements, the modelled meteorology may beused as input to a photochemical smog model.

7. Modelling

72

The use of a meteorological model, in this context,provides a form of “ intelli gent” interpolation oflimited measurements, adding scientifically credibledetail which would otherwise be unavailable.

Any meteorological model is an attempt to simulatethe behaviour of the atmosphere. Normally therepresentation is performed using a computer,although physical models in laboratories also quali fyfor the title.

The atmosphere is treated as a fluid system, obeyingfundamental laws of physics. At points within arectangular grid (see Finite Difference Models,below), values of wind velocity, air temperature, aparameter representing air pressure, and possiblyhumidity and cloud, are stored.

Point-to-point variations of modelled pressure causeaccelerations of the flow, changing modelledvelocities. Variations of velocity develop, so that airconverges at some points, and diverges at others.Where convergences occur, both air density andtemperature increase. These changes feed back intothe modelled pressure, driving further changes towind velocity.

Into this interacting system are added externalinfluences, which vary according to the problembeing studied. These can include topography, whichenters the system through distortion of ground-levelpressure surfaces, and solar heating, which changesair pressure in the heated region.

A typical meteorological model would employ a gridusing 50 to 100 grid points in both horizontaldirections, and 10 to 20 in the vertical. The number of

grid point values whose evolution in time must becalculated can therefore be a milli on or more. Thismeans that such models require sophisticatedcomputers to be used successfully.

Three meteorological models were used during PPSS.One was of local origin, while the others wereemployed by consultants at the CSIRO Division ofAtmospheric Research and Murdoch University. Allwere finite-difference models, representing themeteorology of the region on a three-dimensional gridof points.

The local model was developed originally during theKwinana Air Modelli ng Study (1982), and laterenhanced during the 1986-1987 America’s Cupdefence (Rye 1989). The form used during PPSS wastitled 3DSB (“Three-Dimensional Sea Breeze”). It didnot include topographic effects, but possessed theadvantages of fast execution time, abilit y to accept anon-uniform initial state, and time-dependent externalforcing (Rye 1996c).

The CSIRO model, named LADM (for “LagrangianAtmospheric Dispersion Model” ) was of similarstructure to 3DSB, but included the effects oftopography on the wind field (Physick et al. 1994). Itlacked the capacity for non-uniform initialisation,which affected its abilit y to represent the coastaltrough, the significance of which was explained inChapter 3 and Section 5.3. However, its abilit y toaccept time-varying horizontally homogeneoussynoptic winds was used to largely overcome thislimitation, at least for days when the trough movedthrough the region. LADM also has the facilit y offour-dimensional data assimilation whereby predicted

Finite Difference Models

Fluid dynamic systems, such as the atmosphere, can not be handled exactly by even the most powerful ofcomputers, because processes on scales ranging from global to microscopic are all working at once. Models canrepresent the flow explicitly only at a finite number of points, usually on a rectili near grid.

Typically, a study of an urban region would use a horizontal grid size of a few kilometres. In the verticaldirection, processes near the surface are most variable, so resolution is greatest there − usually about 100metres. At the top of the region, the vertical grid spacing may be close to a kilometre.

The changes of modelled variables at each grid point are modelled using rules based on physics or chemistry.Many of these changes can be related to the finite (non zero) differences between the values of variables atadjacent grid points. These differences are the source of the name of this class of model. For each modelledvariable, the rate of change is multiplied by a chosen time increment, and added to its initial value, to find a setof new values at one increment of model time later.

In a meteorological model, changes to wind reflect the transport of momentum from adjacent grid points, andforces due to the local pressure gradient, the earth’s rotation and friction. Temperature changes result fromexternal heat sources (essentially, the sun), and pressure changes are caused by convergence of air towards oraway from a point. A photochemical model represents only the changes to chemical species in the atmosphere,these being due to transport by wind between grid points, turbulent dispersion, reactions between themselvesand deposition to the surface.

73

winds may be nudged toward individual observations.This was employed for most of the selected modelleddays and had the effect of generally reducing thespeed of the low-level early morning winds and alsoto sharpen the passage of the sea breeze front.

The final model evaluated was named RAMS (for“Regional Atmospheric Modelli ng System”). Themodel was of United States origin (Pielke et al. 1992),and included topographic effects, and initialisationwith a non-uniform windfield. On the debit side werea very slow execution speed, and diff iculties inobtaining initial meteorological fields compatiblewith the needs of the model. When initialised withuniform fields, there was no advantage over the othermodels, and as a result this model was not used inultimate calculations.

7.1.1. Modelling Using 3DSBStudying the meteorology for any day involved twostages. The model was firstly run at low resolution,with a modelled region about 1500 km wide. Theoutput of this run was then used as input for a second,higher-resolution model run.

The low resolution grid normally used a grid cell ofsize 12 km west to east, and 40 km south to north.Different grid intervals in the west-east and south-north directions were employed, because the coastaltrough changed only slowly in the south-to-northdirection, but more sharply from west to east.

The initial state for the first run was derived using thevertical wind and temperature profile for the Perthregion, based on a mix of information from theBureau of Meteorology’s morning sonde, and fromthe study’s own profile measurements, described inChapter 4. Commonly, the Bureau of Meteorologydata were ignored at levels below about onekilometre, because of interference due to flow overthe adjacent scarp. However, at these levels the sodarwinds and radar-measured temperatures were usuallyavailable, and of higher resolution.

When the model was initialised in a horizontally-uniform state, it was found that a coastal trough oftendid not form. When it did, it remained well offshore.

The cause was found to be the lack of realism of thesurface pressures to the west. A typical ozone day inPerth commences with north easterly winds, thepresence of which indicates that pressures shoulddecrease westward. However, the actual pressurepattern offshore usually shows a high pressure cell i nthe centre of the Indian Ocean.

To handle this problem, the initial fields weremodified, giving an approximation to the pressurefield on the morning of the studied day. This fieldtypically consisted of a high pressure ridge to the

south, and a trough, aligned south-to-north, near thecoast (e.g., Figure 7.1(a)).

Commonly, pressures were allowed to decreasewestward again at the western edge of the region (asin Figure 7.1(a)). It was found that, without this,winds to the west of the trough often became toostrong.

During the first 24 hours of model time, the normalstructure of the coastal trough developed (Figure7.1(b)). However, since the modelled winds andtemperatures were still approaching realism throughthis period, the model was run for two days of modeltime. Calculations for the second day were the onesused for smog modelli ng.

The second model run normally used a grid cell sizeof 3 km by 5 km, or 6 km by 10 km. At the higher ofthese resolutions, there were only small changes incalculated smog transport, so the 6 km by 10 kmresolution was more often used. This model run was

(a)

(b)

Figure 7.1. Example of coastal trough modelli ngusing 3DSB. The initial surface pressure field isshown by (a), and the field on the morning of the

second day by (b).

74

normally initialised using data for midnight, 24 hoursinto the low-resolution run. Through the higher-resolution run, values at the grid boundaries werecontinuously drawn towards those calculated duringthe low resolution run’s second 24 hours.

Through the second run, half-hourly horizontal windvelocities, mixing depths and temperatures, at allmodel levels and covering the region to berepresented by the smog model, were written to a diskfile. An example of a surface windfield generated bythe model is depicted in Figure 7.2. The disk file waslater used in preparation of the smog model’s inputfiles.

This approach was used to model several days whenthe trough passed inland. These simulations revealeda number of issues with the potential to reduce thevalue of modelled winds, mixing depths andtemperatures.

Firstly, the higher-resolution model runs were able toresolve pressure changes close to the coast in moredetail than runs at low resolution. This meant thatpressures near the coast fell more quickly than thosecalculated in the low-resolution run. As a result, for aday when the low-resolution run showed the troughmoving inland, the high-resolution run would hold thetrough at the coast.

To counter this effect, pressures inland could bereduced by imposition of an additional pressuregradient. However, this also produced a tendency tosoutherly winds at higher model levels. An unwantedresult of this southerly tendency was the transport ofcooler air from the south, which reduced the stabilit yof the modelled inflow, and increased mixing depths.

Corrections for this effect could also be added to the

model’s data fields, maintaining the correct mixingdepths inland. However, the many correction stepstherefore required made the generation of wind andmixing depth fields for use by photochemical modelsa slow process.

The value of modelled meteorological estimates is,however, ill ustrated by Figure 7.3. The close verticalspacing of the potential temperature contours, bothoffshore and above the mixing depth line inland,shows the capacity of the model to represent thestrong stabilit y which occurs in the onshore flow ondays of smog events. Detailed estimates of bothtemperatures and mixing depths, such as thosedisplayed, are required by photochemical models, inorder to perform their calculations accurately.

6440

6460

6480

6500

6520

334 346 358 370 382 394 406 418 430

Figure 7.2. Windfield for the Perth region calculatedby 3DSB, for 12 noon on 21 March 1994. It shows a

sea breeze which has just arrived at the coast. Data foronly part of the calculation grid are shown here.

328 334 340 346 352 358 364 370 376 382 388 394 400 406 412 418 424 430 0

500

1000

1500

Height(m)

AMG Easting (km)

Figure 7.3. Potential temperature and mixing depth (shaded line) cross-section near Perth, modelledusing 3DSB, for 3pm on 21 March 1994. The location of the coastline is shown by the transition

from shaded to black, at the bottom of the figure.

75

7.1.2. Modelling Using LADMLADM is actually a pair of computer models. Theyare normally operated in sequence, to representpollutant transport in a region of complex winds.These models were developed at the CSIRO Divisionof Atmospheric Research.

The first of the pair, employed in this study, is athree-dimensional meteorological model, which isable to simulate winds, turbulence, mixing depths andtemperatures in regions of complex geography,including coastal and topographic effects. It wastherefore a logical choice as a contributor to thestudy, allowing the effects of both sea breezes and theDarling Scarp to be analysed.

LADM also appeared well suited to modelli ng in thePerth region, where there were clear interactionsbetween large-scale features such as the coastaltrough, and smaller-scale effects associated with thesea breeze. Through the use of nested data grids(Figure 7.4), it was possible to model the behaviour ofthe sea breeze, within the context of the whole south-west region.

A sample surface wind field calculated by LADM isshown in Figure 7.5. It shows morning winds for 19February 1994, a day when high ozone concentrationswere measured at the coast, in association with thepassage inland of the coastal trough. The figure showsincreased wind speeds offshore, due to lower surfacefriction over water, and near the Darling Scarp due to

a katabatic (down slope) flow. West of the acceleratedflow near the scarp is a region of decreased windspeed, indicative of a hydraulic jump.

As previously noted, LADM employed the combinedfeatures of time-varying horizontally-homogeneoussynoptic winds and four-dimensional dataassimilation to simulate the meteorological fields forselected smog event days. For those days on whichthe trough moved eastward through the region, themodel results accorded quite well with themeteorological measurements and appeared toadequately represent the important aspects of themeteorological fields for each modelled day. Even ifthe results did not closely match observations, theywere nevertheless representative of the meteorologyof such events and therefore able to be used to model“generic” smog events over Perth.

According to Noonan (1995), the model had morediff iculty simulating days when a trough remained inthe region than days when a trough moved throughthe region. This was a consequence of the model’scapabilit y to employ only horizontally homogeneous(single direction) synoptic winds. The errorintroduced by this model feature decreases as thespeed of trough passage increases. Given that some ofPerth’s smog events are associated with near-stationary troughs, the model limitation is significant.

Figure 7.4. Calculation grids used by LADM. Thefull region was occupied by the 10 km grid, the pale

shaded region by the 5 km grid, and the darkerregion by the 2.5 km grid. The heavy shaded line isthe south west coast, lighter lines are contours of

heights in metres.

Figure 7.5. Sample wind field calculated usingLADM, for 6am on 19 February 1994. The broad

shaded line is the representation of the coastline usedby the model. Topographic contours are shown as

dotted lines.

76

7.2. PHOTOCHEMICAL MODELLING

In the same sense that a meteorological modelrepresents physical interactions which generatemotions within the atmosphere, a photochemicalmodel represents chemical interactions within analready-moving atmosphere. It uses a similar finite-difference grid, and the transport and dispersion ofmodelled variables by the wind is similarlyrepresented. Emissions of smog precursors arespecified within the model as time-varying inputsfrom each grid cell .

However, while in meteorology we must deal with theeffects of only a few physical laws on a handful ofvariables, atmospheric chemistry involves hundredsof chemical reactions operating on hundreds ofchemical species.

The problem is made even less tractable by poorunderstanding of many of the chemical reactions, andinabilit y to measure economically the concentrationsof many of the species involved.

As a result, many simpli fications are made. Chemicalreactions with littl e effect over the period of a day, orinvolving species known to be present in very lowconcentrations, are omitted from the mechanism.Species undergoing similar reactions and with similarrates are grouped.

The resultant species are then grouped into classes of“slow” and “ fast” reaction rates. Those in the firstclass are treated as time dependent, while those in thesecond are presumed to be in approximately steadystate and so may be evaluated analytically.

However, there is no agreement on the best set ofsimpli fications to use. As a result, there are severalcompeting models of the photochemical smogprocess.

As for the representation of the region’s meteorology,two photochemistry models were used in PPSS. Theuse of two models was considered highly desirablebecause, while both used generally comparableschemes to represent transport, turbulent dispersionand deposition, they differed significantly in the waythe set of photochemical reactions was simpli fied. Byusing two different models, there was a better chanceof identifying errors due to model limitations.

The Carnegie-Mellon/Cali fornia Institute ofTechnology (abbreviated “CIT”) model was appliedby the EPAV, utili sing meteorological fieldsgenerated by the LADM meteorological model. Asexpected, this model has proven the most skil ful insimulating Perth’s smog, reflecting the efforts of theEPAV to improve the model’s meteorologicalcomponents and computational schemes. Apart from

simply obtaining model results, the study team lookedto the specialist staff of the EPAV to present acomprehensive assessment of model performance andcapabiliti es (summarised in Section 7.2.1).

The Urban Airshed Model, as issued by the US EPA,was evaluated by the DEP, utili sing meteorologicalfields generated by the 3DSB meteorological model,and by interpolation of measurements. This model,normally abbreviated as “UAM”, was of earlierorigin. Details of the model’s development have beengiven by Scheffe and Morris (1993).

7.2.1. The CIT Model

The CIT model, incorporating the Lurman CarterCoyner (LCC) photochemical mechanism, isdescribed by Harley et al. (1993). Application of theCIT model by the EPAV is described in detail byCope and Ischtwan (1995).

Before starting modelli ng activities, DEP, EPAV andCSIRO scientists selected five smog events formodelli ng. 4 February 1994 (the day during theintensive experiment on which the mostcomprehensive data set was obtained; see Chapter 4),was selected as the “detailed event day” . 19 February1994, 16 March 1994 and 18 March 1994 wereselected as “generic days” , indicating that they werecharacteristic of well -defined smog events, theimportant features of which the model should be ableto simulate. 12 March 1994 was a reserve day, tostudy if time permitted.

For the detailed event day and each generic day, Copeand Ischtwan conducted a series of sensitivity tests onthe model to ensure that it was stably configured. Incomparing observations to model predictions, thefollowing features and parameters were required toshow reasonable agreement, or fall within acceptablebounds:

• key descriptive (qualitative) features of the event;

• wind patterns;

• mixing heights;

• temporal and spatial distributions of ozone, SmogProduced and Extent (see Section 5.4); and

• statistics of the accuracy with which daily peakozone concentrations were estimated.

For the detailed event day, precursor emissions andthe concentrations of NOx, CO and NMHC were alsotested against measurements for reasonableagreement, and the ozone predictions were subjectedto more intense statistical assessment (residualanalysis).

77

The abilit y of the model to simulate time-varyingconcentrations of ozone at the PPSS monitoringstations on the detailed event day can be seen inFigure 7.6. For this day, the comprehensivemeteorological data set enhanced the predictions fromLADM. Although peak ozone levels tended to beunder-predicted for this and other days, the abilit y ofthe model to simulate the essential features of thesmog formation and recirculation cycle was pleasing.The statistical tests applied to the predictions yieldedgood scores (within acceptable bounds of accuracyfor such models). It is notable that for this day theemissions from Kwinana were predicted not to bemixed with urban emissions to any significant extent.

Modelli ng of winds in the presence of a troughproved to be a problem to a greater or lesser extent forall days. LADM, as constructed, assumes horizontallyhomogeneous synoptic wind across the modeldomain, an assumption clearly contravened when atrough is close to the coast. Errors in wind speed anddirection translate to errors in the location anddispersion of smog plumes; such errors are evident inthe CIT results. This simply highlights the complexity

of meteorological/pollutant interactions and points tothe need for further scientific development.

Cope and Ischtwan (1995) carried out emissionssource sensitivity tests for each of the generic days,by varying one emissions group at a time and re-running the model. An example of results of thesetests, for 18 March 1994, is presented in graphicalform in Figure 7.7 (page 78).

In summary, the graph shows the modelled maximumozone concentration in the region, irrespective ofwhere it occurred, firstly for the “base case” with allemissions as prescribed in the inventory, then for thecase of all motor vehicle emissions excluded, then formotor vehicle ROC emissions increased by 50%, andso on (see the explanatory notes below Figure 7.7).The graph and associated results provided by Copeand Ischtwan reveal important features and raiseimportant questions about Perth’s smog, as follows.

Firstly, the graph indicates that, while settingdomestic area sources (DAS) to zero caused a minorreduction in peak ozone, setting motor vehicleemissions or biogenic ROC to zero yielded a

340 350 360 370 380 390 400 410 420 430 440

EASTING (KM)

6420

6430

6440

6450

6460

6470

6480

6490

6500

6510

6520

NO

RTH

ING

(KM

)

TR

QR

PI

QB

CU

CA

RG

SW

JA

KE

WTHV

RI

RO

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

CULLACABARDEE

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

KENWICK

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

QUINNS ROCKS

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

ROLLING GREEN

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

ROTTNEST ISLAND

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

ROCKINGHAM

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

SWANBOURNE

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

OBS.

EST-INT

EST-BST

MIN

MAX

TWO ROCKS

TIME (HOURS)

CO

NC

EN

TR

AT

ION

(PPM

)

0

0.02

0.04

0.06

0.08

0 6 12 18 24

CAVERSHAM

Figure 7.6. Comparison of observed ozone concentrations with those estimated by the CIT model for 4 February1994. “OBS” stands for observations, “EST-INT” for model estimates interpolated to the location of the AQMS

site, and “EST-BST” for the model estimates in best agreement within a small rectangular area (generallyextending two grid cells east and west of the site, and four grid cells north and south). The “MIN MAX” vertical

bars indicate the modelled range within these rectangular areas.

78

significant reduction. Reducing industry emissions tozero had the opposite effect − a marked increase inpeak ozone.

Given the reasonably high level of confidence in thevehicle emissions estimates, it can be confidentlyconcluded that motor vehicles are a major contributorto peak ozone levels.

The level of confidence attached to the biogenicemissions, derived by the method of Lamb, Gay andWestberg (1993), is very low (see Section 6.4). Copeand Ischtwan (1995) highlight biogenic emissions asan aspect requiring further quantification.

The smog-inhibiting effect of industry (primarilyKwinana industry) seems paradoxical, but it isexplained by the fact that industrial emissions are richin NOx (i.e., a low ROC/NOx ratio).

According to the CIT model’s chemistry, emission ofNOx into low ROC/NOx ratio air suppresses ozoneproduction in a non-linear fashion. Inhibition ofozone production occurs in the following ways:

• a large fraction of ozone is required to oxidise theabundant NO, as per equation 5.1 (page 53);

• the formation of organic radicals required forsmog production will be suppressed. This occursdue to the reaction of abundant NO2 withhydroxyl radicals, OH., to form nitric acid, aprocess which does not contribute to the ozoneformation cycle; and

• a positive feedback process from the decrease inozone concentration. The initial decrease resultsfrom its reaction with NO, and the reduction oforganic radicals. Some atmospheric ozone isnormally photolysed to produce OH., which

participates in further ozone production. With theloss of ozone, OH. production, and subsequentlyO3 production, is decreased.

It should be noted that the non-linear effect describedabove was not observed in the smog chamberexperiments on which the IER and GRS models,described in Chapter 5, were formulated. Whether thesmog chamber processes or the computer modelchemical mechanisms fail to represent the realatmosphere at low ROC/NOx ratios is an issue ofcurrent scientific investigation (e.g., Tonnesen andJeffr ies 1994; Blanchard, Roth and Jeffr ies 1993).

A motor vehicle ROC increase or decrease by 50% isseen to have a significant corresponding effect onpeak ozone. Normally, a change to motor vehicleNOx had an inverse effect, although on this day(unlike the other generic days) a decrease in motorvehicle NOx did not yield a peak ozone increase. Thedifference occurred because the peak smog, asmodelled, was NOx-limited.

The effect of setting industrial ROC emissions(virtually all from the Kwinana region) to zero wasnegligible for this and the other days. This, coupledwith the predicted large impact of Kwinana NOx,points clearly to the need for further work to properlyunderstand the role of Kwinana emissions.

If , as has been indicated by emissions validationstudies (e.g., Section 6.2), the ROC emissions fromKwinana have been underestimated by a factor of twoto four, then the ROC/NOx ratio for Kwinanaemissions will have been underestimated by the samefactor, and the non-linear response of the model couldyield very different predictions for the contribution ofKwinana industries.

BC MV x 0

MVROCx 1.5

MVROCx 0.5

MVNOX x 1.5

MVNOXx 0.5

IND x 0

INDROCx 0

INDNOX x 0

DASx 0

BIO x 0

100

0

20

40

60

80

120

140

Figure 7.7. Estimated maximum ozone concentrations (ppb) in the Perth airshed for various emission scenariosstudied for 18 March 1994. BC : base case inventory used in CIT model simulation; MV x 0 : motor vehicle

emissions scaled by zero (i.e., omitted); IND x 0 : industrial emissions scaled by zero; DAS x 0 : domestic areasource emissions scaled by zero; BIO x 0 : biogenic emissions scaled by zero; similar notation for ROC or NOx

emissions scaled by 0.5 and 1.5. (After Cope and Ischtwan 1995).

79

There was strong supporting evidence for this inmodel estimates for the 4 February 1994 fieldexperiment. Figure 7.8 shows a portion of the flightpath of the FIAMS research aircraft, withmeasurements of ozone concentrations along the path.Separate ozone plumes originating from Kwinana andthe urban area, and being blown inland, are clearlyapparent. The measured ozone peak concentrationswere very similar, whereas the model, while correctlyshowing separate plumes, calculated a significantlylower concentration for the Kwinana plume. The clearimplication is that the inventory ROC emissions forKwinana, as used by the model, are too low.

Cope and Ischtwan also highlighted concerns aboutthe model’s over-estimation of ground level NOxconcentrations close to Kwinana, possibly due toinadequate handling of elevated plumes. Modelimprovements are required to better accommodateindustrial areas like Kwinana.

7.2.2. The Urban Airshed Model

The chemical reaction scheme normally used byUAM groups reactions according to the nature of thecarbon chemical bonds involved. In organiccompounds, carbon atoms may be bound together bysingle, double or triple bonds, or in more complexcombinations involving rings or branches.

In outline, the Carbon Bond model in UAM treatsmany of the reacting species as collections of carbon

bonds, and subdivides their masses into variousfractions, according to the numbers of each type ofcarbon bond contained.

Although the number of reactions is reduced, it stillremains large. The version of the Carbon Bond modelused for the work reported here employed 85reactions between 23 reacting species groups.

UAM was extensively evaluated, using data for oneday during the summer of 1992-1993, and severaldays during the intensive study period in the summerof 1993-1994. Case studies for 8 January 1993, 4 and19 February 1994, and 16, 18 and 21 March 1994were used to investigate the applicabilit y of the modelto the Perth region (Rye 1996d).

For these days, the field measurement programprovided suff icient data to allow creation ofmeteorological inputs directly from measurements.The interpolation scheme employed for this purposeis described by Rye (1996d). Investigations of the useof modelled meteorology showed that, while usefulresults could be obtained, measured data gavesuperior results.

This result was unusual in Australian experience, andreflects both the relative simplicity of themeteorological regime favourable to smog productionin Perth, and the quality of the meteorologicalmeasurements made during the study.

Initially, for all days considered, it was found thatUAM underestimated the rate of formation of smog

80 60 40 20 0

6440

6460

6480

6500

6520

AMGnorthing

(km)

O3

HC/10

NO +20x

Measured concentration (ppb)

Aircraft

path

80 60 40 20 0Modelled concentration (ppb)

O3

HC/10

20

30

40

Model

grid

units

(a) (b) (c)

U U

KK

flight

U

K

NO +20x

Figure 7.8. Ozone, NOx and NMHC concentrations within a north-to-south section of the Perth region’s smogplume, shortly after 1pm on 4 February 1994. Values modelled using CIT are shown in (a), measurements at

equivalent locations in (b), and the flight path in (c). NOx values are increased by 20 ppb in both (a) and (b), toseparate the graphs. Ozone peaks due to the urban and Kwinana plumes, and their approximate paths on the

map, are identified by letters U and K.

80

from urban emissions, but overestimated smogproduction in the south eastern region, and in theperiod before the arrival of the sea breeze.

Errors in the south-eastern region and before seabreeze arrival were both found to be due tooverestimation of the biogenic contribution to smogproduction.

As a result, in consultation with EPAV (Cope andIschtwan 1995), initial emission estimates based onthe work of Guenther et al. (1994) were replaced bymore conservative estimates developed by Lamb, Gayand Westberg (1993). However, even these normallyhad to be halved to generate fair agreement betweenmodelled and measured smog production.

Cope and Ischtwan (1995) found that the CIT modelalso overestimated smog production in similarcircumstances, so errors in the biogenic emissionsestimates were considered likely. While this error wasa significant issue, underestimation of the effect ofurban emissions was of far greater importance.

Measurements indicated that the Kwinana industrialinventory underestimated ROC emissions by a factorof two to four, but a corresponding modelled increaseto Kwinana emissions had only a minor effect onpeak ozone levels. To reproduce measured ozoneconcentrations with any degree of accuracy, it wasalso necessary to increase urban emissions of ROC bya factor close to two (Figure 7.9).

Possible causes for this discrepancy were:

1. an underestimation of ROC emissions in thevehicle or area-source inventory;

2. emissions comprising more reactive species,compared to the normal mix presumed ininventory development; and

3. shortcomings in the chemistry model employedin UAM.

Emissions validation studies did not support the firsttwo possibiliti es although the influence of airtemperature was not well characterised.

The final possibilit y was the most likely, the schemeused in UAM having known limitations in conditionsof low ratio of ROC to NOx emissions.

If , as seems probable, the source of the discrepancy isa shortcoming of the chemistry model, the modelwould overestimate the dependence of the rate ofozone production on ROC concentration. This wouldseverely limit its value in determining techniques formanaging smog concentrations in the Perth region.

The application of UAM in trend forecasting isdiscussed in Section 7.3. For these applications,which are presented for demonstration purposes only,vehicle ROC emissions were doubled, and Kwinana

industrial ROC emissions were three times theinventory estimates.

Comparisons of the UAM and CIT models indicatedthat the latter agreed better with measurements.Figure 7.10 includes contours of base case (emissionsas per the inventory) peak ozone estimates calculatedusing the CIT model, for 18 March 1994. There areseveral issues which arise from an assessment relativeto UAM estimates, as follows.

Firstly, peak concentrations modelled by CIT usingthe inventory emissions estimates were much closer tomeasurements. The contour plot shows peak hourly-averaged concentrations just more than 80 ppb (0.08ppm), inland from Caversham, which were close tothe measured peak of 90 ppb (0.09 ppm) atCaversham. This result was more notable in view ofthe use of the basic Kwinana emissions estimates, inspite of experimental evidence for these being wellshort of actual values (although UAM modelli ngsuggested only a minor role for Kwinana emissionson this day).

Also evident in the contours is a larger modelled peakof more than 100 ppb (0.10 ppm) in the north of themetropolitan area, and a “notch” of low peak ozoneconcentrations extending north to north eastwardfrom Kwinana, caused by NOx-rich Kwinanaemissions. These features were also indicated inmodelli ng of the same day using UAM. Both weretherefore confirmed not to be artefacts of eithermodel.

Overall , comparison of the UAM and CIT modelresults showed that the latter’s chemical mechanismswere better able to handle the mix of emissions whichcharacterised the Perth region.

0

20

40

60

80

100

0 3 6 9 12 15 18 21 24Time (hours WST)

Ozone(ppb)

Figure 7.9. Measured (wide shaded line) ozoneconcentrations for Caversham compared with UAMresults for the doubled urban ROC/tripled Kwinana

ROC inventory (solid line), for 18 March 1994.

81

7.3. METHODS FOR ASSESSINGTRENDS AND CONTROLS

The UAM and CIT models were used to investigatethe effects of possible changes to emission patterns onozone concentrations.

7.3.1. Sensitivity to Possible Trends

The CIT model was applied intensively to the casestudy of 18 March 1994. Figure 7.10 shows a seriesof so-called EKMA (Empirical Kinetic ModellingApproach) diagrams, which are graphs of themaximum 1-hour ozone concentration for identifiedlocations in the modelling region.

These were generated from multiple model runs basedon incremental decreases of ROC and NOx emissionsfrom surface anthropogenic sources, starting from thebase case emission values. The vertical and horizontalaxes show the fraction of the base case NOx and ROCemission, hence the contour value at the top righthand corner of each graph is the base case ozonemaximum for that location. The ozone maximumresulting from various combinations of ROC andNOx reductions can be read from the EKMA contourplot for each site. EKMA plots are used to determinehow peak concentrations of photochemical smogwithin different regions of the airshed would respondto the uniform implementation of control strategies

340 350 360 370 380 390 400 410 420 430 440

EASTING (KM)

6420

6430

6440

6450

6460

6470

6480

6490

6500

6510

6520

NO

RT

HIN

G (K

M)

TR

QR

PI

QB

CU

CA

RG

SW

JA

KE

WTHV

RI

RO

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CT

ION

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CT

ION

0.00 0.20 0.40 0.60 0.80 1.00

ROC SCALING

0.00

0.20

0.40

0.60

0.80

1.00

NO

X S

CA

LIN

G

CA

CUPI

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X S

CA

LIN

G

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CT

ION

SW

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CTI

ON

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CT

ION

0.00 0.20 0.40 0.60 0.80 1.00

ROC FRACTION

0.00

0.20

0.40

0.60

0.80

1.00

NO

X F

RA

CT

ION

TR

KE

Figure 7.10. CIT model estimates, for 18 March 1994, of the response of peak ozone concentrations (ppm) to spatiallyuniform reductions of surface-based anthropogenic emissions of ROC and NOx. Also shown is the estimated spatial

distribution of base case peak ozone concentrations (ppm). Codes used are PI for Pinjar, CU for Cullacabardee, TR forTwo Rocks, CA for Caversham, SW for Swanbourne, and KE for Kenwick.

82

for ROC and/or NOx (e.g., a reduction of motorvehicle ROC emissions by 50%).

For the event as modelled, it is apparent that theresponse of peak ozone concentration variesconsiderably from site to site. However, with theexception of Kenwick, regions of the airshed with thehighest concentrations are more strongly controlledby ROC emissions than NOx emissions.

Consider first the site located to the west of Pinjar.This site has the region’s highest estimated ozoneconcentrations for the modelled event. Peak ozoneconcentrations of more than 90 ppb (0.09 ppm) andpeak SP (Smog Produced) concentrations of morethan 110 ppb were estimated for this site. Moreover,the photochemical smog was diagnosed to be NOx-limited. Despite this diagnosis, the ozone EKMA plotfor this site indicates that ROC controls are still li kelyto be more effective than NOx controls. For example,a 40% reduction in ROC emissions is estimated tolead to a 20 ppb reduction in peak ozoneconcentrations. Conversely, a 40% reduction in NOxis estimated to lead to only a 8 ppb reduction in peakozone concentrations. Moreover, peak ozoneconcentrations are estimated to first increase slightlywith NOx reduction, before this decrease is achieved.

The response at Caversham, which is within theregion of impact of the industrial emissions, is morestrongly controlled by ROC emissions. For example itmay be seen that a 20% reduction of the surface-based anthropogenic emissions of ROC is estimatedto reduce the peak ozone concentration at this site by7 ppb (reduced from 68 to 61 ppb). Conversely, a20% reduction in NOx emissions is estimated to resultin an 11 ppb increase in peak ozone concentrations atthis site. In fact, peak ozone concentrations areestimated to monotonically increase with decreasingNOx until reductions of more than 60% haveoccurred. Ozone concentrations showed littl esensitivity to further NOx reductions. Clearly, theozone EKMA plot for Caversham demonstrates thatconditions for photochemical smog formation at thissite are strongly light-limited, reflecting the influenceof the NOx-rich Kwinana plume.

The EKMA response for Kenwick in the southexhibits the attributes of a NOx-limited plume.Although initially the ozone response favours ROCcontrols (15% reduction in ROC to reduce the peakozone concentration from about 70 to 65 ppb vs a40% reduction in NOx for the same level ofreduction), conditions become NOx-limited forconcentrations of ozone below 65 ppb (78% reductionin ROC to achieve 60 ppb O3 vs a 60% reduction inNOx). Similar results are obtained for SP. Byreference to the previous section on source sensitivity,

it can be concluded that the EKMA response atKenwick is caused by the contribution of biogenicROC emissions from the forested regions on theescarpment to smog generation. As noted above, theROC emission perturbations were applied to surfaceanthropogenic emissions but not biogenic emissions.Thus the reduction in peak concentrations ofphotochemical smog within the Kenwick region arestrongly buffered by smog production from biogenicROC sources and anthropogenic NOx sources. It isimportant to note again the considerable uncertaintyassociated with the biogenic ROC emissions.

In conclusion, it may be deduced that ROC emissionreductions within the range 20-40% would berequired to reduce peak ozone concentrations withinthe majority of the study area below a threshold valueof 80 ppb.

The EKMA results presented on page 81 are for justone particular smog event. Results for a different typeof smog event (e.g., a coastal event) would appearvery different. It will be necessary to compile a seriesof such regional EKMA diagrams, together with otherforms of information, before reliable assessments ofcontrol options or future trends can be undertaken.

7.3.2. Effects of Expected Trends

Due to uncertainty about the validity of the responseof UAM to changes in ROC emissions, it was onlyapplied in a limited number of cases, investigating theeffects of expected changes to emissions patterns onfuture ozone levels.

As explained in Section 7.2.2, vehicle ROC emissionestimates were doubled in order to match theobserved development of smog, and Kwinanaindustrial ROC emission estimates were tripled, asindicated by aircraft measurements.

Since emissions trends are predicted to involve anincrease in NOx emissions, and a decrease in ROCemissions, the future trend is into the range whereUAM is suspect. However, comparative runs at leastprovide a lower limit to any downward change ofozone concentrations. Figure 7.11(a) shows acomparison of peak ozone concentrations calculatedin two model runs, based on the meteorology of 18March 1994. One used emissions estimates for 1991,the other estimates for 2011, based on predictedtrends of population and VKT.

In the region where significant ozone concentrationsdeveloped, 2011 ozone concentrations were modelledas significantly lower than 1991 values. This resultwas generally consistent with the predictions by theCIT model.

83

The greatest reduction was in the area (shown shaded)receiving the largest effect of the city’s morningemissions. Here, increased NOx emissions reactedwith ozone, reducing its concentration. In the outermetropolitan area, the reduction became less, asozone in the 2011 simulation was regenerated bysmog reactions.

A similar result was found for the meteorology of 21March 1994 (Figure 7.11(b)). The zone of reducedozone was larger, but decreases were less, both due tothe more rapid advance inland of the sea breeze.

Both the 18 and 21 March 1994 followed the general

pattern of “ inland” smog events (see Section 5.3.1). Arepeat of the modelli ng exercise for the “coastal”smog event on 19 February 1994 showed a verydifferent pattern.

Figure 7.12 summarises the results of thiscomparison. Rather than decreases, there weresignificant increases of ozone concentration in thearea affected by the smog plume.

The figure shows both ratios and differences. Theformer might be more indicative of trends, because ofthe general underestimation of concentrations by themodel for this particular day.

(a) (b)

Figure 7.11. Difference between peak hourly-average ozone concentrations in 2011 and those in 1991,calculated using UAM and the meteorology of (a) 18 March 1994 and (b) 21 March 1994. Shaded

areas mark decreases over 50 ppb.

(a) (b)

Figure 7.12. Fractional (a) and absolute (b) differences in ppb, between peak hourly-average ozoneconcentrations in 2011 and those in 1991, calculated using UAM and the meteorology of 19

February 1994. Peak ozone concentrations coincided closely with the hatched region, where the ratiowas over 1.2, in (a). The highest factors of increase at the coast were close to 1.3.

84

The reason for the difference in results, from those inFigure 7.11, is the long reaction time experienced bythe smog mass. The large region where the ratio ofconcentrations remained between 1.2 and 1.3,between there and the coast, is consistent with smogreactions reaching early completion in both 1991 and2011 scenarios. In Figure 7.12, a reduction of ozoneconcentrations in a small region well offshore wascaused by higher levels of nitrogen oxides at sunrise,before smog reactions had progressed.

The completion of the smog reactions before thesmog-bearing air returns onshore is characteristic ofcoastal smog events. This ensures that the dominantfactor controlli ng ozone concentrations is the supplyof nitrogen oxides. These are forecast to increasesignificantly from 1991 to 2011.

Also obvious in Figure 7.12 is a reduction of ozoneconcentrations inland, consistent with the resultsshown in Figure 7.11. However, at the coast south ofthe metropolitan area, near Mandurah, concentrationsare largely unchanged. Rye (1996d) highlighted thisregion as an area of potential ozone impact, on anyday when a north-westerly sea breeze develops.

The result suggests that coastal smog events maybecome the most significant aspect of Perth’s smogclimate in the next century, with 20-yearconcentration increases in the vicinity of 20 to 30%likely. This prediction should be considered to betentative in light of the previously mentioneduncertainties.

UAM was also used to study the role of Kwinanaemissions in the generation of photochemical smog.All i nvestigations conducted during the study showedonly a secondary contribution to peaks at inlandlocations, although the role of Kwinana emissions inthe generation of coastal events remains uncertain.

Some analysis of the effect of bushfire smoke, andlikely future trends, have also been conducted. Sincethe reactivity contribution from the smoke isessentially unknown, work concentrated on findingempirical estimates of this reactivity, and projectionof future trends based on these estimates. Theindicated trends were similar to those found for inlandsmog events − namely, a decrease centred in theeastern metropolitan area, with increases furtherinland. The actual magnitude of the trend wassensitive to the presumed smoke reactivity, but thetendency was not.

7.4. SUMMARY OF MODELLING

Computer models for simulating photochemicalsmog, including 3-D meteorology and atmosphericchemistry, are highly complex and still l argely withinthe province of scientific research and development.Nevertheless, they can be applied by skill ed scientists

to yield credible simulations of smog events overcities like Perth and can therefore also be used astools to investigate projected trends and emissionscontrol scenarios.

In total, six 1-day smog events were modelled byDEP, CSIRO and EPAV staff in the current study, notwithout considerable diff iculty and associated effort.The results for these days provide a basis for theinvestigation of trends and control options tocommence, but it is clear that further smog eventsneed to be intensively measured and modelled toprovide a sound basis for smog managementstrategies.

Specific issues requiring further research are:

• simulation of windfields and mixing depths in thecoastal trough/sea breeze conditions whichcharacterise Perth’s highest smog concentrations;

• investigation of the magnitude of ROC emissionsfrom the Kwinana industrial area;

• verification (if possible) of the model response toNOx and ROC emission variations in the contextof the low ROC/NOx ratios present in Perth’s air,notably in relation to motor vehicle and industrialemissions;

• assessment of evaporative ROC emissions frommotor vehicles on very hot days;

• much better quantification of the mass andreactivity of biogenic ROC emissions; and

• better handling of industrial chimney emissionswithin the model.

The model results presented in this chapter are a smallselection of the work carried out by the EPAV,CSIRO and DEP and should not be used as a basis forproposing controls. A full set of model results andanalyses is provided by Cope and Ischtwan (1995),Noonan (1995) and Rye (1996d).

85

8.1. SUMMARY OF FINDINGS

The Perth region experiences photochemical smogduring the warmer months of each year. Nevertheless,the air over Perth on most summer days is relativelyclean, due to the windy climate and isolation fromother cities.

On average, during the three-year period July 1992 toJune 1995, there were 10 days per year on which thepeak hourly ozone concentration exceeded 80 partsper billi on (ppb) somewhere over the Perth Region.This equates to the Canadian goal, and may becompared to the World Health Organisation range of76-100 ppb.

In 1995, the Australian National Health and MedicalResearch Council (NHMRC) set a 1-hour average of100 ppb and a 4-hour average of 80 ppb as goalswhich should not be exceeded. These goals are bothexceeded about two times per year somewhere in thePerth region.

The previous NHMRC goal of 120 ppb for 1-houraverages has been exceeded only twice at Cavershamsince monitoring began in 1989.

Ozone events occur across a large region surroundingthe city. Each of the study’s 11 monitoring sites,covering an area extending from Rottnest to Rolli ngGreen (near Toodyay) and Gingin to Rockingham,recorded exceedances of 80 ppb during the study.

Current health effects of photochemical smog inPerth, including reductions in lung function andincreased risk of asthma attack, are probably mild.

Vegetation impacts from ozone in the Perth region arelikely to be low. However, an 8-hour average of 50ppb, set by the Victorian EPA as the acceptable limitfor protection of vegetation, is exceeded several timesper year in the Perth region. It is possible thereforethat vegetation in the Perth region (includinghorticultural crops) experiences transient stress, anissue which warrants further study.

Based on a comparison of measurement statistics, it isapparent that Perth experiences photochemical smogto an extent similar to Brisbane, greater than Adelaideand somewhat less than Sydney and Melbourne.Given the population difference (people and vehicles)between Perth, Melbourne and Sydney, it may besurmised that Perth has a potential for smog problemssimilar to that of the larger cities.

There is a dominant, well -defined weather patternrelated to Perth’s smog events. The highest smogconcentrations have been found to occur on thosedays during spring to autumn when a weak lowpressure trough is situated very close to the coast andsubsequently crosses the coast in the afternoon.

Emissions from morning peak hour traff ic are blownoff-shore by north easterlies into the light wind regionof the trough, where smog reactions proceed rapidlyunder the typically high temperatures. A strongtemperature inversion, also typical of theseconditions, keeps the smog plume concentrated nearthe ocean surface.

The study has revealed two well -defined classes oftrough-related smog events, namely:

• inland events, which show a pattern of peakconcentrations in the eastern metropolitan area.These correspond to days when the trough movesinland, and recirculates the smog plume backacross the metropolitan area where it receives aboost from fresh afternoon emissions; and

• coastal events, in which high concentrations ofsmog form in a warm, stable air mass offshore,but disperse rapidly after the smog has beenreturned to the coast by the sea breeze.

It has also become apparent in the course of the studythat some smog plumes may cross the coast wellsouth of the monitoring network (e.g., nearMandurah).

The behaviour of the trough on any particular day isdiff icult to forecast with the accuracy needed forsmog event predictions. Nevertheless there is someprospect of generating reliable smog forecasts shouldthis become necessary in the future.

Significant smog concentrations have been observedin other conditions, particularly when bushfire smoke,which contains high concentrations of ROC, is blownacross the metropolitan area during daytime.

Ozone events (except those associated with bushfiresmoke) tend to be of short duration (an hour or two)following the arrival of the sea breeze. This contrastswith regions of the USA and elsewhere whichsometimes experience elevated ozone levels overmany hours or days, caused by the transport ofpollutants from distant cities and industrial areas.

8. Conclusions and Recommendations

86

The inventory of smog precursor emissions, coupledwith modelli ng results, confirmed that motor vehiclesare the dominant cause of Perth’s smog, being thelargest sources of NOx and ROC.

Modelli ng results point to control of motor vehiclekilometres travelled (VKT) and ROC emissions as themost beneficial options for control of peak ozoneconcentrations. According to the more reliable modelused, reduction of NOx from vehicles in the absenceof ROC reductions would generally (andparadoxically) lead to an increase in peak ozone.

The accuracy of the CIT and UAM chemistry modelshas not been properly evaluated (in Australia orelsewhere) for the low ratios of ROC to NOx found inPerth’s air. The models predict a non-linearquenching of ozone production due to the addition ofNOx at low ROC:NOx ratios. The IER model doesnot include such an effect and therefore predicts amore positive consequence of NOx controls.

The Kwinana industrial area was identified within theemissions inventory as a major source of NOx and alesser source of ROC. There are strong indicationsfrom independent measurements that the ROCemissions were underestimated. The resultant effectseen in modelli ng predictions was a significantquenching of ozone across those portions of themetropolitan area impacted by the Kwinana NOxplume, due to the non-linear quenching effect.

Estimation of the magnitude and reactivity ofbiogenic emissions of ROC (from natural vegetation)was necessarily coarse. Modelli ng confirmed that thissource of emissions may be a significant contributorto the magnitude and extent of high ozoneconcentrations.

In all , six one-day smog events were modelled.Configuring and testing of the models to achievevarying degrees of agreement with observations tookseveral person-months of intense effort by Australia’sleading scientists in the area. The model simulationsand associated data sets provide a useful initial basisfor formulating and testing smog managementstrategies.

8.2. RECOMMENDATIONS FORFURTHER WORK

It is clear that improvements and additions will needto be made to the “library” of smog event simulationsto provide a reliable and representative basis forpredicting trends and testing smog managementstrategies. Continued monitoring of regional smoglevels and meteorological parameters will be requiredto characterise individual smog events for modelli ngpurposes.

Continued regional monitoring is also essential tomaintain a record of smog levels over time, fromwhich trends may be determined and the effectivenessof smog management strategies assessed.

Further assessment of the chemistry models isrequired to verify (or otherwise) the non-linearquenching of ozone production due to the addition ofNOx at low ROC:NOx ratios. This requiresfundamental research by the authors of the models,together with further testing by model users such asthe DEP.

The need to monitor smog levels well to the south ofthe city is apparent. As a first step, monitoringequipment has been installed by the DEP at anexisting monitoring station at Rockingham.

Issues in relation to Kwinana emissions which requirefurther investigation are:

• whether, as indicated by independentmeasurements, the ROC emissions from Kwinanaare higher, by a factor of two to four, than thosecalculated in the emissions inventory;

• whether the model correctly simulated thebehaviour of the bulk of Kwinana NOx which ispresent in elevated narrow plumes from tallchimneys; and

• whether the non-linear quenching effect of NOxon peak urban ozone is real.

Further work is needed to improve the accuracy ofbiogenic emissions estimates.

8.3. CONCLUSIONS

Perth experiences photochemical smog levels whichexceed goals set by the NHMRC and other bodies.The potential exists for the problem to grow.

The number of times that the new NHMRC goals maybe exceeded is set to grow rapidly if emissions andthe consequent smog concentrations increase bymodest amounts. From the perspective of publichealth, control of photochemical smog should beconsidered now.

Control of photochemical smog is complex. It is notamenable to simple and uniform solutions. A controlmeasure designed to improve air quality in oneportion of the metropolitan region (e.g., reduction ofNOx emissions) might have a negative effectelsewhere. Given the potentially enormous costs (tothe community, industry and Government) of variouscontrol options, it is critically important to thoroughlyanalyse the cost effectiveness of alternatives beforedecisions are made. Adequate analyses will t ake timeand resources. Fortunately, Perth is not yetexperiencing acute smog problems. The planning,

87

transport and environment agencies therefore havetime to learn from the experience of other cities and,in consultation with affected parties, develop sound,cost-effective smog management strategies.

The modelling capability developed in this study hasbeen immediately applied by Western PowerCorporation to assess the impact of the Pinjar GasTurbine Power Station. This assessment is the subjectof a separate report to the EPA. The DEP has had noinvolvement in its preparation.

Western Power Corporation does not have a centralrole in developing smog management strategies forthe Perth region. This will be progressed by the DEPin consultation with relevant agencies and affectedparties. The Perth Photochemical Smog Study hasprovided a sound foundation on which to build.

88

ACVEN, 1994, “Advisory Committee on VehicleEmissions and Noise, Regulatory Impact Statement −Revision of Australian Design Rule 37/00, EmissionControl for Light Duty Vehicles” , documentAC94/24, June 1994, ACVEN, Canberra.

Adams, M.B., Nichols, D.S., Federer, G.A., Jensen,K.F. and Parott, H., 1991. “Screening Procedure toEvaluate Effects of Air Pollution on Eastern RegionWilderness Cited as Class 1 Air Quality Areas” ,USDA Forest Service, General Technical Report NE-151.

A’Hearn, T.P., Wilson, C.M. and Hardiman, N., 1994,“The Use of Tradeable Permits In EnvironmentalManagement, Some Potential Theoretical andPractical Limitations” , Proceedings of the Clean AirSociety of Australia and New Zealand 12thInternational Clean Air Conference, Perth, WA, 23-28October 1994, 441-462.

Ahmet, S., 1995, EPA of Victoria, personalcommunication.

Air Quality Week, 1995, “Bay Area Victory SendsWholesome Message to Clean Air Opponents” , Issueof May 8, 1995, published by George Spencer, ed.,Washington, D.C., 3.

American Petroleum Institute 1989, “EvaporationLoss from External Floating-Roof Tanks” ,Publication 2517, Third Edition.

American Petroleum Institute, 1991, “EvaporativeLoss from Fixed Roof Tanks” , Chapter 19.1, Bulletin2518.

American Petroleum Institute, 1993, “EvaporationLoss from Internal Floating-Roof Tanks” , Publication2519, Third Edition.

Australian Bureau of Statistics, 1991, “Survey ofMotor Vehicle Usage”.

Azzi, M., Johnson, G.J. and Cope, M., 1992, “AnIntroduction to the Generic Reaction SetPhotochemical Smog Mechanism”, Proc. 11th Intern.Clean Air Conf., Brisbane, 5-10 July, 451-462.

Barnes, J.D., Reili ng, K., Davison, A.W., and Renner,C.J., 1988. “ Interaction Between Ozone and WinterStress” , Env. Pollut. 53, 235-254.

BC MOE, 1995, “Clean Vehicles and Fuels forBritish Columbia”, Policy Paper, British ColumbiaDepartment of Environment, Lands and Parks.

Beard, J.S., 1979a, “Vegetation Survey of WesternAustralia. The Vegetation of the Perth Area”,Vegmap Publications, 6 Fraser Road, Applecross,Western Australia 6153.

Beard, J.S., 1979b, “Vegetation Survey of WesternAustralia. The Vegetation of the Pinjarra Area”,Vegmap Publications, 6 Fraser Road, Applecross,Western Australia 6153.

Beaton, S.P., Bishop, G.A., Zhang, Y., Ashbough,L.L., Lawson, D.R. and Stedman, D.H., 1995, “On-Road Vehicle Emissions, Regulations, Costs andBenefits” , Science 268, 991-993.

Blanchard, C.L., Roth, PM and Jeffr ies, H.E., 1993,“Spatial Mapping of Preferred Strategies forReducing Ambient Ozone ConcentrationsNationwide”. Proc. Air and Waste Man. Assoc.,Denver, Colorado, June 1993.

Bretschneider, B. and Kurfurst, J., 1987, “AirPollution Control Technology” , Fundamental Aspectsof Pollution Control and Environmental Science 8,Elsevier.

Burke, M. 1995. “Are Oxyfuels Good for Us?”, NewScientist, No. 1986, 15 July 1995, 24-27.

Cabrera, H., Dawson, S.V., and Stromberg, C., 1988,“A Cali fornia Air Standard to Protect VegetationFrom Ozone”, Env. Pollut. 53, 397-408.

Carnovale, F., Alviano, P., Carvalho, C., Deitch, G.,Jiang, S., Macaulay, D., and Summers, M., 1991, “AirEmissions Inventory Port Philli p Control Region:Planning for the Future”, Environment ProtectionAuthority of Victoria, December 1991.

Carras, J.N., Clark, N.J., Drummond, M.C., Lange,A.L. and Willi ams, D.J. 1994, “Ground BasedMeasurements in the Perth Region” , CSIRO Divisionof Atmospheric Research.

9. References

NOTE: Reports which resulted directly from work undertaken for the Perth Photochemical Smog Studyare highlighted in the list of references below.

89

Carras, J.N., Clark, N.J., Hacker, J.M., Nelson, P.F.and Willi ams, D.J., 1995, “Airborne Measurements ofUrban and Industrial Plumes in the Perth Region,Summer 1994” , Flinders Institute for Atmospheric andMarine Sciences, Project Report 95-11, July 1995.

Chang, T.Y. and Suzio, M.J. 1995, “Assessing ozone-precursor relationships based on a smog productionmodel and ambient data.” J. Air and Waste Manage.Assoc., 45, 20-28.

Cochran, L.S., Pielke, R.A and Kovacs, E., 1992,“Selected International Receptor-Based Air QualityStandards” , Jour. Air and Waste Manag. Assoc. 42,1567-1572.

Cope, M.C. and Johnson, G.M., 1994, personalcommunication.

Cope, M.C. and Ischtwan, J., 1995, “PerthPhotochemical Smog Study, Airshed Modelli ngComponent, Final Report” , EPA of Victoria.

Cox, R.M., Spavold-Tims, J. and Hughes, R.N., 1989,“Acid Fog and Ozone, Their Possible Role In BirchDeterioration Around The Bay of Fundy, Canada”,Water, Air and Soil Pollution 48, 263-276.

Cox, R.M. and Malcolm, J.W., 1994, “Ozone Effectson Reproductive Processes In Eastern CanadianTrees” , Pollen-Pistil Interactions and Pollen TubeGrowth” , A.G. Stephenson and T-H Kao, eds.,American Society of Plant Physiologists, 296-299.

Crapo, J., Mill er, F.J., Mossman, B., Pryor, W.A.,Kiley, J.P., 1992, “Relationship Between AcuteInflammatory Responses to Air Pollutants andChronic Lung Disease”, American Review ofRespiratory Diseases, 145, 1506-12.

Department of Environmental Protection, 1982,“Kwinana Air Modelli ng Study” , Report 10.

Derwent, R.G., McInnes, G., Stewart, H.N.M. andWilli ams, M.L., 1976, “The Occurrence andSignificance of Air Pollution by PhotochemicallyProduced Oxidant in the British Isles, 1972-1975” ,Report LR 227(AP), Warren Spring Laboratory,Gunnels Wood Road, Stevenage, Hertfordshire SG12BX, U.K..

Dickson, R., Wilkinson, J., Bruckman, L. and Tesche,T., 1991, “Conceptual Formulation of the SARMAPEmissions Modelli ng System”, 84th Ann. Meetingand Exhibition, Air and Waste Manag. Assoc.,Vancouver, British Columbia, 16-21 June.

Dockery, D.W., C.A. Pope, X. Xu, J.D. Spengler, J.H.Ware, M.E. Fay, B.G. Ferris and F.E. Speizer, 1993,“An Association between Air Pollution and Mortality inSix U.S. Cities” , N. Eng. J. Med. 329, 1753-1759.

Dohmen, G.P., 1988, “ Indirect Effects of AirPollutants, Changes In Plant-Parasite Interactions” ,Env. Pollut. 53, 197-207.

Drechsler-Parks, D.M., 1987, “Effect of NitrogenDioxide, Ozone, and Peroxyacetyl Nitrate onMetabolic and Pulmonary Function” , Cambridge MA:Health Effects Institute, Research Report No. 6.

Edwards, R., 1995, “Ozone Alert Follows CancerWarning” , New Scientist, 27 May, 4.

Elsom, D.M., 1992, “Atmospheric Pollution, AGlobal Problem”, Blackwell Publishers, 108 CowleyRoad, Oxford OX4 1JF, United Kingdom.

EPA NSW, 1994, Personal communication.

Evans, L.F., Weeks, I.A., Eccleston, A.J. andPackham, D.R., 1977, “Photochemical Ozone inSmoke from Prescribed Burning of Forests” , Envir.Sci. and Tech. 11, 896-900.

Finlayson-Pitts, B., and Pitts, J.N., Jr., 1986,“Atmospheric Chemistry, Fundamentals andExperimental Techniques. John Wiley and Sons Inc.New York.

Fuentes, J.D. and Dann, T.F., 1993, “Ground-LevelOzone In Canada During 1980-1991” , Report ARD-93-010, Atmospheric Environment Service,Downsview, Ontario.

Galbally. I.E., Mill er, A.J., Hoy, R.D., Ahmet, S.,Joynt, R.C. and Attwood, D., 1986, “Surface Ozone atRural Sites in the Latrobe Valley and Cape Grim,Australia”, Atmos. Env. 20, 2403-2422.

Galbally, I., Bentley, S.T., Elsworth, C.M., Weeks,I.A. and Yee, R.Y., 1995, “A Report on the VolatileOrganic Compound Sampling during the PerthPhotochemical Smog Study” , CSIRO Division ofAtmospheric Research.

Government of Victoria, 1981, “State EnvironmentProtection Policy (The Air Environment)” , VictoriaGovernment Gazette No. 63, 13 July 1981, 2293-2305.

Grieco, A., 1996, “Ambient Air Quality MonitoringData Summary for the Perth Photochemical SmogStudy” , Department of Environmental ProtectionTechnical Series 81.

Grieco, A., Mountford, P. and Kleinfelder, R., 1996,“The Perth Photohemical Smog Study MonitoringNetwork” , Department of Environmental ProtectionTechnical Series 82.

Grosjean, D., Whitmore, PM and Cass, G.R., 1988,“Ozone Fading of Natural Organic Colorants,Mechanisms and Products of The Reaction of OzoneWith Indigos” , Env. Sci. and Technol. 22, 292-297.

90

Guenther, A., Zimmerman, P. and Wildermuth, M.,1994, “Natural Volatile Organic Compound EmissionRate Estimates for U.S. Woodland Landscapes” ,Atmos. Environ. 28, 1197-1210.

Guest, C., Woodward, A. and McMichael, A., 1994,“Air Quality Goals for Ozone, Environmental,Economic and Social Impact Assessment” , Report toThe Environmental Health Standing Committee of theNational Health and Medical Research Council , May1994.

Harley, R.A., Russell , A.G., McRae, G.J., Cass, G.R.and Seinfeld, J.H., 1993, “Photochemical Modelli ngof the Southern Cali fornia Air Quality Study” ,Environ. Sci. Technol. 27, 378-388.

Heggestad, H.E. and Middleton, J.T., 1959, “Ozone inHigh Concentrations as Cause of Tobacco LeafInjury” , Science 129, 208-210.

Hingston, F.J., Dimmock, G.M. and Turton, A.G.,1980, “Nutrient Distribution in a Jarrah (EucalyptusMarginata Donn ex Sm) Ecosystem in South-WestWestern Australia”, Forest Ecol. Manag. 3, 183-207.

Holguin, A.H. and Buff ler, P.A., 1985, “The effectsof ozone on asthmatics in the Houston area”. In: S.D.Lee (Ed) Transactions of the APCA InternationalSpecialty Conference: Evaluation of the ScientificBasis of Ozone-Oxidant Standards. Pittsburgh. AirPollution Control Association 1985, 262-280.

Hurst, D.F., Griff ith, D.W.T. and Cook, G.D., 1994,“Trace Gas Emissions from Biomass Burning inTropical Australian Savannas” , Jour. Geophys. Res.99, 16441-16456.

IUAPPA, 1988, “Clean Air Around the World” ,International Union of Air Pollution PreventionAssociations, 136 North Street, Brighton BN1 1RG,United Kingdom.

James, B., 1993, “Analysis of the 1991 Perth MotorVehicle Fleet for the Perth Air Quality Study” ,Department of Transport, Western Australia.

James, B., 1995, “Motor Vehicle Emissions Inventoryfor the Perth Photochemical Smog Study” ,Department of Transport.

Japan Environment Agency, 1985, “The Quality ofthe Environment in Japan 1984” , EnvironmentAgency, Government of Japan.

Johnson, G.M., 1983, “An Empirical Model ofPhotochemical Smog Formation” , Proc. 6th WorldConference on Air Quality, IUAPPA, Paris, vol. 1,25-32.

Johnson, G.M., 1984, “A Simple Model forPredicting the Ozone Concentration of Ambient Air” ,Proc. 8th Intern. Clean Air Conf., Melbourne, 6-11May, 715-731.

Kiely, P., Yap, D., Misra, P.K., Fraser, D. and Radell ,R., 1995, “A Comparative Study of Toronto’s AirQuality and Selected World Cities” , 88th AWMAAnnual Meeting, San Antonio, Texas, June 18-23.

Kromroy, K.W., Tseng, P.S., Olson, M.F. and French,D.R., 1988, “A Biological System for Indexing AirQuality and Assessing Effects on Vegetation.,Minnesota Bioindicator Study, “Env. Pollut. 53, 439-441.

Lamb, B., Gay, D. and Westberg, H., 1993, “ABiogenic Hydrocarbon Emission Inventory for theU.S.A. Using a Simple Forest Canopy Model” ,Atmos. Environ. 27A, 1673-1690.

Lefohn, A.S., Krupa, S.V. and Winstanley, D., 1990,“Surface Ozone Exposures Measured at CleanLocations Around The World” , Env. Pollut. 63, 189-224.

Lefohn, A.S., Runeckles, V.C., 1987, “Establishing astandard to protect vegetation-ozone exposure/doseconsiderations” , Atmos. Env. 21, 561-568.

Lippmann, M, 1989, “Health Effects of Ozone. ACritical Review” , Jour. Air and Waste Manag. Assoc.39, 672-695.

Lippmann, M. 1993, “Health Effects of TroposphericOzone: Review of Recent Research Findings andTheir Implications to Ambient Air QualityStandards” , Journal of Exposure Analysis andEnvironmental Epidemiology 3, 1-27.

Monk, R., 1994, “Effects of Urban Ozone onAustralian Native Vegetation” , Proc. 12th Intern.Clean Air Conf., Perth, Western Australia, 23-28October, 141-149.

NHMRC, 1990, “Ambient and Indoor air qualityGuidelines” , Australian and New ZealandEnvironment Council and NHMRC, January, 1990.

NHMRC, 1995, Report of the 119th meeting of theNational Health and Medical Research Council ,Canberra, June 7-8, 1995.

Noonan, J.A., 1995, “Predicted Wind FieldSimulations for the Perth Photochemical Smog Study(PPSS)” , CSIRO Division of Atmospheric Research,ref. EPAV95.

O’Connor, J.A., Parbery, D.G. and Strauss, W., 1975,“The Effects of Phytotoxic Gases on NativeAustralian Plant Species, Part 2. Acute Injury Due toOzone”, Environ. Pollut. 82, 197-200.

OECD, 1988, “Emission Controls in ElectricityGeneration and Industry” , OECD/InternationalEnergy Agency, Paris.

OECD, 1990, “Control Strategies for PhotochemicalOxidants Across Europe”, Organisation for Economic

91

Co-Operation and Development, 2 rue André-Pascal,75775 PARIS CEDEX 16, France.

Oge, M., 1995, “Automotive Emissions, Progress andChallenges” , Presented by Margo Oge, Director, EPAOff ice of Mobile Sources, Traverse City, Michigan,August 9.

Olszyk, D.M., Cabrera, H. and Thompson, C.R.,1988. “Cali fornia Statewide Assessment of TheEffects of Ozone on Crop Productivity” , JAPCA 38,928-931.

Olszyk, D.M., Thompson, C.R. and Roe, M.P., 1988,“Crop Loss Assessment for Cali fornia, Modelli ngLosses With Different Ozone Standard Scenarios” ,Env. Pollut. 53, 303-311.

Pearson, R.G., Linzon, S.N. and Donnan, J.A., 1988.“Ozone Effects on Crops in Ontario and RelatedEconomic Impact” , Env. Pollut. 53, 450-451.

Pielke, R.A., Cotton, W.R., Walko, R.L., Tremback,C.J., Lyons, L.D., Grasso, L.D., Nicholls, M.E.,Moran, M.D., Wesley, D.A., Lee, T.J., andCoperland, J.H., 1992, “A ComprehensiveMeteorological Modelli ng System − RAMS”, Met.Atmos. Phys. 49, 69-91.

Physick, W.L., Noonan, J.A., McGregor, J.L., Hurley,P.J., Abbs, D.J. and Manins, P.C., 1994, “LADM: ALagrangian Atmospheric Dispersion Model” , CSIRODivision of Atmospheric Research Technical ReportNo. 24.

Pye, J.M., 1988, “ Impact of Ozone on The Growthand Yield of Trees. A Review” , J. Env. Qual. 17, 347-360.

Rayner, K.N., 1987, “Dispersion of AtmosphericPollutants from Point Sources in a CoastalEnvironment” , W.A. Department of EnvironmentalProtection Technical Series No. 22, December 1987.

Read, C., 1989, “Even Low Levels of Ozone in SmogHarm the Lungs” , New Scientist 123, 14.

Reiss, R., Ryan, B., Koutrakis, P. and Tibbetts, S.,1995, “Ozone Reactive Chemistry on Interior LatexPaint” , Env. Sci. and Technol. 29, 1906-1912.

Richards, J.W., Middleton, B.T. and Hewitt, W.N.,1958, “Air Pollution With Relation to AgronomicCrops, V. Oxidant Stipple of Grapes” , Agron. J. 50,559-561.

RTA, 1995, “New South Wales Road Traff icAuthority. RTA’s Plan for Reducing VehicleEmissions” , NSW Road Traff ic Authority.

Runeckles, V.C. and Krupa, S.V., 1994, “The Impactof UV-B Radiation and Ozone on TerrestrialVegetation” , Env. Pollut. 83, 191-213.

Rye, P.J., 1989, “Evaluation of a Simple NumericalModel as a Mesoscale Forecasting Tool” , J. Appl.Met. 28, 1257-1270.

Rye, P.J., 1991, “The Environmental ProtectionAuthority Telemetry System”, W.A. Department ofEnvironmental Protection Technical Series 44,October 1991.

Rye, P.J., 1996a, “Contribution by Bushfire Smoke toPhotochemical Smog” , CALM Science Supplement 4,Proceedings of an Australian Bushfire Conference,Perth, Western Australia, 27-29 September 1993,W.L. McCaw, N.D. Burrows, G.R. Friend and AMGill , eds., 129-134.

Rye, P.J., 1996b, “Meteorological and SourceAnalysis of Smog Events During the PerthPhotochemical Smog Study” , Department ofEnvironmental Protection Technical Series 83.

Rye, P.J., 1996c, “Modelli ng the Meteorology ofPerth Photochemical Smog Events” , Department ofEnvironmental Protection Technical Series 84.

Rye, P.J., 1996d, “Modelli ng Perth PhotochemicalSmog Events Using the Urban Airshed Model” ,Department of Environmental Protection TechnicalSeries 85.

Scheffe, R.D. and Morris, R.E., 1993, “A Review ofthe Development and Application of the UrbanAirshed Model” , Atmos. Environ., 27B, 23-39.

Schorr, M.M., 1992, “NOx Emission Control for GasTurbines: a 1992 Update on Regulations andTechnology” . IGTI Journal 7, 1-12, American Societyof Mechanical Engineers.

Schwartzel, E. 1994, “Remedies for Canada’s SmogProblem”, Discussion paper developed for theAtmosphere Caucus of the Canadian EnvironmentalNetwork.

Seinfeld, J. 1986, “Atmospheric Chemistry andPhysics of Air Pollution” , John Wiley and Sons Inc.New York.

Streeton, 1990, “Air Pollution Health Effects and AirQuality Objectives in Victoria”, Report to EPAVictoria, September 1990.

Stuart, A. and Carnovale, F., 1994, “PerthPhotochemical Smog Study Area Based EmissionsInventory” , Environment Protection Authority ofVictoria, May 1994.

Tanner, R. and Zielinska, B., 1994, “Determination ofthe Biogenic Emission Rates of Species Contributingto ROC in the San Joaquin Valley of Cali fornia”,Atmos. Environ. 28, 1113-1120.

92

Tattersfield, A. (Chairman), 1993, “Advisory Groupon the Medical Aspects of Air Pollution Episodes(1993) Oxides of Nitrogen” , London: Her Majesty’sStationery Off ice: 104-116.

Tonnesssen, S., and Jeffr ies, H.E., 1994, “ Inhibitionof Odd Oxygen Production in the Carbon Bond Fourand Generic Reraction Set Mechanisms” , Atmos.Environ. 28, 1339-1349.

US EPA, 1985, “Compilation of Air PollutantEmission Factors, Volume 1: Stationary Point andArea Sources” , Report AP-42, 4th Edition, US EPA,Research Triangle Park, N Carolina, USA.

US EPA, 1991, “Non-road Engine and VehicleEmission Study Report” , November 1991, EPACertification Division, EPA-21a-2001.

US EPA, 1993, “Automobiles and Ozone”, DocumentEPA 400-F-92-006, EPA Off ice of Mobile Sources,Ann Arbor, Michigan.

US EPA, 1994, “Vehicle Fuels and the 1990 CleanAir Act” , Document EPA 400-F-92-015, EPA Off iceof Mobile Sources, Ann Arbor, Michigan.

WHO, 1987, “Air Quality Guidelines for Europe”,WHO Regional Off ice for Europe, Copenhagen,Denmark.

White, R., Rappaport, S., Lieber, K., Gorman, A.,DuMelle, F., Maple, D., Bhawnani, M., and Edelman,N., 1995, “Children at risk from ozone air pollution :United States, 1991-1993” , Morbidity and MortalityWeekly Report, 44, 309-312, Air Pollution andRespiratory Health Branch, Center for DiseaseControl.

Willi ams, D.J., Carras, J.N., Hacker, J.M., Clark, N.J.,Weir, P., Johnson, G.M., Azzi, M., Rayner, K.N. andBerko, H.N., 1995, “Field (In)validation of anEmission Inventory for Perth, Western Australia”, tobe published in the Proceedings of the Air and WasteManagement Association Specialty Conference:1995.

Wolters, J.H.B. and Martens, M.J.M., 1987, “Effectsof Air Pollutants on Pollen” , Bot. Rev. 53, 373-409.

Woodward, A.J., Calder, I., McMichael, A.J.,Pisaniello, D., Scicchitano, R., Steer, K. and Guest,C.S., 1993, “Options for Revised Air Quality Goalsfor Ozone (Photochemical Oxidants)” , Project Reportto the Commonwealth Department of Health, Housingand Community Services, August 1993.

Woodward, A., Guest, C., Steer, K., Harman, R.,Scicchitano, R., Pisanello, P. and McMichael, I.,1995, “Respiratory effects of tropospheric ozone:implications for Australian air quality goals” , J.Epidemiol. and Community Health 49, 401-407.

World Health Organisation, 1992, “Acute Effects onHealth of Smog Episodes” , Geneva, World HealthOrganisation, European series, 43.

Yocom, J.E. and Upham, J.B., 1977, “Effects onEconomic Materials and Structures” , Air Pollution,Vol. 3, 3rd edition, A.C. Stern Ed., Academic Press,New York, 65-116.

93

aROC photochemical reactivity coefficient of an organic gas

aROC effective overall photochemical reactivity coefficient in an urban atmosphere

β coefficient in the IER theory which relates maximum smog production to the initial emissions ofnitrogen oxides

Bsp a measure of light scattering by extremely fine particles

CH4 methane

CO carbon monoxide

CO2 carbon dioxide

E extent to which smog reactions have progressed towards the onset of NOx-limited conditions

F coefficient in the IER theory , defining the fraction of oxides of nitrogen emitted as nitric oxide

f(T) function which accounts for the effect of air temperature, T, on the photolysis rate of NO2

G parameter to determine whether smog production is light-limited or NOx-limited

HNO3 nitric acid

kNO2 rate coefficient for the photolysis of NO2

NMHC non-methane hydrocarbons - a group of hydrocarbons which excludes methane

NO nitric oxide

NO2 nitrogen dioxide

NOx oxides of nitrogen, or nitrogen oxides, usually taken to mean nitric oxide (NO) and nitrogendioxide (NO2)

{NOx}em alternative way of writing {NOx}0,t, i.e. the emissions of NOx

NOy represents the full range of gaseous oxygenated nitrogen species including nitric oxide (NO),nitrogen dioxide (NO2), nitric acid (HNO3), peroxyacetyl nitrate (PAN) and other gaseousorganic nitrates

O2 oxygen

O3 ozone

OH⋅� hydroxyl radical

P coefficient in the IER theory , defining the rate of particulate nitrate formation

PAN peroxyacetyl nitrate

PM2.5 mass of fine particulate matter with an aerodynamic diameter less than 2.5 micron

ppb parts per billion

ppm parts per million

RO2⋅� peroxy radical, formed by the reaction of an organic, for example a hydrocarbon (RH) with thehydroxyl radical (OH⋅� ), followed by reaction with molecular oxygen

ROC reactive organic compound - an organic compound which can take part in chemical reactions inthe atmosphere (includes hydrocarbons, carbonyl compounds, alcohols, etc)

AppendixGlossary of Chemical and Mathematical Symbols

94

Rsmog photolytic rate coeff icient for organic species

SMOG a stable measure of photochemical smog concentration, which remains constant when NO is addedto an air parcel thereby producing NO2 at the expense of O3

SNGN denotes the range of stable non-gaseous nitrogen compounds formed in photochemical smog

SO2 sulphur dioxide

SP denotes “smog produced” and represents the gaseous and non-gaseous products of smog reactions,excluding initial concentrations. SP is equivalent to the total nitric oxide consumed plus the ozoneproduced.

SPmax the maximum quantity of smog produced, proportional to the initial supply of nitrogen oxides

SUM0 a parameter representing the sum of all hourly concentrations above zero

T air temperature

τ time

VKT vehicle kilometres travelled

{ } 0 denotes the number of moles present at the start time (i.e., background)

{ } t denotes the number of moles of the particular species present in the air mass at time t

{ } 0,t denotes the number of moles of the species initially present in the air plus the moles added viaemissions up to time t

[ ] denotes concentration of a gas in the atmosphere