Temporal and Spatial Study of Thunderstorm Rainfall in the Greater Sydney Region Ali Akbar Rasuly University of Wollongong

University of Wollongong Research Online

University of Wollongong Thesis Collection University of Wollongong Thesis Collections

1996 Temporal and spatial study of thunderstorm rainfall in the Greater Sydney region Ali Akbar Rasuly University of Wollongong

Recommended Citation Rasuly, Ali Akbar, Temporal and spatial study of thunderstorm rainfall in the Greater Sydney region, Doctor of Philosophy thesis, School of Geosciences, University of Wollongong, 1996. http://ro.uow.edu.au/theses/1986

Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library: [email protected]

TEMPORAL AND SPATIAL STUDY OF THUNDERSTORM RAINFALL IN THE GREATER SYDNEY REGION

A thesis submitted in fulfilment of the requirements for the award of the degree

UNIVERSITY O*

DOCTOR OF PHILOSOPHY from

UNIVERSITY OF WOLLONGONG by

ALIAKBAR RASULY B.Sc. & M.Sc. (IRAN, TABRIZ University)

SCHOOL OF GEOSCIENCES

1996 CERTIFICATION

The work presented herein has not been submitted to any other university or institution for a higher degree and, unless acknowledged, is my own original work.

A. A. Rasuly

February 1996 i

ABSTRACT

Thunderstorm rainfall is considered as a very vital climatic factor because of its significant effects and often disastrous consequences upon people and the natural environment in the Greater Sydney Region. Thus, this study investigates the following aspects of thunderstorm rainfall climatology of the region between 1960 to 1993.

In detail, it was found that thunderstorm rainfalls in Sydney have marked diurnal and seasonal variations. They are most frequent in the spring and summer and during the late afternoon and early evening. Thunderstorms occur primarily over the coastal areas and mountains, and less frequently over the lowland interior of the Sydney basin. Environmental factors, such as the local climatic factors and physiographic parameters may control thunderstorm occurrence and its associated rainfall distribution. More detailed associations, possible causal relationships, using stepwise regression indicate that thunderstorm rainfall frequency could partially be affected by air and sea temperatures, and air humidity.

Accordingly, specific attention was paid to the patterns of the spatial variation of thunderstorm rainfall during the warm months (October to March) over a long time- span (34 years), using data from 191 rainfall stations. Mathematically, the gamma functions (beta and alpha values) describe and summarise the probability distribution of daily thunderstorm rainfall across the Sydney region. The findings reveal the interplay of topographic, coastal and urban effects in controlling the amount of thunderstorm rainfall in both spring and summer.

A "climatologically oriented GIS" (including a Digital Elevation Model (DEM), a proximity map, and a landuse model) together with regression procedures were used to assess the relative importance of physiographic and environmental variables for six of the largest thunderstorm rainfall events. Three patterns emerged. The first is an increase in thunderstorm rainfall occurring toward the coast. The second is an increase in thunderstorm rainfall as elevation increases. Finally, the more compact the urban residential and commercialised areas the greater the amount of thunderstorm rainfall. These variables account for 70 per cent of thunderstorm rainfall variations throughout the Sydney region. ii

ACKNOWLEDGMENTS

I would like express my very special gratitude to Associate Professor Edward Bryant, my supervisor, who gave me encouragement and support throughout the study with ideas, literature, computer programs, proof reading and much more assistance. I am as well grateful to staff and academic members of the School of Geosciences, University of Wollongong for their suggestions and support throughout my period of study. Very special thanks should be given to Professor A. Chivas, Professor M. Wilson, Associate Professor B. Young, Associate Professor G. Nanson, Associate Professor C. Woodroffe, Dr. A. Young, Dr. A. O'Neill, Dr. L. Brown, Dr. L. Head, Dr. J. Formby, Dr. G. Waitt, Dr. R. Wray, Mr. D. Price, Mr. G. Black and Ms. J. Shaw, very kind people, who gave me so much encouragement and provided much more valuable materials during my study. I also wish to thank Mr. J. Marthick for the use of his computing skills, particularly in GIS and Mr. R. Miller and Mr. D. Martin are thanked for advice on cartography. My fellow postgrads were all very helpful and understanding. Thanks must be given to all these friendly people.

Grateful acknowledgment is made also to the people - at the Bureau of Meteorology Sydney Regional Office; the National Climate Centre in Melbourne; Sydney Water; the Australian Oceanographic Data Centre; the Australian Surveying and Land Information Group; and Infomaster Australia (SPANS GIS) - who kindly provided sources of data for this research.

I am also grateful to my parents, relatives, friends and academic members of the Department of Geography in Tabriz University, who consistently encouraged me to finish this study in many letters. Without doubt, I am indebted to my wife and children's infinite understanding, my supporters who shared probably the entire range of emotional states with me in producing this thesis. Appreciation is offered to all.

Finally, I would like to appreciate The Ministry of Culture and Higher Education of Islamic Republic of IRAN, for awarding me a scholarship and providing financial support to make this thesis possible. Yet, I deeply believe that the greatest acknowledgment is for God's support and guidance. iii

TABLE OF CONTENTS PAGE

ABSTRACT i ACKNOWLEDGMENTS ii TABLE OF CONTENTS iii LIST OF TABLES vii LIST OF FIGURES « LIST OF PLATES xi

CHAPTER 1 INTRODUCTION

1.1 Thunderstorm Rainfall in the Sydney Region 1 1.2 Topo-Climatic Characteristics of the Sydney Region 3 1.2.1 Topography of the Region 3 1.2.2 Climate of the Study Area 3 1.2.2.1 Rainfall Characteristics 5 1.2.2.2 Temperature Patterns 8 1.3 Objectives of This Study 9 1.4 Research Significance 10 1.5 Data Management and Modelling Techniques Applied 12 1.6 Thesis Outline by Chapters 15

CHAPTER 2 LITERATURE REVIEW ON THUNDERSTORM RAINFALL

2.1 Introduction 17 2.2 Thunderstorm Characteristics 17 2.2.1 Life-Cycle of a Single Thunderstorm 18 2.2.2 Complex Thunderstorm Systems 19 2.3 Synoptic Weather Patterns Creating Thunderstorms 22 2.4 Climatic Variables and Thunderstorms 24 2.4.1 Air Temperature 24 2.4.2 Sea-Surface Temperature 25 2.4.3 El Nino / Southern Oscillation 28 2.5 Physiographic Parameters and Thunderstorm Rainfall 28 2.5.1 Topography and Thunderstorm Rainfall 29 2.5.2 Effects of Proximity to the Sea upon Thunderstorm Rainfall 33 2.5.3 Impacts of Urban Areas on Thunderstorm Rainfall Distribution 35 2.6 Distribution of Thunderstorms in Australia 38 2.7 Synoptic Patterns Associated with Thunderstorm Activity in Australia 40 2.7.1 Weather Systems and Thunderstorm Activity in NSW 44 iv

2.7.2 Thunderstorm Development in the Sydney Region 48 2.8 Sydney's Physiographic Parameters and Thunderstorm Rainfall 57 2.9 Conclusions 64

CHAPTER 3 TEMPORAL DISTRIBUTION OF THUNDERSTORM RAINFALL IN THE SYDNEY REGION

3.1 Introduction 67 3.2 Data Used 67 3.3 Methods Applied 68 3.4 Yearly Distribution of Thunderstorm Rainfall 71 3.5 Seasonal and Monthly Distributions 76 3.6 Diurnal Variation 79 3.7 Discussion 81 3.7.1 The Role of Synoptic Weather Patterns 82 3.7.2 The Effect of Climatic Factors 83 3.7.3 The Impact of Physiographic Parameters 84 3.8 Summary and Conclusion 85

CHAPTER 4 THUNDERSTORM RAINFALL AND CLIMATIC VARIABLES

4.1 Introduction 86 4.2 Data Sources and Analysis Techniques 86 4.3 Description of Variables 87 4.3.1 Air Temperature 88 4.3.2 Sea Surface Temperature 89 4.3.3 Air Humidity 91 4.4 Correlations Matrices of Variables 92 4.5 Multiple Associations Between Variables 95 4.6 Discussion 97 4.6.1 Effects of Sea-surface Temperature 97 4.6.2 Associations Between Air Temperature and Thunderstorms 98 4.6.3 The Role of Air Humidity 100 4.7 Summary and Conclusion 102

CHAPTER 5 A REVIEW ON GIS TECHNIQUES

5.1 Introduction 103 5.2 What is a GIS? 103 5.3 Purpose of GIS 104 5.4 How GIS Operates 106 5.4.1 Data Structures in GIS 107 5.4.2 Functionality of Data in GIS 108 5.5 Implications of GIS Techniques in Climatology 110 v

5.6 Application of the GIS in Resolving Problems in Rainfall Analysis 112 5.7 Data Sources on GIS System 115 5.8 Methods Used in a SPANS GIS n5 5.8.1 Data Input 116 5.8.2 Model Building I17 5.8.3 Model Analysing I18 5.9 GIS Potential Errors n8 5.10 Summary and Conclusion 119

CHAPTER 6 THE SPATIAL VARIATION AND DISTRIBUTION OF THUNDERSTORM RAINFALL

6.1 Introduction 121 6.2 Data Selection 121 6.3 Techniques Used 126 6.4 Thunderstorm Rainfall Selection Criteria I28 6.5 Spatial Variability of Thunderstorm Rainfall 131 6.6 Spatial Distribution of Thunderstorm Rainfall I35 6.6.1 Average Event Values I3" 6.6.2 The Biggest Events I39 6.7 Discussion *49 6.8 Summary and Conclusion I52

CHAPTER 7 RELATIONSHIPS BETWEEN THUNDERSTORM RAINFALL AND PHYSIOGRAPHIC PARAMETERS

7.1 Introduction I54 7.2 Data Used 154 7.3 Techniques Employed I57 7.3.1 GIS Techniques Applied I57 7.3.1.1 Landuse Map of the Sydney Region 161 7.3.1.2 Advanced SPANS GIS Functions Used 168 7.3.2 Statistical Techniques Used 172 7.4 Topography and Rainfallfrom Thunderstorms 172 7.4.1 Description of Major Topographic Units 172 7.4.2 Association Between Elevation and Thunderstorm Rainfall 174 7.4.3 Association Between Aspect Classes and Rainfall 176 7.5 Proximity to the Sea and Thunderstorm Rainfall Distribution 178 7.6 Landuse Patterns and Thunderstorm Rainfall 180 7.7 Overlay Modelling / Multiple Relations 183 7.7.1 GIS Overlay Modelling 184 7.7.2 Multiple Relations Among Variables 185 7.7.2.1 Stepwise Multi-Regression Technique 185 7.7.2.2 Spatial Distribution of Z Scores Over Sydney 188 7.8 Discussion 191 7.8.1 The Role of Coastal Area 7.8.2 Impact of Topographic Factors 192 7.8.3 Effect of Landuse on Rainfall Distribution 194 7.9 Summary and Conclusion 196

CHAPTER 8 CONCLUSIONS

8.1 Introduction 198 8.2 Major Conclusions of the Thesis 198 8.3 Limitations of the Study 200 8.3.1 Limitations of Data Used 200 8.3.2 Limitations of Techniques Applied 201 8.4 Advantages and Implications of the Study 201 8.4.1 Advantages of the Study 202 8.4.2 Implications of the Study 203 8.5 Suggestions for Future Studies 204 8.6 Concluding Remarks 204

REFERENCES 206

APPENDIX A LIST OF COMPUTER PROGRAMS

A. 1 Computer Program Number 1 230 A.2 Computer Program Number 2 233 A. 3 Computer Program Number 3 234 A. 4 Computer Program Number 4 236 A. 5 Computer Program Number 5 238

APPENDIX B THUNDERSTORM RAINFALL DATA

B. 1 Common Thunderstorm-days in Sydney Region 240 B.2 Monthly Thunderstorm Rainfall Data at Richmond 249 B.3 Monthly Thunderstorm Rainfall Data at Sydney R.O. 250 B.4 Monthly Thunderstorm Rainfall Data at Sydney Airport 251 B.5 List of Rainfall Stations 252

APPENDIX C SYNOPTIC WEATHER CHARTS

Synoptic charts 6.1 from 23th to 25th October, 1987 256 Synoptic charts 6.2 5th to 12th November, 1984 257 Synoptic charts 6.3 9th to 11th December, 1988 258 Synoptic charts 6.4 19th to 22th January, 1991 259 Synoptic charts 6.5 7th to 11th February, 1990 260 Synoptic charts 6.6 10th to 11th March, 1975 261 vii

APPENDIX D DATA USED FOR GIS AND STATISTICAL MODELS

Geographical Location of Rainfall Stations and their Attributes

APPENDIX E EQUATIONS Equations Used in SPANS GIS LIST OF TABLES

TABLE PAGE

1.1 Examples of thunderstorm rainfall events causing flash floods 10 1.2 Gives examples in using thunderstorm data in the region 14 3.1 Represents a detailed description of the codes of present and past weather used in thunderstorm observations 68 3.2 Common thunderstorm-days in the Sydney region See Appendices 3.3 General geographic characteristics of the seven selected stations 69 3.4 Locality of the seven selected stations 71 3.5 Yearly variation of thunder-days frequency and thunderstorm rainfall amounts at 7 thunder-recording stations in the Sydney region 72 3.6 Summary descriptive statistics for yearly thunderstorm rainfall frequency, in the Sydney region 73 3.7 Summary descriptive statistics for yearly thunderstorm rainfall amounts, in the Sydney region 74 3.8 Average seasonal thunderstorm rainfall for selected stations 77 3.9 The percentage of average thunderstorm rainfall to mean monthly rainfall in different stations, in the Sydney region 78 4.1 Monthly thunderstorm rainfallfrequency a t Richmond See Appendices 4.2 Monthly thunderstorm rainfall amount at Richmond See Appendices 4.3 Monthly thunderstorm frequency at Sydney R. 0. See Appendices 4.4 Monthly thunderstorm rainfall amount at Sydney R. 0. See Appendices 4.5 Monthly thunderstorm rainfall frequency at Sydney A. See Appendices 4.6 Monthly thunderstorm rainfall amount at Sydney A. See Appendices 4.7 Description of thunderstorm data 88 4.8 Means and extremes of temperature at three selected stations 89 4.9 Monthly and yearly sea-surface temperature data at Port Hacking 90 4.10 Simple statistics of the relative humidity in the Sydney region 92 4.11 The correlation matrix between dependent variables 93 4.12 Correlation matrix for independent variables 93 4.13 Linear regression coefficients of dependent variables by independent variables 94 4.14(a) Results of stepwise multiple regression analysis of thunderstorm rainfall frequency at the Sydney Airport station 96 4.14(b) Results of stepwise multiple regression analysis of thunderstorm rainfall frequency at Richmond station 96 6.1 Difference between two sets of stations (the Sydney Water and the Bureau of Meteorology) according to their rainfall means 123 6.2 List of stations and the periods from which data were used See Appendices 6.3 Thunderstorm rainfall values extracted from the intersection of two ix

populations using probability excellence graphs 130 6.4 General descriptions for the 6 biggest thunderstorm rainfall events in the region 139 7.1 Origin of the data that used in Chapter 7 157 7.2 Geographical location of rainfall stations and their attributes See Appendices 7.3 Limits of the study area / database 158 7.4 Aspect classes derived from the DEM model, using SPANS GIS 161 7.5 Description of landuse types in the Sydney region. 162 7.6 Equations which were written in SPANS environment See Appendices 7.7 Area cross tabulation results between the topography map of the region and thunderstorm rainfall map. 174 7.8 The areal distribution of thunderstorm rainfall by topographic classes 175 7.9 A linear regression analysis between thunderstorm rainfall amount and elevation of rainfall stations located in the region 176 7.10 Area cross tabulation results between the aspect map of the region and thunderstorm rainfall map 177 7.11 A multiple regression analysis between aspect classes and thunderstorm rainfall amount 178 7.12 Area cross tabulation results between the proximity to sea map of the region and thunderstorm rainfall map 179 7.13 The areal distribution of thunderstorm rainfall by proximity classes 179 7.14 Correlation coefficients between the proximity to the sea and thunderstorm rainfall 180 7.15 Area cross tabulation results between the landuse map of the region and thunderstorm rainfall map 181 7.16 The areal distribution of thunderstorm rainfall by landuse classes 181 7.17 The result of a t-test for rainfall distribution in different landuse classes 182 7.18 A multiple regression analysis between landuse classes and thunderstorm rainfall 183 7.19 Interrelations matrix among physiographic parameters and thunderstorm rainfall 186 7.20 Presents the result of stepwise multiple regression analysis for the average of the biggest thunderstorm rainfall amounts 187 x

LIST OF FIGURES

FIGURE PAGE

1.1 Study area - the Sydney region, NSW, Australia 2 1.2 Topographic and Location Map of Sydney region 4 1.3 Illustrates inter-annual precipitation variation at Observatory Hill, Sydney 5 1.4 Median annual rainfall of THE Sydney region 6 1.5 Average monthly rainfall at three stations in the Sydney region 7 1.6 Average monthly maximum and minimum temperature at Sydney 8 1.7 Percentiles of maximum temperature at Observatory Hill 9 2.1 Schematic representation of a thunderstorm cell 18 2.2 Three stages in the development of a thunderstorm 19 2.3 Schematic visual appearance of a supercell thunderstorm 21 2.4 Average annual thunder-days in Australia 39 2.5 Basic elements in (a) the pattern of pressure distribution and of associated (b) airmasses over Australia in summer 40 2.6 Represent a pre-frontal trough (a), and a line storm associated with an eastward moving trough (b) over south-eastern Australia 41 2.7 Shows a sample of cut-off low in the region 45 2.8 Schematic of the life cycle of the precipitation area of a MCCs 46 2.9 Presentation of the anomaly maps using the Terminal Area Severe Turbulence (TAST) radar data in the Greater Sydney region 50 2.10 Diurnal distribution of thunderstorm occurrence for the different time periods (local time) in the Sydney region 51 2.11 Presents examples of six meso-scale synoptic weather systems causing thunderstorm activity in the Sydney region 53 2.12 Selected MSLP charts illustrating the six synoptic classes over Sydney 54 2.13 Thunderstorm density model based on new radar echoes in Sydney 55 2.14 Lightning density for single thunderstorm events based on data from the NSW lightning detection network 56 2.15 The intensity of annual energy use in the Sydney region 61 2.16 Spatial distribution of nitrogen oxides emissions from all sources in the Sydney 62 3.1 A dendrogram shows the result of the NNA technique 71 3.2 Yearly variation of thunder-days frequency and thunderstorm rainfall at Sydney Regional Office 74 3.3 Yearly variation of thunder-days frequency and thunderstorm rainfall at Richmond 75 3.4 Normalised Residual Mass curves of annual thunderstorm rainfall in the Sydney region 75 3.5 Seasonal distribution of thunderstorm rainfall in different stations 76 XI

3.6 Monthly distribution of thunderstorm rainfall in the Sydney region for different stations 78 3.7 Diurnal variation of thunderstorm rainfall frequency for three thunder seasons at Katoomba 80 3.8 Diurnal variation of thunderstorm rainfall frequency for three thunder seasons at Richmond 80 3.9 Diurnal variation of thunderstorm rainfall frequency for three thunder seasons at Sydney Regional Office 81 4.1 Average monthly variation of the sea surface temperature 90 4.2 Monthly distribution of the mean relative humidity at three stations in the Sydney region 92 5.1 Schematically represents different data structures used in a GIS 108 6.1 Relation between correlation coefficient (r) and interstation-distance 124 6.2 Sydney region - rainfall stations networks 125 6.3 The gamma density function for a and /? values 127 6.4 Probability of exceedence diagrams for 7 selected thunder-recording stations 129 6.5 Geographical distribution of alpha value, Spring (Oct to Dec) 132 6.6 Geographical distribution of beta value, Spring (Oct to Dec) 133 6.7 Geographical distribution of alpha value, Summer (Jan to Mar) 134 6.8 Geographical distribution of beta value, Summer (Jan to Mar) 135 6.9 Average thunderstorm rainfall per event (Oct to Dec) 137 6.10 Average thunderstorm rainfall per event (Jan to Mar) 138 6.11 Thunderstorm rain - Sydney region (23-25 October 1987) 140 6.12 Thunderstorm rain - Sydney region (5-12 November 1984) 141 6.13 Thunderstorm rain - Sydney region (9-11 December 1988) 143 6.14 Thunderstorm rain - Sydney region (19-22 January 1991) 144 6.15 Thunderstorm rain - Sydney region (7-11 February 1990) 147 6.16 Thunderstorm rain - Sydney region (10 and 11 March 1975) 148 7.1 Sydney region - thunderstorm rainfall, the average of the 6 biggest daily thunderstorm rainfall events 156 7.2 Proximity map from average coast-line 159 7.3 Aspect map of the Sydney region 160 7.4 Landuse map of the Sydney region 163 7.5(a-d) Physiographic parameters of the Sydney region subject to the highest daily thunderstorm rainfall amounts 171 7.6 Major topographic units of the Sydney region 173 7.7 The distribution of thunderstorm rainfall in the Sydney region based upon aspect classes 177 7.8 Distribution of thunderstorm rainfall in the Sydney region based upon landuse classes 182 7.9 Spatial Distribution of Z scores over Sydney region 190 xii

LIST OF PLATES

PLATE PAGE

1.1 Gives examples of some extensive and serious damage caused by thunderstorm rainfall in the Sydney region 11 2.1 Shows a coldfront off the South Coast of New South Wales 42 2.2 Displays thunderstorm development over the Sydney region 49 2.3 Shows smog over central Sydney 63 7.1 Closeup view of heavy commercial landuse showing the part of CBD 164 7.2 Closeup view of heavy industrial landuse 165 7.3 View of compact residential landuse in the Sydney region 165 7.4 View of light-moderate residential landuse 166 7.5 View of normal rural /semi-urban area 167 7.6 Shows example of rural / open areas 167 7.7 Closeup view of compact vegetated land cover in the Sydney region 168 CHAPTER ONE Introduction 1

CHAPTER 1

INTRODUCTION

1.1 Thunderstorm Rainfall in the Sydney Region

Sydney, with its sprawling suburban area, and a population of approximately 3.5 million, is Australia's largest city. It is located between the South Pacific Ocean and the mountain ranges in the west. During the last three decades, meteorologists (Williams, 1984; Colquhoun and Shepherd, 1985) and climatologists (Hobbs, 1972; Sumner, 1983b and Linacre, 1992) have indicated that thunderstorm activity is a characteristic feature of the warm summer months in this region and that rainfall from thunderstorms is a major source of moisture for most parts of the study area. Flash floods are usually associated with the type of thunderstorm that produces localised, but very intense rainfall, which damages property and even results in a loss of life (Shanahan, 1968; Riley, 1980; Riley, et al., 1985; Speer and Geerts, 1994). It seems that thunderstorm rainfalls, and occasionally their associated flash floods, are a natural part of Sydney's climatic environment (Bryant, 1991; Johnson et al., 1995).

Information on the variation and distribution of thunderstorm rainfall over time and within the region is, therefore, crucially important in a variety of applications. However, in the Sydney region, the knowledge of the temporal and spatial distribution of thunderstorm rainfall is limited to some case studies of the specific thunderstorm rainfall events which have been considered over a short period of time (Armstrong and Colquhoun, 1976; Morgan, 1979a; Nanson and Hean, 1985; Shepherd and Colquhoun, 1985). Consequently, this study's investigation will involve defining some important aspects of the thunderstorm rainfall climatology of the region, which has not received a similar level of attention. This will be done by analysing the available data over a longer period - from 1960 to 1993 - using appropriate sets of techniques.

There are two main aims of this thesis. The first is to focus attention on the patterns of the temporal and spatial variation and distribution of thunderstorm rainfalls during the warm months (October to March) over the time-span of 34 years. The second is to examine the thunderstorm rainfall patterns in relation to some of the primary climatic variables (air and sea temperatures, for instance) as well as physiographic parameters such as topography, proximity to the ocean, and the landuse of the Sydney region. CHAPTER ONE Introduction 2

In the following section, the study area and the important topo-climatic characteristics of the Sydney region are described. Section 3 outlines the aims of this study. Section 4 offers some important reasons indicating that the anticipated information from the current research will be of value for future investigations in the fields of meteorology and climatology. Section 5 gives a framework for all available data and the techniques which will be used in this study. Finally, in section 6, the whole thesis is outlined.

on ~\«« ^^dJy" w

IHOlAN

OCt AH y.fX *y \. j ^ PACIFIC

j NORTHERN^./ \ ocetH •atra ^—^ | TERRITORY 1 V. I • | \ 1 Troofc i V

' , QUEENSLAND \» W WE5TEF N h I—, X y AUSTRA LIA i i I | SOUTH [_ s—jA 1 AUSTRALIA | / J ^_^ 1 NEW J S . 1 SOUTH / \ SI 1 WALES l 0 S> it) l K yWSTUDY S AR km ,00, ^S ^S1_ —ykf fJ-~~^ / " ^\.

/ THE SYDNEY REGION N^ 33 30 f- CvCT^-'I -^i^S. Gosford «^ tj~ZBROKEN BAY N Richmond _^^*"f A

• Katoomba Penrith J 33 45- • £ T Prospect Dam • Parramatta «, f s ^r$r£Zrj)pORT JACKSON Liverpool Sydney R.O TASMAN • "c / 34 00- Bankstown /^Jjydrfey Airport Camden ^^<-\BpTANY BAY • • ^T^ SEA M Lucas H. C-XL^PORT HACKING Picton Campbelltown / • / 1 | 34 15 / 1 1 / 20 Km

• >Vollongong

-Bowral (i"ORT KEMBLA

•n o a o n e irt

.Figure 1.1 Study area - the Sydney region, New South Wales, Australia. CHAPTER ONE Introduction 3

1.2 Topo-CIimatic Characteristics of the Sydney Region

The Sydney region located on the south-east coast of Australia, in New South Wales (NSW) includes the Sydney Metropolitan area, which is expanding rapidly inland and contains highly industrialised pockets. The study area, as part of the Greater Sydney Region, is bounded in the north by 33° 30* latitude, extending to 150° 15* longitude in the west, and to the south-east of Wollongong as far as Bowral at 34° 30 latitude south. Figure 1.1 indicates the geographical location of the study area within Australia as well as an enlargement of the Sydney region, including the location of all selected thunder- recording stations (used in Chapter 3) in the area.

1.2.1 Topography of the Region

The Sydney region is bowl-shaped with a low plain in the middle of which is effectively walled in on three sides by bills and mountains. In the centre of the region there is the Cumberland Plain opening to the Pacific Ocean from the east. To the north of the plain, the rise is about 450 m to the top of a ridge lying eastward from the Great Dividing Range towards the coast. To the south, the rise in elevation on average is over 350 m. In the south-east of the study area the coastal range rises from 150 m to 500 m, just in the North­ west of Wollongong. To the north of Sydney the land rises from about 150 m near Broken Bay to 450 m on the northern boundary. However, westward, the region rises sharply to over 1200 m at the top of the Blue Mountains, the eastern section of the Central Tablelands. The elevation map of the Sydney region (Figure 1.2) illustrates the topographic features of the region, including the location of the main suburbs. For a more detailed topographic map of the region, refer to Figure 7.6.

1.2.2 Climate of the Study Area

The Sydney region enjoys a temperate climate and generally the broad-scale wind pattern is westerly in the winter, and easterly in the summer. This climate can be classified as being temperate with cool to cold winters and warm to hot summers (Bureau of Meteorology, 1991a). Generally the climate of this region arises from a complex interaction of broad scale, regional and local controls. On the broad scale the region is under the influence of mainly drier westerly airstreams in the winter, and predominantly moist, easterly air streams in the summer months (Linacre and Hobbs, 1977).

On the regional scale, the major influences are physiographic features (for example, topography) in and around the region, sea surface temperature off the coast and the orientation of the coastline (Bureau of Meteorology, 1979). In a such region, local variation in climate may be caused mainly by the topography (exposure to wind direction, elevation), proximity to the sea and other local factors (Cox, 1983). Within this region, ^<5

o B a a a es o S £ s Ln O o o c e o i- CM C*J V B a o 5 a < n IX' T- wieco o CO 1 ' S-* T=- T—- •^ 5"= ir™ T— ?"" O LU g3 T— O LO o a a o o 7o +o CO > i LO ?•» T™ CM m V LO CO CO T- 01 (f) UJ 0 < mm f-HH tr r- > 0 y c < U oe I-SI- > 7 if) LU 1 •£* Sit Lot

00 LSI

oe oe i-

os

3 CHAPTER ONE Introduction 5

which extends to about 100 km inland with some parts reaching elevations of over 1200 m, most climatic elements vary significantly. Therefore, it is not surprising that Sydney's climate has been considered as a complex of many local climates (Paine et al., 1988).

1.2.2.1 Rainfall Characteristics

In Sydney's climate, the primary rain-producing mechanisms are: major storms, cold fronts and thunderstorms. Major storms, which are mainly dependent on the deep low pressure systems in the Tasman Sea, can produce strong winds and heavy rainfall along the NSW coast (Bureau of Meteorology, 1991a). These systems can be classified by their origin into several types which occur at different times of the year. In contrast, cold fronts produce comparatively little rain in the Sydney region, especially in summer, when the flow behind a cold front is most often from the south (Colquhoun et al., 1985). Little moisture is provided unless a following upper-level low pressure trough provides additional instability. Generally, Wilson and Ryan (1987) found that Sydney appears to be a location where only a small amount of the total precipitation is due to mechanisms associated with fronts. Thunderstorm activity is also acknowledged as an important rain-producing system in the region, particularly in late spring and the summer months. Thunderstorms can produce heavy rainfalls resulting in a considerable contribution to the annual precipitation over the Sydney region.

2500 2500

2000 2000 Average = 1182

1500 - - 1500

< LL • 1Q00 §2 10Q0

500 -• - 500

-r- -+- -+- -+- -h -h 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990

YEAR

Figure 1.3 Illustrates inter-annual precipitation variation at Observatory Hill, Sydney.

Although Sydney's annual average precipitation is estimated to be about 1200 millimetres, it tends to be erratic and unreliable. This means that annual rainfall over the Sydney region is extremely variable spatially and temporally, and this may reflect the occurrence of the CHAPTER ONE Introduction 6

large thunderstorm events in the warmer seasons. To gain an appreciation of longer term variability of rainfall, examination of the historical record is useful. The station with the longest record of rainfall in the region was the Observatory Hill station (it has been closed). Figure 1.3 shows the annual totals from 1900 to 1990. Considerable variability is evident, the highest fall recorded being 2194 mm (1950) and the lowest recorded, 625 mm (1976). The totals during the period from about 1990 to 1950 exhibit less inter-annual fluctuation than after 1950 and more recent times.

It is likely that the greatest spatial variation in rainfall can also be associated with changes in topography and distance from the sea, both of which can cause considerable variations in the annual precipitation in the Sydney region (Figure 1.4).

THE SYDNEY REGION

N

S

TASMAN

SEA

Figure 1.4 Median annual rainfall (in mm) of the Sydney region. The isohyets are derivedfrom al l available data from rainfall stations with at least 20 (1832-1986) years of record. The number of years of record varies for different stations in the region (Bureau of Meteorology).

It is evident from the annual rainfall distribution map, for instance, that an area which is located west of Wollongong, over the Illawarra Plateau, receives more than 1600 mm rain annually. While the highest rainfall in Sydney of more than 1300 mm occurs over the more elevated parts of the northern suburbs that form the Hornsby Plateau. The lowest rainfall occurs in low-lying pockets of the western Cumberland Plain. For example, both Windsor and Campbelltown, which are located in lowland areas, receive less than 750 millimetres CHAPTER ONE Introduction 7

per annum. The Blue Mountains, located in the west of the study area, in contrast to the low relief of the coastal plain, receives much more rainfall (more than 1300 mm). This major topographic unit, with more than 1200 m elevation, receives some of its intence rainfalls in the summer months. In addition, in the south-east of the Sydney region, the Ulawarra Escarpment appears to be partly responsible for inducing a minor rain shadow effect over the central part of the Sydney region, as well as increasing local rainfall quite significantly (Bryant, 1982).

It has been suggested that rainfall in the Sydney region falls uniformly throughout the year when compared to other parts of Australia (Bureau of Meteorology, 1991a). As it shown in Figure 1.5, there are, however, considerable variations in monthly average rainfall amounts over the year for the three selected stations located in different parts of the region. These stations are: the Sydney Regional Office located in the east of the region near the coast; Richmond station located in the North-west of the study area (inland); and Katoomba which is located in the Blue Mountains (see Figure 1.2).

Figure 1.5 Average monthly rainfall at three stations in the Sydney region (1960-93).

Figure 1.5 gives the average monthly rainfall amounts during January to December at these three stations from 1960 to 1993. The wettest month, the month in which the average rainfall is highest, is different for the three stations. Generally, during the warm months (November to March) when thunderstorms and easterly airstreams prevail, the monthly rainfall is high. This was noted by Fitzpatrick and Armstrong (1973) who found that in the Sydney region there are clearly steeper gradients of rainfall in summer, a reflection of the influence of prevalent thunderstorm rainfall. Also, there is a secondary maximum in June affecting the region in the cooler months, which may be attributed, in part, to the frequency of east coast lows (Holland et al., 1987). As stated in 1968 by Gentilli, the rainfall of CHAPTER ONE Introduction 8

Sydney can be seen as a complicated regime, with two or three major peaks during the year, and with the driest period in late winter or early spring.

1.2.2.2 Temperature Patterns

Temperatures in the Sydney region vary widely from place to place. As the Sydney region's location is adjacent to a water mass in the east and relatively high ranges in the west, it escapes extremes of temperature. However, within the region the range of temperatures is relatively high. It has been shown by Fitzpatrick and Armstrong (1973) that the variation in temperature can be caused by differences in elevation and distance from the coast, as well as by other factors, such as aspect and slope of a particular site and the surrounding terrain. In Sydney, temperatures are mildest near the coast with a few extremely hot or cold spells during the year. Inland on the plains, temperatures in excess of 35 degrees (°C) occur regularly during the summer. Generally, the highest average temperatures occur in the low-lying central parts of the region, while the lowest occur over the Blue Mountains in the winter. The range of average maximum temperatures across the region is about 7 °C. Figure 1.6 shows the variation of the average monthly temperatures (maximum and minimum) at the three selected stations in the Sydney region throughout the year. More details about temperature patterns of the Sydney region are given in Chapter 4, where a close relationship amongst thunderstorm rainfall and temperature data will be considered.

35

30 f •Sydnry R. 0 Max

Sydnry R. 0 Min •

9 • Richmond Max

ft. 15 Richmond Min 8 \ N A. A Katoomba Max io -- * . -A Katoomba Min 5 --

H 1 1 1- 4 \ h Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 1.6 Average monthly maximum and minimum temperatures (°C) at Sydney.

The variability of maximum temperatures from day to day is also greatest in the summer and increases with distance from the coast. During the summer the difference between the 90th and 10th percentiles (which is a measure of variability) ranges from about 8 °C at CHAPTER ONE Introduction 9

Observatory Hill, Sydney (Figure 1.7) near the coast, to 14 °C at Richmond, about 55 km inland on the plain (elevation = 19 m) and at Katoomba (not shown), about 90 km inland on the Central Tablelands, with 1030 m elevation. Distance from the sea is particularly prominent in respect of maximum temperatures in summer.

During the winter the variability of maximum temperatures is around 6 °C and is fairly uniform across the region, but the difference between the maxima at the three locations varies throughout the year. While in winter the maximum diurnal temperature is quite definite and occurs around 3 pm, in summer it occurs between 11 am to 2 pm, with a slight maximum at 2 pm..

35

-•—Percentile 10 -0— Percentile 50 -A—Percentile 90

5 -:

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months

Figure 1.7 Percentiles of maximum temperature at Observatory Hill, Sydney (1900-1990).

1.3 Objectives of This Study

The main aim of this study is the organisation of thunderstorm rainfall data in time and space within the Sydney region. The second aim is to find any relationships among thunderstorm rainfall and several environmental factors (climatic and physiographic parameters of the region). In particular this study addresses several questions:

1) Are there temporal distribution patterns for yearly, seasonal, monthly and diurnal thunderstorm rainfall in the Sydney region? 2) Are there some possible causal relationships amongst climatic factors of the region and thunderstorm rainfall? 3) Are there recognisable spatial variations in thunderstorm rainfall? CHAPTER ONE Introduction 10

4) What are the average daily thunderstorm rainfall distribution patterns for spring and summer? 5) Do the physiographic parameters of the Sydney region significantly control the spatial distribution of the largest thunderstorm rainfall events?

1.4 Research Significance

The study of thunderstorm rainfall in the Sydney region is important for many reasons. Although the climatology of severe thunderstorms activity in the region has been studied by Griffiths et al. (1993), several features of thunderstorm rainfall climatology in the Sydney region have not yet received a similar level of attention.

Firstly, the temporal distribution of thunderstorm rainfall in the Sydney region, using daily thunderstorm rainfall data, has not been, in detail, studied over long time periods by previous investigators. Secondly, the spatial variation and distribution of thunderstorm rainfall, throughout the Sydney region, has not been analysed completely, for example on a monthly or seasonal bases.

Table 1.1 Examples of thunderstorm rainfall events causing flash floods in the region. Location Subject of Floods and Estimated Date Damage by Thunderstorms Damage, m$ 18 Jul 1965 North and South of Sydney unknown 10-11 Mar 1975 Eastern Parts of Sydney and Illawarra District >20 10 Nov 1976 Metropolitan Area* >40 29 Dec 1980 Sydney and Suburbs >50 5-9 Nov 1984 Most Parts of Sydney Region > 100 18 -19 Feb 1984 Illawarra District (Dapto) >6 25 Oct 1987 Illawarra and Metropolitan Area unknown 3 Feb 1990 Metropolitan Area >30 18 Mar 1990 Liverpool, Ryde and City areas 313 21 Jan 1991 Northern Suburbs of Sydney 560 * The Metropolitan area refers to the entire contiguous built up area in the Sydney region, including suburbs such as Campbelltown, but not separate urban areas such as Wollongong. ** The 'City' is a term often used loosely. Some writers use it in the same sense as 'Metropolitan area', but here it refers to the CBD (Central Business District) and older suburbs.

In addition, previous studies have not emphasised the importance of the effects of some climatic factors and the physiographic parameters upon thunderstorm rainfalls. Some researchers, for example Williams (1991), view an understanding of the physical environment as essential for an understanding of the weather. Therefore, a good CHAPTER ONE Introduction 11 knowledge of the locations of towns, mountains and details of landscape features is very useful when dealing with rainfall modelling. Advantages can be obtained by the studying thunderstorm rainfalls in relation to these parameters.

(a) Roadway at Point Piper (Eastern Suburbs, Rose Bay)

(b) Flood and erosion at Scots College (Eastern Suburbs, Rose Bay)

i _

Plate 1.1 (a and b) Gives examples of some extensive and serious damage caused by thunderstorm rainfall in the Sydney region.

Finally, it is well known that the greatest proportion of the summer rainfall of the Sydney area comes from thunderstorm activities, occasionally causing flash floods in the region (Colls, 1991; Egger, 1991). Flash flooding occurs when the intense sudden rainfall from thunderstorms cannot be absorbed or drained away quickly enough. City areas can also

3 0009 03201123 6 CHAPTER ONE Introduction 12

experience flash flooding when the rainfall is too intense for the thunderstorm water drainage systems to cope with it (Bufill, 1989; Weeks, 1992). Such thunderstorms are most common in the warm months, but little is known about the relative contribution and importance of temporal and the spatial variability and distribution of these thunderstorm rainfalls over a long time-span. Table 1.1 summarises the examples of thunderstorm rainfall events causing flash floods in the Sydney region (Bureau of Meteorology, 1965 to 1991).

Some researchers, such as Colquhoun and Shepherd (1985); Eagle and Geary (1985); White (1985) and Bryant (1991) described the Sydney thunderstorms of November 1984 in which insurance losses from flash flooding exceeded $100m. The damage cost in Sydney from flooding events is particularly high when urban areas are involved. For example, as a result of the flood that affected the Sydney area in August 1986, damage totalling approximately $ 100 million occurred, and six lives were lost. In 1990 and 1991 (Spark and Casinader, 1995), two other thunderstorms caused damage to properties totalling $ 599 million. Rural flooding can result in significant crop and stock losses and increased erosion (see Plate 1.1).

Joy (1991b) has estimated that the total annual cost of flooding to Australia is $ 380 million, and thunderstorms, it is estimated, cause more than 16% of the annual average costs of natural disasters. More recently, Ryan (1993) calculated that since 1967-91 insurance payouts were more than $ 1808 million for severe thunderstorms in Australia. Surprisingly, existing records show that in the Sydney region insurance costs have been more than $ 1100 million in the same period. Most of the damage has been caused by the direct or indirect effects of flash floods, with considerable loss of lives. Urban areas, particularly residential and commercial properties, were the most affected by these severe storm events (detailed inBlong, 1991; Smith, 1993; and Joy, 1993).

Therefore, it is important to obtain temporal and spatial distribution models of this thunderstorm rainfall. Such information is especially relevant, for instance, in city designing programs, channel network, the location of rain-gauge networks (Davidson, 1981), for disaster management, and to the insurance industry (Hobbs and Littlejohns, 1991). As a result, this research may be considered to lie within the bounds of applied climatology

1.5 Data Management and Modelling Techniques Applied

A variety of data with different scales and origins were used to assess the distribution of thunderstorm rainfall distribution in time and space for the Sydney region. These data originated from the National Climate Centres; the Sydney Water Board, and the Sydney Regional Office of the Bureau of Meteorology. Rainfall data, used to investigate rainfall variability, either spatially or temporally, should, as far as possible, be homogenous. Like CHAPTER ONE Introduction 13 most rainfall data collected over time in Australia, that for the Sydney region suffers from inconsistencies and errors in measurement that weaken the absolute confidence that can be placed in the observed data (Lavery et al., 1992; Griffiths et al., 1993). Specifically, inhomogeneities in rainfall records include:

Changes in observing practices; Changes in exposure of rain-gauge; Changes in station location (both in altitude and position); Changes in the type of gauge used; and Missing data (Nicholls, 1995; Karl, 1993).

In addition, data used in the present study were collected by different agencies. Some investigators note that thunderstorm reports are probably the most 'noisy' and biased of all meteorological data (Batt et al., 1995), and that for many types of extreme events, such as thunderstorms, the maintenance of long-term homogeneity in observations is most difficult (Nicholls, 1995).

Few of these limitations could be assessed in detail in the present study because such factors are poorly documented for Australian rainfall stations. As the present study is event- based, some steps could be taken to ensure spatial data continuity for the larger thunderstorm events. First, the raw data on thunderstorm days and associated rainfall amounts were extracted from data tapes provided by the National Climate Centre. This source contained all thunderstorm days from 1960 to 1993 for 30 thunderstorm recording stations. The timing of these events were corroborated using other data sets provided by the Sydney Regional Office of the Bureau of Meteorology. Next, all stations with less than a seven year record and having more than three years of missing data were excluded. This left 15 stations that record thunderstorm events.

Then, rainfall data was extracted from the more that 400 meteorological stations recording daily rainfall in the Sydney region. This data was provided by the Bureau of Meteorology (288) and the Sydney Water Board (112). These stations covered the period from 1960 to 1993 inclusive. To restrict these data, stations with less than a 10 year record were excluded. In addition, stations which had not contributed sufficient thunderstorm rainfall observations, according a criterion (at least 100 observations), were also excluded (see Chapter 6). This left 191 rainfall stations in the data set, 134 from the Bureau of Meteorology and 57 from the Sydney Water Board.

Similar data sets have been used successfully by numerous researchers to provide significant insights into rainfall in the Sydney region. These studies are summarised in Table Introduction

1.2. Further criteria for limiting the data are described in Chapter 3 for thunderstorm observations, and in Chapter 6 for the amount of rain falling at stations during thunderstorm events. These criteria have ensured that only the largest of thunderstorms (1584 events for the study of temporal variation, and 347 events for spatial analysis) were considered in this study.

Table 1.2 Examples of studies that have used thunderstorm data in the region. The Main Aim of the Study Author(s) Period of Data Used Spatial and temporal distribution of Williams, A. 5-9 Nov. 1985 thunderstorm rainfall Climatology offlash-floods i n the Sydney Speer, M., Geerts, B. 1957-1990 Metropolitan area Thunderstorm distribution in the Sydney Matthews, C., Geerts, B. 1965-1989 area Climatology of severe local storms in NSW Batt, K., Hobbs, J. 1991-1995

Following the considerations described above, data were summarised and reduced to managable proportions by writing several specific computer programs (presented in Appendix A). The process of data reduction also involved a number of other steps, some of which, such as data transformation, editing, coding and the generation of new variables, have been done by using a variety of commonly used computer programs. For example, the Microsoft Excel program (Apple Macintosh and PC computers) was widely used for manipulating, summarising and analysing data as well as for graphing purposes. Other computer application programs such as Ms-Dos (for editing data) Clarisworks (as a data­ base) were likewise used extensively in this study.

In the second stage, a major objective was to reduce the complexity of the subject to clearly define the climatological relationships. Because of the data complexity, a set of tools, including statistical, mathematical, and the Geographic Information Systems (GIS) techniques were applied to the available data.

Accordingly, two statistical computer programs, the JMP (SAS Institute Inc, 1989) and SPSS (Norusis, 1994) have been used for statistical analysis. These statistical techniques consist of descriptive and inferential statistics (according to the nature of the variables) are applied to find possible associations between those variables. Both the descriptive statistics and inferential statistics are frequently used in this thesis. Detailed descriptions of these analytical methods are ? in numerous text books.

Using the above-mentioned statistical programs, a wide range of descriptive statistics were applied to the data sets. In most types of climatic analyses these statistics are commonly CHAPTER ONE Introduction 15

used, firstly, to organise large data sets, and secondly to summarise such data either by measuring the central tendency or data dispersion, by using inferential statistical techniques to find bi-variate relationships. For example, the relationship between thunderstorm rainfall and elevation of the study area was considered using a simple regression technique. Other inferential statistical techniques which were used in this study are: the chi-square method, t-tests, and some simple to multiple regression methods.

Each chapter details the data-analysis procedures used. For example, in Chapter 3 the Nearest Neighbour Analysis (NNA) technique was used to establish some possible relationships amongst thunder-recording stations in the Sydney region. The aim was to select the best possible thunder-recording stations as they could represent thunderstorm activity in the region. This technique was again used, in Chapter 6, to determine whether the two networks of rainfall stations are compatible in the Sydney region.

The gamma functions (beta and alpha values) are mathematically used to find the probability distribution of thunderstorm rainfall amounts at each rainfall station in the Sydney region. The purpose of using gamma distribution is to analyse the spatial variation of thunderstorm rainfall on a seasonal basis. Because of its importance, this technique is described in detail in Chapter 6.

Also, in Chapter 6, GIS techniques are used to model the spatial distribution of thunderstorm rainfall over the study area. They are then utilised to find some expected and initial associations among rainfall patterns and physiographic parameters of the Sydney region (see Chapter 7).

1.6 Thesis Outline by Chapters

All relevant material which has been collected using a set of appropriate methods will be presented in seven chapters. Chapter 2 will bring together a general relevant literature review on thunderstorm rainfall. This literature review will consider research questions which can be related to those thunderstorm activities which cause rainfall. The strategy of this literature review is to categorise material which will lead to the development of each question to be tested using various methods.

In this thesis, a methodology chapter was not included because each of the chapters where results are presented (3, 4, 6 and 7), have their own method section, describing a variety of techniques which were applied for different sets of data. Since the use of GIS in rainfall studies, at least in Australia, is a relatively new notion, sources of data and techniques in GIS, particularly its purposes and applications in climatology, are explained in Chapter 5. The relevance of this chapter to the overall thesis can be found specifically in chapters 6 CHAPTER ONE Introduction 16 and 7, where a set of GIS methods were applied to illustrate thunderstorm rainfall data in space in relation to the Sydney's physiographic parameters.

Chapters 3, 4, 6 and 7 will present the various results of this research. The main aim of chapter 3 is to characterise the average temporal distribution of thunderstorm rainfall in the Sydney region. In this chapter the temporal distribution of thunderstorm rainfall, for different time-scales, will be described, while in Chapter 4 the possible causal relationships between climatic variables and thunderstorm rainfall data will be examined on a monthly- basis. The overall goal of this chapter is, therefore, to determine the significant levels of associations among the variables.

In Chapter 6, the spatial variation and distribution of thunderstorm rainfall patterns will be analysed using the gamma distribution technique at each rainfall station. To compare gamma values with actual rainfalls, a GIS method will also be used to visualise the spatial distribution of thunderstorm rainfall amounts over the Sydney region. Then, it will be argued in Chapter 7 that the spatial distribution of thunderstorm rainfall over the Sydney region is largely a function of the interplay and interaction between different physiographic factors such as elevation, proximity to sea, and urban landuse. In order to examine the possible associations between these parameters and rainfalls, a GIS technique and some statistical procedures will be employed to assess the strength and significance of the relationships between variables.

The final chapter, Chapter 8, will, discuss the results obtained and their relationship to those research questions regarding thunderstorm rainfall amounts and distribution in the study area. This will help in returning to the original research questions and unifying the aims of this thesis. This concluding chapter re-states and aggregates all the information from the preceding chapters in terms of the aims of this thesis. Then, based on the results obtained, advantages and disadvantages of all techniques used will be outlined. The last part of the chapter will offer some suggestions for future research.

All computer programs produced in this study to extract data or for other purposes, will be located in the Appendix A. Data that are referenced to different parts of the thesis or presented in summarised form in the text, will also be shown in the associated appendices B to E, in a complete form. CHAPTER TWO Literature Review on Thunderstorm Rainfall 17

CHAPTER 2

LITERATURE REVIEW ON THUNDERSTORM RAINFALL

2.1 Introduction

Chapter 2 brings together a general relevant literature review on thunderstorms. It considers those assumptions which can be related to the thunderstorm activities causing rainfall. The strategy of this literature review is, therefore, to assemble and categorise material which will lead to the development of each question to be tested using various applicable methods. All of the material introduced in the literature review has the purpose, either to develop arguments for use in the analysis to be described later in the following result chapters or to unify these arguments. This chapter is divided into the following sections, beginning with the more general concepts of thunderstorm activities and leading on to the special goals of the thesis.

2) Thunderstorm Characteristics 3) Synoptic Weather Patterns Creating Thunderstorms 4) Climatic Variables and Thunderstorms 5) Physiographic Parameters and Thunderstorm Rainfall 6) Distribution of Thunderstorms in Australia 7) Synoptic Patterns Associated with Thunderstorm Activity in Australia 8) Sydney's Physiographic Parameters and Thunderstorm Rainfall

2.2 Thunderstorm Characteristics

A thunderstorm is defined as a convective cloud or a collection of clouds in which electrical discharges, visible as lighting or heard as thunder, is observed by a person on the ground (Houghton, 1985). Convection systems, which may frequently develop through a considerable depth within the troposphere, are characterised by cumulonimbus clouds and considerable moisture (Lutgens and Tarbuck, 1982; Moran and Morgan, 1991). These are all products of a huge convection system in the atmosphere which can be identified by the towering cumulonimbus cloud (Figure 2.1). Although a thunderstorm cell is defined as a unit of convection circulation, thunderstorms may be composed of single or multiple cells (Oliver and Fairbridge, 1987). CHAPTER TWO Literature Review on Thunderstorm Rainfall 18

Basically, vertical motion in the atmosphere is the key to many of the characteristics of a convection system. Upward motion results in expansion, cooling, and eventual condensation of the water vapour in a stream of air (Wood, 1985). The release of latent heat is often an important factor in accelerating the convection by increasing the buoyancy (instability) of the air (Wallace and Hobbs, 1977). Therefore, the prime prerequisites leading to the formation of thunderstorms are high humidity, high temperatures, an unstable atmosphere, suitable upper wind structure and a lifting mechanism to initiate convective activity.

Figure 2.1 Schematic representation of a thunderstorm cell (Based upon Bryant, 1991).

2.2.1 Life-Cycle of a Single Thunderstorm

Fairbridge (1967) suggested that often a thunderstorm is a small-scale system which affects a relatively small area and is short-lived. The life-cycle of such a thunderstorm cell was summarised by following the three stages that a typical thunderstorm undergoes in its life- cycle.

In the first stage - the developing stage or growing stage - the rising air may cause small cumulus clouds to appear, under daytime conditions of unequal heating, particularly during summer, when convection currents can develop fast. During this stage, strong vertical updrafts (upward moving air) occur throughout the cloud and consequently no precipitation reaches the ground (Tapper and Hurry, 1993).

In the second stage or mature stage, the most important changes take place inside the convection system. Some water droplets begin to freeze, which sets off important drop- growing processes. In this stage, both upward and downward motions occur in the CHAPTER TWO Literature Review on Thunderstorm Rainfall 19 convection cell which reaches its maximum development. The mature stage of the thunderstorm cell may, therefore, be accompanied by violent effects near the earth's surface. These include squalls, often hail and torrential rainfall (Neiburger et al., 1982). Precipitation from the mature thunderstorm is intense and composed of large raindrops, literally a cloudburst (Critchfield, 1983).

In such a situation, the updrafts of a mature thunderstorm produces rain drops through the condensation of moist air which cools as itrises. Whe n rain drops become too large to be supported, they fall, however intense updrafts of a severe storm can suspend huge amounts of rain before releasing a deluge onto the ground. Such rain can reach an intensity of more than 200 millimetres per hour, provided the environment is humid enough to feed sufficient moisture to the storm. Occasionally, these thunderstorms become storehouses for precipitation leading to flash floods (Lilly, 1986, 1990).

(a) Cumulus stage (b) Mature stage (c) Dissipating stage

Figure 2.2 Three stages in the development of a thunderstorm (After Lutgens and Tarbuck, 1982).

In the final stage - the decaying or dissipating stage - the thunderstorm enters the dissipating stage and the updraught currents disappear entirely and the air motion in the convective cloud becomes mainly downwards. During the dissipating stage the thunderstorm cell loses its supply of moisture and energy and disintegrates, but the thunderstorm will continue to exist if new cells are added at its margins (Bradshaw and Weaver, 1993). When the downdraft weakens, the intensity of the rainfall decreases andfinally stops . Surface weather conditions soon revert to their pre-thunderstorm stage. Figure 2. 2 illustrates the three stages in the life-cycle of a thunderstorm.

2.2.2 Complex Thunderstorm Systems

According to the above suggested mechanisms, the life history of a thunderstorm cell used to be classified into the three stages, however, within the atmosphere, in the three-dimensional flowfields associated with a thunderstorm, it is not always possible to distinguish one stage CHAPTER TWO Literature Review on Thunderstorm Rainfall 20 from the other. The complexity occurs when all three stages of development take place in close proximity and nearly simultaneously. In such circumstances, thunderstorms organise themselves into a group of cells where each one is at a different development stage at a specific instant. This multicell stage of a thunderstorm can become severe when it causes surface damages (Oliver and Fairbridge, 1987).

Supercell thunderstorms are also large thunderstorm systems comprising a number of cells probably each at different stages of development (Doswell and Brooks, 1993) which can produce violent and severe weather conditions. The term 'supercell' wasfirst used for such thunderstorms by Browning (1962). Later, the use of radar, scanning the thunderstorms in both the horizontal and vertical planes, has allowed a greater understanding of their structure both in terms of two-dimensional cross-sections and more recently of three-dimensional models. Generally, the supercell thunderstorm is defined as a large and violent storm dominated by one huge cell or supercell in a mature stage of development, which may persist in a steady state for hours, emphasising the fact that such thunderstorms are frequently asymmetric both in shape and in the distribution of their weather elements. These supercell thunderstorms are more highly organised, larger, more persistent and more severe than all other types of thunderstorms (Musk, 1988). The structure of a supercell thunderstorm (a) and its an idealized plan view (b) is shown by Figure 2.3. CHAPTER TWO Literature Review on Thunderstorm Rainfall 21

-OVERSHOOTING TOP 1

BACK-SHEARED, AHV1V MAMMATUS \' '

FLANKING LINE

STORM MOTION

(a) Schematic visual appearance of a supercell thunderstorm

Anvil edge

Light rain Hill Moderate - heavy rain |HI Small hail | Large hail T Tornado

Flanking Overshooting top line

Ob) Idealized plan view of a supercell thunderstorm

Figure 2.3 Schematic visual appearance and an idealized plan view of a supercell thunderstorm (Based on the Australian Bureau of Meteorology and The U.S. National Severe Storms Laboratory publications).

In Austraia, the supercell thunderstorm - another basic type of thunderstorm is far rarer and much more violent. Recently, much research has been undertaken in understanding and modelling much more complex thunderstorms, so called 'supercell thunderstorms' (Bureau of Meteorology, 1995). They could be the subject of much study in the future because of their severe weather characteristics, which are notorious for producing damaging hail and tornadoes (Mitchell and Griffiths, 1993). In case of the Sydney region, occasionally the supercell and multi-cell thunderstorms can be introduced by some of the synoptic weather patterns, generally advancing from the south-east and north-east (Armstrong and Colquhoun, 1976). While these kinds of thunderstorms are rare in the region, they tend to be CHAPTER TWO Literature Review on Thunderstorm Rainfall 22 more severe than air-mass thunderstorms, and they can persist for a longer time, up to several days (Bureau of Meteorology, 1995).

These widespread thunderstorms, introduced by supercell systems, are almost always associated with unstable weather systems (for example, lows and troughs) where they may cause rainfall to develop along the region. Perhaps, the severe thunderstorms over Sydney Metropolitan area on 21st August 1971 (Bahr et al, 1973) or thunderstorm cells which advanced to Dapto in February 1985 (Shepherd and Colquhoun, 1985) are specific examples of such thunderstorms causing damaging flash floods. Two similar episodes took place at 18th March 1990 and on 21 January 1991 producing considerable intense rainfall from thunderstorms (Armstrong and Colquhoun, 1976; Mitchell and Griffiths, 1993). There were some evidence that these storms were supercell thunderstorms. Radar and other meteorological data were supportive of the conclusion that the damages were caused by high precipitation supercell thunderstorms (Spark and Casinader, 1995). It appears certain that in warm seasons (spring and summer) the increased influence of such thunderstorm activity is responsible for some of the greatest and widespread severe events by producing intense and high rainfall amounts.

Severe thunderstorms also impose dramatic environment impacts (Dargie, 1994). These thunderstorms produce hailstones with a diameter of 2 cm or more, wind gusts of 90 km/h or greater, tornadoes, or any combination of the above (Bureau of Meteorology, 1993b and Johnson et al., 1995). Severe thunderstorms are also able to produce very high intensity rainfall causing flash-floods (Elliott, 1994). Such floods from thunderstorms, are exacerbated when the storm moves slowly, so that one small area receives most of the rain. However, the largest amount of rain occurs when organised lines of thunderstorms form and move in such a way that several mature thunderstorms pass over the same location within a short period of time. In such instances, record rainfalls and, thus floods are the result (Bureau of Meteorology, 1993 a).

2.3 Synoptic Weather Patterns Creating Thunderstorms

A great deal of knowledge has been added in the past 40 years to our understanding of the initiation of convective systems and their association with different synoptic-scale circulations (Campbell 1906 and Kessler, 1983). It was found that the distribution of thunderstorms world-wide varies from year to year and reflects the overall synoptic patterns and other affecting factors such as: moist warm waters over oceans, and mountains barriers during the main thunder-producing months. Further studies in thunderstorm distribution have emphasised that the synoptic patterns provide suitable conditions in which thunderstorms can develop easily (Atkinson, 1981 and Mortimore, 1990). CHAPTER TWO Literature Review on Thunderstorm Rainfall 23

Barnes and Newton (1986) suggest that large circulation systems, for example migratory cyclones and anticyclones in temperate latitudes, provide the general conditions necessary for thunderstorm occurrence. Such synoptic-scale systems, whose lives range from days to a week, are very important because their winds transport moist air into the continental regions, where the main thunderstorm activity takes place. In addition to the role played by these circulations in carrying heat and moisture horizontally over long distances, their associated regions of organised ascending and descending motions contribute to thunderstorm development. These motions also affect the vertical stratification of temperature and water vapour in ways that lead to selective occurrence of convective storms in restricted regions. Therefore, most thunderstorms are controlled by the broad-scale circulation systems (anticyclones and cyclones) and their systematicrising motions, particularly over land areas adjacent to the western sides of the oceans in subtropical and temperate latitudes.

In the USA, early investigators (for example, Carpenter, 1913) found that there is an association between thunderstorms and synoptic weather patterns. Later, Blake (1933) noted that two synoptic patterns are associated with thunderstorms. One pattern consists of air approaching from the south and east and it is traditionally called the 'Sonara storm'. The other pattern brings in tropical air from the south and west and is due to a dissipating 'Chubasco' that has penetrated far enough northward to affect southern California. Regardless of large-scale synoptic weather patterns, the resulting thunderstorms are also associated with air-mass systems.

Over the European continent, the origin of sever thunderstorms is correlated to the slow moving low pressure systems and troughs. For example, in the relatively warm summer of 1992 there were several occasions of heavy convective systems with hazardous weather in central Europe (Kurz, 1993). On many of these occasions, synoptic weather patterns such as: the upper troughs moving slowly eastwards, and shallow depressions corresponding to the troughs and associated fronts were responsible for the development of many thunderstorm activities (Andersson et al., 1989 and Prezerakos, 1989). In England, Prichard (1990) found that thunderstorms which occur during summer nights, after a hot day, may be triggered by cold moving fronts driven by fairly sharp upper troughs. These troughs draw hot continental air into a zone where there is sufficient moisture to fuel thunderstorms. However, in Spain Liasat and Ramis (1989) indicated that most convectional generated rainfalls can be controlled by upper cut-off lows. Over this region, thunderstorms can also be encouraged by warm air advection from the south or south-east at low levels.

These studies are only a few examples from different parts of the world, suggesting that within the broad-scale weather systems, thunderstorm activity can take place. Other synoptic weather patterns on a regional-scale such asfronts, lows, troughs and extreme instability in the free upper atmosphere are also favoured systems for the introduction and creation of CHAPTER TWO Literature Review on Thunderstorm Rainfall many thunderstorms. However, they are probably not necessary nor a sufficient reason for the occurrence of thunderstorms. Other trigger mechanisms for thunderstorm initiation are required.

2.4 Climatic Variables and Thunderstorms

It has been suggested that some climatic variables (for example, both air and sea temperatures) are important in creating or affecting a convection system and, as a result, explain thunderstorm variation, in time and space (Willet and Sanders, 1959; Golde, 1977). For example Lutgens and Tarbuck (1982 p:237) suggest:

'All thunderstorms require warm, moist air, which, when lifted, will release sufficient latent heat to provide the buoyancy to maintain its upwardflight. Althoug h this instability and associated buoyancy are triggered by a number of different processes, all thunderstorms need an unstable atmospheric environment in which the instability can be enhanced by high surface temperatures'.

2.4.1 Air Temperature

Generally, the air temperature which was primarily assumed to be a function of the amount of solar radiation received on the ground, has also been known to be one of the main climatic variables causing convection activity (Critchfield, 1983). It has been already shown that thunderstorms generally occur within moist, warm air-masses that have become unstable through surface heating. Because instability is enhanced by high surface temperature, thunderstorms are most common in the afternoon and early evening, particularly in summer months when uneven heating generates vigorous convection which can lead to the growth of storms in a matter of hours.

More recently, Laudet et al. (1994) used lightning data from the Lightning Position Tracking System (LPATS) to derive a preliminary climatology of lightning in NSW. In contrast with the traditional thunderstorm observation (lightning seen and thunder heard by observers), the LPATS system gives real-time lightning and it is becoming an important tool for thunderstorm observation and forecasting. Using this system, it was found that the spatial distribution of lightning (thunderstorms), is seen to be a temperature-related variable more dominant during summer than during the rest of the year. Over the land in summer the diurnal variation in thunderstorm occurrence closely follows the diurnal temperature variation. Therefore, most thunderstorms develop around midday in the spring and summer months when the potential for convection is usually the greatest and adequate water vapour is available.

Areas of high elevation facing the sun - which obtain much more solar radiation (Benjamin, 1983) and, as a result, have high surface temperatures, have been suggested, affect CHAPTER TWO Literature Review on Thunderstorm Rainfall 25 thunderstorm activity, particularly in summer months (see section 2.6.1). However, surface heating is generally not sufficient in itself to cause thunderstorm activity and, therefore, any another climatic or non climatic factor that can destabilise the air, aids in generating a thunderstorm. This simply means that the air temperature should be considered as one of the factors which is able to create or enhance thunderstorm activity.

2.4.2 Sea-Surface Temperature

Surface heating is another climatic factore that may enhance thunderstorm activity by supplying moisture to feed convection (Ramage, 1972). Studies by Rodewald (1963) and Bjerknes (1963) indicate the importance of variations in sea-surface temperature and its associations with other climatic variables over certain regions, particularly in the North Atlantic. Bartzokas and Metaxas (1994) suggest sea-surface temperature is a fundamental parameter in meteorology and climatology of the Mediterranean region. Research has also been focused on sea-surface temperature patterns and their relationship to rainfall in the tropics by Ichiye and Paterson (1963). Their results indicated positive relationships between sea-surface temperature and rainfall amount. Hastenrath (1984) compared selected dry and wet years for the Sahel in Africa, revealing that wet years are associated with warmer than normal surface waters and indicated that sea-surface temperatures can modify rainfall distribution in time and space considerably. In many circumstances a certain relationship between sea-surface temperature - as a fundamental factor in climatology - and rainfall was therefore settled.

In Australia, the possible association between sea-surface temperature and rainfall, but not necessarily thunderstorms, has been well established. For example, Streten (1981 and 1983) demonstrated that wet years over the Australian continent are associated with warmer than normal sea-surface temperature. Whetton (1990), by correlating the Victorian rainfall and patterns of sea-surface temperature anomalies concluded that increased rainfall in Victoria can be related to warm sea-surface temperature off the north-west coast of the continent. Also Nicholls (1984) documented the exitence of a relationship amongst the SST anomalies, the Southern Oscillation, and interannual fluctuations in the Austalian Tropical Cyclone, in a broad band from south-east Australia through the centre of the continent to the north-west coast (Nicholls and Kariko, 1993).

It was also found that the sea surface temperature can affect the rainfall distribution in time and space. For example, in the Sydney region, O'Mahoney (1961) noted evidence of a correlation between Sydney monthly rainfall and sea temperature at Port Hacking. Later, Priestley (1964) showed that there is a positive association amongst monthly anomalies in rainfall, air and sea temperatures along the NSW coast. Subsequent publications by Priestley and Troup (1966) and Priestley (1970) stressed that some of Sydney's rain comes from CHAPTER TWO Literature Review on Thunderstorm Rainfall 26 onshore winds, the moisture content and instability of which is increased by the warm Tasman Sea. However, they indicated that the correlation coefficient was not high (r = 0.2). Hirst and Linacre (1978) examined connections amongst sea-surface temperature, rainfall and prevailing winds. Their evidence indicated that rainfall and sea-surface temperature are positively connected and also that the incidence of onshore winds can increase rainfall amounts. Correlation coefficients were however low (less than 0.24). Hirst and Linacre (1978) concluded that sea-surface temperature and onshore winds individually can control the rainfall distribution in the region. They also suggested that a warmer sea surface and onshore winds may cause more instability and, as a result, enhance orographic rainfall in the coastal hills or convective rainfall by bringing more moist and warm air to the region. Generally, these studies found that the association between sea-surface temperature and rainfall is small for places which are far from the coast.

Later, Fandry and Leslie (1984) found that easterly flows which are located just off the coast and parallel to the coast can move west over the Sydney area and appear as meso-scale phenomena. Leslie et al. (1987) show that these systems can be enhanced by the topography of the region and by the meridional gradient of sea-surface temperature toward the coast. More recently, Hopkins and Holland (1994) found that the combination of the Great Dividing Range, the cooler coastal land mass, and sea surface temperature gradient provides high zonal baroclinicity favourable for formation and intensification of the Australian east- coast cyclones. For example, a 24-hour rainfall event on the 1 st August 1990 was simulated by Golding and Leslie (1993). They showed that falls of over 100 mm were confined to coast facing slopes of the Great Dividing Ranges from Newcastle in the north to Jervis Bay, south of the Sydney region. The output of the model simulations indicated that the precipitation was enhanced about fourfold, with maxima of 90 mm over the sea, and 120 mm over the mountains. It was also found that over half of the precipitation was of a convective nature. Sea-surface temperature data in the region, were already correlated to the rainfall by Bryant (1983a, 1985a, 1988) who suggested significant relationships between monthly sea level, sea-surface temperature and rainfall at Stanwell Park beach just south of Sydney.

In addition, the effect of the difference between sea and air temperatures on the instability of the atmosphere is a very important factor. Such instability, particularly over the coastal areas, can cause thunderstorm activity. Linacre and Hobbs (1977) supposed that in those conditions, when a warm sea makes the air less stable, afree convection in low winds can be expected. This mechanism, which is a result of the instability in the atmosphere, can lead to the growth of tall clouds which are conditionally unstable. Sometimes convection on a vast scale leads to strong updraughts and turbulence within the cloud, which can cause rain, lightning and thunderstorms. CHAPTER TWO Literature Review on Thunderstorm Rainfall 27

The difference between land and sea temperatures may lead to instability and hence influence the growth of thunderstorms in various ways. For instance, when the daytime heating of the ground along the coastal areas causes high temperatures, especially in summer, cold air-mass flows from a cold to warmer surface, from sea to the land (Gentilli, 1971). This mechanism may enhance the afternoon convection activity. In contrast, when the cold air-mass comes from the cold land and the sea is warm, nocturnal thunderstorms may occur over the sea because of oceanic warmth and thus the presence of moisture in the lower layers of the air. Occasionally, thunderstorms can also be expected, because of the passing of cold air over warm sea or warm air. This is relatively common off the coast of NSW in autumn (Linacre and Hobbs, 1977).

All of these studies have generally shown an association between rainfall and sea-surface temperature or air temperature. Although none have established an association directly between thunderstorm rainfall and sea-surface temperature data, Hirst and Linacre (1978 p. 327) announced, 'a high incidence of onshore winds would enhance convective rainfall, by bringing the moist air together with a warmer sea-surface temperature'. Both moist air and a warmer sea can cause greater instability in the coastal atmosphere and, as a result, increase the tendency for convective rainfall. According to this idea Linacre (1992 p:262) suggests,

'It seems plausible that surface conditions which influence evaporation thereby affect the subsequent rainfall, especially in the case of meso-scale convective precipitation'.

He also assumed that in some circumstances, where high temperatures occur with little wind there is more possibility of convective rainfall.

On the other hand, there are some arguments (Linacre, 1992) that surface conditions have little effect on rainfall, except in special circumstances, for at least two reasons. First, variations in surface conditions are usually not felt beyond a few hundred metres from the ground, which is much below the level at which rain is formed. Secondly, the sequence of evaporation, advection, condensation and precipitation normally takes several days. By that time the airflow has separated rainfall from its source by some hundreds or thousands of kilometres. However, it must be noted that in case of convective activity, the precipitation procedure takes place in a matter of hours or within a day. CHAPTER TWO Literature Review on Thunderstorm Rainfall 28

2.4.3 El Nino / Southern Oscillation

In the past, the El Nino / Southern Oscillation (ENSO) phenomenon was correlated to the rainfall variability. The influnce of ENSO on rainfall has been a matter of investigations by many meteorologiests and climatologists in Australia (Allan, 1985). Recent studies questioned the stability of relationships between the Southern Oscilation Index (SOI) and the summer rainfall during the last century (Allan, 1988, 1989; Suppiah, 1992). The instability in correlation patterns is not only apparent in the Australian region, but it is a common feature in the global scale. Certainly, over the Australian continent, the rainfall data indicate long- term variations. For example, Pittock (1975) demonstrated a dry phase between 1913 and 1945 and a wet phase from 1946 to 1978 over the Australian region. In another example, the summer monsoon circulation features were closely linked to various cycles that include the 30-50 day oscillation and ENSO phenomenon over the region. Suppiah (1992) suggested that the SOI could have a strong influnce on local rainfall. An inspection of Bureau of Meteorology (1988a) data - the average number of annual thunder-days in northern Australia - reveals that the areas having large numbers of thunderstorms, show significant correlation between summer rainfall and SOI. Interestingly, these areas indicate a greater number of thunderstorms during the summer season. In the Sydney metropolitan area, Griffiths et al. (1993) correlated the number of severe thunderstorms with SOI. However, coefficients calculated were mostly very small (0.214). This has a probability of occurrence by chance of 0.007, which indicates that the relationship has litle predictive value. Therefore, in the future, studies on the influnces of phenomenon such ENSO on various thunderstorm rainfall systems would be useful for providing further information.

Despite these arguments, it can be concluded that climatic variables on a regional scale have a major influence on rainfall distribution. Ocean waters adjacent to the coast can provide atmospheric moisture and, as a result, affect the rainfall distribution patterns. Thunderstorms can also be enhanced by warm and moist air from the ocean and high temperature of the earth's surface, largely as a result of differential heating.

2.5 Physiographic Parameters and Thunderstorm Rainfall

Many climatologists (Browning and Hill, 1981; Atkinson, 1983) proposed that distribution of precipitation - over a region in a specific time-scale - is largely a function of the interplay and interaction between synoptic air patterns (at several scales), cloud physics and physiographic factors. The influence of each of these factors upon the distribution, amount and variation of rainfall has been explored both temporally and spatially throughout this century (Fogel and Hyun, 1990; Bonell and Sumner 1992). These studies have generally suggested that topography, proximity to sea, and more recently, urban areas are the most important controlling factors in rainfall distribution and its characteristics. CHAPTER TWO Literature Review on Thunderstorm Rainfall 29

However, it has been suggested that the effect of these parameters in a wide variety of latitudes, climates and weather conditions is not the same (Smith, 1982, 1989). Moreover, in contrast with normal precipitation, which is suggested to be of more modest intensity but of longer duration and covering a large area (Oladipo and Mornu, 1985), convectional precipitation has extremely high intensity, a short-duration nature and affects a comparatively small area with considerable spatial variation at the ground surface. Therefore, it is logical in a new region to correlate each physiographic parameter with thunderstorm rainfall distribution rather than rely upon previously defined relationships.

2.5.1 Topography and Thunderstorm Rainfall

Clear relationships between precipitation amount and elevation are now generally established (Bader and Roach, 1977; Browning, 1980; Hill et al., 1981). The effects of topography on annual and seasonal rainfall distribution have long been recognised by Salter and Cale (1921) and Bergeron (1965). Several objective attempts have been conducted to assess statistically the influence of altitude and other topographic parameters on the distribution of precipitation. In the most general terms, the orographic impact on cloud and precipitation enhancement or inhibition is well known (Pedgley, 1970 and Wheeler, 1990).

Orographic precipitation enhancement occurs in a wide variety of latitudes, climates and weather conditions near terrains of differing size and shape (Schermenhom, 1967; Griffiths and Saveney, 1983; Storr and Ferguson, 1983). Over the long term, areas of high relief experience generally greater precipitation amounts and intensities on their windward sides and near the summits, but often produce a rain shadow on their lee side (Craig, 1980; Atkinson, 1983). Topographic features such as spot altitude, rise, orientation and exposure on rain bringing wind, have been suggested to be important topographic factors influencing rainfall amounts and distribution (Balchin and Pye, 1948). Various authors (for example, Browning et al., 1974; Atkinson and Smithson, 1974) have discussed the nature of orographic rainfall in different geographical areas. All these researchers looked at the effects of topographic features upon precipitation distribution. They found that, generally orographical precipitation occurs over, and occasionally immediately downwind of the relief, in a close association with the origin of weather systems.

There is, however, little discussion in the literature dealing with thunderstorm rainfall related to topographic factors. In contrast with normal precipitation, thunderstorm rainfall was suggested to be spatially heterogeneous and highly time dependent (Sumner, 1988). On a global scale, it was reported that the overall distribution pattern of thunderstorms is influenced by three primary elements; the intertropical convergence zone, the solar heating of land masses andfinally warm ocean currents (Brooks, 1925). But, on a regional basis, it CHAPTER TWO Literature Review on Thunderstorm Rainfall 30 has been suggested that the geographical features which encourage convective cloud formation, and as a result thunderstorm development, are: the land-water boundary, land heating, particularly on summer days, and mountainous terrain particularly north-east facing slopes in the Southern Hemispheric (Fuquay, 1962).

Mountains act as a high level heat and moisture source and as a barrier to prevailing air flow that can enhance convective cloud formation (Chacon and Fernandez, 1985). Because the air near the mountain slope is heated more intensely than air at the same elevation over the adjacent lowlands, this may cause a general upward movement during the daytime and the development of thunderstorm cells. For example, mountainous regions such as the Rockies and Appalachians in the USA, experience a greater number of air-mass thunderstorms than the Plains States. (Lutgens and Tarbuck, 1982). Generally, precipitation systems develop easily in mountainous areas which can also be subject to local thunderstorm development during heatwaves. The forced uplift mechanism can provide the final push which causes atmospheric instability. This itself may release massive potential energy. Such a condition can trigger-off damaging thunderstorms, whilst most parts of the region may remain sunny and cloudless (Mortimore, 1990).

In the past, mountain-generated thunderstorms have been studied in several locations. For example: Kuo and Orville (1973) in the Black Hills; Holroyd (1982) in the northern Great Plains; Klitch et al. (1985) in Colorado; and Banta (1984) in northern New Mexico. All these studies have illustrated that convective activity usually occursfirst i n mountainous areas. For example, in a study of a mountain-generated precipitation system in Northern Taiwan, a radar system was used to investigate the effect of terrain on precipitation systems (Chen et al., 1991). They found that mountains can obtain high heat and keep moisture which both are important climatic factors in the producing of thunderstorm activity. The influence of topography and exposure to moisture sources emerges as the major controlling factor of the thunderstorm rainfall amount and its distribution.

In addition to the complex interaction between local winds, topography was hypothesised through the use of many case studies, to be important for generating localised precipitation. For example, in the USA the important influence of topographic features on the distribution of convective rain has long been recognised (Tubbs, 1972). It has been suggested that orographic features such as hills and aspects to the wind direction, contribute to the development of convective clouds for one or more of the following reasons: (1) topographic features may encourage the instability of a conditionally unstable air-mass; (2) the roughness of the terrain results in a series of vertical perturbations, some of which may trigger the formation of cumulus clouds in a conditionally unstable air-mass; (3) the hills and mountains can act as high-level heat sources due to the differential heating of their tops and of the free air at the same altitudes (Byers and Braham, 1949). CHAPTER TWO Literature Review on Thunderstorm Rainfall Jl

Again, in the United States, the effect of mountainous barriers on the distribution and diurnal variations of thunderstorm rainfall was well stressed by Schermenhom (1967), Wallace (1975) and Mass (1982). These studies found that some thunderstorm cells tend to originate in the same place because of topographic effects, and then may follow lower surrounding topography. For example, Astling (1984) studied a relationship between diurnal mesoscale circulations and precipitation in a mountain valley (Utah in USA). He found that the mountainous terrain of Utah is a region where diurnal signatures are present in precipitation occurrences and in local windfields. This study indicated that summer diurnal precipitation modulations are dependent on elevation, with maximum frequencies of measurable events peaking in the early afternoon at high elevations above 2100 m and nearly three hours later in mountain valleys below 1500 m.

The summer thunderstorms over southern California, which were studied by Tubbs (1972), primarily occur over the mountains. It was also evident that the mountains to the south and east receive many more thunderstorms than those ranges to the north and west. Over the years, summer thunderstorms have also hit parts of the Rocky Mountains harder than any other areas of the United States. A study by Easterling and Robinson (1988) and Easterling (1991) showed that parts of the Rocky Mountains have had the highest number of summer thunderstorms. Furthermore, thunderstorm tracks can be frequently guided by topographic features which may locally enhance precipitation when larger weather systems dominate (Berndtsson, 1989).

In other parts of world, for example, over Nigeria, Balogun (1981) found that the orientation of the maximum thunderstorm activity lines along the south-eastern part of Nigeria follows the orientation of mountain ranges. Therefore, he concluded, that even on a localised scale, the degree of instability, and as a result, the intensity of thunderstorm activity, can be largely dependant on the topographic features.

In the United Kingdom, a considerable amount of summer precipitation has been attributed to localised convectional thunderstorms (Mortimore, 1990). In 1962, Shaw indicated that at some rainfall stations more than 90 per cent of summer rainfall comes under the category of 'thunderstorm'. In this region, topographical features play an important part in the more local nature of thunderstorm development. It was suggested that mountain ranges can set-off thunderstorms in potentially unstable airflows and this development can, in some situations, drift away and further develop and affect large areas of lowland Britain. Fogel and Hyun (1990 p:126) have generally suggested that for summer thunderstorms,

'The elevation effect may be caused by: (1) an increase in the rate of arrival of events or equivalent decrease of expected inter-arrival time; (2) an increase in the mean rainfall per event which in turn TWO Literature Review on Thunderstorm Rainfall

may originate from a change in event structure such as changes in event duration or the dependence between duration and rainfall amount. (3) and a combination of (1) and (2)'.

The incidence of thunderstorms in Australia is also higher in the higher parts of the country. Over much of the Australian continent relatively low relief dictates that weather systems are comparactively unaffected by local topographic factors, but where considerable relief is present, such as along the Great Dividing Range and the tablelands of NSW and Queensland, there can be considerable modifications (Sumner, 1983a). As it was indicated by Linacre and Hobbs (1977), in NSW there is a clear correlation between the frequency of thunder-days and elevation. In such area, the occurrence of thunderstorms and rainfall is very much affected by topographic factors (Batt, et al., 1995). This is because the effect of local topography is a major factor in the triggering action. The spatial distributions of lightning in NSW also support the concept that topography is very significant factor in controlling thunderstorms. This effect is evident in areas surronding the Great Dividing Range and over the nearby ranges (Laudet et al., 1994).

On the other hand, there is some argument in the literature illustrating the fact that the enhancement of thunderstorm rainfall is not exactly associated with areas of pronounced or extreme relief. For example, Castro et al., (1992) have attempted to determine whether the topography of the area where storm formation takes place has an effect on the behaviour of storms. The results obtained from an area located in the north east of the Iberian Peninsula, namely Middle Ebro Valley (in Spain), showed how different storms with different internal structure (unicellular or multicellular) and behaviours were differentially affected by topography.

In the USA, (in the Santa Catalina Mountains near Tucson, Arizona) Duckstein et al. (1973) compared precipitation amounts deduced from winter frontal systems with summertime air- mass thunderstorm rainfalls. They noted that winter precipitation was increased more than four-fold at 2100 metres as compared to that found at 1200 metres. For the same elevation (2100 metres), summer rainfall was not quite doubled.

Also, along the Appalachian region, Easterling (1989 and 1990), showed orography plays a substantial role in determining the thunderstorm rainfall regime at a station. This effect was acknowledged by the U.S. Weather Bureau (1947), where a decrease was noted in daily precipitation intensities for the summer months. Also, Easterling and Robinson (1988) have suggested that the mountain areas of the Rockies is a region with a relatively high probability of receiving small rainfall amounts from thunderstorms. In addition, it has been found that in mountainous areas, because of the complex topography, the spatial distribution of thunderstorm rainfall varies greatly (Fuquay, 1962). Therefore, Smith (1989), suggested that CHAPTER TWO Literature Review on Thunderstorm Rainfall 33 the thunderstorm activity in a mountainous area varies not only with elevation, but also with slope angle, orientation and micro topography.

Some researchers, for example Osborn (1982), think that the importance of topography in enhancing the variations in thunderstorm rainfall distribution for each of the weather types such as, frontal systems or troughs, is not the same. Generally, the distribution of thunderstorm rain in mountainous regions reflects variations in wind direction as different slopes and land surfaces induce thunderstorm activity (Smith, 1975). The final role of topography, as a generator of thunderstorms in mountainous regions, is as an initiator of different convective activity between opposing wind streams in different directions. The relative importance of each of these topographic factors in generating thunderstorms may clearly change from day to day, as weather conditions change. For instance, in a study in the Greater Athens area, Amanatidis et al. (1991) found that during summer months the thunderstorm activity is influenced less by the local topographic features.

Worldwide, the positive correlation of increased thunderstorm activity with altitude is well documented, especially on the windward side of mountains (Spreen, 1947; Reid, 1973). In many locations, the effect of mountains on thunderstorm activity is clearly seen on maps showing the correspondence between patterns of thunder-days and terrain height. The results from many parts of the world indicate that thunderstorm rainfall-relief relationships are also positive because topography plays a large part in the formation of heavy showers and thunderstorms in association with advancing airmasses. Therefore, over long-time periods 'classic orographic' enhancement of thunderstorm activity should provide the most suitable explanation for permanent spatial variations in thunderstorm rainfall amounts.

On the other hand, some investigators have pointed out that because of the localised nature of thunderstorms, topography does not always appear to be an important factor affecting thunderstorm rainfall amounts. Thus, attempts should be made by climatologists to look at the thunderstorm rainfall-elevation relationships over both short and long-term periods, for each individual geographic location.

2.5.2 Effects of Proximity to the Sea upon Thunderstorm Rainfall

Proximity to the sea is known to be a very important factor in producing rainfall as well as influencing rainfall patterns (Merva et al., 1976; Berndtsson and Niemczynowicz, 1988). In the literature there have been only few studies that have concentrated on the details of thunderstorm rainfall mechanisms along the coastal areas. However, theoretical studies have generally verified that, in many places, wind circulation may enhance the effects of surface heating and, as a result, initiate a convection system over land near the sea. CHAPTER TWO Literature Review on Thunderstorm Rainfall 34

Estoque (1962) and Findlater (1963), for example, suggested that mechanisms such as surface, upper winds and surface heating together, may produce convection activities adjacent to water bodies, leading to thunderstorms and perhaps rains. Sumner (1983b) found that these meso-scale circulations in the lower troposphere may develop in response to differential surface heating in particular, between the land and the adjacent sea. These mechanisms may cause convectional activity in response to differential solar radiation during the day, depending on the geographic characteristics and weather conditions of each place. Such conditions may produce thunderstorm activity and showers. Simplified and idealised models have been successfully constructed by, for example, Simpson (1964) and Simpson et al. (1977). Further detailed information on the dynamics and theory can be found in Atkinson (1981).

The effect of the distance from the sea on rainfall patterns can be seen on Florida's coastal areas. Perhaps this region provides an ideal example, in which the thunderstorm / local wind systems develop parallel to coast-lines (Byers and Rodehurst, 1948). The close association between the time of occurrence of rainfall and thunderstorms over the region was studied by Gentry and Moore (1954) and L'hermitte (1974). All these studies have emphasised the importance of coastal areas in which the thunderstorm activity can easily be developed.

In Tanzania, in Dar es Salaam, coastal influence on rainfall generation has been illustrated by Sumner (1984) using the spatial correlation of daily data. He found that, because of the development of rainstorms along the coast, rainfall distribution patterns paralleled the coastal area. In other places, locally intense thunderstorm activity, reflecting coastal and orographic influences, have also been highlighted. For example, in Catalonia (in Spain), it was suggested that complex relief and morphology can create unique precipitation areas. One such area is characterised by its vicinity to the sea (Periago et al., 1991). Also, in Israel, a relatively high incidence of convectional rainfall was recognised in the coastal areas (Sharon and Kutiel, 1986). In the western Mediterranean basin, Sumner et al. (1993) found that the heaviest rainfall that had contributed to severe localised flooding, was convectionally generated by upper cut-off lows which were often controlled by proximity to warm Mediterranean waters. More recently, in a study in South Carolina USA, Changnon (1994) indicated that the sea- breeze circulation, which influences convection near the coast, exhibits its strongest influence on heavy rainfalls during the summer months.

On the other hand, it appears almost certain that enhancement mechanisms in coastal areas vary considerably and they may change from winter to summer (Smith, 1985). The effect also depends on the direction of the prevailing wind. In particular, heavy rainfall can occur on steep windward slopes facing the sea, as the hills may trigger thunderstorms and anchor them in the lowlands (Tubbs, 1972). In the USA, Easterbrook and Rogers (1974) concentrated on sea-breeze front thunderstorms along the Georgia coast. They found that CHAPTER TWO Literature Review on Thunderstorm Rainfall 35 generally, thunderstorms occurred in a preferred zone within 50 km of the coast and not exactly near the coastal areas. In this region it was also found that winds blowing parallel to the coast, were seen as important in the generation of thunderstorms.

The proximity factor seems to influence the thunderstorm rainfall pattern, particularly in coastal areas. In these areas, thunderstorms occur in a preferred zone, because in many cases the local winds were thought to be able to increase the instability of the atmosphere. Also the existence of moist winds blowing parallel to the coast had been supposed as important in the enhancing or generation of thunderstorms. In many places decreases may be seen in precipitation as the distance from the coast increases.

By contrast, along the coastal areas, on a smaller scale, the general trend may be spatially reversed particularly during the summer months in which the thunderstorm rainfall patterns display larger variability. Therefore, along the coastal area because of the complex topography and the juxtaposition of land and water over short distances, different thunderstorms rainfall patterns may be experienced in the various seasons.

2.5.3 Impacts of Urban Areas on Thunderstorm Rainfall Distribution

Several climatic studies during the past 100 years have shown that cities develop their own special internal climate, being warmer and less windy than rural areas (Chandler, 1965; Oke, 1977). It has also been established in numerous empirical studies, for example Landsberg (1962) and Oke (1979), that city meso-climates are markedly different from those over surrounding, more natural areas. Moreover, a few key climatic studies in the past 30 years such as Smith (1975), Huff and Vogel (1978) and Lee (1984) found that cities may also produce effects on clouds and precipitation that extend several kilometres out from the city. For several years, the reality of urban effects on precipitation was the subject of considerable debate (Changnon, 1969; Atkinson, 1968, 1971 and Landsberg, 1981). The possible effects of urban areas on precipitation have recently received more attention, particularly in regard to the incidence of short-duration heavy rainstorms of convective origin (Bradshaw and Weaver, 1993). A great deal has been written about the influence of urban areas on summer rainfall distribution and thunderstorm activity in the last decades. For convenience, this section will review thefindings of some investigators throughout the world.

A review of research in the USA concerning the modification of rainfall by urban areas reveals that research was quite poor before Changnon (1968). The 1968 reporting of the LaPorte precipitation anomaly by Changnon (1968) drew a great deal of attention and focused national interest in the USA on the subject of urban effects on precipitation. Results of this study showed about 31 per cent increases in warm season rainfall, more days with moderate to heavy rainfall, 38 per cent more thunderstorms, and even 246 per cent more hail days on downwind of Chicago compared to the surrounding areas. CHAPTER TWO Literature Review on Thunderstorm Rainfall 36

The interest initiated by this study led to intensive climatic studies of other American cities by Changnon (1969), Huff and Changnon (1972 and 1973). In general, relatively strong evidence of urban effects was found in the precipitation distributions for St. Louis, Chicago, Detroit, and Washington (Harnack and Landsberg, 1975). Although in some cities (for example, Indianapolis and Tulsa) evidence was weak, the urban effect appeared to be more pronounced in summer than in winter and usually maximised 50 to 35 miles downwind of the city centre. However, effects were identified within the city also, at Chicago, Detroit, Washington, and New Orleans (Changnon, 1978).

In the USA, concern over the complex problem of urban effects on clouds, precipitation, and related severe weather phenomenonfinally led to a major investigation of the prior research of the Metropolitan Meteorological Experiment (METROMEX) at St. Louis (Changnon et al., 1971). This project was the world'sfirst majo rfield-research progra m with a 5-year field experiment, and it was sponsored by several government agencies, universities and institutes. This study attempted to clearly establish how, when, and where an urban area affects atmospheric behaviour, especially convective rainfall. As a result, summer thunderstorms and associated rainfall have been found to be 25 per cent higher in urban areas.

Other studies such as Huff (1975); Huff and Vogel (1978); and Changnon (1973 and 1978) embraced a variety of cities in different climates of the USA. They mostly reported 5 to 30 per cent local increases in rainfall amounts. These findings were supported by Landsberg (1981) and Changnon and Huff (1986). These studies generally found the urban effects on summer weather conditions included greater convective activity, more thunderstorms, 10 to 30 percent more precipitation, and a greater incidence of hail along storm paths to the lee of the urban area, than over adjacent rural land. According to Huff and Changnon (1986) in urban areas, not only the probability of stormy rainfall is high, but also very heavy rainfall can occur in the late afternoons or nights because of additional urban heating of the lower troposphere.

In the United Kingdom, the possible effects of urban areas on precipitation have been attracting increasing attention in the past years, especially with regards to the incidence of short-duration heavy thunderstorms. Parry (1956) and Barnes (1960) attempted to study single storms over Reading and the Midlands areas, respectively. These studies were the first two examples of the climatological approach to the urban rainfall problems in UK. Although, there was a lack of sufficient data because of less dense rain gauge network, their results indicated the effect of urban areas on the distribution of thunder rain.

Later, Atkinson (1968 and 1969) showed the maximum in thunder rainfall over the central part of London in summer in a feature associated with warm frontal storms. He strongly CHAPTER TWO Literature Review on Thunderstorm Rainfall 37 attributed this pattern to higher daytime temperatures and increased air turbulence over the city centre. In 1971, Atkinson studied a storm formed about 20 miles west of London. He found that when this storm passed over the city, there was rapid cloud growth which resulted in a maximum of precipitation over the local urban area. This case study indicated that moving storms can be influenced by the warm moist air of urban areas.

Atkinson (1977) further demonstrated that convective rainfall over London was enhanced by the presence of the urban heat island. He found that urban-producing heat can increase the incidence of thunder rainfall and thunder itself, especially in the hilly areas of the city. This study suggested that cities which develop an urban heat island also attract thunderstorms, because the heat island tends to initiate updrafts over the city, which then draws in any thunderstorms developing in the area.

In the other parts of the world, the study of urban effects upon rainfall distribution have been followed by several researchers, using available climatological data. For instance, in India, Khemani and Murty (1973) used the rainfall data of three stations in the region downwind of the urban industrial complex at Bombay, and of two stations in the nearby non-urban region. They found that, with respect to the non-urban region, the region downwind of the urban industrial complex recorded an increase of rainfall by about 15 per cent. They attributed this increase in rainfall amount to the high level of industrialisation.

On the other hand, the findings of a number of studies - which were about the relationship between precipitation and urban-dome - have indicated divergent viewpoints. For example, in a study of three Japanese cities (Tokyo, Osaka, and Nagoya), Sekiguti and Tamiya (1970) have noted that it often happens that no rain has been observed in the big cities, but in their outskirts, fairly good amounts of rain (have been) measured. Also, Tabony (1980) in a study of rainfall trends over London, indicated that any feature of rainfall patterns could not be attributed to urbanisation. However, he has not rejected the effect of urban areas upon the frequency of high-intensity, short-duration rainstorms during summer. On the basis of the above-mentioned evidence, the following points appear worthy of emphasis as concluding remarks:

1) Generally, it was found that urban areas can affect incoming solar radiation changing albedo rates and heating processes, so that in the day, there is a greater take-up of solar radiation in the city than in its surroundings (Auer, 1978). Some climatologists such as Vogel and Huff (1978) think that cities decrease wind speeds and, humidity rates but increase cloudiness and precipitation amounts.

2) The materials used in buildings, paved surfaces and the multi-faced nature of the rough urban surface, not only make for increased opportunity for absorption of heat, but also CHAPTER TWO Literature Review on Thunderstorm Rainfall 38 increase heat storage (Henry et al., 1985). The result is that urban areas are appreciably warmer that their surrounding rural areas during the day. This produces the so-called 'heat island' phenomenon which leads to rising vertical motion over the cities and subsequent convectional activities (Hane, 1978). Therefore, cities as warmer locations are often also areas of enhanced thermal process. The magnitude of the urban 'heat island' has been shown to be proportional to city population size for European and North American cities (Oke, 1979).

3) Atmospheric pollution may also increase rainfall (Landsberg, 1962). As a result, it was also suggested that thunderstorm activity may increase, if it is reinforced by increases in the amount of atmospheric aerosols such as smoke from bushfires, pollution or thermonuclear devices (Changnon and Huff, 1973).

4) Finally, there is considerable evidence that the thermodynamic effects of urban environment upon precipitation processes is a very important factor in the initiation of convection cells, and over large cities, aerodynamic roughness of urban structure may enhance the development of severe weather systems such as thunderstorm activity (Landsberg, 1981).

Briefly, the findings of the above-mentioned researchers established the reality of urban impacts upon anomalies of most climatic elements such as temperature, humidity, and precipitation amount. In urban areas, air temperatures are generally warmer than the surrounding areas, so rural cooler air may be drawn inwards to feed the enhanced convection near the urban centre. Although the relative humidity in the city may be lower than that of surrounding areas, but the absolute humidity, which shows the actual moisture in the atmosphere is often higher. This would lead to higher moisture availability for thunderstorms (Lee, et al., 1991). Interactions between thermal conditions and the availability of condensation nuclei sometimes can also trigger convection and the development of thunderstorm clouds. In this situation, there is a slight increase of cloudiness observed and, consequently, more rainfall over or downwind of a city.

2.6 Distribution of Thunderstorms in Australia

In Australia, the distribution of thunderstorms is shown in Figure 2.4 (Bryant, 1991). The greatest intensity of thunderstorms occurs in the tropics and along the Eastern Divide. The highest incidence of thunderstorms occurs in the north, and, not surprisingly, the tropical north of Australia experiences the highest number of thunderstorms. The number of average annual thunder-days increase to above 80 thunderstorm days per year near Darwin (Barkley, 1934 and Crowder, 1995). A similar situation exists over south-east Queensland, where a combination of summer tropical air and the proximity of the Great Dividing Range provide ideal conditions for the breeding of thunderstorms (Colls and Whitaker, 1990). In the eastern CHAPTER TWO Literature Review on Thunderstorm Rainfall 39

part of Australia, along and over the Great Divide the number of thunderstorm-days is more than 40 days per year (Oliver, 1986). A generally low incidence of thunder days, between 10 and 15 thunderstorm per year, can be expected over the southern parts of Australia (Bureau of Meteorology, 1989). It must be noted that 'as currently understood, there are still considerable gaps in our knowledge of the occurrence of such events.

UNOEH «0 D OATS • 0-^9 OATS

£ OATS

60-69 OATS

70-79 OATS

MORE THAN 60 DATS

Figure 2.4 Average annual thunder-days in Australia (After Bryant, 1991).

In Australia, as thunderstorms represent localised areas of instability, their distribution and intensity are dependent upon factors which increase this instability and, as a result, thunderstorm activity. Globally, the overall distribution pattern is influenced by three primary elements: (1) the intertropical convergence zone; (2) solar heating of land masses and (3) warm ocean currents. These conditions provide a favourable environment for thunderstorm development during the entire year (Oliver and Fairbridge, 1987). On a large scale, the common processes that take place from the large scale synoptic weather patterns, for example, a trough line, a low pressure area, or the passage of a cold front, can initiate convection activity and thus thunderstorm rainfall (Kessler, 1986). On a localised scale, the initiation of thunderstorms may be caused by the local physical environment. For example, conditions of instability may be reached when air is forced over a mountain, or the unequal heating of the earth's surface which can appear as a 'trigger action'. In the following sections the role of each the synoptic weather systems, and physiographic parameters in the initiation of thunderstorms will be examined. Thus, the main aim is the understanding of situations that create thunderstorms. CHAPTER TWO Literature Review on Thunderstorm Rainfall 40

2.7 Synoptic Patterns Associated with Thunderstorm Activity in Australia

This section concerns the large circulation systems that bring about the general conditions necessary for thunderstorm occurrence in Australia. Over the Australian continent the anticyclones which move eastwards across the continent and dominate the weather pattern of the whole of country, are a predominating influence (Linacre and Hobbs, 1977). Generally, the seasonal movement of pressure cells - anticyclonic highs and cyclonic lows and associated troughs andfronts - determines the types of air that are drawn towards the region (Tapper and Hurry, 1993). However, this movement is also related to rain-bearing tropical or sub-tropical maritime air-masses and polar maritime airs which dominate over the continent, in summer and winter respectively. Basic elements in the pattern of pressure distribution and associated airmasses over Australia in summer are shown in Figure 2.5.

Figure 2.5 Basic elements in (a) the pattern of pressure distribution and of associated (b) airmasses over Australia in summer (Tapper and Hurry, 1993).

Thunderstorm formation which has been associated with instability of air-masses - in relation to various physical environment (for example, topography) impacts - can also be related to the atmospheric triggering mechanisms (Ludlam, 1962, 1963). Investigation of the respective roles that may have been played by each atmospheric process in the initiation of thunderstorm activity is a very complex task. Meanwhile, in Australia, literature relating the meteorological situation (synoptic patterns) directly to the occurrence of thunderstorm activity may be categorised into four main classes: 1) Tropical cyclones and monsoon depressions of northern Australia, 2) Eastward moving troughs and lows, 3) Frontal activity, 4) Upper atmospheric activities.

1) Tropical cyclones and synoptic-scale depressions are known as major weather-related causes of thunderstorms along the northern coast of Australia. They are generally small intense low pressure cells, often less than 100 km in diameter, associated with stormy CHAPTER TWO Literature Review on Thunderstorm Rainfall 4J_ weather conditions, extremely strong winds and heavy rainfall (Tapper and Hurry, 1993). These tropical cyclones are defined as intense cyclonic storms that originate over warm tropical seas and develop from December to May in this region. They bring intense rainfall to much of northern Australia during this time.

Barkley (1934) mentioned that a majority of storms in the tropics occur just in those portions of the north-east and north-west coasts that are subject to the incidence of tropical cyclones. For example, Willis Island off the North Queensland coast, records many thunderstorms per annum. Also, between Cooktown and Mackay in Queensland, and from Condon to Onslow in Western Australia severe thunderstorms are recorded each year. Occasionally these cyclones can move to the south or south-east of Australia and cause severe widespread rainfall (Whetton, 1988). In western Australia, the cyclones which reach these coasts are usually almost circular storms with well marked discontinuities of pressure in the air streams such as would produce thunderstorms.

(a) (b)

Figure 2.6 Represents a pre-frontal trough (a), and a line storm associated with an eastward moving trough (b) over south-eastern Australia (Tapper and Hurry, 1993).

It was also suggested that some of the thunderstorms over northern Australia may be associated with the monsoon depressions which are responsible for a significant proportion of the seasonal rainfall (Tapper and Hurry, 1993). These systems are normally located over land, but may be intensified when they move over warm ocean waters. Although monsoon depressions are less organised than tropical cyclones, they can often be strong enough to dominate in the region. According to Barkley (1934) along the north and north-west coasts, where monsoons dominate, the maximum number of thunderstorms can be expected. CHAPTER TWO Literature Review on Thunderstorm Rainfall 42

2) Eastward moving troughs are depressions which form in a low pressure trough. They are not a type of thunderstorm, but the most frequent type of thunderstorms which experienced over Australia (Bureau of Meteorology, 1989) may occur in them. Occasionally, ahead of a trough line or east of a low pressure area, there are areas of uplift. This uplift in itself may be sufficient to trigger a thunderstorm. A trough line moving eastward divides the cooler, drier southern maritime air from the warm, moist tropical maritime air-masses (see Figure 2.6). This trough line can enhance the vertical motions just ahead of the trough which leads to thunderstorm development.

Plate 2.1 Shows a cold front off the South Coast of New South Wales.

In summer time, when the troughs exist between high and low pressures, the moist onshore trade winds, orographically uplifted, bring substantial cloud and rainfall to the coastal regions. For example, on Sunday, 11 January 1981, several population centres in Western Australia experienced destructive squalls associated with one or more severe thunderstorms. According to press reports, thunderstorms were associated with an eastward moving trough. Such a trough is a familiar feature of Australian summers (Kingwell, 1982). Intense downpours were experienced with 24-hour total falls of 34 mm reported during thunderstorm activity (Bureau of Meteorology, Perth, 1981). Also, analysis of thunderstorm activity over Tasmania on Monday, 9th, December, 1985, showed that a trough line over CHAPTER TWO Literature Review on Thunderstorm Rainfall 43

Tasmania was responsible for widespread thunderstorm activity with high rainfall rates of greater than 120 mm per hour (Jessup and Hughes, 1991).

3) Cold fronts are found at the boundary between adjacent air-masses. The advancing edge of cold fronts sharply undercuts warmer air, forcing it torise. This lifting mechanism can cause rapid instability resulting in thunderstorm activity with considerable rainfall along the cold front (Hutchinson, 1970; Linacre and Hobbs, 1977). For example, on 22nd December 1990, the study of synoptic patterns showed that thunderstorms which affected Melbourne, were formed in warm moist air, ahead of the approaching cold front moving eastward (Treloar, 1991). Thunderstorms may develop if the air ahead of the front is conditionally unstable and sufficiently moist (Plate 2.1). These thunderstorms tend to occur in long lines parallel to the front. Sometimes, a continuous wall of thunderstorms may stretch over hundreds of kilometres (Tapper and Hurry, 1993). Generally, fronts can introduce thunderstorm activity, particularly over the south-west of Australia where they are common synoptic weather systems. However, it was suggested by Bryant (1991) that the passage of cold fronts plays a minor role in Australia, compared to the United States, in forcing thunderstorm development.

4) Upper atmospheric activity has also been associated with thunderstorm development over Australia. In the upper atmosphere moving air may somehow affect the lower air currents because of complex dynamic process or in response to the thermodynamic activities into the atmosphere. For example, when an area of upper air divergence forms, surface air may be drawn upwards to replace the diverging upper air. Mclnnes et al. (1992) found that upper- level cold pool systems are usually associated with intense thunderstorm activity. For instance, a widespread thunderstorm activity over the south-east of Australia on the 1st March, 1967 and a localised thunderstorm with more than 50 mm rainfall along the east coast, on 9 February, 1990 were both related to upper shear line and cut-off low development. In these situations, warm air advection, in conjunction with the splitting of the jet stream, were also found to be the most significant factors in thunderstorm development (Mottram, 1967). Occasionally, thunderstorm activity can also be initiated by divergence in the upper troposphere causing local convection and representing a localised area of instability (Bryant, 1991).

In short, during the past decades, it has been stressed that, in Australia, thunderstorms can be introduced by a number of above-mentioned synoptic weather systems which are known to be primarily responsible for many thunderstorm developments. These synoptic weather systems also occasionally provide favourable conditions for the widespread development of severe thunderstorms and associated rainfalls. Essentially, the initiation of thunderstorms in conditionally unstable air requires some initial uplift - referred to as 'trigger action' - which CHAPTER TWO Literature Review on Thunderstorm Rainfall 44 may be caused by one or by the accompanying of these synoptic weather systems, synchronously.

2.7.1 Weather Systems and Thunderstorm Activity in NSW

Weather conditions over south-eastern Australia are usually dominated by the generally eastward movement of successive high pressure systems. Troughs or cold fronts form between the highs and are usually associated with low pressure systems over the southern ocean (Bureau of Meteorology, 1991a). However, these normal weather patterns can also be substituted by other atmospheric conditions leading to the development of thunderstorms. In the past, thunderstorm development over NSW has been associated with the following synoptic weather patterns.

Firstly, it was suggested by Williams (1991) that the above-mentioned common synoptic weather situations can introduce thunderstorm activity over NSW. For example, in summer, when the sub-tropical ridge is generally to the south of NSW with centres of high pressure in the Tasman Sea and the lower pressures over the continent, the predominant humid winds from the Tasman or Coral Sea can lead to afternoon and evening thunderstorms, particularly along the ranges or near the trough. At this time the inland trough - sometimes is known as a 'dry-line' - is a boundary between humid air from the east and dry air from the west. This dry-line advances and retreats with heating and cooling of day and night and is known as a line of cumulus cells (as a meso-scale convection system). In this situation, high temperatures and humid air can enhance thunderstorm activity. On average, a large proportion of NSW summer rain comes from these thunderstorms which are occasionally widespread, if there is some upper disturbance present.

Secondly, in the Australian east coast, including the NSW, thunderstorms can be created by cut-off lows which are described as synoptic-scale cyclonic low-pressure systems (Mclnnes et al., 1992). These cut-off lows have been long recognised as major sources of severe weather in south-east Australia (Holland et al., 1987). They are formed when a low pressure system - normally in the upper air - becomes isolated from the main low pressre system by a high pressure system that isridging rapidly east-wards. This process frequently occurs in the south-east of Australia, particularly along the east coast. The includedfigure show s a sample of cut-off low in the region (Figure 2.7). CHAPTER TWO Literature Review on Thunderstorm Rainfall 45

Figure 2.7 Shows a sample of cut-off low in the region which reaches maximum cyclonic development off the southern NSW coast on 11 July 1987 (From Tapper and Hurry)..

Although there is no accepted general terminology to describe these systems, for convenience cut-off lows were classified into two groups namely: coastal lows and blocking lows. Generally they are synoptic-scale systems extending from the surface to the lower stratosphere. Such systems have been recognised as a major source of severe weather in south-east Australia. When the cut-off has no manifestation at the surface, it is referred to by forecasters as an 'upper-level cold pool'. Such systems are usually accompanied with intense thunderstorm activity and often could be associated with widespread rainfall and flooding (Hopkins and Holland, 1994).

The southerly Buster is also an unusual synoptic situation which is the name given to an intense form of cold front that occure along the coast of NSW (Williams, 1991). These fronts, which often herald the 'cool change' for Sydney, occur mainly in the spring and summer months also. The most spectacular casses occur at the of a hot summer day. The wind turns suddenly from the north to the south and blows with some force, with a change in air temperature of 10 degrees or more. There is not a great deal of cloud associated with the initial onset. The intensity of southerly busters and their special characteristics appear to be closely linked with topographic features (Reeder and Smith, 1992). However, the influnces of these unusual synoptic phenomenan on the nature of thunderstorms have been not documanted. CHAPTER TWO Literature Review on Thunderstorm Rainfall 46

(d) DISTANCE FROM RADAR (km)

Figure 2.8 Schematic of the life cycle of the precipitation area of a MCCs as it would appear on rader in horizontal and vertical cross sections during (a) formative, (b) intensifying, (c) mature, and (d) dissipating stages (From Leary and Houze, 1987).

In NSW, it has also been found that meso-scale convective complexes (MCCs) can introduce thunderstorm activities. This phenomenon wasfirst identifie d in the United States (Maddox, 1980). MCCs are a particular class of meso scale (with length scale between 250 - 2500 km with duration > 6 hours) convective weather systems that occur over the central plains of the USA. These systems are generaly much larger than the individual thnderstorms and lines. In fact, they represent the largest member of the family of convective clouds which produce a large proportion of the earth's precipitation and thus are important from a climatological standpoint. MCCs occur in a variety of forms, however, they have several features in common. For example, they all exhibit a large, contiguous area of precipitation, which may be partly stratiform and partly convective (Houze Jr., 1993). The precipitation CHAPTER TWO Literature Review on Thunderstorm Rainfall 47 area of a MCCs exhibits a characteristic life cycle, which is illustrated schematically in Figure 2.8.

In the regional scale a few cases, which occurred over NSW (for December, 1979, December, 1983 and January, 1990) have been documented by James (1992). These thunderstorm-producing systems occur when an instability over a large region is released. Such a self-sustaining system can introduce a heavy rain event or can develop other severe weather systems. A relatively heavy thunderstorm rainfall can develop, lasting for a considerable time (more than 6 hrs) and often reaching a peak overnight (James, 1992). However, it appears that the NSW MCCs are lessfrequent an d generally on average smaller, have a shorter duration and are less efficient in producing precipitation than those documented over the United States. Other conditions necessary for MCCs development include a broad area of high humidity and instability, and the presence of convergence in the lower atmosphere (Hicks, 1984 and Wilson and Ryan, 1987).

In addition, squall lines were suggested by a number of authors (for example, Williams, 1991) to be an important, but infrequent phenomena in causing thunderstorm activities in NSW. Williams (1948) and Tepper (1950) have discussed in detail the most notable features of a squall line. They have attributed to it the following meteorological characteristics: a line of active thunderstorms moving rapidly; brief but sharp pressure changes; very strong wind gusts; brief wind shifts, and abrupt temperature falls. Williams (1991) noted that a squall line isfrequently parallel , or nearly so, to a surfacefront. When thunderstorms develop ahead of a cold front or a trough-line, conditions are often such that the storms develop along a line and propagate as a linear system. The line of storms can be several hundred kilometres long. These conditions were all relevant to the situation on the 21st January, 1977, and the general' weather conditions appeared to have been favourable for squall line formation. Consequently, the line of storms developed ahead of a cold front moving slowly across south-eastern Australia. It wasfirst observed on Sydney Airport radar 55 km to the south­ west of the airport moving rapidly north eastwards. These thunderstorms did extensive damage to property, one person was killed, and twenty three injured (Morgan, 1979b).

Finally, tropical cyclone and other associated weather features, for example troughs, are also responsible for a few thunderstorms in the region. From time to time, across the coasts of NSW, these systems move southwards and cause widespread thunderstorm activity in the Sydney region. For example, in February 1990, considerable thunderstorm activities were introduced by the influence of a tropical cyclone "Nancy" and related troughs over the NSW and Sydney region especially from 7th to 11th. Associated thunderstorms from this activity produced widespread heavy rainfall over the Sydney region causing flooding in the Metropolitan areas. Occasionally, Tasman Sea lows can move to the south-west toward the Sydney region and cause severe thunderstorm activity. For example, between 5th and 12th CHAPTER TWO Literature Review on Thunderstorm Rainfall 48

November, 1984, Sydney had more than 550 mm of rain from thunderstorms. Earlier, in February 1984, under similar synoptic conditions to November, 1984, a centre, in the coastal ranges 60 kilometre south of Sydney, recorded 515 mm of rain in just six hrs and 706 mm in 12 hours, setting new Australian records (Bureau of Meteorology, 1984).

In summary, in producing large thunderstorm activity throughout the State, it could be suggested that the above-mentioned regional synoptic systems are able to make the atmosphere unstable if there is enough humidity and there is a suitable upper wind structure, and finally, if there is a lifting mechanism to initiate convection activities. Generally, positions of highs and lows over NSW, and also moving troughs and fronts across the State can introduce thunderstorm activity, and these thunderstorms are occasionally widespread, if there is some upper disturbance present.

2.7.2 Thunderstorm Development in the Sydney Region

Thunderstorms in the Sydney region can be initiated by local disturbances which occur as warm, moist air is lifted several kilometres into the atmosphere, producing lightning and, at times, strong winds, hail, and torrential rain (Colquhoun, 1972; Mitchell and Griffiths, 1993) (see Plate 2.2).

Nearly all the synoptic weather systems which can introduce thunderstorm activity over NSW, could probably also trigger more widespread thunderstorms over the Sydney region by several mechanisms such as fronts or troughs. More recently, investigations into the occurrence and distribution of thunderstorms within the Greater Sydney Region were undertaken, in relation to the meso-scale synoptic weather patterns. Three major research works which appeared are worthy of emphasis here. Therefore, this part of the literature is fundamentally based upon the following most recent investigations.

The first research was done by Matthews (1993) who examined the spatial distribution and movement of thunderstorms (from 1965 to 1989) within the region using radar facilities. Thunderstorms were found to occur in a number of synoptic situations which have been classified into both frontal and non-frontal systems. In thefrontal category were pre-frontal troughs, pre-frontal and post-frontal systems, while in the non-frontal category were the inland trough, inland low, offshore low and high in the Tasman Sea. The study of seasonal variation of thunderstorms distribution has indicated about 75 per cent of thunderstorms occurred in warm months - spring (September - November) with 30 per cent, and summer (December - February) with 45 per cent. This study also found that spatial thunderstorm distributions for the above-mentioned synoptic classes were significantly distinct, physically plausible and to some extent internally consistent. CHAPTER TWO Literature Review on Thunderstorm Rainfall 49.

Plate 2.2 Displays thunderstorm development over the Sydney region.

Figure 2.9 presents the anomaly maps of thunderstorm occurrence in the Sydney region, from Autumn to Summer seasons. Generally, in winter, thunderstorms occur in the east of the region, particularly over the Tasman Sea. In contrast, in summer they mainly occur over the land, especially over the elevated parts in the west of the region.

In addition, it was investigated by the above-mentioned study that thunderstorms tracks within the Sydney area are almost always from a westerly direction - starting mostly in early afternoon over higher terrain in the west and reaching over the east of the Sydney region (near and over the City and the Tasman Sea) in the late evening (Figure 2.10). Although there is a general eastwards movement from the early afternoon to the early evening in the diurnal distribution of thunderstorms, there are still isolated thunderstorm occurrences that are independent of time. CHAPTER TWO Literature Review on Thunderstorm Rainfall 50

Anomaly plot [or autumn Anomaly plot lor winter

Anomaly plot for spring Anomaly plot for summer Jl rTTrTWTfTnfTrT

Figure 2.9 Presentation of the anomaly maps - using the Terminal Area Severe Turbulence (TAST) radar data - from Autumn, Winter, Spring, and Summer in the Greater Sydney Region. Note that the shaded areas represent areas of low occurrence of thunderstorms with respect to the all years distributions (from 1965-89), while the unshaded areas represent greater occurrence than the all years distribution (After Matthews, 1993).

Another point of interest was that thunderstorms tracks were found to be largely independent of the synoptic systems (the surface pressure patterns). Matthews (1993) has, therefore, assumed that thunderstorm movement may be related to the dynamics of the atmosphere at all levels, particularly the upper-level atmosphere. Thunderstorms occurrences are also assumed to be related to main topographic factors of the region. This latter point was again emphasised by May (1995, personal conversation). In spite of the topographic controls upon the distributions and movement of thunderstorms, figures 2.9 and 2.10 suggest that thunderstorms also follow the sources of heat and moisture in the region in their seasonal and diurnal distributions. CHAPTER TWO Literature Review on Thunderstorm Rainfall 5/

Figure 2.10 Diurnal distribution of thunderstorm occurrence for the different time periods (local time) in the Greater Sydney Region. Shaded areas represents areas of low occurrence with respect to the all years distribution (1965-89) (After Matthews, 1993).

In the Sydney region, Speer and Geerts (1994) have more specifically presented a synoptic climatology of flash-floods-producing storms for the period 1957 to 1990. 94 'flash flood' events - many of which were associated with thunderstorms - have subjectively been classified into four synoptic-meso-scale groups: 1) Easterly troughs; 2) Pre-frontal systems; 3) Lows; and 4) Post-frontal systems.

1) Easterly troughs events are known to be the most common (39 per cent), and usually the longest in duration due to the quasi-stationary nature of the surface trough and upper-level trough. Two obvious synoptic types within this category are known. Thefirst typ e occurs when the surface trough moves east, from a quasi-stationary position mostly on the western side of the ranges. The second type, referred to as the offshore trough, occurs when a quasi- CHAPTER TWO Literature Review on Thunderstorm Rainfall 52 stationary surface trough in the easterlies is located just off and parallel to, the coast. Shanahan (1968) found that associated thunderstorms with these systems can move over the Sydney region and they can produce heavy rainfalls causingflash floods. Figure 2.11 gives examples of easterly troughs: a) an onshore trough occurred on 25 March 1968, one and a half hours before flash flood rainfalls, and b) an offshore trough occurred on April 1985 two and a half hours before the flash-flood rainfall.

2) Pre-frontal systems (31 per cent) occur typically in a north-westerly flow with a mid- tropospheric short wave to the west. They are typical of late spring or summer and are more common in the afternoon. These systems are usually associated with intense though brief bursts of precipitation. Speer and Geerts (1994) distinguished two types in the pre-frontal category. The first is characterised by a meso-scale surface low ahead of the main front. The second type, with a pre-frontal surface trough through Sydney is more common. Associated thunderstorms with these systems develop in a convergence zone between north-east and north-west winds at the surface and low levels. They are generally common features in eastern NSW over the warmer months producing thunderstorms when high pressure in the Tasman Sea has been advancing moist north east winds. These systems may persist for up to several days introducing thunderstorm activity with heavy rainfalls and, as a result, flash floods in the region. Figure 2.11 shows two examples of pre-frontal systems: c) 14 December 1971, two hours before the flash-flood rainfall, and d) 9 March 1989, before flash-flood rainfall.

3) Lows - which developed from low pressure systems - the majority of events (17 per cent) classified as lows, evolved from a depression developing on an old front in the Tasman Sea. The coastal south easterly winds provided low-level moisture flux. Based on location of the low with respect to the Sydney region, and as for easterly troughs, two types were known.

The first type is characterised by a slow-moving single or complex system of lows to the west of Sydney over south-east of Australia, and was referred to as an onshore low which causes 3 per cent of flash floods. The second type occurs with a single or complex low pressure system over the adjacent Tasman Sea with aridge along the coast, and was referred to as the offshore low (14 per cent). For example, on the 23nd November, 1979, such systems passed over Sydney and the flash-flood was produced from thunderstorms as the surface low moved across Sydney. Generally, lows exhibit the weakest diurnal and seasonal variation. Figure 2.11 gives an example (e) of the offshore low, case of 25 October 1960.

4) Finally, Post-frontal systems are rare (13 per cent) and strongly modulated diurnally and seasonally, towards the warmer periods. They are associated with a southerly change, aligned roughly parallel to the Great Dividing Range. These synoptic categories were also known to be important in producing some of thunderstorms, and as result, introducing flash CHAPTER TWO Literature Review on Thunderstorm Rainfall 53 floods in the Sydney Metropolitan area. Such thunderstorms are the most common phenomenon, especially in spring and summer. An associated synoptic chart for this class is given in Figure 2.11 which shows an example of a post-frontal system (f) causing flash-flood on 18 March 1990.

Figure 2.11 Presents examples of six meso-scale synoptic weather systems causing thunderstorm activity in the Sydney region. The thick broken line indicates the trough on the mean sea level pressure chart (After Speer and Geerts, 1994). See text for details.

Most recently, the spatial distribution of deep convection in the Greater Sydney area was examined by Matthews and Geerts (1995), using an archive of 25 years of radar data located at Sydney's Mascot Airport. This research studied a sets of different characteristic thunderstorm distributions in Sydney under different synoptic conditions. It was found that thunderstorms occur in a number of distinctively different synoptic settings, both frontal and non-frontal. Three types of frontal settings (pre-frontal trough, pre-frontal and post-frontal), and also three types of non-frontal situations (inland trough, offshore low and offshore high) were distinguished. Figure 2.12 gives six examples of selected mean sea-level pressure (MSLP) patterns in which thunderstorms occurred. CHAPTER TWO Literature Review on Thunderstorm Rainfall 54

Figure 2.12 Selected MSLP charts illustrating the six synoptic classes: (a) pre-frontal trough (6/5/85 at 0200 UTC); (b) pre-frontal (16/12/85 at 0100); (c) post-frontal (9/1/87 at 1800); (d) inland trough (8/1/85); (e) offshore low (20/6/85 at 2300); (f) offshore high (14/1/86 at 0700). Note that only the last two digits of MSLP in hPa are shown (After Matthews and Geerts, 1995).

Although there are some distinguishing differences between the two above-mentioned classifications, the differences are of little importance, it can be accepted that all synoptic weather systems are able to introduce thunderstorm activity and therefore introduce their associated rainfalls over the region. In one step further, Matthews and Geerts (1995) compared the findings with data obtained from much more recent thunderstorm detecting facilities, such as, a lightning network and a new automatically radar system, as independent sources. They constructed a thunderstorm density model based upon the new radar data for above-mentioned synoptic conditions. Because of their importance, the thunderstorm density and lightning density maps are given in Figure 2.13 and Figure 2.14 respectively.

The results (normalised storm probability and density maps) indicate that the distribution of thunderstorm activity is not the same for all synoptic situations. Although there are distinct differences in synoptic conditions by which thunderstorm occur, general patterns can be understood in terms of low-level flow, topography and land-sea differences. CHAPTER TWO Literature Review on Thunderstorm Rainfall 55

Figure 2.13 Thunderstorm density model based on new radar echoes, under the following synoptic situations: (a) pre-frontal trough; (b) pre-frontal; (c) post-frontal; (d) inland trough; (e) offshore; (f) offshore high (After Matthews and Geerts, 1995).

The more general findings of these investigations indicate that thunderstorm activity is stronger near the coastline, especially the southern coastline and the northern beaches, due to the coastal topography. Places such as the Hornsby Plateau, Illawarra Plateau and the Blue Mountains are also subject to the highest number of thunderstorms. In winter, thunderstorm cells are generally more common offshore (over the coastal areas and over the Tasman Sea). In summer, on the other hand, thunderstorms are relatively more common over land, except over the Southern Tablelands. Matthews and Geerts (1995 p:133) suggest:

'In many cases for example offshore high, topographically controlled, thermally forced convergence is a primary trigger of convection systems in the region.'

In the Sydney region, although thunderstorms are known to be important components of many weather systems such as activefronts, trough s or squall-lines (Morgan, 1979b), in turn thunderstorms can also be initiated by local disturbances mainly caused by climatic- environmental factors (Linacre and Hobbs, 1977). Broadly speaking, thunderstorms may be, CHAPTER TWO Literature Review on Thunderstorm Rainfall 56 with caution, classified into two general categories in terms of their origination and organisation; 1 - the air-mass thunderstorms (thermal) which are more likely to have an environmental origin, and 2 - big and multi-cell or supercell (Dickins, 1994) thunderstorms (dynamically) which are largely of synoptic derivation (Alford, 1994). Simply, it can be viewed that the region can be overwhelmed by both systems during the thundery seasons of the year.

Figure 2.14 Lightning density for single thunderstorm events based on data from the NSW lightning detection network, in units of number of strikes per km^ per event, under the following synoptic situations: (a) pre-frontal trough (22/12/92); (b) pre-frontal (09/11/92); (c) post-frontal (23/2/93); (d) inland trough (06/01/93); (e) offshore low (06/12/92); (f) offshore high (23/12/92). After Matthews and Geerts (1995).

In brief, the data from Sydney thunderstorms during past decades may indicate that thunderstorm development is the result either of the larger synoptic weather systems or, more specifically, synoptic-meso-scale systems. Both systems are potentially able to provide an unstable environment in which thunderstorms occur and track. Although, development of thunderstorms over the Sydney region varies from month to month and reflects the overall impact of above-mentioned synoptic weather patterns, they may also reflect the effects of climatic factors (discussed in Section 4) and physical environment parameters by which CHAPTER TWO Literature Review on Thunderstorm Rainfall 57 thunderstorms may be developed or controlled. In the next section, the role of physiographic parameters will be examined literally .

2.8 Sydney's Physiographic Parameters and Thunderstorm Rainfall

In the Sydney region there is some evidence indicating the fact that spatial distribution of rainfall can be influenced by some physiographic parameters such as elevation and distance from the sea. For example, Linacre (1992) stressed the effect of landform on rainfall distribution patterns in the Sydney region by showing that isohyets closely paralleled the height contours. However, in terms of convectional rainfall, little, if any work has been done relating thunderstorm rainfall to physiographic parameters such as topography, proximity to the sea and urban centres.

Topography has been suggested to be an important factor in affecting thunderstorm development in the NSW coastal areas (Sumner, 1983a). In terms of topographic effects, Williams (1991) mentioned that places along the Illawarra Escarpment experience very high orographic rainfall, illustrating the effect of local topography and exposure on rainfall. For example, in an area of more rugged terrain, such as the Illawarra escarpment, just south of Sydney, intense rainfalls with much longer duration causes flash floods which may be correlated with thunderstorms (Foreman and Rigby, 1990). In 1983, Cox supposed that in the Illawarra there are pronounced differences in rainfall totals between wet and dry years, but the distribution pattern remains quite stable because of the topographic effects.

In the north-west of the Sydney region, it was suggested that, one of the most regular and predictable types of orographic rain may occur, particularly in warm season conditions, near, over and adjacent to the Blue Mountains (Gentilli, 1971). This happens, because the general meridional alignment of this relief along the NSW eastern margins causes an orographic uplift of the moist air streams which is clearly reflected in the amount of rain. The amount is sharply increased on the windward side, and gradually decreased on the leeward side of mountains. Also, it may happen because the daily heating of the hillsides generates warm upslope winds which continue rising after reaching the mountain ridge-top and trigger deep convection. These convective clouds can produce thunderstorm rainfall in the afternoons over the peaks, or downwind if there is cloud drift. This behaviour is shown clearly in the results of some research (Morgan, 1979a) and the daily thunderstorm patterns of the Sydney region.

However, these mechanisms may also produce statistically verifiable night-time thunderstorm activity over the City and the coast line. It can be supposed that thunderstorms build over the mountains and then travel to lowland areas. For example, during the late afternoon and evening of the 10th November 1976, thunderstorms developed over elevated terrain to the south-west of Sydney and then moved over the urban area (Morgan, 1979a). CHAPTER TWO Literature Review on Thunderstorm Rainfall 58

Rainfall from the thunderstorms was very heavy in some parts of the Metropolitan area. Both the highest rainfall and rainfall rate were reported from Observatory Hill, where 30 mm of rain fell in 11 minutes, a rate of 164 mm per hour (Bureau of Meteorology, 1976).

The topography of the coastal region also plays an important role in enhancing surface convergence near the coast. Holton (1992) found that if orography slopes locally upward in a downstream direction, vortex lines in the lower layer are compressed and the flow in the lower layer must move towards the equator to conserve potential vorticity. Also, Leslie et al., (1987), in modelling east coast cyclogenesis, suggested that without topography no such convergence and convective concentration occurs and the cyclone development is retarded.

In addition, Speer and Geerts (1994) using radar data, presented examples in which quasi- stationary thunderstorm cells developed over relatively high topography in the Sydney region. They concluded that topography would, therefore, aid convective systems to produce higher rainfall totals over higher ground. More recently, using data recorded by the LP ATS system (Laudet et al., 1994) it was found that the spatial distribution of lightning (associated with thunderstorms) is closely related to the topography of the region. The effect of topography was pronounced as flash data showed marked concentration over the mountains of the region particularly on and east of the central part of Range. The results support the concept that topography is a very important physiographic parameter in controlling thunderstorm occurrence.

At the same time, there is some evidence which shows that coastal areas - proximity to sea as an important physiographic parameter -can affect the distribution of thunderstorm rainfall in the Sydney region. Speer and Geerts. (1994) found that the south-easterly winds, oriented by the coastalridging, ca n enhance the low inflow of moisture to the storm. Also, in some postfrontal cases, the coastal ridging is known to be responsible for the stationary convergence zone causing thunderstorms to propagate over the coastal areas (Speer and Leslie, 1994). It was found that, the interaction of fronts with coastal dividing ranges in the presence of more humid air to the east of the ranges, can lead to severe thunderstorms along the south coast of New South Wales (Reeder and Smith, 1992). Another important point is that thunderstorm rainfall of longer duration than 6 hours was recorded in areas close to the coast. (James, 1992). For example, during a series of thunderstorms on 10th and 11th March, 1975, thunderstorms were extended over the coastal areas of the Metropolitan and Illawarra districts and intense rainfalls were recorded in costal areas (Armstrong and Colquhoun, 1976). Meanwhile, some researchers such as Abbs and Physick (1992) believe that the topography of the coastline has an extra important controlling effect upon thunderstorm activity in the region. CHAPTER TWO Literature Review on Thunderstorm Rainfall 59

Finally, it is more likely that, in the region, urban areas affect the distribution of thunderstorm rainfall to some degree. Although there is no data correlating the thunderstorm rainfall patterns to Sydney's environment directly, Linacre (1992) mentioned that there is more rainfall over the City. Investigation of thunderstorms, over Sydney during the past years, has shown that a sequence of thunderstorm cells originating over the elevated terrain may, subsequently, enter the severe stage prior to moving (tracking) over the City (Matthews, 1993). The resultant heavy rainfall, large hail and strong wind gusts can cause extensive damage throughout the Sydney Metropolitan area. Speer and Geerts (1994) studied some of the heavy rainfall from thunderstorms during the period 1957 to 1990. They showed that the slow movement of storms over Sydney causing flash floods, can produce much more rainfall over the Metropolitan areas.

An important question is why the City is a preferred (favoured) area for more thunderstorms. In recent years, Tapp and Skinner (1990) and Tapper and Hurry (1993) examined some aspects of urban climatology and have suggested that in many large cities in Australia, noticeable heating of air over the urban centre occurs relative to adjacent suburban and rural areas at night, particularly in winter. Schwerdtfeger (1982) gave evidence (a map) showing the heat island effect over central Melbourne on a winter night. The warmest part (8.9°), took in the north-eastern end of the CBD, the western part of Fitzory, and much of the suburb of Carlton. More recently, Crowder (1995) in comparing rainfall distributions over Sydney and Melbourne, says that rainfall over the Sydney Metropolitan area varies from less than 700 mm near Campbelltown to more than 1600 mm on the coast just north of Stanwell Park. Also, there are wet spots over 1400 mm near Katoomba and southeast of Hornsby and another one over the centre of the Sydney (CBD) (see Figure 14). He concluded that Sydney is prone to severe thunderstorms and associated very heavy rainfalls. Although in Sydney there is no measured data for the urban heat island over a long time span, the effect of such phenomena upon rainfall distribution is highly reasonable. Several authors have over the past years approached the modification of Sydney's atmospheric boundary layer by thermal and mechanical turbulence. For example, Fitzpatrick and Armstrong (1973 p:18) in a study of effect of the urbanisation on climate in the Sydney area wrote:

'Although the maps of mean maximum and minimum temperature do not reveal any clearly identified effect of the urbanisation, this cannot be taken to indicate that such effect must be small. All investigations of temperature have shown that one heat island effect over cities is best developed under calm, clear sky conditions that favour maximal daytime heating and the development of strong temperature inversions at night'.

The detailed spatial pattern of temperature within the Sydney area was carried out by McGrath (1971) using mobile temperature recording equipment at about 1.00 am. on April CHAPTER TWO Literature Review on Thunderstorm Rainfall 60

24th, 1971. He found a difference of about 5 °C between the City and the outlying rural areas. Although he did not attribute the difference found to a true heat island, a steep temperature gradient was observed between closely built-up areas and those having extensive parkland and open spaces.

In another study, Kemp and Armstrong (1972) examined temperature trends at Observatory, Hill, Sydney for the period from 1859 to 1971. They concluded that there has been no change in maximum temperatures over the past years. However over the same period they indicated that minimum temperatures increased by about 0.6 C - an increase which may possibly be explained as a 'heat island' effect resulting from increasing industrialisation of the City.

A study has also been made by Kalma et al. (1973) in spatial and temporal aspects of energy use by domestic, industrial and commercial sectors and in transport, in the Sydney Statistical Division (SSD), in 1970. They found that there is a great spatial variation in energy use, and it was estimated that intensity of energy use ranged from 1 * 1010 BTU/sq mi/yr in rural areas such as Camden and Windsor to 143 * 10*0 in Parramatta and 382 * 1010 in the City of Sydney. This study indicated that energy use on average days in July was about 20 per cent greater than on average days in January. In this research they also gave evidence showing the intensity of annual energy use across the Sydney region. In areas such as the CBD and surrounding sectors, energy use ranged between 300 and 400 * 10*0 BTU ye/sq/mile. In contrast, in suburban areas further out, the rate was a minimum (1 * 10 ^ BTU ye/sq/mile) as shown in Figure 2.15. This high spatial variation in artificial heat generation of Sydney region may be correlated to the urban heat island phenomena. Examples of estimates of artificial heat released from various urban areas (in 1970s) were given in Bridgmam (1990). It was expected that artificial heat released would be increased with the growth of population and density in many cities around the world, including Sydney specifically in central urban area. Identification of the magnitude of these artificial energy use is obviously of fundamental importance in understanding of heat island progress over the Sydney region (Kalma and Byrne, 1976). CHAPTER TWO Literature Review on Thunderstorm Rainfall M.

Figure 2.15 The intensity of annual energy use in the Sydney region (After Kalma,, et al., 1973).

In addition, Linacre and Edgar (1972) reported evidence (visibility and suspended particulate as atmospheric pollution) on the surface configuration of Sydney's heat island. Their work has given improved evidence of the influence of urban development on the climate of the Sydney area. For example, they gave a typical isotherm map showing the temperature difference within the City. In another work by Kalma (1974) it was again shown that, in the Sydney region, considerable spatial and temporal variation in energy use exists. These studies provided evidence that artificial heat generation is a significant factor in energy exchange processes over the urban environment when compared with the surrounding less build-up areas. These studies also discussed the primary processes involved in the formation of the urban heat islands and they generally concluded that artificial heat generation in the Sydney region was largely responsible for the downwind temperature increase over the City. This effect may be maximised by high density building within City centres and may create a heat island (Davey, 1976) with greater cloudiness and, as a result much more rainfall.

Undoubtedly under specific conditions, for example, calm weather conditions, especially in summer, the influence of urban aerosols on cloud condensation nuclei or ice nuclei also appears to be an important factor in the modification of the City's atmosphere. In an extensive study of the climatology of air pollution by Carras and Johnson (1982) and Literature Review on Thunderstorm Rainfall

Leighton and Spark (1995), it was found that the Sydney's climate is subjected to serious pollution events and there are different sources of pollutants emitted to Sydney's atmosphere (Moss, 1965). Industrial and commercial activities, including the usual forms of transport, are amongst the important man-made local sources of emission (Linacre, 1970). Although trends of air pollution levels in Sydney since 1950 showed a general decrease in traditional pollutants, such as dust and sulfur dioxide, because of less use of coal as a fuel in large boilers and for transport, new emerging pollutants including lead, hydrocarbons, and oxides of nitrogen have become of increasing concern in the Sydney region (Paine et al., 1988) (see Figure 2.16).

HYDROCARBONS EMITTED PER 3-2km GRID SQUARE |kg/h|

Figure 2.16 Spatial distribution of nitrogen oxides emissions from all sources in the Sydney region (After Carras, et al., 1982).

All above-mentioned studies showed a great spatial variation in the distribution of some pollutants, such as nitrogen oxides emissions from all sources in the atmosphere of areas located within the City, which has much more pollution than the suburbs. The areas such as CBD, and south of Parramatta river were among the worst. In 1992 once again, the problem associated with the polluted atmospheric environment was highlighted by Taylor. He found that as the Sydney region expands, the air pollution levels increases, and smog moves west CHAPTER TWO Literature Review on Thunderstorm Rainfall 63

as Sydney grows due to the increasing concentration of chemical materials such as nitrogen oxides and hydrocarbons in the atmosphere of the Sydney (see Plate 2.3). Also, Cohen, et al., (1994) indicated that spatial and temporal concentrations of particles such as lead in Sydney is higher than other areas. This study emphasised the important role of motor vehicles in generation of lead in the region.

All aerosol particles, may contribute in formation of cloud condensation nuclei and, as a result, rainfall occurrence. The pollution products may contribute to cloud formation and to changes in the drop-size spectra. As such they can either promote or inhibit thunderstorm rainfall. In case of ice nuclei, particulate matter, especially hygroscope particulates, might initiate the precipitation process in supercooled clouds. A study of rainfall in Melbourne has shown that the average rainfall on Sundays is 1.7 mm/d, but during the week the average rainfall is 2.2. So, weekends are significantly drier than weekdays, presumably because of air pollution volume changes (Linacre, 1992). Despite intensive efforts devoted to the understanding of Sydney's atmospheric environment in the past (State Pollution Control Commission of NSW, 1974 and 1975), the role of pollutants upon the urban precipitation process has so far remained difficult to identify.

Plate 2.3 Shows smog over central Sydney.

Although, in the Sydney region, there has been some progress in understanding meteorological processes in the urban boundary layer in the past years, there are still large CHAPTER TWO Literature Review on Thunderstorm Rainfall 64

areas of uncertainty. Experimental studies of radiation and energy balance at both meso-scale and micro-scale are limited to the few above-mentioned studies, largely as a result of the complexities of the urban atmosphere and different landuse patterns with dissimilar surface materials. More importantly, the relative importance of the urban heat island and surface roughness in modifying the thunderstorm rainfall distribution is still unclear.

In summary, in the Sydney region, the coastline and nearby hills and high elevated areas, it has been suggested, have sufficient encouraging physiographic factors in the development of convective clouds. They may also be supposed to be important parameters in the triggering of thunderstorm activity and, as a result, more thunderstorm rainfall. It is also more likely that the urban area (heat island) may affect the special distribution of thunderstorm rainfall. Because, urban heat island impacts are most noticeable on cool, clear, stable spring and summer evenings, when most convection cells advance over the urban environment and the moist coastal margins.

2.9 Conclusions

The thunderstorm is a much more familiar weather event worldwide. It is a combination of many things all occurring together. Usually, strong gusty winds, vertical currents at higher levels, heavy precipitation with thunder and lighting, even more spectacular scenes and occasionally with distractive consequences, are products of thunderstorm activity. Although there is wide agreement that all thunderstorms require warm and moist air (as the prime gradients leading to the formation of thunderstorms) in the atmosphere, other suitable conditions are needed to increase instability and, as a result, to initiate a convective activity. The roles that may be played by atmospheric instability are very important factors in thunderstorm development.

Synoptic weather patterns such as fronts, lows, troughs and extreme currents in the free upper atmosphere are among conditions which are the main factors responsible for the introduction of the instability in the atmosphere and, thus, the creation of many thunderstorms. However, synoptic weather systems are probably not always necessary nor sufficient conditions for the occurrence of thunderstorms. Other trigger mechanisms for thunderstorm initiation are also important. Over the last few decades, it has widely been stressed that the development, occurrence and distribution of thunderstorms over a region, also largely depends upon some climatic factors (air and sea surface temperatures, for example) and physiographic parameters such as topography and proximity to sea.

As with other places in the world, thunderstorms in Australia can be introduced by the presence of larger-scale synoptic weather systems and the nature of the prevailing air­ masses, for example, lows and fronts. In addition, thunderstorms can be enhanced by CHAPTER TWO Literature Review on Thunderstorm Rainfall 65 physiographic parameters which have been suggested to be more important parameters in the occurrence and the controlling of thunderstorms.

On a regional scale, over the NSW and the Sydney region in particular, researchers have found that the seasonal movement of pressure cells - anticyclonic highs, cyclonic lows and troughs, determine the type and nature of air that is drawn towards the region and, as a result, also in part determine thunderstorm occurrence. Meanwhile, when these air-masses encounter the physical environment with various physiographic parameters, uneven spatial distribution of thunderstorm patterns can be expected.

While synoptic weather systems control the availability of moisture and other gradients needed for thunderstorm occurrence, and thus the actual amount of thunderstorm rain that can fall, site and physiographic characteristics of each specific geographical location could encourage and determine the spatial distribution of thunderstorm rainfall. Both climatologists and meteorologists have emphasised that, in the Sydney region, climatic and physiographic features play an important part in the more local nature of thunderstorm development in different ways as follows:

Firstly, these investigations suppose an interaction between surface heating and source of moisture and its subsequent impact on the thunderstorm activities. This may occur in response to the combination of solar heating of surface layers to a critical temperature, with the forces of air motions associated with synoptic weather systems and climatic factors of the region. In Chapter 4, the close relationship between thunderstorm rainfall and climatic factors will therefore be examined.

Secondly, it was also suggested that hills and mountain ranges can set-off thunderstorms in potentially unstable airflows and these developments can, in some situations, drift away and further develop and affect large areas of lowland in Sydney. For example, mountainous areas, located in the west of the Sydney region, can be subjected to local thunderstorm development during the warm seasons (late spring and summer months). These areas can introduce convection systems and release massive potential energy, triggering-off thunderstorms with intense rainfalls, while most parts of the region remain sunny and cloudless.

In addition, coastal areas, particularly the high ground near windward coasts, are also subject to thunderstorms. Areas near the coast may play a large part in the attraction of thunderstorms deducing heavy rainfalls in association with advancing moist winds from the ocean. CHAPTER TWO Literature Review on Thunderstorm Rainfall 66

Finally, it may be proposed that heat-island effects and other physical and thermodynamic characteristics of the Sydney region, such as pollution and roughness of the City itself, also help in the development of thunderstorm rainfalls over and near the Sydney Metropolitan area or over the City with large industrialised and commercial-residential areas. There is some evidence that thunderstorm rainfall over the City is greater than for nearby suburban areas. Therefore, in Chapter 6, the spatial variations of thunderstorm rainfall in the Sydney region will be studied. Then, existing associations amongst above-mentioned physiographic parameters and thunderstorm rainfall will be examined in Chapter 7.

To sum up, a complex interaction and relationship between synoptic weather systems, climatic factors and the physiographic environment could be responsible for much of the thunderstorm activity in the Sydney region. At times that the larger or meso-scale synoptic weather systems are dominant, widespread thunderstorm activity is pronounced and, consequently, high rainfalls can be expected. Although local climatic and physiographic factors, which influence rainfall distribution, may tend to be masked by the nature of such widespread thunderstorm activity, even in this situation, some degree of regionalization of thunderstorm rainfall patterns can be climatologically distinguished. CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 62

CHAPTER 3

TEMPORAL DISTRIBUTION OF THUNDERSTORM RAINFALL IN THE SYDNEY REGION

3.1 Introduction

The main aim of this chapter is to characterise the general behaviour of thunderstorm rainfall in the Sydney region, over time. Thunderstorm activity can be addressed using a variety of time scales. The pattern of thunderstorm frequency and rainfall amounts in the study area are examined at yearly, seasonal, monthly and diurnal levels using measures of central tendency or dispersion of data. As the distribution of thunder-recording stations reflects the distribution of major cities, suburbs and dams, as afirst approximation , spatially the sampling network of thunderstorms is uneven. This chapter attempts to understand the behaviour of thunderstorms using the better thunder-recording stations (with longer more complete records) in the Sydney region. The NNA technique, defining significant thunderstorms in the Sydney region, was therefore applied to thunderstorm data.

The sources of the data and the choice of data analysis techniques will be explained in sections 2 and 3 respectively. Section 4 identifies the distribution of thunderstorms on a yearly basis. In section 5, the seasonal and monthly distribution of thunderstorm rainfall is analysed. In section 6 the diurnal variation of the thunderstorm rainfall frequency will be determined. In thefinal section ,findings o f the temporal distribution of thunderstorms in the Sydney region, can be discussed.

3.2 Data Used

The National Climate Centre (Bureau of Meteorology, Melbourne) provided the raw data on thunderstorm activity in the Sydney region. Firstly, thunder activities data, which have been recorded on three magnetic tapes were loaded on a PC computer system. In the second stage, the University of Wollongong's main frame computer system was utilised to extract the thunderstorm observations for all thundery days in all thunder-recording stations. They were summarised for different time-scales - such as diurnal, monthly, seasonal and yearly time spans. For each thunder-recording station a thunderstorm day was considered to be a day for which at least one thunderstorm observation was reported.

For this study, according to the World Meteorological Organization (WMO, 1975 and 1988) a thunder observation is defined as the occurrence of a thunderstorm when one or CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 63 more sudden electrical discharges are manifested by a flash of light and a rumbling sound. The traditional method of recording thunderstorm occurrence is to simply note whether thunder is heard during the day. All thunderstorms observed during the last hour (past weather) and also during the observation period (present weather) were counted in this study, according to the present guide of the Bureau of Meteorology, Australia (Table 3.1).

Table 3.1 Represents a detailed description of the codes of present and past weather used in thunderstorm observations.

Past Weather Code Present Weather Code (Thunder was heard) Description (Thunder is heard) Description 9 Thunderstorm with or 17 Thunderstorm without without precipitation precipitation 13 Lightning seen, 95 Slight or moderate no thunder is heard thunderstorm with rain 29 Thunderstorm 96 Slight or moderate (thunder is heard) thunderstorm with hail 91 Slight rain at the time of 97 Heavy thunderstorm observation with rain 92 Moderate or heavy rain at 98 Thunderstorm with the time of observation dust and sand storm 93 and 94 Slight, moderate or heavy 99 Heavy thunderstorm hail at the time of observation with hail

Therefore, for each thunder-recording station a thunderstorm day was considered to be a day for which at least one thunderstorm observation was reported. A thunderstorm may or may not be accompanied by precipitation. In this case, code 17 refers to instances where thunder was heard at the station but no precipitation occurred. Again, if lightning is observed (code 13) without thunder being heard at the station, the event is not considered in this research to be a thunderstorm because this can occur at distances remote from the station, especially at coastal locations with flat topography. Using these codes the daily thunderstorm rainfall amount (more than 0.1 mm) and its frequency data were collected for the period 1960-1993 for each individual thunder-recording station. From these data sets, the mean monthly, seasonal and annual time-series have been derived. It must be noted that rainfall on a thunderday is not necessarily produced all or in part by a thunderstorm. This was addresed by selecting the better thunder-recording stations (see Section 3) and defining a thunderstorm-day for the Sydney region, using data from a set of selected stations. At least three thunder-recording stations had to record a thunderstorm on a thunderstorm-day throughout the region (see Chapter 6).

3.3 Methods Applied

It has been found that many thunderstorms seem to occur in an independent manner in time and space (Duckstein et al., 1973). Also, the results of researchers such as Sharon and Kutiel (1986) and Sharma (1987) have indicated that thunderstorm rainfall values are typically strongly skewed in time and space, occasionally with extremely intense localised CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 62

rainfalls. Therefore, the estimation of the temporal and spatial distribution of thunderstorm rainfall can be biased by these cases, particularly where there is missing data.

In the case of the Sydney region, this bias was assessed by examining the similarities (associations) amongst different stations of thunderstorm rainfall. Figure 1.2 shows the location of the 15 thunder-recording stations in the Sydney region. To find the relative interdependence and associations among all existing thunder-recording stations, the following steps were undertaken.

In the first stage, a computer program was written to find common thunderstorm-days between 15 key stations listed in Table 3.3 (the computer program number 1 is located in Appendix A). These stations were selected because they had data for at least seven years continuously and the fewest missing values for the period 1960 to 1993. The specific days, totalling about 1584 thunderstorm-days, are listed in Table 3.2 (see Appendix B).

Table 3.3 General geographical characteristics of the thunder-recording stations. No of Stations Latitude Longitude Distance Altitude Period stations names from sea inm of data in Km used** 1 Katoomba 33.72 150.30 93 1030 1987-93 2 Richmond 33.60 150.78 54.5 19 1960-93 3 Camden A. 34.05 150.68 47 70 1972-93 4 Bankstown 33.93 150.98 26 9 1969-93 5 Sydney Airport 33.93 151.17 8 6 1960-93 6 Sydney R. Office 33.87 151.20 7.3 42 1960-93 7 Wollongong 34.40 150.88 3.4 30 1972-93 8 Prospect Dam 33.82 150.92 29.3 61 1965-92 9 Liverpool 33.92 150.92 29.3 21 1962-92 10 Lucas Heights 34.05 150.98 16 140 1962-82 11 Bowral 34.48 150.40 47 690 1975-92 12 Parramatta 33.80 151.02 24 60 1967-92 13 Penrith 33.75 150.68 56 25 1967-85 14 Campbelltown 34.08 150.52 31 75 1962-84 15 Picton 34.18 150.62 37.3 171 1965-75 * The nearest distance from the average coastal line. ** See Table 3.2 (Appendix B).

In the second stage, a clustering analysis technique based on the Nearest Neighbourhood Algorithm (Tversky, 1983) was used to group similiar thunder-recording stations. The technique of NNA is discussed at some length by many geographers including, for example, Cliff, et al. (1975), Cliff and Ord (1981) and Unwin, (1981). The technique has been used specifically in regionalising climatic variables and clustering point observations of rainfall (Theakstone and Harrison, 1971). In nearest neighbour clustering, the optimality condition is for the generated clusters to give the least possible distance amongst all possible cluster combinations (Dasarathy, 1991). The computer program for the NNA technique uses an iterative procedure, where at each iteration each data sample is compared to all other sets of randomly chosen seeds. After each iteration the set of seeds with the minimum distance is grouped and the central (centroied) of each cluster is calculated. The iterations are CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall W

stopped when the ratio of overall improvement compared with the previous step, is less than a predefined value. The criterion for clustering is based on Euclidean Distance defined by the magnitude of cluster members for a particular event for each thunder-recording station relative to the cluster centre, where N is the number of stations which could be clustered into K groups. The distance can be written as:

1=1

where D shows the overall distance and D[ shows the distance in cluster /'. Distance in each cluster can be calculated by summing the distance of all the cluster members from the cluster centre. The definition of distance depends on data characteristics. A simple definition is the Euclidean Distance is defined as:

2 2 2 2 Euclidean Distance = •yj(x1 -x2) + (yx -y2) +(*, ~x„) +(yf-y„)

where jq and y\ are the data components. In clustering all thunder-recording stations, all possible physiographic components of each station such as latitude, longitude, distance from the sea, and altitude, were used (see Table 3.3). These parameters were selected because they have been linked to the distribution and variation of thunderstorms over the study area (see Chapter 2).

Using the NNA technique, seven main clusters (A to E) were found representing distinct areas of thunderstorm activity in the Sydney region. As it can be seen from a dendrogram (Figure 3.1), for example, group six includes the Sydney Regional Office, Sydney Airport and Parramatta stations. Katoomba and Bowral stations which are located in mountainous areas, have been classified as two separate groups, as has Wollongong located in the south­ east of the study area.

Locations such as coastal strips or tablelands in the region, with different geographic and physiographic characteristics have already been suggested by Williams (1991) to affect the occurrence of thunderstorm activity. The NNA clustering results support this assumption. For instance three stations, Camden, Campbelltown and Picton, close to each other in the southwest of the Sydney basin, group together, as do Sydney airport, Sydney Regional Office and Parramatta in the centre of the study area. The stations with the longest and most complete record of data collection in each group were selected for further analysis. These stations are listed in Table 3.4. CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 11

Index Of Similarity (Euclidecen Distance Score)

- Katoomba L_

Bowral L_

— Parramatta

"Sydney Airport

1— Sydney R.Offiee i

Liverpool

Bankstown

- Lucas Heigths i

- Camden

Picton

CampbeUtoum i

- Richmond

• Penrith

Prospect Dam I

Wollongong L Figure 3.1 A dendrogram shows the result of the NNA technique in grouping thunder-recording stations.

Table 3.4 Locality of the seven selected stations.

Selected Stations Groups Locality Katoomba C Mountain Bowral A Mountain Richmond D Near-mountain Camden Airport B Far-inland Bankstown E Near-inland Sydney R.O. G Coastal Wollongong F Coastal

3.4 Yearly Distribution of Thunderstorm Rainfall

The broadest time scale over which thunderstorm rainfall varies is the year-to-year variation in total amounts. In many respects this is also the most important as it represents the changes of thunderstorm rainfall frequency over the 34 year period from 1960 to 1993. CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 22

Table 3.5 Yearly variation of thunder-daysfrequency an d thunderstorm rainfall amounts (rainfall is in mm) at 7 thunder-recording stations inth e Sydney region. Stations Katoomba Bowral Richmond Camden A. Bankstown Sydney R.O. Wollongong

Year TDF Rain TDF Rain TDF Rain TDF Rain TDF Rain TDF Rain TDF Rain 1960 11 118 22 136 1961 23 100 22 216 1962 15 86 19 112 1963 22 125 24 231 1964 10 133 17 196 1965 15 167 17 109 1966 32 90 22 133 1967 12 58 13 229 1968 16 96 14 71 1969 34 312 7 21 227 1970 33 264 9 20 26 310 1971 20 128 6 42 14 185 1972 32 161 8 48.4 7 70 20 158 6 1973 28 172 14 182 12 124 17 271 4 54 1974 14 76 7 67 6 69 9 87 3 48 1975 21 170 25 207 14 115 10 111 16 322 1976 25 102 21 105 9 55 17 129 28 273 5 83 1977 21 110 29 193 9 85 12 97 21 120 10 75 1978 23 130 29 186 9 40 9 132 20 113 9 210 1979 24 101 25 87 14 147 11 114 25 115 9 52 1980 12 68 18 70 6 34 10 140 11 113 4 43 1981 16 105 23 135 12 78 12 93 13 110 5 49 1982 19 67 14 111 6 13 3 10 7 46 3 10 1983 18 98 26 115 11 87 20 180 15 125 1984 21 204 26 126 10 75 28 399 23 487 14 311 1985 15 102 33 254 14 93 25 221 21 116 9 98 1986 12 79 11 63.2 11 30 16 195 15 77 5 40 1987 13 190 17 160 22 138 13 73 14 249 15 196 6 115 1988 24 205 21 188 25 186 20 157 22 215 23 274 8 142 1989 22 139 19 122 22 136 3 51 18 196 20 217 7 381 1990 29 350 26 201 34 369 1 9 25 297 27 251 14 202 1991 31 212 27 175 26 170 12 86 16 146 21 261 12 235 1992 30 300 25 140 34 183 7 38 8 23 114 9 43 1993 25 163 41 199 2 18 213 13 161 Total 174 155SI 362 2322 801 5119 212 1562 323 3249 639 6214 155 2352 Average 25 223 20.1 129 24 157 10 74.4 13.5 135 19 183 8 125

Table 3.5 summarises the total annual thunder-days over this time-span. TDF represents Thunder-Days Frequency, and missing years are left as blank cells. The Sydney Regional office and Richmond stations, recorded data continuously over this period. Annual thunderstorm rainfall amounts over the same period are also shown in Table 3.5. The highest mean number of thunderstorm days were observed at Katoomba, Richmond with 25 and 24 thundery-days per year respectively while the minimum average frequency occurred at Wollongong and Camden stations with 8 and 10 thundery days per year respectively.

To see the variations of thunderstorm rainfall over the study area, the coefficient of

variation, cv = (s /x )*ioo was used. Where x = the average annual thunderstorm rainfall, and s = standard deviation. CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 23

This technique was not used to find one-to-one links between thunderstorms and their associated rainfalls at each thunder-recording station. Thus, it could not be assumed that the rainfall falling at a station comes from the storm where thunder in heard. It should be regarded as a descriptive statistical technique, simply representing means and the coefficient of variations of the yearly thunderstorm frequency and rainfall values in different stations (given in Tables 3.6 and 3.7 respectively). The greatest coefficient of variation in yearly thunderstorm frequency (50.2 %) was observed at Bankstown station about 26 Km far from the coast. In contrast, the lowest coefficient of variation in thunderstorm frequency (25 %) occurred at Katoomba located in the Blue Mountains. Also, Sydney Regional Office showed a low coefficient of variation with 27 %, located in the east of the study area about 8 km distance from the Tasman Sea. The highest number of thunderstorm days was observed at Katoomba and Richmond stations located in the north-west of the Sydney region, 93 and 55 km respectively from the ocean.

Table 3.6 Summary descriptive statistics for yearly thunderstorm rainfall frequency, in the Sydney region, from 1960 to 1993. No. Station Name No. of Mean Max. Min. Range Coefficient Years of Variation (%) 1 Katoomba 7 25 31 13 18 25 2 Bowral 18 20.1 27 12 25 28 3 Richmond 34 23.6 41 10 31 34.1 4 Camden Airport 22 9.6 20 1 19 47 5 Bankstown 24 13.5 28 3 25 50.2 6 Sydney R.O. 34 18.8 28 7 21 27 7 Wollongong 20 7.8 14 3 11 45.6

The highest thunderstorm rainfall values occurred at Katoomba (223 mm per year) and 183 mm at Sydney Regional Office. Also later station recorded the highest annual rainfall values (see Table 3.7). CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 14

Table 3.7 Summary descriptive statistics for yearly thunderstorm rainfall amounts, in the Sydney region, from 1960 to 1993. No. Station Name No. of Mean Max. Min. Range Coefficient Years of Variation (%) 1 Katoomba 7 223 350 139 211 34 2 Bowral 18 129 204 67 137 29 3 Richmond 34 150.5 369 58 311 46.8 4 Camden Airport 21 74.4 182 9 173 61.5 5 Bankstown 22 147.7 399 10 389 62.5 6 Sydney R.O. 34 182.8 487 46 441 50 7 Wollongong 19 123.8 381 10 371 82.6

The annual variation of thunderstorm frequency and rainfall for the two stations, with a complete 34 year record, Sydney Regional Office (Figure 3.2) and Richmond (Figure 3.3) were graphed in more detail. It should be noted that these stations are not representative of the thunderstorm variation over the entire Sydney Region.

Sydney Regional Office

Rainfall • Thurxfer-days

Figure 3.2 Yearly variation of thunder-days frequency and thunderstorm rainfall at Sydney Regional Office station (1960-93).

Comparison of the number of years above and below average in all stations shows that there is a fluctuating pattern with high and low years. It was found, on average, for example that in 1969, 1975 and more significantly in 1984, there were some considerable thunderstorm rainfalls in the Sydney region. Distribution of Thunderstorm

Richmond

a & O) « g

Figure 3.3 Yearly variation of thunder-days frequency and thunderstorm rainfall at Richmond station (1960-93).

To gain an appreciation of the longer term variability of thunder-day frequency and rainfall, the data in Figures 3.2 and 3.3 were analysed using Normalised Residual Mass curves (NRM), used by the Bureau of Meteorology (1991a). The NRM can be defined as the accumulated difference between the actual annual thunderstorm rainfall for each year and the mean annual thunderstorm rainfall over total years of the record, divided by the mean of these factors. The NRM for the Sydney region (the average of two above-mentioned stations) from 1960 to 1993 is shown in Figure 3.4. This graph clearly shows sequences of wet or dry thunderstorm rainfall years.

•NRM Rainfall NRM'Thunder-day

100.0 T

80.0 -•

it •o s H

Ol w .a .9 •a •a s I -40.0 -•

-60.0 Figure 3.4 Normalised Residual Mass curves of annual thunderstorm rainfall for three important thunder-recording stations (1960-93). CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 76

The results of this section indicate, the average frequency of thunder-days and thunderstorm rainfall amounts over the mountainous and urban areas adjacent to the coast are much greater than over the inland areas in the Sydney region. This result confirms the work of Griffiths, et al.(1993) who found that the number of severe thunderstorms in New South Wales is increasing. They also give evidence showing a pronounced maxima in the distribution of thunderstorms in recent decades, particularly in the Sydney region. However, these results are not yet valid indications of thunderstorm rainfall variations over the study area, because the increase might be a function of greater population densities and better of data reporting (see Chapter 6 for spatial variations). Overall, the pattern for yearly variation of thunderstorms at different stations are not similar. Comparison of the number of years above and below average in all stations show that there is a fluctuating pattern with high and low years.

3.5 Seasonal and Monthly Distributions

In the Sydney region there is a considerable seasonal variation in thunderstorm rainfall throughout the year as shown in Figure 3.5.

Figure 3.5 Seasonal distribution of thunderstorm rainfall in different stations in the Sydney region.

As expected, in response to the warm environment and unstable atmosphere, thunderstorm activity is greatest during late spring (October to November) and summer (December to February) and weakest during autumn (March to May) and winter (June to August). For the whole region the maximum falls are in late spring and summer, however there are considerable differences. Some stations, for example Wollongong, show a maximum in spring rather than in summer. The graphs for the Sydney Regional Office and for Bankstown are very similar with a peak in summer and a secondary peak in spring. In contrast, Katoomba shows a peak in summer and a much smaller secondary peak in CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 71

Spring. Camden and Richmond stations which are located in inland areas, experience a summer maximum with little thunderstorm activity in autumn and winter.

The range of seasonal values are generally considerable at all stations. Average seasonal thunderstorm rainfall amounts are given in Table 3.8 for select stations. Both Richmond and Katoomba, which are located in the west of the Sydney region receive a high percentage of rainfall in summer rather than in other seasons. In contrast, stations which are located in the east of the Sydney region near the coast, for example Wollongong and Sydney Regional Office, have considerable thunderstorm rainfall (on average 32 per cent) in both autumn and winter seasons.

Table 3.8 Average seasonal thunderstorm rainfall (in mm and %) for selected stations.

Spring (SON) Summer (DJF) Autumn (MAM) Winter (JJA)

Stations Rain mm % Rain mm % Rain mm % Rain mm %

Katoomba 43.5 19.1 141 61.5 32.3 14.2 11.4 5.6 Bowral 32 24.1 60 4 45.6 30.2 22.7 10 7.5 Richmond 46.9 31.1 65.1 43.1 25.8 17.1 13.1 8.7

Camden 27 36.3 33.3 44.8 11.7 15.7 2.4 3.2 Bankstown 43.5 29.4 60.5 41 32.8 22.1 11 7.4 Sydney R.O. 58.5 32 62.5 33.6 40 22 21.5 12.4 Wollongong 39.5 37.5 46.5 32 21.5 18 15.5 12.7 Average 41.5 28.7 67.03 46.32 24.07 16.6 12.12 8.4

In light of these comparisons, it appears that the seasonal response of thunderstorms varies across the region. Despite this variation, in general, late spring and summer are the peak seasons of the year for thunderstorm activity, 72 per cent of all thunderstorms occurred at these times. Thunderstorms also tend to account for a higher percentage of rain-days toward the end of the thunderstorm season.

The monthly distribution of thunderstorm rainfall at different stations is shown in Figure 3.6 in more detail. Thunderstorms occur mostfrequently i n November and December, and least frequently in May, June and July. Generally, the warm summer months, October to March, clearly dominate. However, the peak months for thunderstorm rainfall differs amongst stations. For example, stations such as the Sydney Regional Office, Camden and Bankstown, have peak thunderstorm rainfall in November. Unlike these stations, Katoomba, because it is a mountain station, receives more thunder rainfall in January and February. Both coastal and inland stations exhibit two maxima, in November and February during the course of the year. CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 18

Bankstown -A Camden Katoomba Bowral Sydney R.O •Richmond • Wollongong

Oct Nov Dec

Figure 3.6 Monthly distribution of thunderstorm rainfall in the Sydney region for different stations.

The percentages of the average monthly rainfalls due to thunderstorms are shown in Table 3.9. The highest proportion occurs, on average, in the Sydney region in November with 31.4 per cent, but this rate varies for different stations.

Table 3.9 The percentage of average thunderstorm rainfall to mean monthly rainfall in different stations, in the Sydney region.

Stations Katoomba Bowral Richmond Camden Bankstown Sydney R.O Wollongong Average Month (1987-93) (1975-92) (1960-93) (1972-92) (1970-91) (1960-93) (1973-93) Jan 32.2 30 23.5 14.4 18.3 17.7 8.0 20.6 Feb 29.7 28 18.2 11.4 20.1 21.6 6.0 19.3 Mar 13.3 13 17.3 4.5 15.1 18.6 8.0 12.8 Apr 4.4 5 9.0 7.5 11.3 7.5 1.2 6.5 May 1.5 1.2 4.9 2.4 5.9 4.7 4.1 3.5 Jun 3.3 3 7.1 0.0 4.7 8.2 8.4 5 Jul 0.3 0.3 12.9 4.6 5.4 2.3 1.1 3.8 Aug 7.9 7 9.6 1.4 8.3 11.9 5.4 7.4 Sep 9.1 8.2 22.5 5.3 10.9 13.1 12.4 11.6 Oct 6.5 7 18.9 13.5 13.4 18.9 4.7 11.8 Nov 26.3 25 31.6 18.8 36.7 42.3 27.4 29.7 Dec 27.8 26 29.9 16.7 28.2 22.3 15.0 23.7 Yearly 15.3 12.8 18.0 9.0 15.6 14.9 8.4 13.42

For example, in November, in the Sydney Regional Office about 42 percent of monthly rainfall is obtained from thunderstorms. In contrast, Katoomba receives only 26.3 per cent of its rainfall from thunderstorms at this time of the year. In the winter months, May to July, all stations obtain less than 6 per cent of their rainfall from thunderstorms. Distribution of Thunderstorm Rainfall

3.6 Diurnal Variation

The importance of the diurnal cycle in the Sydney region is clear, and this work is in line with that of earlier investigators. During the past decades, this most regular and predictable diurnal distribution of thunderstorm activity has produced studies all over the world by Brooks (1925), Tubbs (1972), Wallace (1975) and Tucker (1993) for USA, Oladipo and Mornu (1985) in Zaire, Barkley (1934) for Australia, Grace et al. (1989) for South Australia, Treloar (1991) for Melbourne and Williams (1991) and Griffiths et al., (1993) for the Sydney region. Generally, all these investigators indicated that in many parts of the world, during the summer months, convectional processes predominate and can produce a distinct diurnal distribution and variation in thunderstorm activity and an induced precipitation amount mainly in the afternoon or early evening.

This section will describe the diurnal variation of thunderstorm rainfall at various parts of the study area during the spring, summer and autumn seasons. In the Sydney region, there are some problems with the data (provided by the Bureau of Meteorology) as they relate to thunderstorm observations. Firstly, there are insufficient thunderstorms to define a clear pattern for winter. Second, for some diurnal time spans there are not enough data for all seven stations considered. Also, some stations report every 3 hours, some twice and others only once a day. In data-set, there are no thunderstorm observations for 1, 4, 7, 10, 13, 16, 19, 22 and 24 hours, based on the NSW Local Standard Time (LST). These problems with data caused some spatial and temporal gaps in illustrating the diurnal variation of thunderstorm rainfall throughout the Sydney region.

Therefore diurnal patterns are only, shown for Katoomba, Richmond and Sydney Regional Office stations. These are representative of three different geographical locations located in the Sydney region, including mountainous, inland and coastal areas respectively. They also have sufficient data for the purposes of this study in comparison with other thunder- recording stations. Moving eastward across the region, there is a gradual transition to a late afternoon and early evening maximum over the City and coastal areas. This diurnal pattern is shown by data for the three stations located in the region (figures 3.7-3.9). CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall EQ

katoomba (1987 -93)

40

co co un to oo oi *-•

Figure 3.7 Diumal variation of thunderstorm rainfallfrequency for three thunder seasons at Katoomba station.

At Katoomba, over the mountains, the highest thunderstorm activity occurs in the afternoon and a second highest in late afternoon (see Figure 3.7). However, this station has the fewest number of thunderstorms at midnight and early morning.

Richmond 0960-93)

BSj?ig03SD •SbnrH?(DaF) S AiiumflVlAM)

40

Figure 3.8 Diumal variation of thunderstorm rainfall frequency for three thunder seasons at Richmond station.

The Richmond station, which is located near the base of the mountains, has a maximum thunderstorms between 1700 and 1800 LST, but there is an additional one in the late afternoon or early evening, between 1400 and 1500 (see Figure 3.8). CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 81

Finally the Sydney Regional Office station (Figure 3.9) has one maxima in the late afternoon (1700-1800 LST). However, it is clear from the results, that a considerable number of thunderstorms have a nocturnal nature in the western and in the eastern parts of the study area. There are also some differences in diurnal patterns regarding different seasons.

Figure 3.9 Diumal variation of thunderstorm rainfallfrequency fo r three thunder seasons at Sydney Regional Office station.

It can be seen that in all cases at Sydney Regional Office, there is an afternoon-evening diurnal pattern between 17 and 21 LST. In contrast, Katoomba shows a distinct afternoon pattern (14-17) during the summer months. January and February, in particular at Katoomba, seem to be more characterised by an afternoon (12-18) pattern than the other months.

On average, the results of the diurnal variation analysis indicate that in most of the region to the west and central parts of the Sydney region, thunderstorms exhibit a strong late afternoon maximum, particularly during summer months. In the eastern part, thunderstorms show marked diurnal distribution with maxima during the late afternoon and earlier evening, however, thunderstorms may also occur at any time of the day or night, in the Sydney region.

3.4 Discussion

This chapter has presented a temporal distribution of thunderstorm frequency and rainfall amount in the Sydney region for the period 1960 to 1993 using data from the seven selected thunder-recording stations. It is clear from the results that the temporal variation of thunderstorm frequency and rainfall amount over the Sydney region varies from year to CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 82

year. During a calendar year, two seasonal periods with different thunderstorm rainfall characteristics were distinguished: September-March and April-August. It was found that in the September-March period thunderstorm activity predominates, particularly during November through to February.

Thunderstorms are most common in the spring and summer during the afternoon or evening hours. Generally, thunderstorm weather begins in late October and increases quite abruptly in November as was shown by Ryan (1992) who found that in NSW, both severe and less severe thunderstorms are most common in the summer months. There are, however, alternating periods of high and low thunderstorm rainfall amounts during the spring and summer seasons. The results of the diurnal variation analysis indicated that the thunderstorm regime in the Sydney region is an 'afternoon/early evening' type. This result agrees with the work of Griffiths et al. (1993) and Batt (1994).

In the summer months, the maximum values of thunderstorm frequency are observed west of the study area over the mountains. It is evident from the data that the mountains to the west of the Sydney region receive many more storms than lowland areas between the mountains and the coastal areas. Thunderstorms are least frequent in these lowland areas. The areas near the coast and City, however, generally received greater thunderstorm rainfall amounts on average in the same period (1960 to 1993). Explanations for the temporal distribution and variable nature of thunderstorms over the Sydney region are complex. Various mechanisms, which can introduce or enhance thunderstorm activity, have been proposed by different authors for the observed thunderstorm patterns in the region. These may reflect the overall impact of three important controlling factors that include: 1) the synoptic weather patterns, 2) the local climatic factors and 3) physiographic parameters of the Sydney region.

3.7.1 The Role of Synoptic Weather Patterns

In the past, atmospheric conditions leading to the creation of thunderstorms have been the subject of detailed investigation by different authors (Hales, 1978; Winkler, 1988). They have suggested that convection rainfall often develops when synoptic weather patterns and meso-scale mechanisms promote instability in the atmosphere or enhance present unstable conditions. More recently, Konrad and Meentemeyer (1994) found that various synoptic scale features such as lows and fronts, can be connected with heavy rainfall from thunderstorms over the Appalachian region. These synoptic features provide a supportive environment on the meso-scale (10-100 km) for the development of convective cells which produce heavy rainfall as they move and interact with one another (for more details see Chapter 2). CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 8J

In the Sydney region, while local convection as a result of maximum solar heating occurs mainly in the afternoon or evening, the timing of thundery weather still depends upon timing of the passage of synoptic weather systems which may occur during the day or night. These close relationships between thunderstorm activity and weather patterns have been studied in the past by several researchers such as Morgan (1979b); Griffiths et al. (1993); Matthews (1993); Speer and Geerts (1994). These studies indicate that there are some synoptic conditions which favour thunderstorm development in the Sydney region.

Firstly, occasional lows, formed over the north of the Tasman sea, move close to the NSW coast and produce considerable thunderstorm activity, causing very heavy rainfall in the region. Secondly, it was found that in some cases thunderstorms can occur with, or ahead of an active front leading to thundery showers and that these thunderstorms are occasionally widespread if there is some upper air disturbance present (Williams, 1991)

Finally, investigators such as Matthews (1993) and Speer and Geerts (1994) - who more recently studied the formation and structure of thunderstorm events in the Sydney region - have found that the occurrence of thunderstorms in the Sydney region can also be correlated to the synoptic-meso-scale weather systems such as: easterly troughs; pre-frontal systems lows which developed from low pressure systems, and post-frontal systems. In such situations, thunderstorms can move over the region, particularly in the spring and summer months, when the predominant winds are humid north-east or south-east originating from the Tasman or Coral Seas. To relate some specific thunderstorm events to type of synoptic conditions, several examples are given in Chapter 6 in Table 6.4 and Appendix C.

It is also possible that the effectiveness of night thunderstorms in generating rainfall is probably different from daytime thunderstorms, due to the expected differences in humidity and temperature patterns produced by the above mentioned types of synoptic systems. Although the synoptic weather systems are very important in the production and enhancement of thunderstorms, this study will not analyse the effects of these weather patterns upon the thunderstorm activity in the region. The association between synoptic weather patterns and the development of thunderstorms in the Sydney region was discussed in more detail in Chapter 2.

3.7.2 The Effect of Climatic Factors

Synoptic scale weather patterns allow for a broad understanding of thunderstorm rainfall occurrence over the study area. However, the fact that all stations in the Sydney region have at least a diurnal maximum in the afternoon or early evening, suggests that the diurnal distribution of thunderstorms may be controlled by local climatic factors such as air and sea-surface temperatures and air humidity. Explanations for the prevalence of afternoon- CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 8J evening thunderstorms over a region such as the Sydney area (with variable physiography) are complex. Various mechanisms can be proposed for the observed diurnal thunderstorm patterns.

One such group of mechanisms are those based on thermodynamic actions (for example, solar radiation). It seems that most thunderstorms are due to local convection induced by diurnal surface heating particularly in summer months. This occurs because high temperatures are available from the heating of ground surfaces and there is a great deal of moisture in the air, originating from a warm, nearby ocean.

This mechanism can be applied to the Sydney region as a simple explanation for the maximum thunderstorm activity in the late spring and summer. That is 1) a diurnal heating over the land and 2) an abundant supply of moisture from the Tasman Sea, which not only increases the amount of precipitation produced, but also the degree of conditional instability in the atmosphere. This may suggest an investigation of the possible relationships between thunderstorm rainfall and some climatic factors such as, the sea-surface and air temperatures and relative humidity in the region.

3.7.3 The Impact of Physiographic Parameters

More likely, as was discussed in Chapter 2, it is also possible that some types of thunderstorms could be controlled by physiographic parameters such as topographic features, proximity to the ocean and landuse patterns. These parameters play an important part in the more local nature of thunderstorm development in the Sydney region. Previous studies such as Astling (1984), and Smith (1979 and 1985) have already shown that hills and mountain ranges can set off thunderstorms in potentially unstable airflows and these developments can, in some situations, drift away and further develop and affect large areas of lowland. The Sydney region is walled by a mountain range to the west, so, it is possible that the most regular and predictable types of orographic thunderstorm activity may occur in the warm season conditions to the west of Sydney. In this area, the daily heating of the hillsides may generate warm up-slope winds which continue rising after reaching the mountain top. In this situation, the heated airrises to form convective clouds which can trigger deep convective systems. Then, this diurnally forced convection may produce more thunderstorm activity when the thunderstorm is forced to travel some distance away from the mountains towards the coastal areas (Morgan, 1979a).

Alternately the combination of mountainous terrain and moist, warm and unstable air masses may provide the most favourable conditions for thunderstorm development. This is why they are more common over mountains and about the coastal areas particularly in summer. In this case, the violent thunderstorms which can be triggered by, for example, the CHAPTER THREE Temporal Distribution of Thunderstorm Rainfall 85 physical environment also may be enhanced by convergence, a high atmospheric advection or an active front.

The resulting thunderstorms which occur primarily over the mountains and coastal areas, are most important evidence signalling that topography and proximity to the sea can control the temporal thunderstorm occurrence and also associated rainfall distribution. This may reflect a tendency for some of thunderstorms to develop over the mountains, then move eastward towards the coastal area. It seems quite likely that the afternoon maximum in summer thunderstorms along mountainous parts of the study area and the late evening and night-time maximum on coastal areas can be explained by this mechanism.

3.8 Summary and Conclusion

Careful study of thunderstorm rainfall amounts and frequency at the different stations located in the Sydney region, indicated that thunderstorms show marked diurnal and seasonal variation. They are most frequent in the summer months and during the late afternoon and early evening, but there are some recognisable differences between stations in the region. Thunderstorms are most frequent over the west of the region where they may be initiated by the airrising ove r the mountains, and less frequent over the lowland interior of the Sydney region. However, it is evident from the above results that the stations which are located in coastal areas, near the ocean, receive more thunderstorm rainfall than those located inland or in the nearby mountains. This result does not hold, however, for thunderstorm frequency, because the periods of monthly maxima of thunderstorm activity do not necessarily coincide with the periods of rainfall maxima in all of the study areas.

It is obvious from the evidence that there is no single hypothesis that is capable of explaining the nature of temporal variation of thunderstorm activity in different parts of the Sydney region. As the results of this study indicated, the distribution of thunderstorms over the Sydney region varies over time and space. On one hand, these results may reflect the overall effects of some of the synoptic scale weather patterns upon thunderstorm activity which have been widely studied by many researchers in the region.

On the other hand, because the temporal and spatial variation of thunderstorms in the Sydney region is relatively high, they may be affected by the above-mentioned climatic factors and physiographic parameters of the region. Therefore, in Chapter 4, the more detailed associations among additional climatic factors and thunderstorm rainfall will be analysed. Chapters 6 and 7, will then address the spatial variation and distribution of thunderstorm rainfall with an emphasis on the possible associations between thunderstorm rainfall and physiographic parameters of the Sydney region. This will open a new approach for future studies on a regional scale. CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 86

CHAPTER 4

THUNDERSTORM RAINFALL AND CLIMATIC VARIABLES

4.1 Introduction

In Chapter 3, it was argued that the distribution of thunderstorm rainfall in the Sydney region may be affected, to some degree, by some important climatic variables such as air and sea-surface temperatures. The overall goal of this chapter is, therefore, to determine the association of three main climatic background variables, sea-surface temperature, daily air temperatures (maximum and minimum), and mean relative humidity with thunderstorm rainfall amount and its frequency. This will be done using monthly data mainly from Richmond, Sydney Airport and Sydney Regional Office. These stations, having the longest data in the region.

In section 2 the data sources and analytical techniques are given. In the first stage of data analysis, in section 3, descriptive statistics describe and summarise single variables in order to demonstrate the general characteristics of these variables. In the second stage of the data analysis, in section 4, some types of statistical techniques, such as simple correlation techniques, determine the significant levels of associations amongst variables. Section 5 examines the effect of independent variables separately upon the dependent variables, using a stepwise multiple regression technique. Finally the reasons, for statistically significant comparisons, are outlined in section 6.

4.2 Data Sources and Analysis Techniques

To find the possible associations between thunderstorms data and climatic factors, data were obtained from the Sydney Regional Office on a monthly basis. Sydney's air temperature and humidity records for three synoptic stations, namely: Sydney Airport; Sydney Regional Office, and Richmond, were available (this is the most appropriate range of stations with data available for this analysis). The data werefirst describe d on a monthly basis, and were then calculated using mean daily maximum and minimum values (one-half of mean daily maximum plus minimum) from 1960 to 1990.

Tables 4.1 to 4.6 (see Appendix B) show the monthly thunderstorm rainfall frequency and thunderstorm rainfall amounts for the three above-mentioned stations. These stations were selected because: (1) they are the main synoptic stations recording thunderstorm events; (2) they have the longest time-span records; and (3) they have complete records of the CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables -XI

necessary data. These data were restricted to the period 1960 to 1990 because sea-surface temperature was available only for this period.

The sea-surface temperature data was obtained from the Australian Oceanographic Section (Sydney Division). The data was collected at a station off of Port Hacking (34° 05' S, and 151° 12' E) which performs measurement on a semi-regular basis. Most, but not all the years contain observations for each month of the year. Sea-surface temperature observations are usually taken weekly. Therefore, monthly means were calculated by aggregating the weekly data and then averaging these values. Where a single month was not observed, its value was assessed by a linear interpolation and then used as part of the time series.

To analysis the data and to find the possible associations among these data sets, some simple to complex statistical techniques were used. The following methods form the basis for all analysis in the current chapter:

1) A simple correlation technique was employed in order to determine the possible causal relationships between dependent (thunderstorm rainfall) and independent variables (air and sea-surface temperatures) and the extent to which the variables are interrelated. This statistical technique is used when both the independent variables and the dependent variables are measured on a ratio scale.

2) A correlation coefficient test was undertaken for preliminary hypothesis testing. This technique was also employed for the purpose of determining the reliability of the variables.

3) For the last stage of analysis, a stepwise regression was used. By using this technique, independent variables were entered into the equation separately according to the strength as predictors of the dependent variables. Consequently, the regression coefficients provided estimates of the effect of each of the independent variables on thunderstorm rainfall frequency, holding statistically constant the effects of the other variables included in the equation.

4.3 Description of Variables

In this chapter, both thunderstorm rainfall amount and its frequency have been assumed to be dependent variables affected by above-mentioned climatic variables. Addressing this assumption, and to find possible associations between these independent and dependent variables, it isfirst necessary to describe each of these variables using monthly-based averages. This will facilitate analysing, comparing, and measuring the correlations between CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 88 the different variables. Table 4.7 gives simple descriptive statistics for the monthly distribution of thunderstorm data.

Table 4.7 Description of thunderstorm data.

Stations Sydney Airport Sydney R.O. Richmond Statistics TRF TRA TRF TRA TRF TRA Minimum 1 0.1 1 0.1 1 0.1 Maximum 9 265.5 9 334 8 126 Range 8 265.3 8 333.8 7 125.8 Mean 2.1 21.9 2.1 23.4 2.5 20.1 Std. Dev. 1.4 29.5 1.4 34.5 1.7 20 Variance 1.9 868.9 1.9 1187.4 2.9 405 Sum 531 5533 501 5626 561 4568 N 252 252 237 237 227 227 N= The number of months with thunderstorm rainfall (> 0.1 mm) in the sample TRA = Thunderstorm Rainfall Amount, TRF = Thunderstorm Rainfall Frequency

It is evident from Table 4.7 that the total number of thunderstorm rainfalls recorded (1960- 90) at the Richmond station with 561 thunderstorms is greater than the Sydney Regional Office and Sydney Airort stations with 501 and 531 thunderstorms respectively. In contrast, the total amount of thunderstorm rain at the Sydney Regional Office (5626 mm) and Sydney Airport (5533 mm) is higher than at the Richmond station (4568 mm).

4.3.1 Air Temperature

Air temperature was primarily assumed as a function of the amount of solar radiation received on the ground which can be an important climatic factor affecting thunderstorm activity (Critchfield, 1987). Table 4.8 summarises air temperature data on a monthly basis between 1960 and 1990 only for those months having thunderstorms with at least more than 0.1 mm rainfall. The data are measured for the three selected stations.

Table 4.8 lists average maximum and minimum temperatures and extremes for three stations in the Sydney region. It is evident that there are considerable differences in temperature between these three stations. Generally, the highest extremes and variance occur at Richmond, while stations, which are located in the east of Sydney, show less variability. Figure 1.5 shows the average daily temperature for different months in the study area. CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 89

Table 4.8 Means and extremes of temperature at three selected stations.

Line 1: Mean Daily Max. Temperature Line 2: Mean Daily Min. Temperature Line 3: Average Daily Temperature N= Number of months

Stations Mean Min. Max. Range Std. Dev. Variance Sydney Airport N=252 1 22.91 16 29.2 13.2 3.43 11.8 2 14.10 5.2 21 15.8 4.22 17.82 3 18.51 10.6 24.6 14 3.78 14.33 Sydney R.O. N=237 1 22.88 16.2 28.6 12.4 3.17 10.07 2 15.15 7.1 21.1 14 3.73 13.94 3 19.01 11.65 24.4 12.75 3.43 11.77 Richmond N=227 1 25.33 16.6 33.3 16.7 3.94 15.56 2 13.18 2.8 19.5 16.6 4.33 18.83 3 19.26 9.95 25.65 15.7 4.04 16.36

4.3.2 Sea Surface Temperature

In the Sydney area, it has been assumed that the sea-surface off the coast has a major influence on rainfall on a regional scale (Priestley, 1964, 1970; Hopkins and Holland, 1994). This hypathesis has never been tested for thunderstorms alone. Also, it was assumed that the ocean waters adjacent to the coast can provide atmospheric moisture, and as a result, affect the temperature patterns in the region. Table 4.9 gives the average monthly sea-surface temperature. CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables _20

Table 4.9 Monthly and yearly sea-surface temperature data (°C) at Port Hacking. Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Yearly 1960 21.4 22.0 21.7 20.3 18.3 17.0 17.8 17.1 16.9 18.1 19.1 19.0 19.1 1961 20.8 21.7 22.1 18.2 18.5 17.1 15.9 15.4 15.4 18.1 18.0 19.5 18.4 1962 21.6 22.0 21.8 20.6 19.2 17.6 16.7 15.5 16.0 16.9 19.0 19.5 18.9 1963 19.3 22.1 22.2 21.7 19.1 18.2 17.3 15.9 16.5 16.5 18.3 19.8 18.9 1964 21.5 22.3 20.8 20.5 19.2 18.7 16.9 16.9 16.5 17.3 18.8 19.3 19.1 1965 20.5 21.8 21.0 19.3 19.3 18.4 17.7 17.2 16.7 17.5 18.9 19.7 19.0 1966 21.1 21.4 20.9 21.1 19.8 19.0 16.6 16.7 17.6 18.2 19.1 20.1 19.3 1967 20.8 21.7 21.4 21.0 19.4 17.5 16.0 16.4 16.5 17.1 18.3 19.6 18.8 1968 22.2 20.5 22.7 21.4 18.9 17.4 16.0 16.1 16.5 16.1 17.3 20.4 18.8 1969 21.7 19.9 19.9 20.1 19.3 17.4 17.3 16.2 16.0 16.5 20.1 20.3 18.7 1970 21.9 21.4 22.4 21.1 18.3 16.6 15.1 14.8 16.5 17.6 18.5 20.2 18.7 1971 21.9 22.8 20.4 20.3 18.9 17.2 15.9 16.5 16.0 16.4 16.9 19.9 18.6 1972 19.9 20.4 21.4 20.4 19.0 17.8 17.4 16.9 16.6 17.3 19.1 21.8 19.0 1973 21.9 21.6 22.4 20.9 19.4 17.4 17.3 18.3 16.8 18.1 19.4 21.0 19.5 1974 21.4 22.4 21.5 22.4 19.5 17.7 16.7 16.8 16.3 17.3 18.7 20.4 19.3 1975 20.7 21.5 21.9 20.0 19.7 17.8 16.3 17.0 17.7 17.7 18.8 19.7 19.1 1976 21.7 23.1 23.4 20.9 20.6 19.5 18.8 18.6 18.3 17.8 19.6 20.9 20.3 1977 20.6 20.6 21.4 20.9 19.0 16.9 15.4 15.2 17.7 17.8 17.2 19.6 18.5 1978 21.6 22.3 21.3 21.4 20.1 17.6 16.9 16.3 16.2 17.4 18.5 19.6 19.1 1979 21.0 21.5 21.4 20.2 18.8 17.9 17.0 15.7 16.1 16.4 18.5 20.1 18.7 1980 21.4 22.1 22.3 21.1 19.0 18.1 16.2 16.8 16.4 17.0 18.0 19.7 19.0 1981 21.8 23.0 22.6 21.6 20.4 18.1 16.7 16.0 17.1 17.2 18.3 20.6 19.5 1982 21.6 21.3 22.0 20.9 20.9 18.7 17.4 16.6 17.2 18.0 18.4 21.6 19.5 1983 21.0 22.3 19.7 20.4 18.9 17.1 15.5 16.3 16.1 18.3 19.4 20.1 18.8 1984 18.6 22.5 21.9 19.9 18.9 17.9 15.9 15.4 16.7 16.8 20.8 20.5 18.8 1985 21.7 20.6 20.3 21.3 19.2 18.0 16.7 16.2 18.0 18.3 19.3 21.9 19.3 1986 22.5 22.9 21.3 20.4 20.4 17.8 17.2 16.3 16.9 18.4 19.0 19.3 19.4 1987 21.3 23.8 21.0 19.8 19.0 17.7 18.4 16.4 17.0 18.5 18.9 21.1 19.4 1988 20.4 22.5 21.4 19.9 19.5 18.0 18.0 16.0 16.5 19.1 19.6 20.0 19.2 1989 20.5 22.1 21.1 21.0 20.6 18.5 17.3 15.7 16.0 19.4 18.2 18.9 19.1 1990 21.0 22.3 24.1 23.1 19.9 18.2 16.6 16.0 16.8 16.4 16.2 22.1 19.4 Mean 21.1 21.9 21.6 20.7 19.4 17.8 16.8 16.4 16.7 17.5 18.6 20.2 19.1

SST - at Port Hacking

24T 0 22± U 20+ tn 18-- 16-- 14-- + + + + + + 12-Ja- n Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Average Monthly Variation of SST (1960-1990)

Figure 4.1 Average monthly variation of the sea surface temperature in °C. CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 91

Figure 4.1 plots the average monthly variation of sea-surface by calender month over the period 1960 to 1990 (372 months), at Port Hacking, with 34°05 S and 151°12 E for 0 depth. It is clear that sea-surface temperature had a maximum of 22 °C in February and a minimum of 16.40 °C in August.

4.3.3 Air Humidity

Humidity, or the water vapour content of the air, it is suggested, is an important meteorological element both in terms of the development of weather patterns and in terms of the efficiency of living systems such as convection cells (Lutgens and Tarbuck, 1982). Both the absolute humidity and relative humidity are known as important indicators in computing atmospheric moisture amounts in creating or affecting convection systems (see Chapter 2). Many climatologists recommend that the absolute humidity, as measure of independent variable, should be used in correlating air humidity to thunderstorm variations. They regard the relative humidity as more or less meaningless. Unfortunately, the absolute humidity, as an important climatic variable, was not availble in the Bureau of Meteorology sources. Thus, for estimation of absolute humidity, there is a need to have other humidity parameters such as: dry-bulb and wet-bulb temperatures, atmospheric pressure or saturation vapour pressure at the wet-bulb temperature (Abbott and Tabony, 1985). Still, the calculation of absolute humidity is a complex task and there is a variety of methods, using a set of different equations. Without such data, this is very costly in terms of computing time, with difficulty in evaluating of results (Sargent, 1980).

Therefore, in this study the relative humidity was used, which is the maximum amount of water vapour that the atmosphere can hold. The relative humidity of the air at a given temperature is the ratio (expressed as a percentage) of the actual vapour pressure to the saturation vapour pressure. Relative humidity is defined (World Meteorological Organization, 1988) by

U= 100 (e/ed) per cent, where e = ambient vapour pressure in millibars, and ed = saturation vapour pressure in millibars with respect to water at the same pressure and temperature.

Generally, this is dependent on the temperature of the air and increases with increasing temperature. Consequently, in this study, if temperature is to be used as a measure of convection activity, relative humidity must also be taken into consideration.

According to the report of the Sydney's Bureau of Meteorology (1991a), the maximum relative humidity nearly always occurs near dawn and the minimum at about noon during CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 92

summer, and at 2 pm or 3 pm during winter, at about the time of maximum temperature. The mean relative humidity at 3 pm in the three stations in the Sydney region is given in Table 4.8. This time of day was adjusted in the analysis because many of the convection activities take place in the afternoons. This matter was discussed in Chapter 3.

Table 4.10 Simple statistics of the relative humidity (in per cent) in the Sydney region, from 1960 to 1990 (372 months). Station Mean Min. Max. Range Std. Dev. Variance Sydney 55.22 30 78 48 7.7 59.4 Airport Sydney 56.5 33 72 39 7 48.9 RO. Richmond 47 21 71 50 8.7 75.3

Also, Figure 4.2 shows the average monthly variation of the relative humidity at 3 pm by the calender month over the period of 31 years. The period of analysis is 1960-1990.

Figure 4.2 Monthly distribution of the mean relative humidity at three stations in the Sydney region (1960-90).

4.4 Correlations Matrices of Variables

To find possible associations between the different variables some analytical procedures have been established. First, in order to determine the associations between thunderstorm rainfall frequency and thunderstorm rainfall amount as dependent variables, and air and sea- surface temperatures and air humidity as independent variables. An initial correlation has been separately introduced for each group of variables. The correlation found does not depend on data distribution normality. Because of the nature of thunderstorm data no CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 9J

transformation procedure was applied. Table 4.11 summarises the correlation matrix of thunderstorm items in the sample for the period beginning from 1960 and ending to 1990.

Table 4.11 The correlation matrix (associations) between dependent variables.

Y Variables i Y2 Y3 Y4 Y5 Y6

TRA Sydney Airport Yi 1

TRA Sydney R.O. Y2 0.82 1

TRA Richmond Y3 0.40 0.43 1

TRF Sydney Airport Y4 0.64 0.56 0.44 1

TRF Sydney R.O. Y5 0.57 0.61 0.48 0.80 1

TRF Richmond Y6 0.45 0.46 0.72 0.64 0.67 1 All associations are at 0.05 significant level TRA = Thunderstorm Rainfall Amount, TRF = Thunderstorm Rainfall Frequency

In this correlation matrix, each of the thunderstorm items was correlated against the others. They were interrelated in the range of approximately 0.4 to 0.82. These correlation coefficients may not show, on average, high correlation, but they indicate that all dependant variables are positively associated, and all associations are significant at 0.05 level. Also, an attempt was made to see if there are any associations between air, sea- surface temperatures and humidity among the three above-mentioned stations. Therefore, again, a simple correlation method has been employed in order to estimate the associations between independent variables.

Table 4.12 Correlation matrix for independent variables. X X Variables *i x2 x3 x4 x5 x5 7 x8 9 Xio

SST at Port Hacking *i 1

Sydney Airport Max. Tem. x2 .75 1 Min. Tem. x3 .80 .95 1

Rel. Hum. x4 .50 .22 .44 1 Sydney R.O. Max. Tem. x5 .77 .99 .95 .25 1 Min. Tem. x6 .81 .96 .99 .42 .97 1 Rel. Hum. x7 .56 .40 .56 .92 .40 .56 1 Richmond Max. Tem. x8 .44 .60 .58 .11 .60 .59 .20 1 Min. Tem. x9 .78 .94 .98 .47 .94 .98 .60 .57 1 Rel. Hum. Xio .30 -.12 .13 .79 -.08* .09* .70 -.13 .14 1 * Non-significant at 0.05 level

Table 4.12 shows that there are generally high positive correlations among independent variables. It also may indicate that the association between sea-surface temperature and air temperatures at Sydney Airport and Sydney Regional Office are higher than the association CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables JL4 at Richmond station which is the weakest. It must be noted that the relationship between temperature and air humidity is not entirely independent. According to these correlations, it can be concluded that all independent variables which have been used in this study have a common relationship to each other.

In one stage further, to examine the coefficient of correlation between the thunderstorm data and the independent variables, it was decided to find the possible correlations, However, a problem which may arise in the study of thunderstorms on the basis of monthly rainfall is brought about by the inclusion of the data from those months without any thunderstorm rainfall. If these months are included in the computation they will increase the correlation coefficient significantly without entailing any increase in information on the thunderstorm rainfall incidence. This problem has been identified by Cornish, Hill and Evans (1961) and Sumner and Bonell (1990). For this reason, months with zero values were excluded from the analysis.

Table 4.13 summarises the regression analysis between the climatic variables and thunderstorm rainfall values for the 3 stations. This table reveals distinct associations between independent variables and thunderstorm rainfall frequency and thunderstorm rainfall amounts in the study area. However, there are two points to consider. Firstly, in some cases associations are not at significant levels. Secondly, as can be seen from Table 4.13, there is a general decline of relationships between dependent variables and sea surface temperature with the distance inland. For example a 0.0009 significant level at Sydney Airport declines to 0.05 at Richmond for thunderstorm rainfall frequency values.

Table 4.13 Linear regression coefficients of dependent variables by independent variables. Dependent Variables Stations and Thunderstorm Thunderstorm Independent Variable Rainfall Frequenc Rainfall Amount R* p** R P Sydney Airport SST .21 .0009 .17 .008 Max. Air Tem. .29 .0001 .12 .05 Min. Air Tem. .32 .0001 .17 .008 Rel. Hum. .20 .001 .19 .002 Sydney R.O. SST .14 .05 .13 .05 Max. Air Tem. .22 .0005 .09 NS Min. Air Tem. .25 .0001 .12 .05 Rel. Hum. .22 .0005 .17 .007 Richmond SST .12 .05 .05 NS Max. Air Tem. .29 .0001 .07 NS Min. Air Tem. .34 .0001 .16 .01 Rel. Hum. .13 .04 .20 .002 * Regression coefficient, ** Probability level NS = Non-significant correlation at 0.05 level CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables £5

4.5 Multiple Associations Between Variables

At this stage of the data analysis, to ensure that each major independent variable makes a statistically significant contribution to the predictable variance of thunderstorm rainfall amount (here the focus is on how much rather than how often), and in order to examine the relative importance of each climatic variable, a stepwise multiple regression technique was introduced. This technique was considered a 'stepwise solution' which is common computational procedure in regression analysis having been used in many studies (Ohring, 1972). For example, Bryant (1985a) used this technique for correlating beach erosion with sea-levelrises. Therefore , this technique, as a rank ordering of the total correlations of major independent variables with thunderstorm rainfall, was separately applied to two stations (with the longest records), Sydney Airport station (the nearest station to Port Hacking) and Richmond station (the furthest station from Port Hacking).

Generally, this technique can predict as much variance in the dependent variable as is possible from the composite of independent variables. This process is complicated by the fact that the independent variables may be correlated with each other and, consequently, each predicts the "same part" of the variation in the dependent variable. For example, sea- surface and air temperatures may correlate with each other (see Table 4.12).

By using a stepwise multiple regression procedure it was therefore possible to evaluate systematically the relative contributions of important variables in the explanation of thunderstorm rainfall. This statistical technique was also used in order to measure whether there is a cumulative effect of several variables on thunderstorm behaviour. Independent variables in this regression equation were mean daily maximum and minimum temperatures, sea-surface temperature and mean relative air humidity, with thunderstorm rainfall as the dependent variable. The independent variables were selected according to the literature framework of this study, because in Chapter 2, it was hypothesised that all of these four factors would contribute significantly to the explanation of thunderstorm rainfall.

Each independent variable was entered into the regression equation in order to determine its unique contribution in relation to the other three. The order in which the independent variables are entered into the equation had no impact on the outcome because each variable is treated as though it is the last variable to be entered. The stepwise regression procedure selected the strongest independent variable in the first stage and at each stage a new variable was added to the equation. The results of the stepwise regression are presented in Tables 4.14 (a) and 4.14(b).

The independent variables were introduced into the regression equation, summarised in Tables 4.14 (a) and 4.14 (b), in the order in which they increased the explained variance in M thunderstorm rainfall. In both stations, the rank ordering of variables in terms of their predictive strength are: mean daily minimum air temperature; sea-surface temperature; mean relative air humidity; and mean daily maximum air temperature.

The first step identified the minimum air temperature as the best single predictor of thunderstorm rainfall. Sea-surface temperature is the second strongest predictor of thunderstorm rainfall ( F[30.8], p< 01). This variable adds 6 per cent to the explained variance in thunderstorm rainfall at the Sydney Airport station.

Table 4.14 (a) Results of stepwise multiple regression analysis of thunderstorm rainfall at the Sydney Airport station (n=252 ).

Step Multiple R Variance F ratio Number of Number Predictor Variable R Square Added in to enter * Variable in the % Equation

1 Minimum Air Temperature 0.35 0.12 12 30.8 1 2 Sea-Surface Temperature 0.42 0.18 6 24 2 3 Mean Relative Humidity 0.44 0.19 2 18 3 4 Maximum Air Temperature 0.45 0.20 1 14 4 Total 21 All F values are significant at 0.01 level

Table 4.14 (b) Results of stepwise multiple regression analysis of thunderstorm rainfall at Richmond station (n= 227).

Step Predictor Variable Multiple R Variance F ratio Number of Number R Square Added in to enter* Variable in the % Equation

1 Minimum Air Temperature 0.32 0.11 11 29.4 1 2 Sea-Surface Temperature 0.34 0.12 1 16.2 2 3 Mean Relative Humidity 0.35 0.13 1 11.4 3 4 Maximum Air Temperature 0.36 0.137 1 10.1 4 Total 14 * All F values are significant at 0.01 level

Tables 4.14 (a) and 4.14 (b) also show a significant relationship between the relative air humidity at each station and thunderstorm rainfall. This variable is recognised as the third major predictor of thunderstorm rainfall and adds about 2 per cent to the explained variance in the equation. When the 4 variables are included in the regression equation, the amount of explained variance in thunderstorm rainfall behaviour increases to about 21 per cent at the Sydney Airport and about 14 per cent at the Richmond station. The stepwise CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 91

relationships are not high, and despite the significance, the variance explanied is very low, therefore, other factors must be operating.

As the result of the stepwise regression technique indicated, all 4 independent variables have significant relationships with thunderstorm rainfall. A comparison of these tables clearly indicates that both sea-surface and air temperatures can affect the thunderstorm occurrence throughout the Sydney region and they can explain some of the variance statistically. However, the percentage of variance explained is not high generally, and it is not the same for the two selected stations which are located in the east, near the coast, and in the west of the Sydney region.

4.6 Discussion

In this chapter - using the available data taken from the limited of stations - first, some descriptive statistic techniques have been used to summarise, present and compare the distribution of thunderstorm data on a monthly basis. Then, to find the possible associations among some of the climatic variables which may affect thunderstorms, different kinds of statistical procedures were considered. Some simple correlation methods and a stepwise multiple technique were used to find the percentage contribution of each independent variable upon thunderstorm rainfall. Results indicated that there are possible causal relationships between the above-mentioned climatic variables and thunderstorm data specially for coastal stations. This relationship becomes weaker further inland.

The evidence involves relatively low correlation coefficients, similar to those reported by Priestley (1964) and Hirst and Linacre (1978) for monthly rainfall values (see Chapter 2). The results of the stepwise regression technique indicated that thunderstorm rainfall amount (explaining about 21 per cent of the variance at the Sydney Airport station and 14 per cent at Richmond) is associated with three main climatic factors; air, sea-surface temperatures and air humidity. Three distinct effects come to mind as likely to cause these associations between variables in the region.

4.6.1 Effects of Sea-Surface Temperature

The direct effect of sea-surface temperature upon the rainfall process was shown by Priestley and Troup in 1966. The ocean waters adjacent to the coast can provide atmospheric moisture and moderate temperatures and therefore affect rainfall patterns in the region (Rochford, 1977). Apparently the importance of sea-surface temperature to the climate of the Sydney region has been recognised for a long time. For example, investigation has revealed that the east coast current which carries warm tropical water southwards along the New South Wales coast can affect the climate of the region (Lough, 1992). The positive correlations between rainfall in eastern Victoria and the warm sea- CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 28

surface temperatures from the Coral Sea, which was found by Whetton (1989), also supports this positive correlation. Therefore, although the ocean temperature varies in both space and time, it can also significantly influence the distribution of rainfall patterns in the region.

Supposing other affecting factors to be equal, the warmer the water, the warmer and more moist will be the lower layers of air reaching the coast. In addition, there will probably be a greater tendency for convection activity. In 1978, Hirst and Linacre indicated that the onshore winds, which can control the sea-surface temperature, may also enhance convective rainfall by bringing in moist warm air to coastal areas. In this case, a warmer sea surface would cause instability of the coastal atmosphere, increasing the tendency to convective rainfall.

Although it is evident from the result of this study that the association between sea-surface temperature and thunderstorm rainfall is positive during the calendar year, it seems this association is stronger in autumn/winter than the spring/summer seasons. This was emphasised by Colquhoun and Batt, in a personal conversation (Bureau of Meteorology, NSW Regional Office, 1994). This is when the land-sea temperature difference is greatest (Holland et al. 1987). Also, Hopkins and Holland (1994), found that the East-Coast Cyclones show a preference for formation in the autumn/winter months which occasionally create very heavy rainfalls along the east coast ranges.

On the other hand, the results of this study indicate that the association between coastal sea-surface temperature and thunderstorm rainfall at the Richmond station (which is more than 55 Km inland) is less or even non-significant. A simple explanation may be that in the west of the Sydney region, because there is less moisture and because it is furtherfrom the warm easterly winds off the ocean, the chance of thunderstorm occurrence with intense rainfall is lower than in the coastal areas.

4.6.2 Associations Between Air Temperature and Thunderstorms

A relatively high association between air temperatures (minimum and maximum) and thunderstorm rainfall amount probably indicates that the incidence of high air temperature can cause high thunderstorm activity in the region. This simply means that the air temperature should also be considered as one of the factors which is able to create or enhance thunderstorm activity.

Many researchers in the field of thunderstorm activity have found that air temperature is an important climatic factor in creating or enhancing a convection system. For example, Lutgens and Tarbuck (1982 p:237) wrote: CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 92

"All thunderstorms require warm, moist air, which, when lifted, will release sufficient latent heat to provide the buoyancy to maintain its upwardflight. Although this instability and associated buoyancy are triggered by a number of different processes, all thunderstorms need an unstable atmospheric environment in which the instability can be enhanced by high surface temperatures."

To these features can be added the effects of unequal heating of the land surfaces, particularly in summer months when the clear areas can be warmed rapidly by solar radiation (Baines, 1990). Probably this uneven heating can generate vigorous convection which leads to the growth of storms in a matter of hours. It has been shown in chapter 3 that most thunderstorms develop in the afternoons in the spring and summer months, when the potential for convection is usually the greatest and adequate high air temperatures are available.

In the USA, Benjamin (1983) found that some severe thunderstorms were the result of differential heating, differential advection and local topography. In addition to these factors, Golde (1977) has shown that vigorous thunderstorms can occur along an active cold front or in squall lines in the warm air ahead, at any time of the day or night.

It may be supposed that surface heating is generally not sufficient, in itself, to cause thunderstorm activity, and any factor that can destabilise the air, aids in generating a thunderstorm. As Smith (1975 p:13) mentioned:

"The importance of surface features increases markedly as the scale of climatic reference clirninishes and it is only at the very lowest levels of the atmospheric boundary layer that surface influences become strong enough to create really special phenomena".

A high air temperature may indirectly cause, or enhance, other associated atmospheric phenomena which should be considered as important factors in introducing or causing a convection activity in the region. Many investigators (for example, Simpson, 1964 and Atkinson, 1981) have highlighted the importance of unequal heating in coastal plains. They linked these phenomena to both local convection and to the role of sea breeze fronts in the generation and enhancement of meso-scale systems such as thunderstorms.

In Australia, sea-breezes have been studied extensively by some researchers who have found that the summer months are times of the greatest sea breeze development (Lyons, 1977). Hobbs (1971) noted that for the NSW coast the incidence of sea breezes increases as the summer progresses. Clarke (1955 and 1960) and Drake (1982) suggested that sea breeze penetration is greatest in southern of New South Wales. For example, at Nowra, CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables 1M about 15 km north-west of Jervis Bay, Mathews (1982) found that the magnitude of the sea-breeze component depends on the land-sea temperature difference. The occurrence of temperature differentials leads directly to pressure variations that giverise t o air movement. At the mesoscale level, this effect may cause local sea-breezes and formation of thermal lows over land masses in summer. Also, Sumner (1983b) proposed that local winds and proximity to the sea may cause storm development which is dependent largely on the presence of an escarpment. These studies generally indicate that the diurnal variation of land and sea heating is the cause of sea-breezes.

In the Sydney coastal areas, because of the apparent difference between the temperatures of the land and of the adjacent ocean, sea-breezes are generated by cool air from the ocean replacing warmer air rising over the land. In 1974, Linacre and Barrero showed the positions of the sea-breeze front at various times in the Sydney region. They concluded that, although the sea-breeze front moves inland to a distance which depends on the day- length and the speed of thefront, i t is stronger in the mid and early-afternoon, particularly in summer months. Therefore, the sea breeze is generally most pronounced in the late spring and summer, and during the early afternoon hours which have the highest daily temperatures. This is an atmospheric phenomenon which may contribute, in a general way, to the convection mechanisem and as a result to thunderstorm enhancement (Abbs and Physick, 1992). This may be one of the reasons thunderstorms are most common in the afternoon and warm months of the year.

4.6.3 The Role of Air Humidity

Finally, the role of the moist air available in the surrounding atmosphere can also be statistically seen to be an important factor in initiating convection activity, because it makes the atmospheric environment more unstable. As Moran and Morgan (1991) indicated, thunderstorms usually develop in unstable atmospheric environments as a consequence of uplift caused by one or more of the following: (1)frontal activity , (2) orographic effects, (3) surface convergence, or (4) intense solar heating of the land surface.

According to these mechanisms, it was suggested that the available moisture in the air can help a convection development when dense cold air overlies warm, moist air which is less dense. Therefore, many thunderstorms require warm, moist air which will release sufficient latent heat to provide the buoyancy necessary to maintain its upward flight.

Although this instability and associated buoyancy are triggered by a number of different processes, all thunderstorms need a moist atmosphere to keep their life cycle. A trigger such as solar heating, a front, or a trough-line can then begin the development of a thunderstorm. Thus, high heat energy and water vapour stored in the air can be converted CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables Ml into wind and electrical energy introducing much more instability into a convection system, which may then organise and produce much rain.

Eagle and Geary (1985 p:2) both point to the importance of moist air in the increasing of rainfalls from well-organised and widespread thunderstorms in the region. They suggested that:

'The atmosphere is modified by the ocean surface over which it moves. Given a sufficient transport time over the sea, air in the lower layer will achieve a balance with the temperature and vapour pressure at the underlying surface. With favourable conditions the water vapour may be adverted vertically through the air column. Ocean conditions have no obvious immediate influence to the coastal location but if winds are onshore a short term effect may ensue, as is the case with coastal shower situations,.

They found that during early November, 1984 a north easterly airstream originating in low latitudes, was moving coastward across the Warm East Australian Current. At this time the waters of this current were one to two degrees above average for the time of year. As a result, the temperature and humidity of the airstream which affected the coastal areas, was largely sustained.

In coastal locations, this positive temperature anomaly was very favourable to the maintenance of a warm moist air mass and then conducive to thunderstorm rainfall with the presence of uplift mechanisms. All previous evidence has indicated that, on average, coastal areas in the Sydney region experience many more gradient winds which are both stronger during the day or night, and which have a higher percentage of humidity.

As the results of the statistical analysis have indicated, there is a considerable amount of unexplained variance which may suggest other independent variables need to be incorporated into the regression model. It seems certain that suggested climatic variables are not the only factors which explain all the variation of thunderstorm rainfall in the region. This occurs because, in a complex three dimensional atmospheric environment, in which convection activities take place, there must be several independent variables affecting the development of a thunderstorm system.

More importantly, it is well known that the occurrence of a thunderstorm also depends upon the vertical distribution of temperature and humidity in the atmosphere. Most often, moving convection systems track from places where they create and affect the other surrounding low land areas. Therefore, it is clear that accurate atmospheric information on the convective motions within, about and beneath a thunderstorm system, is necessary to CHAPTER FOUR Thunderstorm Rainfall and Climatic Variables M2 understand the nature of convective activity and, as a result, to explain the associated rainfall amount from thunderstorms for a specific location.

While the quality of data and the number of stations used in this study may not be precise enough to confidently take the calculated coefficients as predictive values of thunderstorm activity in the region, the results obtained statistically, indicate that the air and sea-surface temperatures, and air humidity can be linked to thunderstorm development and amount of precipitation. These factors seem to be more effective climatic factors, particularly in the coastal areas where moist air currents and warmer sea-surface temperatures can cause instability of the atmosphere, increasing the tendency to convective rainfall. This last factor is probably less important further away from the coast where the moist local winds arrive later or not at all.

4.7 Summary and Conclusion

Results obtained statistically demonstrate that the associations between some of the environmental-climatic variables; such as air and sea temperatures and relative air humidity and thunderstorm rainfall, are mainly positive. There are some possible causal correlations among independent and dependent variables, particularly in the east of the study area, near the ocean. They may reflect the combined effects of all these environmental variables upon thunderstorm rainfall. It appears that the correlation between sea-surface temperatures and thunderstorm rainfall roughly decreases with distance from the coast. In contrast, the correlation between air temperature and thunderstorm rainfall amount becomes stronger with increasing distance from the coastline.

In summary, it is more likely that the effects of above-mentioned climatic factors upon thunderstorm rainfall patterns result from a complex climatic interaction. It is clear from the total variance discovered that there are definitely other independent variables which are also responsible for spatial variation of thunderstorm rainfall amounts in the region. Therefore, to visualise these variations in space, the variation and distribution of thunderstorm rainfall will be examined in Chapter 6. CHAPTER FIVE A Review on GIS Systems Ml

CHAPTER 5

A REVIEW ON GIS SYSTEMS

5.1 Introduction

GIS can be used to visualise the spatial pattern of thunderstorm rainfalls by emphasisng the physiographic features of the Sydney region. The current chapter reviews an overal GIS methodology which could be applied for data with a spatial nature, and follows a progression of topics, becoming more specialised in the following chapters. Chapters 6 and 7 will, thus, deal with analysis and modelling of thunderstorm rainfall data and physiographic parameters of the study area.

The literature on GIS is vast and spread over a large number of areas, representative of many disciplines and covers an enormous number of applications. The most relevant sources for the material presented in this thesis are, therefore, selected and explained. Sections 5.2 to 5.6 bring together the relevant conceptual issues of GIS. First, some of the currently used definitions, the common purposes, the principle mechanisms and operating systems of GIS are defined. Then, the use of GIS for geographical applications and its use in climatology are reviewed. The application of GIS techniques in spatial modelling of thunderstorm rainfall is also examined. Data sources and those technical aspects of the SPANS (which is an acronym and stands for SPatial ANalysis System) are explanied in sections 5.7 and 5.8 respectively. Finally, GIS potential errors are outlined in section 5.9.

5.2 What is a GIS ?

GIS is a computer technology consisting of hardware and software that is used to produce, organise, and analysis information (Aronoff, 1989). In fact, GISs are computer software for managing data that are spatially distributed over the Earth (Bonham-Carter, 1994). Maguire (1991) states that GISs are computer systems capable of storing, analysing, manipulating and displaying spatial data from the real world which can be represented spatially in a computer environment (Dangermond, 1986).

Accordingly, GIS is able to provide natural resource managers with the tool to merge spatial data and their attributes into computerised data base systems allowing input, storage, retrieval and analysis of geographically referenced data (Calkins and Tomlinson, 1977). With this capacity for spatial and temporal modelling of the real world, GIS as a technology has been developed to accomplish the complicated tasks which are grouped CHAPTER FIVE A Review on GIS Systems WI together as 'GIS' functions. Two selected definitions of GIS are as follows; Aronoff (1989 p:39) "any manual or computer based set of procedures used to store and manipulate geographically referenced data". Koshkariov et al. (1989 p:259) "a system with advanced geo-modelling capabilities". Many of the definitions are relatively general and cover a wide range of subjects and activities (Tomlinson et al., 1976; Moore et al., 1981).

GISs have three important components- computer hardware, sets of application software modules and a proper organisation context (Burrough, 1989). These three components need to be in balance if the system is to function satisfactorily. Maguire and Dangermond (1991) believe that four basic elements of GIS, which operate in an institutional context are: computer hardware, computer software, data and liveware. However, some researchers think that GISs are the result of linking parallel developments in many separate spatial data processing disciplines (Cliff and Ord, 1981).

Essentially, all these disciplines are attempting the same sort of operation, mainly to develop a powerful set of tools for collecting, storing, retrieving, transforming, and finally displaying spatial data from the real world for a set of particular purposes. These sets of tools were combined to constitute a GIS environment (Burrough, 1989). In other words, GIS should be thought of as being very much more than a means of coding, storing, and retrieving data about aspects of the earth's surface (Goodchild and Kemp, 1990). In fact, GISs are designed to bring together diverse spatial data sources into a unified framework, often employing a variety of digital data structures, and representing spatially varying phenomena as a series of data layers as models from the real world (Prisley, 1986; Rhind, 1988).

5.3 Purpose of GIS

The purpose of using a GIS system for geographical and other applications can be reduced to about six activities dealing with spatial data: 1) organisation, 2) visualisation, 3) combination, 4) analysis , 5) modelling, and 6) query (Bonham-Carter et al., 1988; Burrough, 1989; Goodchild and Kemp, 1990)

1) Organisation is the ordering of information according to logical links (Bonham- Carter, 1994). Anyone who has collected a large mass of data for a particular purpose knows that data organisation is essential. Data can be arranged in many different ways, but all the data has to be spatially referenced in GIS. For example, a table of geographic data may be interesting for viewing relationships between elements, but without knowing the locations of samples, the interpretation of spatial patterns and relationships with other spatial data, such as geographic features, cannot be made and understood (Johnston, 1987). A GIS must be concerned not only with location, but must also organise data to allow the extraction of other types of information (Aronoff, 1989). Because, the GIS can CHAPTER FIVE A Review on GIS Systems Ml

organise data both by spatial and non-spatial attributes, the efficiency and type of data organisation effects all the other five activities and is therefore of fundamental importance (Maguire, 1989).

2) Visualisation is an important technique for analysing, explaining and understanding the distribution of a phenomenon on the surface of the earth (Buttenfield, 1987). Using new technology capacities, the graphical capabilities of computers are exploited by GIS for visualisation (Dangermond and Smith, 1988). Generally, visualisation is the assessing of information through the use of sight which is normally carried out using the video monitor, but other output devices such as colour printers are used for hard-copy displays (Intera Tydac, 1992a). Often, visualisation is enhanced in a GIS system by specialised methods using colour, perspective, shadowing and other means. One of the immediate benefits of this function of GIS is that visualising data stimulates the mind in ways which are different from traditional data analysis procedures (Cuff and Mattson, 1982).

3) In a GIS, combination is the bringing together of data sets. Data used in GIS often come from many different sources, are of many different types (even with different spatial nature) and are stored in different ways (Flowerdew and Bantin, 1989). GIS provides the tools and method for combining, or integrating, these data into a format which allows the data to be compared. This process of creating a common form of the data or the bringing together spatial data from a number of sources, is described as data integration. The role of the GIS as an 'information integrator' was examined by several researchers on various approaches. DoE (1987 p:2) states that, 'The benefits of a GIS depends on linking different data sets together.' Dangermond (1989 p:25) said that:

'A GIS brings information together, it unifies and integrates that information. It makes available information to which no one had access before, and places old information in a new context. It often brings together information which either was not or could not be brought together previously'.

This is one of the really powerful features of GIS in which the ability to link several maps together provides various kinds of models. The benefits that follow the integration of diverse information are widely recognised.

4) One of the important stages in the GIS environment is the analysis of the results of previous stages or the process of inferring meaning from data (Berry, 1986). In fact, analysis is the interpretation and the study of data and information that have been collected. With GIS, the relationships between different spatial data and their associated features can be measured and understood. Spatial analysis in a GIS simply means, the analysis of spatial data. For instance, the area cross-tabulating of two maps may lead to A Review on GIS Systems 106 useful conclusions about the relationship between the two map layers. Therefore, with GIS the relationships between different spatial data and their associated features can be measured and understood (Samet, 1989).

5) Just as it is possible to analyse spatial information to extract knowledge, it is also possible to use known relationships to model geographically the outcome of a set of conditions (Intera Tydac, 1992b). This function of GIS is helpful for assessing models from patterns in the data. Normally, thefinal purpos e of many GIS studies is often for the prediction and modelling of data. For example, a number of data layers can indicate new sets of maps which could be combined to predict the suitability of thefinal desire d model (map). Such a map may then be used as a basis for making exploration or landuse decisions (Dickinson and Calkins, 1988). In other words, prediction is sometimes a research exercise to explore the outcome of making a particular set of assumptions, often with the purpose of examining the performance of a model (Alberti, 1991).

6) Finally, a strong feature of GIS is the ability to query intellectually the underlying data simply by moving a pointer around on a map. Since all data in a spatial database are geographically referenced, a pointer to location means access to all data associated with that location (Intera Tydac, 1993). Spatial query is a complementary activity to data visualisation, because it would permit the user to find the special circumstances of each case, by searching the name and other particulars of characteristics of individual geographic features in the selected locations of interest. Generally, GIS provides tools for two types of interactive query: geographical information of a location and its attributes (Unwin, 1981). This powerful function of GIS allows the user to enjoy the dynamic query of attributes of up to 19 map layers, simultaneously.

5.4 How GIS Operates

At its most basic level, a GIS can be viewed as a simple input / output process. Data goes into the GIS (such as collected data), some form of processing occurs (averaging of data for different areas), and information comes out (perhaps in the form of a map). Regardless of its complexity the input / output view of GIS is a useful starting point from which to examine how the technology actually works. However, in order to understand the basic operations in a GIS environment, it isfirst necessary to understand the main structure and functionality of the GIS in which the data must be processed. CHAPTER FIVE A Review on GIS Systems Ml

5.4.1 Data Structures in GIS

There are currently three common data structures used by geographical information systems; 1) vector, 2) raster, and 3) quadtrees (Ibbs and Stevens, 1989). Each structure has an associated set of characteristics, some good, some bad (Bonham-Carter, 1993).

1) The Vector format was defined as positional data in the form of co-ordinates of the ends of line segments in a point, line or polygon format (Intera Tydac, 1993). This is the most common method for representing spatial data in which, 2-D space is assumed to be continuous and allows very precise representation of locations, lengths, distances and areas. Locations are described by coordinate pairs, and these pairs are the fundamental building blocks from which spatial entities such as points, lines, and areas are composed. In a vector structure, points are represented by a single x, y coordinate pair, while liner entities and area entities (polygons) are composed of straight line segments joining two coordinate pairs (vertices). The attribute of the values for point, line, and polygon entities are typically stored independently of the entity's spatial representation. Generally, the vector structure is ideal for representing point (rainfall stations) and linear features such as rivers, and for cartographic map production. This structure is also very useful for topological relations, but is very limiting for overlay modelling procedures (Cook, 1978)

2) The Raster format is spatial data expressed as a matrix of cells or pixels, with the spatial position implicit in the ordering of the pixels. The simple raster data structure represents 2-D space as an array of matrix of square or rectangular grid cells. Each grid cell represents a square or rectangular portion of the Earth's surface. The resolution of raster data is determined by the size of the cell on the ground, thus, raster data represent a discrete space where the locational precision is dependent upon the size of a grid cell (Brown and Norris, 1988). Each grid cell is assumed to have only one value for any given attribute. A grid cell attribute value may represent a point measurement (for example, elevation) or an integrated areal measurement (for example, landuse map). In a raster data structure, points are represented as individual cells, while lines and areas are represented as clusters of adjacent pixels. The coordinated precision of raster data is constrained by cell size. Generally, the raster structure is ideal for representing continuous data, such as elevation and is excellent for multiple map overlays, but it is poor for certain data approximation (Knaap, 1992).

3) Finally, the Quadtree format is a data structure for thematic information in a raster database that seeks to minimise data storage. In fact, this kind of data structure is a hierarchical grid based data structure which is used to improve the storage efficiency of either its raster or vector counterparts (Ibbs and Stevens, 1989). A hierarchical spatial data structure is one which is developed through a process of regularly subdividing the space A Review on GIS Systems occupied by geographical entities on a map layer into regular spatial units (Intera Tydac, 1993). This process continues until each unit produced by the subdivision is occupied by spatial entities with similar attributes (see Figure 3.1).

Each data structure has its merits and its pitfalls (Ibbs and Stevens, 1989). Generally, vector data structure are used for digitising data and cartographic purposes which use (x, y) coordinates to describe point, line and area features. In this format, data structure retains information about the consecutiveness and adjacency of features, but are computationally more demanding. The raster data structure is, however, useful when combining satellite imagery, which is already in raster format, into the database and this is used for analysis (Johnston, 1987). A raster data structure is formed by a matrix of regular cells, each a specified size and area (Knaap, 1992). Many GIS have the capacity to use both data structures. The quadtree structure is ideal for representing both continuous data and discrete polygonal data. In other words, it can be thought of as a raster structure with the ability to have a variable sized grid cell (Webster, 1992).

Figure 5.1 Schematically represents different data structures used in a GIS: (a) raster and quadtree (b) points and lines (vector) and (c) polygons.

5.4.2 Functionality of Data in GIS

In GIS, realistic spatial models of the world, called entities, can be developed using these structures. Entities are points, lines, areas, surfaces and networks (Martin, 1982). An entity has a spatial dimension which identifies its geographical location. GIS data structures are able to accept both spatial and non-spatial data in any GIS project. Therefore, identification and collection of relevant structure and data are essential (Webster, 1990). Data used in GIS often come from many different sources, are of many types, and are stored in different ways. These mechanisms should be summarised into 6 stages as follows: CHAPTER FIVE A Review on GIS Systems Ml

1) Geographic data sources which can be imported into the GIS environment include: paper maps; aerial photographs; satellite images, and digital data from other areas which can be combined to create new complex maps or tables (O'Neill et al. 1992). This is another source of geographic information which is not often thought of as being geographic. These data are mainly tabular databases orfiles o f records such as weather station observations (rainfall records) or water samples records.(databases) which are often geographically referenced. If the underlying structure of the geographic reference system is known (latitude and longitude), it is possible to transform and integrate this information into thematic data which can then be processed in the GIS environment (Bonham-Carter, 1994).

2) After the data are collected and integrated, the GIS must provide facilities which can contain and maintain the data (Brown and Norris, 1988). Effective data management has many definitions but should at least, include all of the following aspects: data security, integrity and maintenance abilities. In fact, data management refers to the ability of a GIS to manage functions efficiently, the ability to link to other data types and transfer data in compatible formats (Davis, 1991).

3) Data processing operations are those performed on the data to produce information. In GIS, data on its own may be impossible to interpret and data processing is not an end in itself. It should turn data into a form that is informative, that helps the user decide what to do next and whether more data processing or qualitative analysis should be done. Data processing produces images, reports and maps.

4) Data integration and conversion is only part of the input phase of GIS. What is required next is the ability to interpret and analyse, quantitatively and qualitatively, the information that has been collected. This ability to analyse and manipulate spatial data that has led to the use of GIS for both statistical and deterministic modelling (Cressie, 1991). Analysis is carried out on data organised as maps, and also on data organised as tables. Using the analysis function of GIS, it is possible to explore existing relationships between the data sets.

5) The ability to model geo-referenced information is critical in a GIS (Webster, 1992). In the geosciencefields, especially geographical exploration, this type of overlay modelling has been done for years, typically with several maps and a light table. The main objective is to create a new map which highlights areas which meet a certain set of criteria favourable for modelling. The GIS allows geographers to combine maps to produce new maps, without struggling with the variable scale and projection problems (Rasuly, 1991). CHAPTER FIVE A Review on GIS Systems LM

6) Finally, one of the most exciting aspects of GIS technology is the variety of different ways in which information can be presented, once it has been processed by the GIS. Traditional methods of tabulating and graphing data can be supplemented by maps and three dimensional images. Also, tables andfigures, havin g results, can be transformed into maps which reveal spatial or non-spatial entities. The use of GIS technology allows information to be viewed on the computer screen, plotted, as paper maps, captured as a image or slide and used to generate a computerfile. Generally , visual communication which is the most important aspect of GIS technology, can be enhanced by the diverse range of output options (Webster, 1990).

5.5 Implications of GIS Techniques in Climatology

The GIS has been widely used in recent years for natural resource planing and management (Alberti, 1991 and Davis, 1991) and solving complex problems associated with multiple-use of land resources (Martin, 1985). Initially, the origins of GIS lie in environmental management (DoE, 1987), but uses of GIS have expanded to incorporate private and government planning in areas such as: property and land parcel data; transport, and distribution networks; civil engineering; defence; industrial site selection; and water supply application (Tomlinson, 1987; Johnston et al., 1988).

In addition, GISs are used in many environmental spatial analysis and modelling situations. Technical and applications-oriented workers from many fields (for example, ecology, hydrology and geography) are interested in the use of GIS (Ferrier and Smith, 1990). Recent environment applications can be expanded to include: survey design, dynamics and distribution of soil (Moore et al. 1981), individual species and soil-climate modelling (Duff and Eamus, 1992), vegetation communities (Head et al., 1992), bushfire patterns (O'Neill et al., 1993) and habitat modelling (Marthick, 1995). All these studies found that the GISs can be used in the handling of environmental problems. But these are only a few applications within the general GIS literature in climatology which is both highly disparate and complex.

Currently, there is a fast growing interest in using GIS methodology within physical geography (Rasuly, 1993) and environmental science, which can be characterised as "Physical-Environmental GIS" (Riddle, 1991). For example, Maguire (1989 p:222) said that:

"The synthesis of geographical facts relating to the locational properties of spatial entities and their associated attributes is a necessary counterbalance to analytical studies carried out in physical and human geography'. CHAPTER FIVE A Review on GIS Systems Ul

From 1991 there were rapid increases in further developments in the use of GIS for research in physical geography and the environmental sciences (Raper, 1993). This interest is growing fast, because a GIS can store both cartographic data, showing topography or individual themes, such as soils or rainfall distribution, and attribute data associated with the spatial entities (points, lines and polygons), that were represented in Figure 3.1. Therefore, in many respects a set of disparate data can be only linked by GIS techniques.

The methodological problems and applications of this new sub-field have resulted in a number of publications. For example, an application research was introduced to estimate crop yield in south western Ethiopia (Simmons, 1986). Using GIS it was possible to perform a series of map overlays of climatic and soils factors from which predictions of crop yields were calculated. In this study three input maps (climate zones, elevation and soil types maps) were used to produce different classes of climatic suitability.

Although, GIS has been used for a variety of projects, many with environmental themes, there are examples of GIS techniques being used in climatic studies. Recently, its use in the atmospheric and climaticfields hav e been concentrated on the modelling of spatial impacts of climatic events or conditions. For example, Michener (1991) assessed the ecological disturbances due to hurricane Hugo in 1989 by integrating a large quantity of data with different sources.

The topic of GIS and climate is very new, but, because of the ideal application of the GIS technology to environmental subjects, there is already a strong tendency to use the GIS for climatic purposes. In many circumstances, new technology allows the rapid mapping of point or polygon climatic variables, the correlation of maps, and the use of maps as variables in computer models.

In the literature, there are some examples of the use of GIS which can demonstrate its suitability to Climatology. For example, Johnson and Worobec (1988) used GIS techniques in the study of spatial analysis of insects in relation to weather conditions. In this study, the abundance of adult grasshoppers was correlated to monthly rainfall, monthly hours of sunlight and annual grasshopper counts. The grasshopper distribution was estimated from the previous year's grasshopper population in a close association with climatic variables which were successfully constructed using GIS techniques. In another attempt, in Italy, a GIS application for climatological analysis and productivity estimation was applied (Ciaramaglia et al., 1992). This paper described the research that was aimed at developing climatologically based rainfall-landscape planning models, using the GIS technology. CHAPTER FIVE A Review on GIS Systems 112

Also, GIS techniques were used in some studies of climate change. For example, in 1990, Aspinall and Miller described a modelling procedure which mapped climate change scenarios on a national and regional scale. The procedure was applied through a raster- based GIS system which allowed integration of land cover data from remotely-sensed sources with scenarios of climate change for impact assessment. Using climatic data such as an accumulated growing temperature and the length of the growing season, a variety of agricultural land-suitability assessments were derived for both current conditions and for a future scenario of climatic change. Therefore, attempts to assess the possible impact of climate change on agriculture and natural ecosystems are increasingly drawing upon GIS in order to gain the regional and national picture required for policy-relevant results (Brignall et al., 1991).

GIS, as a tool, was similarly used to bring together different elements of the climate of a region and its physiographic attributes. For instance, Strobl (1992) modelled the spatial distribution of climatic elements in high-relief terrain using GIS techniques. Various topographic, atmospheric and surface data are combined in the SPANS GIS environment to asses the climate variations in the Alpine Regions. In Australia, a GIS system was also applied for visualisation and demonstration of some environmental factors such as sea- surface temperatures and rainfall distribution to estimate drought scenarios (Beswick et al., 1993). It was generally suggested that ultimately, spatial modelling of climate elements should necessarily replace the use of old hand-drawn maps (they have been good - but their contents are non-reproducible) which give limited results.

5.6 Application of the GIS in Resolving Problems in Rainfall Analysis

Clearly, there is no apparent relationship between the thunderstorm rainfall - the main topic of this thesis - and the GIS techniques discussed here. However, various aspects of the GIS technology can be orientated towards to solving some problems in the study of the rainfall distribution. There are some distinct advantages using a GIS in the study of the spatial distribution of rainfall described below:

1) For a climatologist the understanding of the spatial distribution of a climatic variable, say rainfall variation, over a specific area, is a very important task (Berry and Marble, 1968; Rasuly, 1993). For elements such as rainfall maps, GIS produces good and satisfactory information of the spatial distribution of rainfall, which is of interest not only from a climatological viewpoint, but also for its importance in different fields such as agriculture, hydrology, water resources, atmospheric pollution or even in flood control. The estimation of the spatial distribution of rainfall is a complex and lengthy task. But, when detailed information concerning the rainfall records is available, the use of GIS in constructing the distribution maps, it is a matter of only several hours. The contouring CHAPTER FIVE A Review on GIS Systems 113 approach is probably the most common and thus the most familiar to climatologists and hydrologists. During the last two decades, the need for an efficient and rapid method of contouring and computing areal estimates of rainfall from rain gauge data has been a demanding task (Chidley and keys, 1970 and 1972). Currently, a GIS system, for example SPANS software, creates a Triangular Irregular Network (TIN) between the points (rainfall stations) and interpolates a surface (rainfall map) model in a very short time (Intera Tydac, 1993).

In the literature there are a few examples of direct and indirect use of GIS in the study of rainfall distribution. For example, in 1990, Eklundh and Pilesjo suggested that it is possible to create a rainfall data base explaining the variation of mean rainfall in Ethiopia, using a GIS including a digital elevation model. Currently, at the Canadian Climate Centre preparations are being made for the production and publication of long term monthly climatic variables such as rainfall data. For example, Sajecki (1991) used SPANS GIS to produce a set of sample maps for the temperature, precipitation and sea level pressure elements so that they may be included in climatic atlases. Finally, Bryceson and Bryant (1993) created the continental rainfall maps for Australia by interpolating between the sparse rainfall-recording stations. They suggested that GIS techniques can be used to mitigate the climatic variables, for example rainfall, by a better supply of information. This is a significant improvement in the ability to accurately interpolate point rainfall data and allows a greater confidence in using GIS techniques to create rainfall maps from point rainfall data for modelling purposes.

2) In all practical spatial analysis of rainfall distribution, climatologists and others have to estimate the areal distribution of rainfall from point measurement. This can be done with methods ranging from simple arithmetic averaging to sophisticated computerised interpolation and extrapolation techniques (Watson, 1992). The development and spread of personal computers, equipped with GIS software - for example the SPANS GIS - provides an ability to maintain and exploit the climatologists ordering of information in ways never before attempted. In this way, different rainfall maps can be drawn and compared with each other by computer and summaries provided by whatever set of areal distributions seems necessary to the investigator.

3) Historically, the study of the spatial organisation and distribution has always been an important factor to many climatologists, especially when rainfall maps can be correlated to the main physiographic parameters, for example, the topography or landuse patterns of a specific area. The area based statistics and a standard overlaying feature of a GIS allows for the estimation of the areal distribution of rainfall based on physiographic parameters (Webster, 1992). In this way, a GIS system can be employed to the data due to the subsequence analysis for integrating, constructing and exploring relationships between the CHAPTER FIVE A Review on GIS Systems 1L4. different variables involved. Generally, the functionality of a GIS system allows for the construction of the elevation, aspect, distance and landuse maps. These maps can be compared with maps of rainfall distribution based on rain-gauge observations. Therefore, describing and explaining all possible variables which may contribute in the distribution and variation of such rainfall, is made possible by GIS.

4) More importantly, overlaying techniques provided by GIS can also be used to create a set of new maps with specific aims (Bernhardsen, 1992). This should be done in the GIS environment, according to the rule based combination of maps and some specific overlay modules, it is possible to evaluate a set of values of maps. Berry (1993:111) states that, 'in GIS, overlaying maps go beyond traditional procedures of "sandwiching" map sheets on a light-table'. In a GIS, procedures for point-by-point, regionwide, and mapwide summaries can be described. Using such overlaying techniques in the GIS environment, for many climatological purposes, a series of further digital distributed maps of the physical environment can therefore be correlated to the rainfall distribution models. In GIS it is possible to specify, analysis and display several raster-base maps simultaneously. For example, the topographic information, which is referenced as the primary requirement for rainfall distribution information, is the essential database to be included in the GIS structure.

5) Finally, all maps provided by GIS can very efficiently convey information about the earth's surface with an adequate selection of the colour palette. By the assignment of colours to values of the variable and gradually varying the number of colours, considerably interesting effects can be achieved in the displaying and visualising of the topographic and rainfall maps (Max et al., 1993; Kelly, 1994).

Nowadays, GISs have enormous scientific importance and, more significantly, they are already being used to make valuable contributions to the understanding and solution of environmental problems. Currently, interest in GIS is expanding rapidly and it is therefore reasonable to expect that GIS should also be carefully used to solve some of the problems in climatology by a better modelling of information. In this way, although the climate will never be controlled, the use of accurate data and powerful computing technology and sophisticated software such as SPANS GIS may allow greater access to methods of monitoring rainfall distribution patterns spatially.

To sum up, the GIS as a representative of recent technology, can not only be used to analyse climatic variables, but can also be adapted to examine the spatial aspects of rainfall distribution. Therefore, some geographers (for example Maguire, 1989) think that there could be more advantages in using GIS in the study of the climatic variables such as CHAPTER FIVE A Review on GIS Systems 111

rainfall distribution, if they can be related to the other climatic or physiographic factors simultaneously.

5.7 Data Sources on GIS System

Although the learning of a GIS technique seems a difficult task, in the current study only a short time was spent establishing a database and in converting / translating existing maps and spatially-referenced data into a SPANS GIS system. The various types of data created for this study included elevation data, rainfall data including the geographic locations of rainfall stations, a basemap, a proximity map and a landuse map. These are outlined below together with a brief description of their sources.

The Digital Elevation Model (DEM) data was obtained from the Australian Surveying and Land Information Group (AUSLIG, 1993). These data are basically produced for mapping and geographic information systems. DEM data were collected by digitising all spot heights on 1:100 000 maps and selected points from 20 metre contours. Heights together with location information (on AMG Easting and Northing) were recorded in ASCII format. Using an excel computer program, the elevation data were first prepared in a specific text format. Then they were imported to the SPANS GIS environment to establish a digital elevation model which represents a continuous property of the topography in the region.

To model the general distribution of thunderstorm rainfall patterns in the study area, a sets of point data from rainfall stations (see Chapter 6) were entered into the GIS environment. Within the GIS, the rainfall data were analysed and integrated with other physiographic data, in the modelling of thunderstorm rainfall distribution in the Sydney region.

To evaluate rainfall maps by a number of physiographic parameters, some GIS internal methods have been used to construct the basemap, the proximity and landuse maps of the study area. For this purpose, a database was created for the study area and the digitised data was converted into raster or vector format and entered into the SPANS GIS (see Chapter 7 for more details).

5.8 Methods Used in a SPANS GIS

This section will explain the SPANS GIS module and all methods used in which the data have been analysed to construct GIS models. SPANS is a microcomputer-based geographic information system which were developed by Intera Tydac (established in 1982). SPANS GIS is currently being used and supported world wide by professionals and decision makers attempting to solve complex spatial problems. Therefore, the SPANS line CHAPTER FIVE A Review on GIS Systems 116 of software products, as a tool, was selected for the organisation, integration and analysis of the geographic information obtained for the spatial study of thunderstorm rainfall distribution in the Sydney region.

In this study, the power of the analytic and modelling capabilities of the SPANS GIS allowed the researcher to work with a climatic phenomenon of spatial nature (for example, modelling the distribution of rainfall). The SPANS version 5.3.1 was used for many spatial analyses in this study. Therefore, the following stages have been proposed as three more general categories: data entry, model building and model analysing procedures.

5.8.1. Data Input

Two GIS systems have been employed to enter the data and subsequent analysis. First, an Environmental Resource Mapping System, E-RMS (1989) was used for digitising a basemap of the Sydney region. The basemap boundary was entered by manually digitising from a 1:250,000 scale map of the Sydney region. The accuracy of digitising is estimated to being within 80 m2 of the indicated location on the map. The E-RMS system was developed by the National Parks and Wildlife Service of New South Wales. Using this system, a basemap of the Sydney region was entered in digital form, then edited and converted into a grid cell format to be exported. After that, the data export module of E- RMS allowed data to be exported to the SPANS GIS. Because many SPANS operations require a basemap, this studyfirst establishe d a basemap to define the boundaries of the study area in the SPANS GIS system. The basemap must be a binary map, that is, it could not contain classes other than 0 and 1.

In the second stage, a SPANS raster module was used to transform the basemap to a raster-base format to be imported into the GIS environment. Then, a SPANS was also used to enter the DEM data. Generally, this system accepts any ASCIfile, thi s can be data related to the location of rainfall stations and their associated attributes (rainfall amounts or other statistic values). A digital elevation data set consisting of approximately 20741 points (Australian Map Grid coordinates), together with elevation in meters was imported into a SPANS GIS. Six 1:100,000 scale maps covered the study area, four of which extend beyond it. The data approximates a 20 m grid which roughly covers an area of 9170.36 km2. These data were imported into a GIS coverage system showing point data and displayed on a computer screen to provide a visual impression of the distribution of sample elevation points. Imported data were then checked for possible errors or corrections. CHAPTER FIVE A Review on GIS Systems 111

5.8.2 Model Building

Once the data were integrated into SPANS, various techniques were used to analyse the data sets. A major part of the analysis involved the generation of elevation and rainfall maps from the point data sets. Several functions of GIS, for example, a contouring method, were used in SPANS to convert the data to thematic maps.

Firstly, a set of elevation data (with point structure) was used to create a digital elevation model by establishing topological relations between the elements using a rectangular grid (or elevation matrix) with a Triangulated Irregular Network method (TIN). TIN structures are based on triangular elements, with vertices at the sample points. Generally, the TIN surface can be constrained to pass through the point data. In this case, the contouring program was used to convert point data representing spatially continuous phenomena into classified, trend surface maps such as elevation maps which were used then for further analysis. After the TIN was created, some classification schemes, for each specific data, were applied to produce the desired classifications. The accuracy and reliability of this technique has been computed by Weibel and Heller (1991). They found that the surface models can be used to create, analyse and display surface information.

Also, the SPANS contouring module which interpolates a surface map from a point data set through a process of triangulation, was used to generate maps of thunderstorm rainfall distribution based on rain-gauge station observations. This surface was constrained to pass through the data points. Generally SPANS GIS supports both linear and non-linear implementations and it allows extrapolation outside the convex hull defined by the data points. In this study, a linear interpolation model which computes a linear interpolation surface, was applied for the data. During data analysis stages, a query module containing a query capability was used to verify thefinal results. The query function of SPANS GIS was also used to perform and confirm all geographic information in relation to locations specified on the map layers.

Secondly, some of the information related to topography such as aspect and elevation maps were automatically produced in the SPANS environment. The DEM quadtree was used to create an aspect map which is measured in azimuth degrees. In SPANS GIS, a map of the aspect is computed from a grid elevation map. In fact, the aspect is the orientation of the steepest slope with respect to north and is computed as an angle clockwise from north. A slope facing north has an aspect of 0°, facing east, 90°, facing south 180°. If it is a flat surface (no slope) it has the value 360°. The aspect map derived was used to analyse and identify the relationship between rainfall distribution and exposure. CHAPTER FIVE A Review on GIS Systems JM

The next important specific aim was the production of a proximity map. Proximity to the sea was suggested to be an important physiographic variable affecting the distribution of thunderstorm rainfall patterns in the region. A distance map was therefore generated by creating concentric buffers, with 10 km distance (a arbitrary classification), around the average coastline. In this map 10 different classes have then been expressed as buffer zones which have been used in a proximity analysis of the thunderstorm rainfall distribution.

A set of satellite images and also hard-copy maps of the Sydney region were used to create the landuse map of the Sydney region which covers the whole of the study area. All procedures, which were taken to establish the 'specific landuse' map the Sydney region, are described in Chapter 7.

5.8.3. Model Analysing

Within a SPANS GIS there are several analytical functions which allow a user to explore the possible relationships between the data sets and associated map layers. One such function was, for example, used to determine the average thunderstorm rainfall amounts for each of the topographic, proximity and landuse classes.

An area-based analysis function of a GIS was also used to analyse single map characteristics or determine the content of each area covered by different topographic classes. A statistical report was then produced, for example, to give the average rainfall for each class.

In addition, using an area cross-tabulation technique it was attempted to find the extent of the correlations between the digital elevation map, the aspect map, the proximity classes and the landuse patterns with thunderstorm rainfall distribution. Generally, the results of the cross tabulation can indicate the possible correlations between the two map layers. Statistically, chi-square coefficients are used as the measure of the degree of correlation, association or dependence of a thunderstorm rainfall map to the topographic maps. Some examples of the GIS functions which have been used for data building or data analysing have been given in the current chapter, other advanced GIS functions, which could be used in the analysing of thunderstorm rainfall data, are given in detail in Chapter 7.

5.9 GIS Potential Errors

There are some advantages in using the GIS method in evaluating the spatial distribution of rainfall from thunderstorms. One of the advantages is the ability to model and display the results as colour maps which show the spatial pattern of rainfall variation over the study area. The contouring approach is probably the most used and thus the most familiar CHAPTER FIVE A Review on GIS Systems 119 to those who are interested in studying rainfall distribution in space. The most significant advantage of using a SPANS GIS, in this study, lies in its modelling capabilities. With SPANS, simple to complex models have been generated, modified, and regenerated to be compared to the originals in a matter of hours.

However, there are two main potential errors or problems with using a SPANS GIS. Firstly, SPANS's TIN method, as an interpolation technique, was required to approximate the surface behaviour between sample points. The SPANS contouring module interpolates a surface from a point dataset through a process of triangulation which honours the data points. In the triangulation method, the surface passes exactly through each known data value, and interpolation is only affected by the heights at the three vertices. Thus, the size of the zone of influnce of a point is affected by the density of the surrounding points (rainfall stations). In small triangles (dense points) the effective zone of influnce of a single observation is correspondingly small, whereas in large triangles (sparse points) the zone of influnce is large. TIN method is desirable in cases where the values at the data points are known to have relatively small errors, such as elevation data. However, where the samples of a surface are associated with errors due to sampling and measurement, such as rainfall observations, there are relatively large errors as compared with the overall spatial variation. In such situation, an alternative interpolation method that produce smooth surfaces, and do not necessarily honour the data points, should be used (Bonham-Carter, 1994).

Secondly, because the interpolation technique in SPANS GIS is based on a raster based format, the resultant rainfall maps may not be completely smoothed to reduce noise in the data. It is because, surface modelling of spatially continousfield variables (such as rainfall values) involves interpolation from the irregularly-spaced samples to a raster format. Each interpolated point is simply a cell over which the variable (rainfall or topographic values) is constant. The resultant raster map has a relatively continous values dependent upon classes used rather than discrete values of observed in thefield. It must be noted that, in this study, the data taken from rainfall stations for interpolation purposes had discrete nature. GIS technique was, however, successfully used (Skidmore 1989 and 1990) to interpolate the rainfall data, digital terrain data, and to identify terrain position and to calculate aspect values from a girded digital elevation model.

5.10 Summary and Conclusion

The SPANS GIS can helpe to visualise, organise, combine, analyse, model and question the real data from Sydney's climatic environment which has been spatially organised in a computer environment. In other words, the power of a GIS is in its ability to integrate, manipulate, and process data from different sources. Data with spatial and non-spatial CHAPTER FIVE A Review on GIS Systems 120. nature could be handle to provide information and models which aid in the understanding of the Sydney's geographic location and its attributes. In Chapter 6, the GIS interpolation techniques could be used for mapping of thunderstorm rainfall variations throughout the Sydney region.

The combination and display of map layers in pairs is also an important aspect of GIS, because it allows the examination of spatial relationships between spatial phenomena such as a rainfall map and a topographic map, for example. Although the ultimate goal of most GIS studies involves multiple data layers, the relationship between map pairs is often an exploratoryfirst step, and may determine how features of one or both maps are to be enhanced or extracted for subsequent analysis (see Chapter 7).

Multiple maps could be obtained using overlay techniques in a GIS environment. The ultimate purpose of this study would be to combine spatial data from diverse sources together, inorder to describe and analyse interactions, to make models, and to provide support for decision-makers. Chapter 7 will, therefore, present some models of interst to climatologists, to show how they can be implemented in a GIS environment. Multiple maps also help to illustrate the models with reference to two applications: selection of geographic areas which are suspect for the highest rainfall values, and their associations with physiographic maps of the Sydney region. These techniques could be also supported using statistical procedures. It may be concluded that although the topic of GIS and the climate is relatvely a new notion in Australia, some functions of GIS can be applied to the raw data sets to create new products such as rainfall maps. Other products such as landuse, aspect and elevation maps could also be suited for climatic applications. CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 121

CHAPTER 6

THE SPATIAL VARIATION AND DISTRIBUTION OF THUNDERSTORM RAINFALL

6.1 Introduction

Information on the distribution of thunderstorm rainfall in space is very important in a variety of applications. In the Sydney region, the knowledge of the spatial variation and distribution of thunderstorm rainfall is limited to a few case studies (Bahr et al., 1973; Morgan, 1979a; Nanson and Hean, 1985). Although various aspects of the thunderstorm activity of the Sydney region have been examined over the years, two aspects that have not been thoroughly examined are: the long term variability of thunderstorm rainfall of the region over a long period of time, and the relationships between the local physiographic parameters and the distribution of thunderstorm rainfall patterns for the whole of the Sydney region. The former is examined in the current chapter. The latter is the subject of the next chapter.

The purpose of this chapter is to analyse the spatial variation and distribution of thunderstorm rainfall in the study area. First, in sections 2 and 3 the data and methods used are described respectively. Section 4 examines the methodology developed for the generation of reliable data on daily thunderstorm rainfall events, based on specific criteria. Then, in section 5, the spatial variation of thunderstorm rainfall in the Sydney region, using gamma distributions, is analysed. The gamma distribution is used to find the probability distribution of thunderstorm rainfall amounts at each rainfall station. Finally, in section 6, the spatial distribution of thunderstorm rainfall patterns are constructed using a GIS technique. Thunderstorm rainfall maps are based on average seasonal values and the biggest thunderstorm rainfall events for each month used in the study.

6.2 Data Selection

Unlike most previous studies that examined single but major thunderstorm rainfall events over a short period in the region, (for example, Williams, 1984; Colquhoun and Shepherd, 1985), the present study views the distribution of thunderstorm rainfall at two long time- scales namely Spring (October, November, December) and Summer (January, February, CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 122

March) and over 34 time period. The study also considered the largest thunderstorm rainfall events for each of the warm months (October to March). The definition of seasons is not conventional by Bureau of Meteorology standards. There are two reasons for this choice. Firstly, as it was shown in Chapter 4, three important climatic factors (air, sea temperature and air humidity) are high during October to March in comparison to the other months of the year. This seasonality affected the temporal and spatial distribution of thunderstorms considerably (see Chapter 2). As shown in Chapter 3, the warm months clearly dominate thunderstorm rainfall with the most thunderstorm activity, and more than 72 per cent of thunderstorm rainfall. Secondly, in the cold months of the year including September, thunderstorm rainfall is not common enough to give a reliable indication of thunderstorm distribution throughout the region.

The precipitation data were provided in three data sets supplied by the Bureau of Meteorology and the Sydney Water Board. The original data set was collected on a daily basis and was converted to monthly or seasonal records, where needed. Thunderstorm-day records were extracted from the original 3 data sets, using three different computer programs written for this purpose (see Appendix A, computer programs 2, 3 and 4).

A prime consideration of the present study was to determine the spatial variation and distribution of thunderstorm rainfall over a region that extended beyond the Sydney Metropolitan area for as long a time-span as possible. During the processing of the data set, it was noted that for some stations there was a considerable amount of missing data. Stations with less than 10 years of records were excluded from the analysis. Mooley and Crutcher (1968) in a study of rainfall in India investigated the number of years of record needed to stabilise the gamma parameters. Although Weisner (1970) indicated that from 25 to 50 observations of precipitation data are needed to give a stable frequency distribution, Bridges and Haan (1972) estimated that with 100 observations, there is a negligible 0.6 percent chance of error (see Section 3 for functionality of gamma distribution). Therefore, in this study, in order to ensure stability in the statistics, stations with fewer than 100 thunderstorm observations for the entire period were excluded. The resulting data set consisted of 191 stations (134 from the Bureau of Meteorology and 57 from Sydney Water Board) covering the period 1960-1993.

To show that the two networks of rainfall stations were comparable the data were subjected to statistical techniques. First, to find any possible difference between the means of the 2 data sets, an analysis of the variance (ANO VA) technique was applied (Webster and Oliver, 1990). An F value of 2.36 was calculated showing that there is not a significant difference between the means of the two sets at the 0.05 level of significance (see Table 6.1). CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 123

Table 6.1 Difference between two sets of stations (the Sydney Water and the Bureau of Meteorology) according to their rainfall means. __ Number of Mean in F-test Groups Stations mm Std. Dev. (SWs vs. BMs) Sydney Water stations (SWs) 57 14.57 1.769

Bureau of Meteorology stations 134 14.99 1.761 F = 2.362* (BMs) T = 1.537 Total 191 * non-significant at 0.05 level

To confirm this result, the NNA technique (see Chapter 3) was applied to the data. This technique takes into account the range and structure of the data at each rainfall station. All rainfall stations from the Sydney Water Board and the Bureau of Meteorology clustered based on 8 optimum clusters. The results indicate that the Sydney Water Board rainfall stations (57 stations) were randomly interspersed with the Bureau of Meteorology stations (134 stations) at the 05 level of significance.

Finally, to test the results of the NNA technique,the nearest Bureau of Meteorology station was found to each of the 57 Sydney Water Board stations (see Table 6.2, Appendix B) using the SPANS GIS Spatial Query function (see Chapter 5). For each pair of those stations, the commonality statistical association in thunderstorm rainfall was then evaluated using the correlation coefficient (Hutchinson, 1970; Davis, 1973). Only rainfall grater least 0.5 mm at each pair of stations for Spring and Summer thunderstorms from 1960 to 1993 was used in this analysis. These correlations were then plotted against distance and a linear regression performed on the results (Figure 6.1). This model describes the relation between the correlation coefficient and the interstation distance of rainfall stations (Stol, 1972). Figure 6.1 indicates that the correlation coefficient between the pairs of stations decreases with increasing distance (n = 0.957) in a linear fashion. The best corresponds occurs when stations from the two data sets lie within 5 Km of each other. Within that distance the correlation between paired stations approximates 0.9. CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 124

Regression Fit Y= 0.956853-1.64E-02X R-Squared = 0.957 95.0% Confidence Bands 95.0% Prediction Bands

1.0 - 0.9 -

•s °-7 - tx °-6 - 0.5 - 0.4 - 0.3 - v '-••"fv-™"mwr "• r'"'"1 i r n T™ m'T 0 5 10 15 20 25 30 35 Distance Figure 6.1 Relation between correlation coefficient (r) and interstation-distance (in Km) of pairs of thunderstorm rainfalls in the region.

The results of the ANOVA, the NNA and the correlation coefficient techniques make it clear that there is no significant deifference, at the 0.05 level of significant, between the two networks of stations, in their recording of thunderstorm rainfall in the region. Accordingly a dense network (combining the Sydney Water Board and the Buerau of Meteorology stations) ensures that a reliable spatial distribution of thunderstorm rainfall over the study area could be constructed. The combined rainfall records covers an area of 9172.21 square kilometres.

It was impossible to separate the precipitation amounts for thunderstorms from the daily rainfall total received on any thunderstorm-day. As the works of Sharon and Kutiel (1986) suggest, most rainfall from an individual event containing a thunderstorm is likely to come from the convection associated with that thunder. Hence in this study, for each individual thunderstorm-day event which matched the criteria, it was assumed that all the rainfall in that thunderstorm-day was the result of thunder activity (the validity of this assumption will be discussed latter in Section 4). Three hundred and forty seven relatively intense and widespread events between October and March occurred in the Sydney region for the period 1960-1993.

The list of the rainfall stations used, together with their geographical coordinates and their elevations above mean sea level, are given in Table 6.2 (see Appendix B). The spatial 125

%

0£ IQl <_ 8 Z (fl c 0 13 00 tSJ a *— c "5 DC i c oe 09 0 ? > si- ost ID c o 10 D O d T- CO (0 CO o 0) CO CO CO CO h. CO 3 CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 126 distribution of these stations is shown in Figure 6.2. Generally, the distribution of stations reflects the distribution of major population consentrations, suburbs, dams, post offices and rail stations. The possibility of spatial bias exists with this type of sampling network, but no better one was available.

6.3 Techniques Used

Two sets of techniques were used to analyse the variation and spatial distribution of thunderstorm rainfall in the Sydney region. Firstly, in order to allow direct inter-station comparisons and tofind the probability distribution of thunderstorm rainfall amounts at each station, the gamma distribution was used. Thom (1958) introduced the gamma distribution with two estimators - which are in fact a minimal number of summarising measures. The mathematical functionality of the gamma distribution has been widely discussed by, for example, Thom (1958, 1968) and Shenton and Bowman (1970).

In fact, the probability density function is one of the statistical characteristic measurements of the spatial distribution of thunderstorm rainfall that should be determined. The gamma distribution with two parameters is therefore the most flexible class of probability density functions and has extensive applications in the analyse of rainfall data (Bridgman,1984) and thunderstorm rainfall modelling (Easterling and Robinson 1988).

This method was first introduced by Thom (1958) as a frequency analysis:

*, ^ l r-i ~XIP P>0 m f(x) = -z^—-xrle ; ' 0) P r(y) 7 > 0 wherex is thunderstorm rainfall amount, beta (fi) is the shape parameter of the distribution, gamma or alpha (y ) is the slope parameter and r is the gamma function of ? • These parameters were estimated for each station by the maximum likelihood method (Thom, 1958). In this method the best estimate of gamma is given by

(2) A = Inx--Unxi n 1 + JTT473A 1 AA

_! + [! + Ajlnx -1 / rillnXi) 13]1/2 0r 7" Wnz-UnHnXt) (4) CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 127 where x is the mean thunderstorm rainfall amount and n is the number of thunderstorm rainfall occurrences on a daily basis in the data set. The best estimate of beta is then

P=X/Y (5)

Some curves from the family of gamma distributions, and probability density functions, are graphically illustrated in Figure 6.3 for different values of a and p. a is the shape parameter which determines the peakedness of the curve. Figure 6.3 (a) shows that the probability density function can take the shape of an exponentially decaying curve (a = 1) or the shape of a normal distribution (a = 4). A value below 1 indicates an exponential decrease in the probability density function from a maximum of infinity at zero x. When a equals one, the exponentially decreasing curve has a zero x intercept at 1/p. A value greater than unity produces a curve with zero probability density function at zero x, a rapid increase to the maximum probability density function and a slow decay thereafter. As a increases, the peakedness and skewness of the curve decreases, and the curve approaches that of the normal Gaussian distribution. This implies that as a increases the range of x commonly occurring also increases. The skewness of the probability density function increases and the peakedness decreases, as P increases.

0.8 0.2 •

\ cc = l

^ 0.4 *** 0.1 /Ty=2 0.2 • / \A~N?=4 0.05 R = 4 -^^a = 8

0 5 10 15 2 0 0 2 4 6 8 1 0 X X

(a) (b)

Figure 6.3 The gamma density function for (a) four a values, P = 1 and (b) three P values, a = 4.

To calculate the alpha and beta values, a computer program was thus written using MATLAB, the MATrix LABoratory programming language (1994). Both alpha and beta values were extracted for 191 rainfall stations in the Sydney region. Then alpha and beta values were contoured over the study area using the SURFER computer program, version 5.01 (1994). This is an objective mapping technique which interpolates to a grid by fitting CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 128 a polynomial to a set number of control points surrounding each grid point. Results were subdivided by season - spring (October to December) and summer (January to March), representing the two main seasonal thunderstorm rainfall regimes in the region for the period 1960-93. There are some distinct advantages for using gamma function. For example, it is one of the best probability models which could be fitted to data with a skewed nature, as it summarises thunderstorm rainfalls data by a minimal number of measures. Also, using gamma function, it is possible to group all stations by similar alpha or beta values, with direct comparisons between stations.

In the second stage, the thunderstorm rainfall data were mapped using the SPANS GIS interpolation module. This is an objective mapping technique which interpolates data with a spatial nature using a "Triangulated Irregular Network" (TIN) interpolation method (Intera Tydac, 1993). It is less likely that areas with sparse distribution of rain-gauges can lead to some overestimated problems using this interpolation technique. For obtaining the general distribution of thunderstorm rainfall models (and evaluating rainfall by a number of physiographic parameters in the following chapter) a standard GIS linear interpolation method was used. In this way, an average depth for two seasons, mainly Spring (October to December) and Summer (January to March) from 1960 to 1993, were determined. Also, for each month of the warm season (October to March) the biggest thunderstorm rainfall events from the record were analysed. These maps are based on data from 191 rainfall stations (see chapters 5 and 7 for more details about GIS techniques).

6.4 Thunderstorm Rainfall Selection Criteria

Thunderstorms occur randomly in time and space (Duckstein et al., 1973; Fogel and Hyun, 1990). The amount of rainfall shown in chapters 3 and 4, is strongly skewed over time and space, low rainfall occurs often during thunderstorms while heavy and intense rainfalls are rare. Most previous work considered thunderstorm rainfall to be convective no matter what the rate was, or the season of occurrence. For example, the U.S. Weather Bureau (1947) examined rainfall totals for days with thunder and daily totals for the summer months in the USA. In another work Sharon and Kutiel (1986) isolated heavy convective rainfall by assuming a rainfall rate of 20 mm/hr or more. Both of these studies assumed the majority of rainfall to be convective in nature by the season of occurrence.

In order to study only significant thunderstorm rainfall events across the whole Sydney region, the data were constrained using a number of analytic stages and associated criteria as follows:

In the first step, by using the NNA method, the associations between 15 thunder-recording stations, in the Sydney region were found for thunder-days in which thunderstorms have been observed (see Chapter 3 for more details). The results have indicated that in the CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 129

Sydney region, there is a definite statistical relationship between many of the stations when there is a thunderstorm-day. These associations are strong among some stations. According to these associations, all 15 stations were then clustered into seven groups. Each similar group may represent an area for the development or occurrence of thunderstorm activity, in the region. In fact each cluster can be considered to be a representative of an area for thunderstorm activity based on general geographic characteristics (for example distance or locality). Associations obtained may indicate a relative interdependence among different clusters statistically, and as a result, suggest a common thunderstorm activity for the whole of the Sydney region.

Camden Airport Wollongong Bankstown •«.» 0.1 •9.9 0.1 • 9.9 ' 0.1 1972-93 (172 Events) % laaathaji 1972-93 (139 Events) %L*a»m*n 1969-92 (2S7 Events) «lMI Owl M.a i.e •9.0 1.0 •9.0 1.0 • • " M.O 3.0 OS.O . 3.0 •3.0 Population 1 • S.O •0.0 10.0 90.0 to.o • 0.0 10.0 M.O 20.0 •0.0 -20.0 10.0 -20.0 Tfl.l! V Population 1 30.0 70.0 I Population 1 .30. 0 70.0 30.0 M.O 40.0 •0.0 .40. 0 •0.0 .40. 0 M.O so.o 50.0 50.0 30.0 50.0 40.0 so.o 40.0 GO.O 40.0 CO.O M.O 70.0 30.0 • 70.0 30.0 - \^"***^V 70.0 20.0 so.a 20.0 M.O 20.0 • 1 ^^^^^ Population 2 •0.0 10.0 - ^^»w Papulation : 90.0 10.0 \ ~^»-— Population 3 -90 0 10.0 " ' ^"***^^ •0.0 S.O 95.0 9S.0 os.o 3.0 • 3.0 1.0 so.o 1.0 99.0 1.0 9t.O XEttmrnimiKM 0.1 % Excaadanea *!. Exoaadanca 99.0 0.1 -99.9 0.1 o*.9 I , I •

14.0 2S.0 42.0 56.0 7 .0 40.0 SO.O 120.0 isoo 2C 0.0 14.0 2B.0 42.0 56 0 70.0 Thunderstorm Rainfall in mm 1 Katoomba Richmond •f.fl 0.1 19.9 0.1 1987-93 (152 Events] % Law Dun 1960-93 (650 Events) % Lasa than " M.O 1.0 99 0 1.0

M.0 5.0 95.0 . 5.0 •0.0 . Population 1 10.0 •0.0 10.0 •0.0 20.0 10.0 , Population 1 20.0 70.0 30 0 70.0 10.9 •0.0 40.0 10.0 .40. 0 50.0 SO 0 50.0 30.0 40.0 BO 0 40.0 eo o 30.0 30.0 70.0 • \\ 70.0 20.0 \ ^*^,^ Population 3 eo.a 20.0 •0.0 ro.o 90.0 to.o _'90. 0 S.O 95.0 5.0 95 0 _ \ ^^-^. Population 2 " Thunder-recording Extracted rainfall 1.0 1.0 99.0 99.0 Values XExcaadanca % Eicaaoinca "°~~ — Stations 99.9 0.1 99.9

12.0 24.0 38.0 4S.0 • .a 2o.o 40.0 so.o ao.o %100. 0 Richmond 11.0

-1 i 1 Katoomba 10.0 Bowral . Sydney Regional Office 0.1 99.9 0.1 Bowral 9.0 1905-93 (105 Events) *. k«" 0i»" 1960-93 (556 Events) % La«a thu 17.0 99.0 1 .0 Sydney R. 0. 99.o • Bankstown 7.5 as 0 - S.O 93.0 5.0 ao -10 0 •0.0 . 10.0 Camden Airport 7.0 ft Population 1 70.0 • 0 M.O Pop ula Hon 1 20.0 30.0 Wollongong 16.0 70 D -\ ; 70.0 30.0 to 0 - V so!o 80.0 40.0 so o - vv 60 0 so.a SO.O 40 ! 40.0 \ 60.0 70.0 SO.O Average Value 11.0 30 ° - \ *v . 70.0 20 o0 "- \•\-^7*'" ^v v 20.0 •0.0 10 RO'D . vy on X 10.0 90.0 s 0 - * ^^^^ as.a s.a : rv 93.0 • \ 09 0 -\ ^^"^, Population 3 1 0 1.0 —• —^^ 99.0 % £•itm*A«» MB 0 %ExcaManca ^— - 0.1 99 9

10.0 20.0 30.0 40 0 JO 0 40.0 10 0 120.0 150.0 200.0

Thunderstorm Rainfall In mm Thunderstorm Rainfall In mm

Figure 6.4 Probability of exceedence diagrams for 7 selected thunder-recording stations located in the Sydney region.

In the second step, daily thunderstorm rainfall amounts were studied in order to determine a standard value to be known as, a thunderstorm rainfall event, in the region. For this purpose the 'probability of exceedence' technique, modified by Bryant (1991) for the CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 130

Macintosh computer, was used. Probability of exceedence diagrams could be constructed by the following equation:

Exceedence probability = M(N+1)_1 100 % where N = the number of ranks M = the rank of the individual event (highest =1)

This technique was applied graphically using the amount of rainfall, recorded during thunderstorm events, on a special semi-logarithmic paper for each station. Note that the amount of rainfall values (in mm) from thunderstorms was plotted along the x- axis, while the probability of exceedence (in per cent) is plotted along the y-axis.

It is generally assumed that straight line segments plotted on this paper form normal populations. The intersection of the two straight lines forms the boundary between two normal sub-populations (see Figure 6.4). All thunderstorm rainfalls plotted in this fashion show two distinct populations on straight line segments. An advantage of this technique is that the probability of both low and high rainfall values from thunderstorms can be determined. The average value of these line intersections using data from 7 selected thunder-recording stations was 11 mm (see Table 6.3). Also, the average rainfall for all 3070 thunderstorm events (taken from all thunder-recording stations) was 10.4 mm for the region. This analysis indicates that there appears to be 2 types of rainfall associated with thunderstorms in the Sydney region. In thefirst type , small thunderstorms with less than 11 mm of rainfall are very common, representing 95 per cent of all thunderstorms. In the second type, rainfall exceeding 11 mm was rare but copious. This latter event represented 5 per cent of all thunderstorms, accompanied with extreme rainfalls.

Table 6.3 Thunderstorm rainfall values extracted from the intersection of two populations using probability exceedence graphs, for the Sydney region.

Row Thunder-recording Year Extracted rainfall

Stations From- To values 1 Katoomba 1987-93 10 2 Bowral 1975-92 9 3 Richmond 1960-93 11 4 Camden Airport 1972-93 7 5 Bankstown 1969-92 7.5 6 Sydney RO. 1960-93 17 7 Wollongong 1972-93 16 Average 11.0

For this work, a thunderstorm-day event was, therefore, defined as the occurrence of any storm with at least 11 mm of rain or more in at least one of these 7 stations. Accordingly, CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 13J. three criteria had to be met before a thunderstorm's rainfall was considered a significant event.

Criteria 1. Three thunder-recording stations had to record a thunderstorm on the same day - at least 2 of the clusters of stations, defined in chapter 3, had to be represented.

Criteria 2: At least one of these stations had to be a main station: Richmond and Sydney Regional Office. The Sydney Airport station could be also the main station, because it does appear to associate well with the Sydney Regional Office station.

Criteria 3: At least one of the main stations had to have 11 or more millimetres of daily rainfall on the day of a thunderstorm.

In this way a thunderstorm rainfall event could be characterised by above-mentioned criteria. Thus, all rainfall data recorded by any rainfall station - on a thunderstorm day which met this study's criteria - were considered to come from a or more convection systems which introduced thunderstorm activity for whole of the Sydney region. Accordingly, three hundred and forty seven common thunderstorm-days were selected for the period 1960 to 1993 to be used in a spatial analysis.

6.5 Spatial Variability of Thunderstorm Rainfall

The spatial variation of thunderstorm rainfall in the Sydney region is illustrated best by values of the gamma distribution. The gamma distribution has been used in several research works associated with rainfall and thunderstorms (for example, Simpson, 1972; Robinson and Easterling, 1988). A similar approach, using the two-parameter gamma distribution (the empirical distribution of Thorn's maximum likelihood estimators), was adopted for use in this study. Estimates for alpha and beta were estimated using only those observations where measurable rainfall occurred. Hence, the two parameters, alpha and beta, describe the probability distribution of rainfall amounts only from thunderstorm observations giving measurable rainfall which match with this study's criteria.

In spring, the spatial distribution of the alpha parameter is shown in Figure 6.5. The lowest alpha values (which indicate a high probability of rainfall amounts from thunderstorms), less than 0.8, can occur in the eastern parts of the Sydney region, over the Metropolitan area and to the north of the City. Alpha values between 0.7 and 0.9 can be also seen over the two major topographic features of the study area, mainly over the Blue Mountains and the Illawarra Plateau.

In the central part of the region and to the south-west (south of Bowral), alpha values increase to greater than 1.1. The geographical distribution of the values for alpha, shown CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 132 by Figure 6.5. The lowest values occur south-east of the region, where alphas with less than 0.70 can be seen. Areas which are located in the southwest of the Richmond and Camden station appear to have the highest alpha values (alpha > 1.2).

Longitude On decimal)

Figure 6.5 Geographical distribution of alpha value, Spring (Oct to Dec), over the Sydney region. H designates areas of alpha > 1.1, in contrast L represents areas of alpha < 0.8.

The geographical distribution of beta for spring (October to December) is shown in Figure 6.6. Beta values greater than 32 - which indicate a high probability of rainfall amounts from thunderstorms - can be seen for the topography of the south-east of the study area, just over the Illawarra Plateau. In comparison with the lowlands of the Sydney region. Over the Blue Mountains, beta values are also relatively high, more than 22 for Katoomba. Again, beta values more than 23 can also be seen over the eastern part of the City. The central part of the Sydney region, for example in the south-west of Richmond, and areas located in the south-western corner of the study area (at Bowral), show very low beta values, less than 12 on average.

The summer situation is considerably different, because the alpha values, to some degree, increase over the Illawarra Plateau, and an unclear pattern dominates the non-coastal areas. The highest values for alpha occur in the middle portion of the Sydney region, for example near Campbelltown (alpha > 1.2), and in the south of Camden Airport, where alpha values exceed 1.4 (see Figure 6.7). CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 133

In contrast, along the coast, beta values show an extension of coastal influence, particularly over the Illawarra Escarpment, over the City and north of Sydney just over the Homsby Plateau (beta > 27). These values generally decrease moving towards the west and south­ west, over the central parts of the study area. The values for beta are generally quite low in this area, being less than 10 near and south-west of the Camden Airport (Figure 6.8).

Figure 6.6 Geographical distribution of beta value, Spring (Oct to Dec), over the Sydney region. H designates areas of beta > 20, in contrast L represents areas of beta < 11.

Generally, the distribution patterns indicate four main thunderstorm rainfall areas in the Sydney region, better shown by the alpha and beta values. Over the eastern part of the study area extremely high beta values can be seen (a high beta value represents a high probability of rainfall amounts from thunderstorms) indicating that, the coastal location is very important in the distribution of thunderstorm rainfall amounts.

In contrast with the centre of the study area, which shows a greater amount of spatial stability for low beta values, the Metropolitan area and the nearby northern suburbs show sharply increasing or decreasing beta values. For example, over the City there are high beta (or low alpha) values indicating the significant centres for the occurrence of thunderstorm rainfalls. Also, the areas most likely to produce high rainfall from a thunderstorm event during the warm months include the northern suburbs of the City. Therefore, considering CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 134 both the alpha and beta values, the chance of high rainfall from thunderstorms over an area located in the centre of the Sydney region (for example, Prospect Dam) is much less than thunderstorm rainfalls which could occur over the Metropolitan area or its north-western suburbs.

Lougltudo (In decimal)

Figure 6.7 Geographical distribution of alpha value, Summer (Jan to Mar), over the Sydney region. H designates areas of alpha > 1.1, in contrast L represents areas of alpha < 0.8.

Over the Blue Mountains, in the north-west of the study area, the probability of the rainfall amount from thunderstorms is high. On average, over the Illawarra Plateau, located in the south-east of the region, the probability of the rainfall amount from thunderstorms is approximately twice the amount recorded over an area in the central low lands of the Sydney region. These areas tend to experience the lightest rainfall from thunderstorms.

The rest of the region including the central part of Sydney and the foothills and higher areas to the south-west comprise the fourth area, on average a low probability in thunderstorm rainfall. The lowest beta (or the highest alpha) values occur through a large portion of the study area, in the centre of the Sydney region with a north-east to south­ west direction. CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 135

Longitude (in decimal)

Figure 6.8 Geographical distribution of beta value, Summer (Jan to Mar), over the Sydney region. H designates areas of beta > 20, in contrast L represents areas of beta < 10.

The probability distributions of thunderstorm rainfall amounts at each station were summarised by use of the gamma distribution for total events for the spring and summer seasons. These probabilities have established different spatial patterns throughout the study area. In other words, the curve-fitting process and calculation of alpha and beta values, allows us to map and describe the probability distributions, but does not reveal how much rain may fall from an individual thunderstorm. To test the reliability of this technique in modelling thunderstorm rainfall distribution, there is a need to compare results with surface rainfall maps. Thus, in the following section the 'average thunderstorm rainfall value' is the object of primary interest, not only for comparison purposes with probability patterns, but also for mapping its variations in rainfall over the Sydney region. The basic question, from a climatic view point, is' what is the spatial distribution of average thunderstorm rainfall patterns, taking into account the associated data which were used for gamma distributions?

6.6 Spatial Distribution of Thunderstorm Rainfall

The study of the spatial organisation and distribution of thunderstorm rainfall has always been an important factor to many climatologists and meteorologists (Hobbs, 1972; Sharon, 1983; Bryant, 1991 and Batt, et al., 1995). This section generalises the point observations CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 136

(rainfall data) into a spatial maps of thunderstorm rainfall (mean area rainfall values) over the Sydney region. Thunderstorm rainfall maps were constructed using a standard GIS interpolation method. Constructed GIS maps successfully show the apparent detail of rainfall variation over the study area described by alpha and beta values. These should provide a logical and effective approach to the derivation of an area mean. The mean is obtained by applying an area weighting (TIN) interpolation method to the mean rainfall between each pair of consecutive isohyets. Spatial distributions of two seasonal averages and the biggest daily events for each thundery month, from 1960 to 1993, are described in separate sub-sections.

6.6.1 Average Event Values

The distribution maps obtained from the available data are most effective in showing the general trends in the distribution of thunderstorm rainfall, over the study area. The average thunderstorm rainfall map for spring is shown in Figure 6.9. The wettest part of the region is found in the south-east of the Sydney region at the top of and over the Illawarra Plateau where the rainfall per event averages between 20 and 23 mm. Just to the west of this region, in the Burragorang Valley, averages are less than 10 millimetres. Areas with thunderstorm rainfall in excess of 20 mm are found near and over the City. Generally, the coastal areas show high thunderstorm rainfall at this time of the year, and the high elevated areas, such as a section of the Blue Mountains, have more rainfall from thunderstorms.

Figure 6.10 displays the average thunderstorm rainfall per event for the summer season. During the period, the Blue Mountains (Katoomba) and the Wollongong Escarpment show distinct locations for thunderstorm rainfall distribution. In the east, over the Sydney Metropolitan area, there is a high average rainfall (more than 22 mm per event). Another location for rainfall maxima is the Hornsby Plateau which is located to the north of the Sydney. The driest areas are located in the low-lying central plain, with a north -south path in the region between the mountains in the west and the coastal areas. The south-west of the region shows an average of less than 10 mm rainfall from thunderstorms.

On average, figures 6.9 and 6.10 indicate that the maximum rainfall occurred over the eastern parts of the Sydney region, and more precisely on the north-west part of the City near Turramurra, over the Illawarra Plateau and Escarpment, and over a small section of the Blue Mountains. The minimum rainfall occurred in the south-west of the Sydney region and over inland areas. Of particular interest, are the low rainfall amounts over the Southern Tablelands (for example, at Bowral located in the south-west corner of the study area) and near the Campbelltown basin (just the west of Lucas Heights). 137

00 CD t o CJO CD 0 £1

O LU k. CD Q. c DC 00 1-Sh E <_ 0 c h oeost CD » CD > St-091- < to

O) 138

co I o Lfl » i» V J= •P™ 7™ C\i C\J o esj i i i i f"™* if—i ,"M ^f .iJCT 1— CO 1 C\j LCI 03 r-^ o CT) F» T™ T- CVJ CVJ

QE&-SI-

00 l-St^

Q£ (SI-

SI-OS I- CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 139

Although thunderstorm rainfall distribution is especially characterised by an extremely irregular area pattern, a marked gradient in rainfall from east to west can be observed. At the same time, these isohyet maps feature a number of cells or centres produced by thunderstorms over the City, northern suburbs and elevated areas. Generally, it can be seen that these rainfall distribution maps agree with gamma distributions patterns. However, some researchers, such as Chuan and Lockwood (1974) and Sumner (1988), have already suggested that individual thunderstorm events analysis may offer the soundest basis for determining the spatial distribution of thunderstorm rainfall. Therefore, in the second stage, the individual biggest events for the warm months, October to March respectively, have also been considered. In this way it was expected that the more intense thunderstorm rainfall events may be distinguished from the more uniform average thunderstorm rainfall patterns.

6.6.2 The Biggest Events

The impact of thunderstorm rainfall, from the point of view of flooding processes taking place in the Sydney region from time to time, is very important (Colquhoun and Shepherd, 1985). Therefore, the purpose of this section is to model the amount of rainfall from individual thunderstorms likely to be recorded on thunder-days with flash flooding in the region. Rainfall events, in this study, were defined by the daily occurrence of measurable precipitation for periods (days) with available data. This definition provided some heavy thunderstorm rainfall event samples that displayed relatively widespread characteristics.

Table 6.4 General descriptions for the 6 biggest thunderstorm rainfall events in the region. Events by Main synoptic Areas subject toflash flooding Month Date weather patterns

October 23-25 1987 Fronts and Low East of Sydney, Metropolitan area pressure systems Georges, Nepean and Hawkesbury Rivers November 5-12 1984 Troughs and Low Rose Bay, Kensington, Metropolitan area

pressure systems Hawkesbury Rivers and Mawaira area December 9-11 1988 High pressure system Sydney Metropolitan area, Illawarra are

and Fronts north-east of Sydney January 19-22 1991 Low pressure Turramurra, Parramatta areas

system and Fronts and north-west of City

February 7-11 1990 Tropical Cyclone Metropolitan area, north-western

"Nancy" and Troughs suburbs March 10-11 1975 Tropical Cyclone Dapto, Wollongong City, Metropolitan area

'Alison' and Troughs Mount Keira

All thunderstorm events, included in Table 6.4, were not defined according to their percipitation produced. They have been selected, because they were the biggest 140

CO 0)

CD s 0 Q 0 0 '—> •—i i 0 I I I I I i E Of-smoLnoLTj z B M LO i— r-CsJCMflO") 0 '5

oei-st

001-91=

3 CD 141

CO ^3- i^ <™°i- 13 i=°g 0) oo E E LO CSJ CSJ co i i i V LO LO I CD E LO CSJ CT7 CID OIO iI— CSJ e i- N M PO T LO £ CO < > LU o CO z >- z oei-si- z . 2 CM O fa B . Si s >• -> z2 c 1- .Is Xo 0 h- cc DC O 5- » ST ? s" CD TO c 8, QOS-SI- • §: * IX \999Wr~ ^W"""^M hb. o O 7? I _l X C m ,CC DC a.

o < 2 lQ£OSi= c 03 C .St OSt- so

CO

3 O) LL CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 142 thunderstorm rainfall events, in the data-base for each month, introducing floods in the region. The biggest events are, therefore, presented here to characterise thunderstorm rainfall in a climatologically more meaningful way. This also includes reference to the synoptic patterns leading to the situations in which these thunderstorm events occurred. The biggest and most widespread events for each of the 6 months, October to March are presented in Table 6.4.

October 1987: In this month five cold fronts moved through NSW and caused some widespread rainfalls (Bureau, of Meteorology, 1987b). However, the major features of the month were the low pressure systems that intensified some thunderstorm activities, bringing widespread rain to the region. These occurred most frequently in the eastern half of the State. On the 22nd and through to the 25th, a low pressure system moved from the west of the State to the north coast (see Appendix C, synoptic charts 6.1 from the 23nd to the 25th October, 1987b) and brought wide-spread thunderstorm rainfalls to the study area.

The rain was often heavy, particularly in the south-east of the study area. The heaviest rainfall for October occurred on, and adjacent to, the Illawarra Escarpment with Cataract Dam recording 600 mm rain which was a very high rainfall, well above the monthly average. The heaviest fall in a one day period during the month was 343 mm at Cataract Dam in the 24 hours to 9 am on the 25th. Another heavy thunderstorm rainfall occurred in the Metropolitan area. Severe thunderstorms brought minor to moderate flooding in the Metropolitan area overnight on the 24th / 25th, and, as a result, a number of deaths were attributed to the thunderstorms. At the same time, minor to moderate flooding occurred in the Georges River (north-east of Lucus Heights), Nepean-Hawkesbury River (west of Sydney Basin). Further details on the spatial distribution of rainfall from these thunderstorms are shown on Figure 6.11.

November 1984: On the 14th of this month, moving troughs, upper air disturbances, and a coastal low over the Tasman Sea (Jessup, 1985) together produced some very intense thunderstorm rainfalls, and as a result, floods in many parts of the Sydney region (synoptic charts 6.2 from the 5nd to the 12th November, 1984, see Appendix C). During this period atmospheric soundings showed that the airmass above Sydney was moist and unstable up to 300 mb. Also the temperature sounding had warmed moticeably. This event was marked by a period of intense thunderstorm rainfall activity which led to severe flash flooding in the Sydney Metropolitan area (Bureau of Meteorology, 1985). Flooding was most severe at Rose Bay and Kensington and other inner City areas. In the period from the 9th to the 12th, major flooding was reported at the Hawkesbury River, to the west of Sydney area. Rainfall was, however, very much above average in many coastal districts. 143

11 — i i •• II

%• <3 y**, CO CO a> T" LO O LO S Q '— O CO LO us eo en E as E CM 1 i i i i ^ £ 2 1 i™ ,_ ,„ ,_, ,_ CO E CSJ < c S CO LO O LO GO £ *— CM CO LO CO CO CD "5 0 CO < CC ••••• CD LU 0 IS T" z I * 0£!-9t 0) *»-^ . z . c i —> o < 2 r- X <& "o0^ Z X CD £ 1- DC >. 3" CO g CO . S . . . . C . B!" OOhSI- 1 Ifah, •X H - • C/5 5 •fi* O" i ' _1 X J fl e- -C • ^i .A jBE Jj :,'^W Q"C3 • ^ J^ i E pk • 0) r 0 •IPS? Fo< d QCOSt rt 1 (. J 13 • j^ o < h- .1 ~> - '' ~" 1st OS!- HH JO 3 V 8 8 •o CO CO CO CO CD CO 0) k. 3 O) LL 144

0)

LO GO LO GO CO E GO CO LO CO CO SO rt < ^ o E I 3 c *~• CD r™ CO T" CM LO CM CO LO CO CO C CsO < 03 rt LU CC i 2 CO CM QEISL CM

ootsL

oeoet

soost CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 145

During this time, Sydney received one of its highest rainfalls, more than 500 mm. The monthly rain totals for this month exceeded 300 mm in the Metropolitan and Illawarra areas. Extremely heavy rainfalls of over 580 mm occurred in Sydney's City area and the adjacent eastern suburbs and over the Illawarra Plateau in the south-west of the Sydney region. At Observatory Hill in the City, 235 mm was recorded in the period from 9 am on the 8th to 9 am on the 9th. This was the highest 24 hour November rainfall on record. The driest part of the Metropolitan area was in the outer western areas with rainfall totals of less than 150 mm. Sydney's rain was the highest for the State. By contrast the rainfalls in the western half were mostly well under 120 mm, except for Katoomba which had more than 300 mm of rain for the same period. Full details of the spatial distribution of thunderstorm rainfall for this event indicated that in the Sydney Metropolitan area the rainfall from thunderstorms on the 8th and the 9th gave a record-breaking intense rainfall for November over the City and near the City areas, particularly the suburbs just east of the City (see Figure 6.12).

December 1988: This month was dominated by a series of high pressure systems that moved into the Tasman Sea and directed warm, moist, unstable air across the State (Bureau, of Meteorology, 1988). By the 8th, coldfronts ha d moved into the Sydney area bringing thunderstorm activity (see Appendix C, synoptic charts 6.3 from the 9th to the 11th of December, 1988). Rainfalls from these thunderstorms caused considerable damage in the Sydney region. For example, on the 9th, a thunderstorm in the Metropolitan area dropped 35 mm of rain in 30 minutes at the Sydney Regional Office, where flash flooding resulted. During this event the Metropolitan area and the North-east of Sydney received more that 110 mm rain. The areas just to the west of Bankstown and the Illawarra Escarpment were subject to high intense rainfalls from these thunderstorms (see Figure 6.13).

January 1991: Between the 19th and the 22nd of January, 1991, a low pressure system accompanied by a series offronts located at the western side of the Sydney region - was developed over inland NSW bringing widespread thunderstorms over the Sydney region and causing considerable damage in the Metropolitan area from 19th to 22 January 1991 (Appendix C, synoptic charts 6.4). Some of the meteorological conditions such as low- level moisture, thermodynamic instability, and a lifting mechanism, reduced the atmospheric stability throughout the region, and as a result produced some thunderstorms. It was reported that a severe thunderstorm moved from the south-west of Camden, over Parramatta and through Turramurra and Palm Beach before moving out to sea (Bureau of Meteorology, 1991b). Very heavy rain, up to 90 mm, in a severe thunderstorm was reported in the north-west of the City. The heaviest rainfall occurred to the north-east of the Turramurra area under the growing thunderstorm cells (Spark and Casinader, 1995). CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 146

This caused extensive damage and minor flash flooding in the areas, just north of Parramatta. Generally, during this time, the central parts of the Sydney region, from the south-west to the north-east, received a large amount of rain from moving thunderstorms. Other parts of the study area had, on average, less than 20 mm of rain (see Figure 6.14)

February 1990: Between the 7th and the 11th February, 1990, after tropical cyclone "Nancy" passed southwards from the Coral Sea (where it had developed) over the Sydney region, a trough caused unstable conditions bringing thunderstorm activity (see Appendix C, synoptic charts 6.5 from the 7th to the 11th February 1990). These unstable conditions caused further thunderstorms and rains, particularly in the Metropolitan area and flash flooding in the northern and north-west suburbs (Bureau of Meteorology, 1990). Over this period, other parts of Sydney, specially the southern parts of the Sydney region, had less than 40 millimetres of rainfall (see Figure 6.15).

March 1975: Tropical cyclone "Alison" and other associated weather features such as upper level currents and a trough line dominated during thefirst half of this month (Bureau of Meteorology, 1975). As a result, thunderstorms developed in the Sydney area on the 10th and the 11th March, 1975. A moist air mass had been advected to the coast by a tropical cyclone well to the north. This air mass was almost saturated. An upper level low moved from the west across the area (see Appendix C, synoptic charts 6.6 from the 10th to the 11th March, 1975). Temperature and dew-point temperature soundings at Sydney Airport showed that the instability in the upper atmospheric was high enough to provide enouph buoyancy for thunderstorm development.

All these weather situations caused torrential rain of more than 440 mm to occur in the Illawarra district and Sydney Metropolitan areas. The isohyet Figure 6.16 shows March thunderstorm rainfall as an extremely localised event, with maximum falls over the south­ east of the study area. This event was described by Armstrong and Colquhoun (1976). They showed that the thunderstorm rainfall was concentrated over Kiama and Mount Keira just west of Wollongong. There was major flooding in Dapto, located south of Wollongong, and in the Sydney Metropolitan area which resulted in flash flooding. Except for a small area over the Blue Mountains, much of the western parts of the region, particularly over the central plains at Richmond and Camden, had rainfalls of under 50 millimetres.

In brief, from the general climatological point of view, thunderstorm rainfall distribution is meaningful in terms of the synoptic processes leading to such high variations in space. Spatial variations of the seasonal averages and thunderstorm rainfall values for the biggest events may, however, suggest that the great spatial variability of thunderstorm rainfall, in the Sydney region, can not only be attributed to the synoptic weather patterns. 147

O 0) CD CO GO LO CO ro r*. T" CM CM E 1 1 i i > E 1 I ,_ f- ,_ T~ eo (d CO f» *™ LO CO CO CM e CO P^ T"~ r- r-CM 3 •— "5 m co < LL < LU CO i N oei-si- C g "5 QC & c OOVQi-

c rt CC E

£_ o •#-r CD ID c 3 H I- U) T" (O V w. 3 DI­ LI 148

oe \-9\-

OO 1-S4-

oeosi-

SI OS I CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 149

6.7 Discussion

This chapter has presented details of spatial variation and distribution in thunderstorm rainfall resulting from the spring and summer averages, and from the biggest thunderstorm rainfall events for each of the six months (October to March for a 34 year period, from 1960 to 1993). The results described in this chapter, are only representative of the Sydney region with its prevailing climate type. Because of the arbitrary criterion for selection of the thunderstorm events , the results can only be generalised for relatively big and widespread thunderstorms.

The criteria outlined above (in section 6.4) were employed in the determination of the spatial distribution and variability of thunderstorm rainfall observations using data supplied by the Sydney Water and the Bureau of Meteorology of the Sydney region. Using these criteria, it was therefore suggested that over a long period of time there will be a series of thunderstorms passing over each individual station, on Sydney's thunderstorm-days, producing various precipitation amounts. Given a sufficient time period (at least 10 years in this study and more than 100 observations) this will produce general thunderstorm rainfall statistics. An immediate benefit of using this criteria is that it is possible to consider many thunderstorms which developed further, and became larger and lived longer, over the study area in the past 34 years. Examples of these types of events occurred in March 1975 and November 1984.

In order to allow direct inter-station comparisons, a gamma function was applied to thunderstorm data. The probability distribution of thunderstorm rainfall amounts at each station can be expressed by two summarising measures, the alpha and beta values. During the last decades, this technique has been widely used by many investigators in meteorology and climatology. For example, Mooley and Crutcher (1968) used it in India for rainfall analysis, while Simpson (1972) used the gamma distribution in single-cloud rainfall analysis in the south of Florida. This technique was also applied to the study of the spatial distribution of rainfall in the Hunter Region (located at the north of the NSW) by Bridgman (1984). Richardson (1982) modelled the distribution of daily rainfall amounts from 10 locations in the United States. Later, Easterling (1989) used this technique to differentiate between different regions in the USA in terms of thunderstorm rainfall patterns. Finally, Fogel and Hyun (1990) applied the gamma distribution technique to data to simulate the spatial variation of thunderstorm rainfall in the USA. They all concluded that, unless other models can be shown to have a clear advantage over the gamma distribution for a given application, the gamma distribution should be the appropriate choice of models for most applications. CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall 150

In the present study, the results obtained by using the gamma distributions have established three main spatial patterns in the Sydney region, indicating topographic, coastal and non- coastal (inland) areas. The distribution of mean thunderstorm rainfall amounts confirms the reliability of these patterns. The geographical distribution of beta and alpha values illustrate that, in total, coastal areas are subject to thunderstorms with a high probability of rainfall. In the south-west extension of the coastal area, over the Illawarra Escarpment, topography influences thunderstorm rainfall amounts. The relationship between rainfall and topography of the Illawarra Escarpment was already detected by several other researchers (Bryant, 1982 and Cox, 1983), where a distinct daily thunderstorm high rainfall amount can be also seen in both spring and summer season averages and even for some of the biggest event values. High thunderstorm rainfall totals in the vicinity of elevated topography of the Illawarra Escarpment suggest orographic enhancement of instability, particularly for sites facing the east. At these times of the year the prevailing easterly moist winds provide much of the moisture needed for thunderstorm activity in the region (Sumner, 1983b).

In the area west of Sydney, over the mountains, there appears to be at least two different patterns to thunderstorm rainfall. The Blue Mountains area, located in the north-west of the study area has high rainfall, particularly in the summer months. This may be due to orographic influences. Over the Southern Tablelands located in the south-west corner of the Sydney region, however, the topographic influence disappears, showing considerably lower rainfall amounts from thunderstorms. This low annual rainfall was confirmed by the Sydney's Weather Bureau (1979). A more recent investigation by Matthews and Geerts (1995) suggested that, in summer, thunderstorms were relatively less likely over the Southern Tablelands.

Along the coast, in the eastern part of the Sydney region, the greatest proportions of rainfall from thunderstorms occurs over the Central Business District (CBD) and over the Turramurra area just north of the City. The increased roughness associated with variations in tall buildings could also affect the spatial distribution of thunderstorm rainfall. The pronounced highest rainfalls over and nearby City may support the theory of the heat island, and particularly the mechanical effect of an extended urban area on rainfall enhancement by promoting atmospheric instability (Goldreich and Manes, 1979). The surface roughness, caused by many tall buildings mainly located in the Sydney's Metropolitan area, may increase mechanical turbulence, thus increasing the instability. In a case study over London it was suggested that the mechanical effect of the urban area may be of prime importance in the process of urban enhancement of precipitation (Atkinson, 1975).

As discussed in Chapter 2, these areas are also under the influence of sea-breeze circulations, which work with relatively high surface temperatures and readily available CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall Ml moisture from the nearby ocean to enhance convective rainfall. This influence may extend northward and southward along the coastlines during the summer months. The effect may also be due to the effect of urban heating or particles provided by pollution, both factors can enhance thunderstorm rainfall. These will be discussed in the following chapter.

In addition, the gamma parameters indicated that there was a relatively low incidence of high-intensity thunderstorm rainfall in the Sydney basin centred on the Hawkesbury - Nepean river between the Blue Mountains to the west and the coastal areas in the east. This low can extend over the adjacent mountain region and to the south-eastern parts of the study area. There, thunderstorm rainfall is predominantly low, especially for the biggest thunderstorm events. These general patterns of thunderstorm rainfall distributions may be attributed to the following phenomena.

First of all, the selected events can occur when some kind of low level cyclones, fronts, troughs or local convection systems are active overhead or in nearby areas. These systems are known to initiate or enhance thunderstorm activity in the region (Bryant, 1991). This is of course not surprising, since synoptic systems can potentially produce heavy thunderstorm rainfall over large areas (Speer and Geerts, 1994). The systems which are favoured for thunderstorm activities and associated rainfalls, have a maximum frequency in spring and summer (Matthews and Geerts, 1995). However, during the late spring and summer months, isolated convective thunderstorms may occur in any part of the study area. These events have also been known to contribute to the production of a high portion of total thunderstorm rainfall observations. Thunderstorm development in the Sydney region which was discussed in Chapter 2 supports this idea.

Secondly, the main rain-bearing depressions moving from the north-east to the south-west, or from the east to the west, can be influenced by the ocean when they are passing over the coast. Generally, very moist air over the NSW coast are advected from warm tropical and sub-tropical waters associated with the moisture of the East Australian current. These conditions occur predominantly in the period of January to May when the sea-surface temperatures are higher and the prevailing and saturated easterlies winds cause thunderstorms to move inland to adjacent coastal areas from this convection (Eagle and Geary, 1985). The passage from the sea to the land, together with the forced ascent due to physical environment effects, are thought to be responsible for some of the spatial patterns observed.

Superimposed on the above pattern are the effects of localised topography. It appears that the coastal area, where 'the Metropolitan area' is located, particularly on the Hornsby Plateau has experienced much larger thunderstorm rainfall amounts. Also, the uplands to the west of the Sydney region (Katoomba), and the Illawarra Escarpment located in the CHAPTER SLY Spatial Variation and Distribution of Thunderstorm Rainfall HI south-east of the region, show clear topographic enhancement of rainfall during thunderstorm rainfall events.

Although this work has used data from selected thunderstorm days for the Sydney region and also the six biggest major thunderstorm rainfall events from each month, the results obtained could be linked closely to the major synoptic weather patterns and physical features of the Sydney region. Because of the large number of observations used to determine thunderstorm rainfall values, if provided the best available mathematical estimate of the probability of thunderstorm rainfall amounts and, as a result, the true average amount of rainfall from thunderstorms could be expected. Generally, most of the studies in the region, cited above or in the literature, have explicitly concentrated upon one or two short terms, severe thunderstorm events (Batt, et al. 1995), or thunderstorm rainfall (Armstrong and Colquhoun, 1976). The present chapter, however, has focused on the study of thunderstorm rainfall in general during the warm seasons over a 34 year period in order to define more precisely the regional patterns of thunderstorm rainfall distribution over the longer period in the Sydney region.

6.8 Summary and Conclusion The purpose of this chapter was to focus attention on the patterns of the spatial variation and distribution of thunderstorm rainfall during the thundery months of the year (October to March). According to the applied criteria, the amount of rainfall recorded during a thunderstorm-day event was derived for 191 stations in the Sydney region by season for a 34-year period (from 1960 to 1993).

The probability distribution of thunderstorm rainfall amounts was described using the gamma distribution. This distribution provided two measures which describe the patterns of thunderstorm rainfall in the Sydney region well. The gamma distribution was therefore found to be a suitable technique for characterising the distribution of thunderstorm rainfall amounts at individual observation sites (rainfall stations). This method also allows specification of the probability distribution of rainfall amounts and has the potential to be a predictive tool. As a result, the gamma distribution, as a summarising function, could be regarded a basis tool for defining a thunderstorm rainfall climatology in the Sydney region. To compare these values with actual rainfalls, a GIS method was used to characterise the spatial distribution of thunderstorm rainfall patterns over the Sydney region.

The geographical representation of alpha and beta values, including the mean rainfall values from thunderstorms, indicated that there is considerable spatial variability in rainfall related to Sydney's physical environment. It seems that while the spatial distribution of thunderstorm rainfall follows a gradient between inland and coastal areas, it is also CHAPTER SIX Spatial Variation and Distribution of Thunderstorm Rainfall 153 influenced by the topography of the region. Two major topographic features of the region, such as the Illawarra Plateau and the Blue Mountains, are seen as areas with high rainfall.

Although the development of thunderstorms over different areas suggests that there are synoptic-scale processes which cause such thunderstorm rainfalls, these processes are most likely not solely responsible for the resulting variation and distribution patterns. More detailed analysis of thunderstorm rainfall amounts are required to explain the relationship between amounts and such parameters as proximity to sea, topography and landuse patterns. The next chapter pursues these relationships. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters ISA

CHAPTER 7

RELATIONSHIPS BETWEEN THUNDERSTORM RAINFALL AND PHYSIOGRAPHIC PARAMETERS

7.1 Introduction

The spatial analysis of thunderstorm rainfall has indicated that the distribution of thunderstorm rainfalls is highly variable over the Sydney region. This should be apparent from the results of the previous chapter, particularly in the case of widespread thunderstorm rainfall events. It was also argued that despite the role of the different synoptic air patterns, the distribution of thunderstorm rainfall in the region would be largely a function of physiographic parameters such as elevation, aspect, proximity to the sea and landuse patterns of the study area. The influence of each of these factors upon the spatial distribution of thunderstorm rainfall is examined in more detail throughout this chapter.

In sections 2 and 3 respectively, the data and methods used are described. In section 4, the elevation and aspects throughout the study area are consideredfirst, a s they are the most important physiographic parameters affecting the distribution of the thunderstorm rainfall event. Then, the relationship between the proximity to the sea and thunderstorm rainfall is analysed in section 5. After that, section 6 examines the possible spatial relationship between the landuse patterns and the distribution of rainfall from thunderstorms over the study area. In section 7, the areas affected by high thunderstorm rainfalls are highlighted firstly by using an overlay modelling technique of GIS. Then, a 'stepwise multiple regression' technique is applied to explore the statistical relationships amongst these variables. Finally, in section 8, the results obtained are discussed.

7.2 Data Used

In this chapter, some of the Geographical Information Systems (GIS) techniques and also simple-to-complex statistical methods (for example, a t-test or regression techniques) were used in order to describe the spatial nature of the data, and to explore the possible associations between thunderstorm rainfall amounts and several of the important physiographic parameters of the Sydney region.

Firstly, thunderstorm rainfall data sets, from the previous chapter, have been reassembled. Then, the average thunderstorm rainfall map was constructed by using data from the six CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 151 largest thunderstorm rainfall events which were taken for each thundery month (October to March, 1975 to 1993) (see Figure 7.1). These largest thunderstorm rainfall events were selected because they are quite important in the region (Colquhoun and Shephred, 1985; Bryant, 1991). There are several reasons for this criterion.

In the literature, the idea has been well established, that the importance of topography in enhancing the variations in thunderstorm rainfall distribution for each of the weather types (for example, frontal systems or air masses), is not the same (Osborn, 1982). Some researchers, for example Amanatidis et al., (1991) think that the relative importance of each of the topographic factors - such as elevation and aspect to the wind direction, or facing the sun - in generating thunderstorms may change from day to day, as weather conditions change. Climatologists such as Chuan and Lockwood (1974) and Passarelli and Boehme (1983) and Smith (1989) have pointed out that because of the localised nature of thunderstorms, topography does not always appear to have the same effect upon the distribution of thunderstorm rainfall over long and short time-spans. Analysis of the relationships between the average daily thunderstorm rainfall amounts (as they have been shown in Chapter 6) and physiographic parameters would be helpful in the understanding of the role of the Sydney's physical environment upon the rainfall distribution pattern. However, they are not ends in themselves. They are intended to provide a start for the analysis of individual widespread thunderstorms which are the soundest basis for determining topographic effects on precipitation deduced from such thunderstorms. This last aim is within the scope and content of the current chapter.

More importantly, the impact of large thunderstorm rainfall events - from the point of view of flooding processes which take place within the Sydney region from time to time - is a very serious concern, because of these disastrous consequences (Riley et al., 1985; Speer and Geerts, 1994). Therefore, the purpose of this chapter is to discover the relationships between physiographic parameters and the average of large thunderstorm rainfall events likely to be recorded on thunder-days with flash-flooding in the region (all thunder-days and associated floods have been mentioned already in Chapter 6). This procedure will, therefore, provide opportunities by which the relative effect of each physiographic parameter upon thunderstorm rainfall data, of a widespread nature, could be characterised. 156

to CO -*—I CD c I CD LO CO > o~> CD LU CC CD CD c CD CD CCS F CO CC E CD E CO *-" o o3 M— to i_ O) o CD E CD c CO ~o c Q. E CO —J O i_ C/O o O o sz o O CD o CM LO CO r- OJ T— OJ > CO CD 1 < "•w I 1 1 T— | M— CJ LO 00 CD LO to CD co c CO h- Q 'co DC Cf) US < CO < LL oe isi < DC

DC o f- C/) or LU Q z => X H oI LU DC > LU Z Q >• C/O -SI osi Cl) 3 LL CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 157

Other data sets and their sources which were used in this study, such as proximity to sea or landuse data are shown in Table 7.1.

Table 7.1 Origin of the data that used in Chapter 7.

Variables Data Sources and Origin Thunderstorm rainfall data a) Maps See Chapter 6 b) Point data-sets See Appendix D (Table 7.2)

Elevation Map Digital Elevation Model (DEM)

Aspect Map Created from DEM

Proximity Map Created by GIS Internal Functions

Landuse Map Using Landsat, Hard-Copy Maps and GIS Functions

7.3 Techniques Employed

To analysis the above-mentioned data sets, a GIS method and a set of statistical techniques were used in the study of thunderstorm rainfall in the region.

7.3.1 GIS Techniques Applied

Increasingly, GIS systems are currently becoming popular in many academic centres, for example, in universities, for storing, integrating, analysing and displaying various types of climatic data (Sajecki, 1991; Brignall, et al., 1991). There is a large amount of GIS software being used for spatial analysis purposes by geographers. In thefirst phase of data analysis, GIS techniques were, therefore, applied to explore the spatial characteristics of the variables.

The SPatial ANalysis System (SPANS), which is a microcomputer-based geographic information system, was used to analyse data in relation to the Sydney region. SPANS GIS is a software product of Tydac Technologies (now Intera-Tydac). The SPANS 5.3.1 version, using Operating System 2 (OS2) version Warp, is currently being used by the Geosciences School at the University of Wollongong. SPANS works with point, vector, raster and quadtreefiles (fo r more details see Chapter 5). In applying the SPANS GIS, for the purposes of this study, several steps had to be taken before using the other advanced GIS functions.

As a first step, the study area - the boundaries of the Sydney region - had to be established in a SPANS directory. This was done by means of a set-up menu in SPANS GIS. Using SPANS projection function, a projection (from the master projection list) for Thunderstorm Rainfall and Physiographic Parameters the Sydney region was defined. The geographic location of the Sydney region, located in the Southern Hemisphere, was assigned according to the Equator and the Prime Meridian.

The term "Study Area" is used in SPANS to define the location and description of the current project. Both SPANS and E-RMS, which was used for digitizing the basemap and landuse maps of the study area, require that the geographic co-ordinates of a square or rectangular boundary be defined for the study area / database (Table 7.3). These co­ ordinates define the rectangle in the projection plane containing the study area / database. Both systems - SPANS and E-RMS - however allow for irregular areas to be defined in the study area / database as a study site for analysis and modelling purposes. These are called the "Basemaps" and domain in SPANS and E-RMS respectively.

Table 7.3 Limits of the study area / database. Geographical Coordinating of the Study Area Min. Max.

Easting 247529 * Easting 353852 Longitude 150° 15' Longitude 151° 25'

Northing 6289469 Northing 6178739 Latitude -33° 30' Latitude -34° 30' * Easting and Northing in meters

The second step, was the preparing and formatting of data sets on thunderstorm rainfall in SPANS formats for the average daily thunderstorm rainfall from the six largest rainfall events (the biggest event of each thundery month was taken from October to March, from 1975 to 1993) using data from 152 rainfall stations located by latitudes and longitudes in Table 7.2 (see Appendix D). The results of this phase of data analysis have already been applied to create the thunderstorm rainfall maps, described in Chapter 6. Also, the digital elevation data (DEM) set was added at this stage to the SPANS GIS system. DEM data consisted of height digitised at 20 m contour intervals from 1:100,000 scale maps. Over 20741 points were used to describe the topography over the study area, with 80 m2 resolution in SPANS raster format.

At the third step, the SPANS contour module was used to produce raster-based maps. An analysis was made by means of a Triangular Irregular Network (TIN) grid between sets of data points. Interpolation could be linear or non-linear. A non-linear interpolation usually provides smoother rounded contours which are more visually appealing. However, there is a high possibility that with a non-linear interpolation, an isohyet line, for example, may be projected to an unrealistic value exceeding the observed 159

(1) E *£ v> to IP c o 03Q0QQ9DDP 05 N OCMrO^LOtDI^-COCni-A < C-K i i i i i i i i i y .C0Cni- X c JO CO < LU CO OS 191 CD c

o o CD U> CC 00 f-SJ- I < E 2 LL OS est Q. CC g SI-OS I-

£ O) • ••» LL 160

to CD to to as E O .^ o o *- 1C OTOT -3D C«O CD _ro CD OJ Q. 0 CO O LU c/> CO <

oe LSL o 'CD CD DC fc- CD [> C •o >. CO CD sz

a. co

o CD a. to < co CD

CD 91 091- LL CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 161

values. For this reason, linear interpolation was used to interpolate rainfall values on a raster-based format.

At the fourth step, the "proximity to the sea" of each rainfall station was created using automatic compilation of buffer zones. For this study, 10 kilometre wide zones were subjectively defined away from the coastline (see Figure 7.2). Also, using the digital elevation model (DEM), an aspect map was automatically produced in the SPANS GIS environment. In SPANS GIS, a map of the aspect, which is measured in azimuth degrees, is computed from a grid elevation map. The orientation of the steepest slope clockwise with respect to North is computed (Figure 7.3). A slope facing the sky has an aspect of 0°, facing East, 90°, facing South 180°. If it is facing the west, it has the value 270° (see Table 7.4).

Table 7.4 Aspect classes derived from the DEM model, using SPANS GIS

No. of Digital Numbers Classes Aspect Class Degree ° Lower Limit Upper Limit

1 Flat (None) 255 255 2 North (>337.5 to 22.5) 239 254 0 15 3 East (>22.5 to 157.5) 16 111 4 South (>157.5 to 202.5) 112 142 5 West (>202.5 to 237.5) 143 238

The fifthste p involved the addition of a landuse map of the SydneI y region to tl GIS model. However, no landuse map exists that is primarily concerned with climatological aspects.

7.3.1.1 Landuse Map of the Sydney Region

According to the literature (for example, Atkinson, 1969 and 1977) different landuse patterns create different meso-climates. Large urban areas, with their characteristically warmer urban climates and other thermodynamic effects, are able to enhance convectional rainfalls (Changnon, 1978; Landsberg, 1981). Therefore, the following procedures were taken to establish a 'tailer-made landuse map' for the Sydney region:

1) A set of satellite images were used to find the boundaries of different landuse patterns. 'Landsat' satellite images of the Sydney region cover the whole of the study area, and were recorded by USA (NASA) satellites on December 1972, October 1986 and November 1990. They use false colour composites of bands having different wavelengths. Such false colour images indicate variations in vegetation types and vigour and they have a Rainfall and ground resolution of 30 metres and measure the reflectance of light from the surface at 7 different wavelengths or band-passes.

2) The impact of man on the natural terrain of the Sydney region is clearly shown on these images. Urban development which is encroaching upon rural areas and natural wilderness reserves can easily be recognised using these reflectance colours; water appears as black to very dark-blue, depending on the amount of sediment and depth of water; vegetation is highly reflective in the infra-red band and shows as red. Natural bushland, in and around Sydney, appears red-brown. The areas of cultivation, especially along the Nepean - Hawkesbury river, are light-red. It was possible to delineate the textural variations and the fine detail of residential development in contrast to adjacent larger paddocks and cultivated lands.

Table 7.5 Description of landuse types in the Sydney region. Description _ Type Landuse Patterns Use / Structure Area Km2

1 Central Business District Metropolitan natural 7.9 0.09

(CBD) very dense built-up areas with skyscrapers 2 Industrial areas Airports, factories, 120.17 1.36 (IND) refineries 3 Urban-Residential, Barren Compact residential 807.75 9.13 area (URB) with separated treed areas 4 Urban-Residential, Treed Dispersed residential 784.48 8.87 area (URT) with intense treed area 5 Rural / Semi-Urban area Agricultural rural and 885.57 10.01 (RUS) light urbanisation 6 Rural / Open areas Agricultural rural 1292.87 14.61

(RUO) Grass, trees 7 Treed area, National / Dense vegetated areas 4947.67 55.93 Urban parks (TNP) (forest, grass lands) Dams, lakes and rivers of the study area have not been regarded as part of the landuse map.

3) Digital data produced for the urban development program by the GIS cartographic section of the Urban Planning Centre (1993) was used to refine the boundaries of forests and the existing urban and non-urban areas. 163

OT -1*— CO s ^ ti­ c r OT to (D rt CO nas er _Q CD OT CD CO CD CO CO CO co CO -Q D CD CO ' J c CO CEO 05 "cO E CD Q a." Q. rt CD 08 o c z CD CO 4-* o CD '*— _ o- OT CD •o U OT CD •a CO CO 3 Q CO CD OT 1— CO •a CQ CO Z3 OT CD CD CC CD CD i_ CC •c c O _J •o co 1- c DC CO • ••: _l c oe L9t o CD CD DC >* CD C "O (f) CD

CL CO

CD to •a c CO

-3- h-' CD i_ CD 91- 091- LL CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 161

4) The Sydney region outline plans, produced by the State Planning Authority of NSW for 1970 and 1994, were used to locate the main industrial areas.

5) These landuse maps were combined and digitised as impute data, using the E- RMS computer program. Then, the landuse data was imported into the SPANS GIS program and converted into a raster format. Seven classes, as shown in Table 7.5 and Figure 7.4, were delineated for thisfinal map .

Using the landuse map which was created for this study, the Central Business District (CBD) was known as landuse 'type 1'. The central portion of the City is the most developed, containing most of the major commercial developments. The centre of Sydney, with very dense built-up areas and a rough surface topography with tall skylines (see Plate 7.1). There are some small CBD nodes, for example Parramatta's CBD. Because of the scale of model used in this study, these areas could not be shown in the landuse map of the Sydney region.

Plate 7.1 Closeup view of heavy commercial landuse showing the part of CBD.

All major industries, factories and airports (type 2) have been categorised as 'Industrial areas (IND). Plate 7.2 gives an example. The term 'Urban-Residential Barren' (URB) area has been assigned to that area encompassed by the out-lying boundary of the dense residential areas (type 3) generally with less tree coverage (see Plate 7.3). CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters Ml

Plate 7.2 Closeup view of heavy industrial landuse (type 2).

Plate 7.3 View of compact residential landuse (type 3) in the Sydney region. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters J66

In contrast, the term 'Urban-Residential Treed' (URT) was applied to those areas (Plate 7.4) with less dense residential area and much more tree cover (type 4). It should also be pointed out that there are a few small recreational parks and cemeteries throughout the Metropolitan area.

Plate 7.4 View of light-moderate residential landuse (type 4). CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters J61

Plate 7.5 View of normal rural / semi-urban area (type 5).

Plate 7.6 Shows example of rural / open areas (type 6). CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 168

Plate 7.7 Closeup view of compact vegetated land cover (type 7) in the Sydney region.

If rural areas are being encroached upon by urban development, they were categorised as type 5 and they were simply called 'Rural / Semi-Urban' areas (RUS) (see Plate 7.5). But, cultivated areas or productive rural areas (type 6) with considerably more natural reserves were classified as the 'Rural / Open' areas (RUO). An example of this kind of landuse is shown by Plate 7.6.

Finally, all areas with considerable natural vegetation, including State forests, National parks, urban water catchments, and many major parks located in the region (type 7), were traced and defined as 'Treed National and Urban parks' (TNP) (see Plate 7.7).

As a result, seven landuse types comprise the major portion of the Sydney region which can be seen in Figure 7.4 and the associated Table 7.5. Also, plates 7.1 to 7.7, as representative of all landuse classes, illustrate a close-up view of different landuse patterns from the region

7.3.1.2 Advanced SPANS GIS Functions Used

The following analytical functions of the SPANS GIS were utilised to analysis the data and to produce thefinal products. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 162

1) A GIS allows simple statistical comparisons between maps by arranging the data in a contingency table (Bonham-Carter, 1994). In the GIS, a contingency table was, therefore,first use d to test the hypothesis that the area-based distribution of rainfall in the various categories of one map are independent of, or dependent on, the classes of the physiographic maps (as independent variables). Statistically, chi-square coefficients are used to be the measures of the degree of correlation, association or dependence of a thunderstorm rainfall map to the physiographic maps. In practice, because the contingency table varies in dimension, SPANS GIS uses three measures of association - such as Contingency coefficient, Tschuprow's T and Cramer coefficient - to measure the degree of correlation between two map layers as follows:

First, the contingency coefficient, C , is estimated by

ifo^+ril(GFX 2 + n

where

Theoretically C lies between 0 and 1 but does not always reach 1, even when the variable seem completely associated. In square tables (that is, I = J), for instance, its maximum value is ^(1 -X)lI.

Then, Tschuprow's T can be estimated by SPANS GIS. This estimator varies between 0 (for independence) and 1.0 (dependence), but it can only attain its maximum in square tables. T is calculated by

• O2 J GFX2 T"^|)(I-IXJ-I)""H(I-IXJ-I)

Also, the Cramer coefficient (V) corrects for some of the deficiencies of the contingency coefficient C and Tschuprow's T in that it achieves its maximum in asymmetric arrays (Intera Tydac, 1993). v varies between 0 (no correlation between maps) to a maximum value of 1 and, then, it is determined by

GFX2 (3) ~&in m Where m equals the number of classes in column and it is the smaller of (I - 1 or ( J -1), while n is the number of classes in row. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 170

In fact, the degrees of freedom for a two-sample test are given by the number of rows in the contingency table minus one multiplied by the number of columns minus one (Ebdon, 1985). In this study, the above-mentioned three measures of association were found to be effective analytical techniques in comparing the spatial distribution of thunderstorm rainfall amount (as a dependent variable) by the number of physiographic parameters (as independent variables).

2) In the second stage, the 'Map Analyse' functions of GIS were used to calculate the average areal distribution of rainfall from point measurement (rainfall stations). The averaging technique was found to be extremely useful when averaging rainfall values over fixed physiographic units such as topographic, the proximity zones or landuse classes. The results produced by this function display area-based mean thunderstorm rainfall values for each class of physiographic maps, for example, the landuse map of the Sydney region. Such procedures can also help to find new statistics (such as attribute means or totals) derived from the areal distribution of thunderstorm rainfall based on different classes of each physiographic parameter.

3) In the third stage, GIS internal reclassification methods and a multi-overlaying technique were used to analyses all maps (rainfall and physiographic maps) in the GIS environment. By using GIS reclassification techniques, it wasfirst determined which of these areas in the Sydney region, had the highest amounts of rainfall distribution, (more than 120 millimetres). This rate indicates the medium class on the thunderstorm rainfall map, and it was subjectively selected (see Figure 7.1). This arbitrary rainfall value was selected because in Chapter 6, the study of the 6 largest thunderstorm rainfall events indicated that there is not a definite rainfall value for the onset of floods in the region. It is possible that the conditions, by which these events occur, differ from one weather system (with different rainfall intensities and associated amounts) to another, and possibly, it also depends on the ground conditions at the time offlooding. Armstrong and Colquhoun (1976) defined a daily heavy thunderstorm rainfall with more than 100 mm isohyet extended over the Sydney region in March 1975. Therefore, the current study supposed that areas in the region which have rainfall amounts above the average values of 120 mm, are more prone to floods and, as a result, they must be visualised.

Then, the SPANS GIS modelling language, as a powerful analytical mapping tool, was employed to produce new maps which overlayed the rainfall map and all the independent physiographic models, simultaneously. This was made possible by writing equations to be understood by the SPANS modelling language system (see Table 7.6 located in Appendix E). The final productions are new maps each with a specific aim using the visualisation capabilities of GIS techniques. By this, it was possible to show areas, for example on a landuse map, subject to the highest rainfall from thunderstorms (see Figure 7.5 (a-d)). C , M 11 Oil S c Ss aj £ nj IB iici c 3 ™^2 D.D.' oooooooH jr «J D — '— oOi-rMrotriDU)«oa » T3 TJ — "D Q 3-13-s 2 y CJ _= cc a: oc I-

*

CO o O

LU LJJ >

M N •s; s ^ 3 ^ ^ oooooo jg o o ra o o o - ) OJ co ^r co cn T^ la. Z Z LU CO CO CO , 'oooooo !••• i— C*M• CO f^f- C 5O 8CT )

o o a> a Thunderstorm Rainfall and Physiographic Parameters

4) Finally, the 'Visualise' menu of SPANS which is an important tool in GIS, was used to annotate all maps and slides it produced. Rainfall maps and topographic features have been annotated with complete captions, titles, legends and coordinates with desired pallets. Colours were usually considered for each map as an advantage of using SPANS GIS. For numerical and interval values the accuracy of the data were tested using a query module.

7.3.2 Statistical Techniques Used

In the second phase of data analysis, statistical analysis techniques, consisting of the t-test and simple regression coefficient techniques, were used to find possible associations between thunderstorm rainfall amounts and physiographic parameters, individually. In addition, a stepwise multiple regression (Hauser, 1974; Bryant, 1985a) was used to construct a statistical model explaining the average thunderstorm rainfall as a relative function of elevation, aspect, proximity to sea and landuse pattern. These analytical methods were not only applied tofind possibl e associations between physiographic factors and thunderstorm rainfall, but also to test the results obtained from the GIS overlaying map techniques.

7.4. Topography and Rainfall from Thunderstorms

To investigate any possible associations between the topography of the region and thunnderstorm rainfall, it is necessary to describe the main topographic features of the Sydney region in detail.

7.4.1 Description of Major Topographic Units

Generally, the Sydney region can be characterised by five topographic units. Most of the Sydney region is spread out along the gently undulating Cumberland Plain, on average, less than 100 meters above sea level. This basin is surrounded by four other topographic units rising to elevations up to 1200 m in height. Sydney covers part of the Hornsby Plateau in the north-east of the study area,rising to nearly 250 meters above sea level. The Blue Mountains is located to the west and north-west of the region with elevations higher than 1200 meters. To a lesser extent the Illawarra Plateau to the south-east of the region has an average elevation of 350 to 450 meters with a relatively sharp escarpment facing the Tasman Sea. Finally, the south-western part of the study area can be characterised by a relatively flat landscape which is a part of the Southern Tablelands. This flat relief lies approximately 700 meters in elevation on average, above sea level. The elevation map for the study area contains elevations that range from sea level to 1200 meters at the top of the Blue Mountains (see Figure 7.6). Also Table 7.8 summaries the area based on topographic characteristics of the Sydney region. CO +-• 'c Z>

D. CO i_ CO o Q. o E o o o o o o o o o O o o o o o o o LO o CO LO CO + o SI o 1 1 1 1 1 1 OJ CO L1O 1 OJ C J CO r> O LO o o o o o C_) o Q 'CD l LD CM CO "<* in CO ou T~ X oS q co" c < o 'co CD oe 1-5L tr CD C >> CO H— ? O to Cj c Z> o 1c Q. CO I— CO o d CoO

CD

CD L- CO fm *&?* 91. 091- LL CO <* co CO CO CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters L7A

The major river network in the region is the Hawkesbury - Nepean River which flows through the study area south to north. In the east, the coastline is crossed byrivers such as the Parramatta and George Rivers. At the coast theserivers form significant estuaries such as Broken Bay in the north and Botany Bay in the south. Zigzag coastline which is a boundary between the Sydney region and the Tasman Sea, cuts the region in the east.

7.4.2 Association Between Elevation and Thunderstorm Rainfall

In the literature (reviewed in Chapter 2), a positive correlation of increasing thunderstorm activity with altitude is well documented (Spreen, 1947 and Reid, 1973). The effect of mountains on increasing thunderstorm activity is most clearly seen on thunder-days and rainfall maps (Court, 1960; Duckstein, et al., 1973). Results from many parts of the world also confirm that thunderstorm rainfall increases with relief (Cheong and Tay, 1982). The association noted here is broadly true for NSW and whole of Australia (Hobbs, 1972). In instance, for Hunter Valley (NSW) Hutchinson and Bischof (1983) used a new method (Laplacian Smoothing Spline Function) for estimating the spatial distribution of mean seasonal and annual rainfalls. The rainfall maps show that the areas with higher elevations (for example, the Barrington Tops) received much more rainfall than to the low-lands (the Goulburn River). The stronger influnce was evident in summer reflecting the inflow of warm saturated air of equatorial origin from the norttheast.

The thunderstorm rainfall-elevation relationships over the Sydney region were defined using an area cross-tabulation technique between the digital elevation map (Figure 7.6) and the thunderstorm rainfall map (Figure 7.1). The GIS technique indicated that there is a spatially significant association between the topography of the region and the rainfall distribution map. Chi-square coefficients are given in Table 7.7.

Table 7.7 Area cross tabulation results between the topography map of the region and thunderstorm rainfall map.

Thunderstorm Contingency Tschuprow's T Cramer's V Rainfall Map Coefficient value value

Topographic 0.429* 0.179 0.213* Map * significant at 0.05 level.

The chi-square value is at a significant level of 0.05 which can confirm the existence of a correlation between variables. However, it does not give any information about the effect of the topography of the region upon the rainfall distribution pattern. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 175

The areal based distribution of mean rainfall which was calculated using a GIS technique, indicates that the distribution is not the same for all topographic classes and the area distribution itself varies according to the topographic units. It seems that the highest amounts of rainfall from thunderstorms (more than 120 millimetres) is limited to the 400 - 500 elevation classes.

Table 7.8 shows the areal distribution of thunderstorm rainfall amounts for each topographic interval. The advantage of this kind of calculation is that it shows simply the average area of each topographic class from the total area of the region. At the same time, this estimation gives the rainfall amounts (in percentage) for each individual corresponded topographic classes.

Table 7.8 The areal distribution of thunderstorm rainfall by topographic classes in per cent Thunderstorm Rainfall Classes in mm Topographic 35-60 61-90 91-120 121-150 151-180 181-200 Total Area Classes in m Km sq Per cent 0-50 3.1* 36.3 27.2 24.5 7.4 1.3 1802.7 19.7 7.7** 18.2 18.2 25.3 35.3 76.2 51 -100 0.8 60.8 19.6 15.6 2.8 0.6 1313.2 14.3 1.4 22.2 9.5 11.6 9.6 23. 101 - 150 4.3 44.8 26.6 22.5 1.9 0.0 942.0 10.3 5.7 11.8 9.3 12.0 4.7 0.0 151 - 200 7.6 31.6 36.9 23.3 0.6 0.0 588.6 6.4 6.3 5.2 8.0 7.8 0.9 0.0 200 - 300 10.0 36.8 34.0 15.8 3.4 0.0 1060 11.6 14.9 10.9 13.4 9.5 9.7 0.0 301-400 7.0 28.9 25.9 26.4 11.7 0.0 959.5 10.5 9.6 7.7 9.2 14.3 29.7 0.0 401 - 500 5.9 37.2 32.8 20.2 3.9 0.0 715.9 7.8 6.0 7.4 8.7 8.2 7.3 0.0 501 - 600 8.9 51.3 29.8 9.6 0.4 0.0 590.5 6.4 7.4 8.5 6.5 3.2 0.6 0.0 601 - 800 32.0 33.2 22.3 11.7 0.8 0.0 871.7 9.5 39.2 8.0 7.2 5.8 1.9 0.0 801 -1000 5.6 0.3 83.6 10.1 0.4 0.0 251.5 2.7 2.0 0.02 7.8 1.4 0.2 0.0 above 1000 0.0 0.0 77.5 22.0 0.4 0.0 74.1 0.8 0.0 0.0 2.1 0.9 0.07 0.0 Total km^ 711.8 3588.0 2697.8 1766.1 376.2 30.7 9170.4 100 7.8 39.1 29.3 4.1 0.4 % 19.3 4.1 0.4 * The area of each topographic class in per cent ** The areal distribution of thunderstorm rainfall for each topographic interval in per cent.

Because the initial analysis of thunderstorm rainfall distribution over the study area indicated that local spatial variations could be influenced by small-scale topographic features, an attempt was made to look at the thunderstorm rainfall - elevation relationships using statistical techniques. To do this, the Sydney region has been divided into four sub- regions according to the main topographic units of the region: namely, the Blue Mountains, Hornsby Plateau, Southern Tablelands and the Illawarra Plateau (see Figure CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 126

7.6). The relationships between thunderstorm rainfall amounts and the elevation in each of these four topographic units were analysed using simple linear regression (Table 7.9).

Table 7.9 A linear regression analysis between thunderstorm rainfall amount and elevation of rainfall stations located in the four topographic units of the Sydney region.

Topographic Units n F value Blue Mountains = A 26 .82 .67 49.6 .0001 Hornsby Plateau = B 74 0.002 .004 0.33 .0.56* Southern Tablelands = C 21 -.42 .17 4.06 .05 Illawarra Plateau = D 31 .67 .45 30.2 .0001 * Not significant at 0.05 level

The liner regression analysis indicates that there is a close relationship between average thunderstorm rainfall and elevation. Over the Blue Mountains and Illawarra Plateau, the relationship between elevation and rainfall amounts for the average of the largest thunderstorm rainfall are generally high. However, over the Southern Tablelands, correlation coefficients between rainfall and elevation are negative. Also, over the Hornsby Plateau, the effect of elevation upon rainfall amounts is not at the significant level of 0.05. These results imply that the relationship may be masked by other controlling parameters.

7.4.3 Association between Aspect Classes and Rainfall

The gradient of thunderstorm rainfall with elevation depends not only on the height of station, but also its aspect (Reid, 1973). In fact, slope, in relation to prevailing wind direction, provides a basis for the identification of zones of potential relative thunderstorm rainfall in the region (Sims, 1981; McCutchan and Fox, 1986).

To find an association between the aspect map - as an important topographic feature - and the distribution of the thunderstorm rainfall map (Figure 7.1)firstly th e aspect map of the region was derived from the DEM model (Figure 7.3). There are 5 main aspect classes: 1) aspects facing the sky (flat), 2) north, 3) east, 4) south and 5) west directions (see Table 7.4). This procedure helped to analyse and identify the spatial correlation between the thunderstorm rainfall map and the exposure of each rainfall station to one of the main aspect classes.

An area cross tabulation technique, was therefore, introduced between above-mentioned maps. The results are shown in Table 7.10. The association between maps is significant at 0.05 level. SEVEN Thunderstorm Rainfall and Physiographic Parameters 177

Table 7.10 Area cross tabulation results between the aspect map of the region and thunderstorm rainfall map.

Thunderstorm Contingency Tschuprow's T Cramer's V Rainfall Map Coefficient value value

Aspect 0.52* 0.37 0.41* Map *significant at 0.05 level.

Accordingly, the position of each rainfall station with respect to these aspect classes was found using the query function of the SPANS GIS. The average distribution of thunderstorm rainfall amounts for each individual aspect class was then found using all rainfall stations data located in the region. As it is clear from Figure 7.7, the distribution of thunderstorm rainfall amount, based upon aspect classes, is not the same. There is a considerable difference between the different aspect classes in obtaining rainfall from thunderstorms. It seems that stations which are exposed to the west and east receive the most rainfall from thunderstorms.

E E

e '3

IB U

West Flat North South East Aspect Classes

Figure 7.7 The distribution of thunderstorm rainfall in the Sydney region based upon aspect classes.

In order to prove statistically the association which was found by GIS technique, and to calculate the different distribution pattern which is graphically shown in Figure 7.7, a multiple regression method was introduced between all aspect classes (as nominal independent variables) and thunderstorm rainfall amount as a dependent variable (Table 7.11). In this statistical procedure the aspect of west was entered into the model as a constant variable. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 128

Table 7.11 A multiple regression analysis between aspect classes and thunderstorm rainfall amount. n= 152 Beta value t-value P Aspect Items (Std. Coef.) East-West 0.08 0.66 0.51* South-West -0.25 2.45 0.01 Flat-West -0.37 3.34 0.0009 North-West -0.22 2.74 0.007 * not significant at 0.05 level, r2 = 0.22

In Chapter 2, it was generally argued that the study area can be affected by two main weather systems associated with thunderstorms. First, air masses thunderstorms which are created mainly over the elevated terrain of the Sydney region and, then, move from west to east (Matthews 1993). Second, some weather systems, for instance, Tasman Sea lows and associated weather features, occasionally, move from the sea over the Sydney region and causing wide-spread thunderstorm activity. As Figure 7.7 shows, stations facing the west and east were more exposed to rain-bearing thunderstorm systems. So, using a multiple regression method (with some nominal variables), just one of these geographic directions could be keept as a constant parameter, in the analyse of thunderstorm rainfalls.

Clearly, the relationship between west and east-facing aspects is not significant. Both aspects are subject to the same amount of thunderstorm rain. The Beta value (Std. Coef. in Table 7.11) and the associated t-value confirm the above-hypothesis. Also, this statistical technique indicates that there are significant differences between stations which face the west and stations which are exposed to other aspects, such as the south and north. Flat topography is of minor importance in controlling thunderstorm rainfall. In total, aspect classes can explain about 22 per cent of variance in the distribution of thunderstorm rainfall in the region.

7.5 Proximity to the Sea and Thunderstorm Rainfall Distribution

As a physiographic parameter, proximity to the sea is also known to be a very important factor in controlling thunderstorm rainfall amounts, especially near the coast (Merva et al. 1976; Berndtsson, 1989). For the Sydney region, it was suggested by several researchers, for example James (1992), that proximity to coastal areas can increase the amount of thunderstorm rainfall considerably. In this study, two methods were introduced tofind the possible relationships between proximity to the sea and thunderstorm rainfall distribution. First, a GIS technique was produced to show the spatial correlation between thunderstorm rainfall and proximity maps. Then a simple regression method was applied to assess the statistical significance of the relationship. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 122

As a first step, using a GIS buffering technique, a proximity map was constructed. Then, an area cross tabulation was performed between the proximity and thunderstorm rainfall maps. Results are shown in Table 7.12 and statistically indicate that, in the Sydney region, there are high positive associations between the spatial distribution of rainfalls from thunderstorms and the proximity to the coast.

Table 7.12 Area cross tabulation results between the proximity to sea map of the region and thunderstorm rainfall map.

Thunderstorm Rainfall Contingency Tschuprow's T Cramer's V Maps Coefficient value value

Distance Map 0.723* 0.394 0.468*

* significant at 0.05 level

Table 7.13 The areal distribution of thunderstorm rainfall by proximity classes in per cent. Thunderstorm Rainfall Classes in mm Proximity 35-60 61-90 91-120 121-150 151-180 181-200 Total Area Classes i Km sq Per cent Km 1-10 0.0* 0.0 14.4 59.1 23.5 3.0 1030.7 11.2 0.0** 0.0 5.5 34.5 64.4 100.0 11- 20 0.0 2.1 34.6 53.1 10.1 0.0 1188.5 13.0 0.0 0.7 15.3 35.8 31.9 0.0 21- 30 0.0 26.5 49.0 24.3 0.3 0.0 1235.7 13.5 0.0 9.1 22.4 17.0 0.8 0.0 31- 40 1.4 81.3 16.7 0.6 0.0 0.0 1234.1 13.4 2.4 28.0 7.6 0.4 0.0 0.0 41-50 13.5 78.2 8.3 0.0 0.0 0.0 1241.3 13.6 23.5 27.0 3.8 0.0 0.0 0.0 51- 60 36.4 58.7 4.9 0.0 0.0 0.0 1223.4 13.3 62.6 20.0 2.2 0.0 0.0 0.0 61- 70 9.4 60.9 29.7 0.0 0.0 0.0 871.1 9.5 11.5 14.8 9.6 0.0 0.0 0.0 71- 80 0.0 2.2 80.7 17.0 0.0 0.0 584.2 6.4 0.0 0.4 17.5 5.6 0.0 0.0 81- 90 0.0 0.0 72.2 25.4 2.5 0.0 380.0 4.1 0.0 0.0 10.2 5.5 2.5 0.0 91- 100 0.0 0.0 87.0 12.4 0.5 0.0 179.8 2.0 0.0 0.0 5.8 1.3 0.3 0.0 101- 110 0.0 0.0 100.0 0.0 0.0 0.0 3.3 0.04 0.0 0.0 0.2 0.0 0.0 0.0 Total km^ 711.6 359.0 2699.3 1766.5 376.2 30.7 9172.4 100 7.8 39.1 29.4 % 19.3 4.1 0.3 * The area of each proximity class in per cent ** The areal distribution of thunderstorm rainfall for each proximity interval in per cent.

In addition, areal distribution of thunderstorm rainfall amounts, which were calculated based on the classes shown on the proximity map, are summarised in Table 7.13. According to these tables, all rainfall in the highest class (180-200 mm) was distributed within 10 km of the coast (zones 1). Also, more than 95 per cent of rainfall (in class 151- CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 1M

180) fell on zones within 20 km of the sea. Except for parts of the Blue Mountains, the percentage of high rainfall decreased sharply westwards from the coast.

To verify these results statistically, a simple regression model was applied to the data to find out the possible relationships. A computer program wasfirst written to calculate the shortest distance of each station to the average coastline (see the computer program 5, Appendix A). The results of this phase of the analysis are summarised in Table 7.14.

Table 7.14 Correlation coefficients between the proximity to the sea (in Km) and thunderstorm rainfall (average of the biggest thunderstorm rainfall events). n r r2 F-test Probability Distance 152 -0.61 0.37 87.6 0.0001 (proximity to sea)

As it is evident from Table 7.14, the proximity to the sea, as a physiographic parameter, is negatively correlated with thunderstorm rainfall amounts (r2 = 0.37) at 0.0001 significant level. There is a high possibility that the addition of other independent variables such as, distance from the mountain ranges, might substantially improve the results of this kind of analysis. The GIS map also shows small variations in thunderstorm rainfall along the coast implying the influence of other factors.

7.6 Landuse Patterns and Thunderstorm Rainfall

The purpose of this section is to find and describe possible relationships between urban- rural landuse patterns and maximum thunderstorm rainfall distribution in the Sydney region. Recently, a great deal has been written about the influence of urban areas on climatic factors (Auer, 1978; Henry et al., 1985; Bradshaw and Weaver, 1993). All suggested that many climatic factors, for example temperature and rainfall patterns, can be affected by city environments. Other investigations in urban climatology such as those by Changnon (1973), Changnon and Huff.(1973), Landsberg (1981) and Houghton (1985), have indicated that surface conditions such as the heat island effect or physical features of a city, which influence most weather elements, also affect the subsequent rainfall, especially in the case of meso-scale convective precipitation (see literature chapter).

In the Sydney region, where about 3.5 million people live (Department of Planning, 1995), the ground surface is covered by houses, paved roads, factories, warehouses and tall office and apartment blocks. These structures contrast with the ground cover of the surrounding rural areas, such as forests and open rural areas, and they may produce local differences in Sydney's climatic environment. In such an environment, land cover may affect, to some degree, the distribution of thunderstorm rainfall patterns throughout the region. To CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 181

correlate the distribution of thunderstorm rainfall over the region with landuse patterns, both GIS methods and statistical techniques were again used.

Firstly, an area cross-tabulation was used to find the correlation between the landuse and thunderstorm rainfall maps. It was found that chi-square coefficients are significant at 0.05 level.

Table 7.15 Area cross tabulation results between the landuse map of the region and thunderstorm rainfall map.

Thunderstorm Rainfall Contingency Tschuprow's T Cramer's V Map Coefficient value value

Landuse Map 0.58* 0.30 0.32*

* significant at 0.05 level

Secondly, a GIS technique was used to calculate the details of the areal distribution of thunderstorm rainfall based on different landuse classes. For more details in the areal distribution of thunderstorm rainfall based on landuse classes see Table 7.16.

Table 7.16 The areal distribution of thunderstorm rainfall by landuse classes in per cent. Thunderstorm Rainfall Classes in nun Landuse 35-60 61-90 91-120 121-150 151-180 181-200 Total Area Classes Km sq Per cent CBD 0.0* 0.0 0.0 0.0 13.1 86.9 7.9 0.09 0.0** 0.0 0.0 0.0 0.3 24.0 IND 0.0 39.5 18.0 35.6 6.9 0.0 120.17 1.36 0.0 1.4 0.8 2.5 2.3 0.0 URB 0.0 44.0 21.9 20.0 12.2 1.8 807.8 9.1 0.0 10.2 6.8 9.6 27.6 50.7 URT 1.5 15.5 25.4 49.7 7.5 0.5 784.5 8.9 1.8 3.5 7.7 23.0 16.5 12.9 RUS 9.5 88.3 2.1 0.0 0.0 0.0 885.6 10.0 12.7 22.3 0.7 0.0 0.0 0.0 RUO 10.5 63.4 21.4 4.6 0.2 0.0 1292.9 14.6 20.3 23.4 10.6 3.5 0.7 0.0 TNP 8.8 27.9 38.5 21.0 3.8 0.1 4947.7 55.9 65.2 39.4 73.3 61.4 52.5 12.44 Total km^ 665.4 3507.6 2597.5 1691.3 356.0 28.5 8846.4 100 7.5 39.7 29.4 19.1 4.00 0.0.33 % * The area of each landuse class in per cent. ** The areal distribution of thunderstorm rainfall for each landuse interval in per cent.

The comparison of the rainfall amounts in the different landuse classes indicates that these stations which are located in built-up areas have more rainfall from thunderstorms, on average. The highest thunderstorm rainfall amounts are measured in the centre of the Sydney (CBD). CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 182

200 E 180 E 1 60 140 1 20 OS 1 00 E t- 80 o •** 60 (A L. Ol 40 •O e 20

3 : 0 •ttffll i laaa • Ivffifl • E:::: :l • Eaal | CBD IND URB URT RUO RUS TNP Landuse Classes

Figure 7.8 Distribution of thunderstorm rainfall in the Sydney region based upon landuse classes.

Also, the average rainfalls in residential (barren and treed areas) were higher than other landuse classes. The uneven distribution pattern of thunderstorm rainfall is clearly shown by Figure 7.8.

Table 7.17 The result of a t-test for rainfall distribution in different landuse classes. Landuse CBD IND URB URT RUS RUO TNP Classes CBD

IND 6.7*

URB 4.3* -1.18

URT 4.65* -2.68* -1.67*

RUS 25.7* 4.26* 4.61* 7.25*

RUO 9.29* 2.65* 4.68* 7.34* -1.25

TNP 4.24* -0.61 0.73 2.4* -3.65* -3.60*

Significant Difference at 0.05 level

Statistically, to examine any possible difference among the means of the rainfall stations, located in different landuse classes, a t-test technique was applied (Shaw, Wheeler, 1985). The null hypothesis is that there is no difference in the mean thunderstorm rainfall

population between each paired sets of landuse classes, H0:X = Y. The alternative hypothesis is that mean rainfall differs by a degree that is too great to be attributed to random sampling variations from a common thunderstorm rainfall population. Thus, the CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 183

mean for the stations located, for example in the CBD, should be higher than for those

stations located, for example in the industrial areas, HX:X>Y Thunderstorm rainfall data were considered for the average of the biggest rainfall events.

The results are shown in Table 7.17. A significant level of 0.05 was set as the criteria for accepting or rejecting the null hypothesis. The results of the t-test which was introduced among all landuse classes adds support for the presence of significant differences between different landuse classes concerning thunderstorm rainfall distribution in the Sydney region.

Table 7.18 A multiple regression analysis between landuse classes and thunderstorm rainfall, r2 = 0.39

Landuse Beta value Classes (Std. Coef.) t-value P IND-CBD -0.56 4.92 0.0001 URB-CBD -0.84 4.70 0.0001 URT-CBD -0.72 3.97 0.0001 RUO-CBD -0.1.03 6.97 0.0001 RUS-CBD -0.82 7.01 0.0001 TNP-CBD -0.91 5.09 0.0001

Clearly, stations located in the CBD and urban-residential areas receive much more thunderstorm rainfall amounts. To test this idea and to see the total effect of landuse classes upon the distribution of thunderstorm rainfall, a multiple regression technique was again introduced between all landuse classes (as nominal independent variables) and thunderstorm rainfall amount as a dependent variable (see Table 7.18).

Statistically, it was found that there is a significant difference between the centre of Sydney (CBD) and other landuse classes in obtaining rainfall from thunderstorms. Also, the r2 value indicates that landuse pattern of the region is a very important parameter in explaining about 39 per cent of variance (in total) in terms of the biggest monthly thunderstorm rainfall events from 1975 to 1993.

7.7 Overlay Modelling / Multiple Relations

So far, by using the SPANS GIS software, which was employed for pre-processing of two-map layers analysis, some initial associations among different variables with a spatial nature, have been established. At the same time, some simple statistical techniques were used to analyse and find correlations between data sets with a spatial context. In the final stage, more procedures would be employed to consider the associations among CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters ISA physiographic parameters (as independent variables) and maximum thunderstorm rainfall (as a dependent variable) synchronically. In fact, this is the main problem, and there are two possible approaches to its solution.

The GIS overlaying techniques and the multiple stepwise regression method were applied to find possible associations between all the variables. These techniques were also used to consider three important questions (with a climatic content). These were:

1) where are the locations most likely in the region to be subject to the highest rainfall from thunderstorms? 2) what are the physical characteristics of these locations? 3) and finally what are the statistically significant relationships between the physiographic parameters and rainfall amounts?

7.7.1 GIS Overlay Modelling

At the first stage in the GIS environment, an overlay modelling technique was introduced to visualise the area's different physiographic characteristics, for example, topography or landuse with the highest thunderstorm rainfall. The main aim, was to derive a set of new maps by imposing the areas of high rainfall (more than 120 m) over all the independent maps, in a process known as overlay modelling. For this work, from the rainfall map the highest rainfall classes werefirst differentiate d using GIS reclassification techniques. These classes were then spatially imposed over all the physiographic maps of the study area by writing some equations in the SPANS GIS system (see Appendix E, Table 7.6). The resulted Figure 7.5 (a to d) can easily be used to achieve the following information:

1) the amount of the highest thunderstorm rainfall at a specific distance from the average coast-line (distance zones); 2) topographic classes in which the highest rainfall occurred; 3) variation of rainfall amount with aspect classes; 4) character of a landuse pattern in relation to the highest rainfall amounts; 5) and finally all classes from the physiographic maps exposed to the highest thunderstorm rainfall amounts which can be visualised simultaneously on the computer screen.

Results show that areas closer to the sea (zones 1 to 3), about 30 kilometres from the average coastal line, had high rainfalls. Also, it is evident from Figure 7.5 (a) that zones 8, 9 and even 10, which were located in the west over part of the Blue Mountains, had been subject to the highest thunderstorm rainfalls. Closer examinations of distant zonesfrom th e coastal areas indicates that there is a considerable variation in rainfall distribution. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 181

Clearly, high elevations near the coast, for example, the Illawarra Plateau located in the south-east of the region, and parts of the Hornsby Plateau located in the north-east of the study area, received the maximum amount of rainfall from thunderstorms. Also, parts of the Blue Mountains located in the west of the Sydney region, were also susceptible to the highest rainfall amounts from thunderstorms. However, Figure 7.5 (b) indicates that the Southern Tablelands in the south-west of the region have not been susceptible to the highest thunderstorm rainfalls.

A close examination of the aspect classes of the Sydney region shows that generally, in those areas which were subject to the highest thunderstorm rainfall amounts, aspects facing the east and west are dominant, except for areas with flat topography which are located in the east near the coast.(Figure 7.5 (c) ). Exact view of the landuse map of the region (Figure 7.3) indicates that these areas are used for urban purposes.

Finally, it is clear that built-up areas, for example residential areas, and in particular, areas which are located in the centre of the City (CBD) and east of the Metropolitan area were much more subject to the highest thunderstorm rainfall amounts. It can be also seen from Figure 7.5 (d) that some areas in the National parks had higher thunderstorm rainfalls.

7.7.2 Multiple Relations Among Variables

The initial results of GIS and also the statistical techniques have indicated that there are statistically significant associations among variables explaining the thunderstorm rainfall of the region. Certainly, a stepwise multi-correlation regression technique and then a Z score method are needed to verify the results obtained by GIS and statistical techniques to determine the relative effect of each individual physiographic parameter upon the spatial distribution of maximum thunderstorm rainfall in the model.

7.7.2.1 Stepwise Multi-Correlation Regression Technique

Before introducing a stepwise multi-correlation regression technique, a correlation matrix between items of each scale of independent variables and dependent variable was applied. A correlation matrix is therefore found to be a statistical technique to see the interrelationships among all items.

The results show that the relationships between many physiographic items and thunderstorm rainfall are significant at 0.05 level (shown by asterisk). In general, the interrelations between all items presented in Table 7.19 confirm the reliability of data used. Accordingly, it can be concluded that most physiographic items which have been used in SEVEN Thunderstorm Rainfall and Physiographic Parameters 1M this study are relatively homogeneous items and therefore they have a common relationship to thunderstorm rainfall.

Table 7.19 Interrelations matrix among physiographic parameters and thunderstorm rainfall (n=152).

Variables *1 X2 X3 X4 X; X{ X7 X8 X9 X10 Xn X12 X13 X14 X15

X\ = Proximity to Sea \

X2 = Elevation 53* \

X3 = Aspect Flat 07 .13 j

X4= East .24* ..05 -.49* 1

X5= South ,12 ..05 -.29* -.35* 1

Xg= North 04 .08 -.13 -.16* -.09 1

X7 = West .06 .25* -.24* -.26* -.18* -.08 1

X8 = Landuse CBD . 15* ..JQ ..10 .04 .14 -.03 -.06 1

Xo= URB ..19* ..35* .32* ..04 -.08 -.11 -.21* .09 1

Xl0= URT ..23* -.08 -.26* .21* .06 -.04 .0 -.09 -.32* 1

Xtl= IND ..os ..17* .21* -.14 .03 -.05 -.1 -.04 -.14 -.14 1

Xi2= RUS .23* ..14 .19* ..21* .09 -.05 -.03 -.04 -.14 -.16* -.07 1

Xi3= RUO .25* .21* .02 -.06 -.13 .22* .08 -.06 -.21* -.23* -.1 -.10 1

XM= TNP 16* 48* ..25* .05 -.01 .05 .24* -.09 -.31-.31* -.32* -.14 -.15 -.22* 1 x15 = Rainfall -.61* .16* -.31* .38* -.14 -.17* .17* .35 .20* .27* -.06 -.31* -.36* -.03 1 * significant at 0.05 level

Independent variables which have been entered into a stepwise regression equation were - proximity to the sea (in Km) and the spot elevation (in m) as interval variables. Other variables (aspect and landuse classes) were entered on a nominal scale. These two last variables were transformed into dummy variables, and then applied with thunderstorm rainfall amounts as the dependent variables. Because there were five levels of aspect classes as nominal scale variables, four dummies were required in the regression model. In terms of landuse classes, there were seven levels. In general, L-l dummies were required, where L is the number of levels of the variable to be represented by them (Zar, 1984).

Each independent variable was entered into the regression equation in order to determine its unique contribution in relation to the other variables. The order in which the independent variables were entered into the equation has no impact on the outcome because each variable is treated as though it were the last variable to be entered. The stepwise regression procedure selects the strongest independent variable in thefirst stag e and at each new stage, the next most significant variable is added to the equation. The results of the stepwise regression are presented in Table 7.20. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 181

Table 7.20 Presents the result of stepwise multiple regression analysis for the average of the biggest thunderstorm rainfall amounts (n=152 ). Step Multiple R Variance Number of Numbe Predictor Variable R Square Added in F Value Variable in the %

1 Proximity to Sea 0.607 0.369 37 87.62 1 2 Landuse 0.735 0.54 17 42.3 2 3 Aspect 0.810 0.656 12 24.5 3 4 Elevation 0.835 0.697 4 22.3 4 Total 70 * All F values are significant at 0.0001 level

The stepwise regression procedure automatically selected the strongest independent variable (proximity to sea). At thefirst stage , and at each subsequent stage a new variable was added to the equation in the order in which they increased variance (up to 70 per cent) in thunderstorm rainfall amounts. Using a F test, statistically significant variables were determined at less than 0.001 level of significance.

The rank ordering of variables in terms of their predictive strength for thunderstorm rainfall is: distance to the sea; landuse pattern of the region; aspect classes; and the elevation of rainfall stations. Generally, the following results were found in this analysis:

1) Despite the presence of variations in maximum thunderstorm rainfall amounts in coastal areas (see Figure 7.5 (a)), the correlation coefficient between thunderstorm rainfall and the distance from the sea is quite high (r2 = 0.37), indicating that proximity to the sea is the main predictor of thunderstorm rainfall distribution in the region.

2) Although the urban area does not appear to be the best predictor of thunderstorm rainfall amounts, it increases the variance added in the model significantly, (17 per cent). This suggests that there is a strong difference between urban (CBD, URB and URT) and non-urban (RUO, RUS and TNP) areas in the amount of rainfall during major thunderstorms.

3) Exposure to rain-bearing winds (particularly the east and west aspects), appears to be the most important topographic factor in explaining statistically some of the distribution of thunderstorm rainfall in the region. It increases about 12 per cent of variance in the model.

4) The elevation of the study area is the final important physiographic parameter in explaining the spatial distribution of thunderstorm rainfall amount (about 4 per cent). It CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 188 can therefore be concluded that the spot elevation does not appear to be the only topographic feature affecting thunderstorm rainfall distribution over all of the region and more likely the amount of thunderstorm rainfall does not increase with elevation, as it was shown in section 7.3.

In brief, these findings support the effects of different independent variables upon the spatial distribution of thunderstorm rainfall (as a dependent variable) and together explain about 70 per cent of the variance in thunderstorm rain. The remaining 30 per cent unexplained variance may be attributed to the other parameters.

7.7.2.2 The Spatial Distribution of 2 Scores Over Sydney

The preceding stepwise analysis provides only summary statements for the entire study area. Statistically, it summaries the relationships between thunderstorm rainfall and the four above-mentioned independent variables. Therefore, it is impossible to distinguish where, in the Sydney region, these relationships hold best.

However, Z scores, derived from the stepwise regression model, provide a means by which the appropriateness of the model can be assessed spatially. (Berry and Marble, 1968; Shaw and Wheeler, 1985). Z scores were calculated using the following formula: x - x z = SD where x = observed thunderstorm rainfall values x = mean of variable x SD = standard deviation of variable x z = obtained Z scores

A map was then prepared using GIS interpolation techniques to visualise the Z scores spatially. The more positive the Z score the better the model fits for that site. For generalisation purposes,five class intervals were used ranging from > -2 Z to < +2 (see Figure 7.9).

The spatial distribution of Z scores suggests there are a considerable number of areas above the mean (Z > 0). As Figure 7.9 shows, over the elevated areas, such as the Blue Mountains, Illawarra Plateau and over a small part of the Southern Tablelands, Z values are highly positive. Likewise, Z scores over the Metropolitan area, north of the Parramatta River and over small areas in the centre of the Sydney region (one located between Camden and Picton and another located in the northwest of Hornsby Plateau), show positive spatial variations. These patterns of Z values may indicate that distribution of CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 182 thunderstorm rainfall is, in general, highly correlated positively to many of the parameters in the stepwise regression.

The remainder of the study area, particularly the Hawkesbury - Nepean Valley and some small pockets located in coastal areas, have negative Z values (Z < 0). These areas do not fit the model well for one reason or another. Consequently, it is possible to examine only those variations in the spatial pattern of potential thunderstorm rainfalls which have a close relationship with physiographic parameters considered in this study. More importantly, the produced map may also provide valuable insights as follows:

1) establishing and modifying regional boundaries of thunderstorm rainfall distribution in the Sydney region, 2) selecting unit areas in which to conductfield work , 3) and identifying additional independent variables to be included in future investigations. 190

•^5 CO CD LO CD ¥LO co co co CM 1 o CO + + V V CD N N N N V ll » CO ll II N > V V ll •£• CM + + CM T- V N I I o

c o V5 O) 2 CD CC oe 1.91. >> CD C "O >> Cf) CD > o CO CD i— o o Cf) N c o

CO b

+-> as Q_ CO 91 09V OJ N." CD i_ O) CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 191

7.8 Discussion

To date, most studies in the Greater Sydney region (for example: Colquhoun and Shepherd, 1985; Speer and Geerts, 1994) have concentrated on thunderstorm activity over a few days, examining the synoptic weather patterns which caused the thunderstorm rainfalls. Although none have concentrated exclusively on the effect of the local physiographic parameters upon the distribution of thunderstorm rainfall, they have already acknowledged the importance of the Sydney 's climatic-environmental factors.

The present chapter has addressed this approach by examining the detailed spatial analyses of thunderstorm rainfall (for the 6 biggest short-term events) with respect to important physiographic parameters (elevation, aspect, proximity to the sea) and landuse patterns of the region. These parameters were chosen according to the results of the spatial analyses in Chapter 6. They were also selected because studies elsewhere indicated they might be important in controlling the spatial distribution of thunderstorm rainfalls (see Chapter 2). Two different methods; GIS techniques, and statistical procedures have been used, one following the other, to analyse the spatial data across the Sydney region. The information provided using these techniques confirm and extent the results which have been found by other researchers.

7.8.1 The Role of Coastal Area

It is evident from the results of this chapter that the coastal areas in the east of the Sydney region receive much higher thunderstorm rainfall amounts than those located inland in nearby high relief areas. Over the coastal areas, the following mechanisms are supposed to be more importance in the controlling of the spatial distribution of thunderstorm rainfall.

It is possible that, meso-scale circulations in the lower troposphere over the coastal areas develop in response to differential surface heating. This mechanism, in particular, between the land and the adjacent sea, can cause convectional activity in response to differential solar radiation during the day, depending on the geographic characteristics of each place and on weather conditions. Gentry and Moore (1954) and L'hermitte (1974) stressed mechanisms by which thunderstorm rainfall can increase along Florida's coastal areas. For the NSW coasts, intense locally thunderstorm activity reflecting coastal influences has been emphasised by several researchers.

Hobbs (1971), using harmonic analysis, investigated some spatial characteristics of the rainfall regimes in north eastern NSW. The results indicated that there are four major terrestrial determinants of the spatial variations in rainfall over the study area. The four factors concerned were: distance from the coast; distance from the scarp; relief (the scarp) CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 122 and latitude. It appeared that from the four factors mentioned above, the most influential was distance from the coast. Hobbs (1971) also mentioned that in summer - when thunderstorms are the most important source of rainfall - although thunderstorms are experienced over the Tablelands, they are more likely to be accentuated in coastal areas, particularly when orographic influences operate on air streams with strong easterly components. Also, Sumner (1983a) highlighted the importance of the coastal plains with reference to both local convection and the role of sea-breeze fronts in the generation and enhancement of meso-scale systems such as thunderstorms.

An alternative mechanism is the possibility that occasionally, some weather systems, for example lows, are accompanied by convection activity enhanced by nearby seas or evaporative heating processes over by advection at the coastline. In such circumstances, onshore winds can supply the moisture to the thunderstorm systems evaporated from the Tasman Sea (Speer and Geerts, 1994). It was also found by Reeder and Smith (1992) that the coastal areas acts as a stationary convergence zone causing longer duration thunderstorms. These thunderstorm systems can extend over the coastal areas with intense rainfalls (James, 1992). However, it was found in the current research that, even when rainfall from thunderstorms is general over the coastal areas there are still isolated centres of high rainfall with sharp isohyet gradients.

In addition, thunderstorms which most often develop over the relatively high topography west of the Sydney region (over the Blue Mountains) can move towards the east of the region over the coastal areas (Matthews, 1993). Speer and Geerts (1994) supported the findings of Matthews (see Chapter 2, section 2.6.2). They also found that, in a close relation with synoptic systems (for example quasi-stationary or eastward moving troughs), convective systems typically start around midday to the west of Sydney and reach the east of Sydney in the afternoon to evening. It is therefore more likely that these systems follow the sources of moisture, available mostly over the coast and nearby Tasman Sea which aids the convective systems to produce higher rainfall totals over the coastal zones.

7.8.2 Impact of Topographic Factors

The results in this chapter also indicated that both aspect and elevation influence the amounts of thunderstorm rainfall in the study area, particularly in high relief areas over the Blue Mountains and the Illawarra Plateau. This occurs mainly because both factors have a strong effect on the initiation of thunderstorms. The effect of Sydney's high lands on the distribution of thunderstorm rainfall amount is most clearly seen on maps (figures 6.7, 6.8 and 7.1) showing the relationship between rainfall patterns and terrain height. There are three possibilities to explain how the region's terrain is able to influence thunderstorm rainfall so considerably. CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 121

Firstly, one of the most regular and predictable types of thunderstorm activity can occur in warm seasons over the Blue Mountains (Gentilli, 1971). The daily heating of the hillsides generates warm up-slope winds which continuerising afte r reaching the top of the ridges and trigger deep vertical convection (Maine, 1962). The thunderstorm rainfall patterns over the Blue Mountains throughout the warm months may be dominated by this mechanism. More recently, using data recorded by the Lightning Position Tracking System (see Chapter 2) Laudet et al., (1994) found that the spatial distribution of lightning (associated with thunderstorms) is closely related to the topography of the region. Their results support the concept that elevation is a very important physiographic parameter in controlling thunderstorm occurrence. The relatively high rainfall over parts of the Blue Mountains from thunderstorms which was clearly shown in the results in Chapter 6 and the current chapter, may be explained by the above-mentioned mechanism.

Secondly, convection systems can occasionally be developed over the Tasman Sea (Bureau of Meteorology, 1989) during unstable conditions. These systems may move toward the west of the region, and as a result, they may be cut off by the elevated terrains due to an air-mass modification effect. Low-level air may be scavenged of its water by drops falling from a seeder cloud above. It is likely that as the air descends beyond the high elevated areas it is dry and cannot restore its water vapour by evaporation from the Tasman Sea, therefore, without the low-level moisture, convective precipitation would be suppressed. In fact, this mechanism may explain some of the thunderstorms with high rainfalls over parts of the Blue Mountains (for example, Katoomba), which is another sign of orographic control upon the distribution of thunderstorm rainfalls over the region.

Finally, topographic units which are located near the coast have an extra influence upon thunderstorm rainfall amounts. For example, places along the Illawarra Plateau or Hornsby Plateau experience very high rainfalls from thunderstorms, illustrating the effect of elevation and exposure on wind directions. It is more likely that the height and exposures of these topographic units to the Tasman Sea acts as a barrier to thunderstorms moving from the east, and isolates coastal areas from those to the west of the plateau. For example, Shepherd and Colquhoun (1985) studied the meteorological aspects of an extraordinary flash flood event (17-19 February 1984) near Dapto just south of Wollongong. They found that a trough had moved slowly from east to west over the Illawarra area and formed several convective cells. These systems produced maximised rainfalls during the event. In such a situation, the orographic lifting mechanism clearly contributed to the short-duration heavy rainfalls along the escarpment. While an extraordinary amount of rain (in excess of 200 mm) fell over the region in a band from Stanwell Tops to Jervis Bay, the heaviest point rainfall recorded was 803 mm at Wongawilli located along the escarpment facing the east. This topographic effect causes CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters L9A different rainfall distribution patterns in the region, as is clearly evident from the results of rthe current chapter. The Illawarra Plateau provides an ideal example of thunderstorms producing more rain parallel to the inland and coast-lines.

How important the above-mentioned mechanisms are, however, can not be clearly identified. This is because in mountainous areas, for example over the Blue Mountains, thunderstorm rainfall distribution patterns are very complicated, showing strong differences over short distances. In contrast with the Blue Mountains or the Illawarra Plateau, over the Southern Tablelands (located on the south-west of the study area) the results showed that there is a general decrease in the vigour of thunderstorm rainfalls at all seasons. This occurs because, a large number of local factors may reduce the amounts of rainfall received from thunderstorms, and most of these local factors vary greatly with relief. They include elevation, steepness of slope, relief forms which cause convergence or divergence of air streams, and aspects, that is, exposure to the rain-bringing winds. As it is evident from the aspect map (Figure 7.3), over the Southern Tablelands - where a relatively flat aspect dominates - topography may not strongly affect the distribution of thunderstorm rainfall in the region.

7.8.3 Effect of Landuse on Rainfall Distribution

Results of the various analyses may finally lead to the conclusion that different landuse patterns, for example, residential or CBD areas, are able to affect the spatial distribution of thunderstorm rainfalls. Evidence of a localised increase in total rainfall from thunderstorms in spring and summer, over the CBD and generally over the Metropolitan area, can be seen fromfigures 6.7 and 6.8 in Chapter 6. This increase was found stronger for the biggest thunderstorm rainfall events (Figure 7.1). Statistical tests, consisting of t-tests and a stepwise analysis - which were applied for the point data sets - confirmed the assumption that in the Sydney region, the effect of 'built-up areas' upon the distribution of thunderstorm rainfall is real.

More recently, Speer and Geerts (1994) gave examples of high rainfall from thunderstorms in the study of flash-floods. Radar images, taken from the three storm events (namely for the 9 March 1989; 10th February 1990; and 2nd April 1992 at the time of the flash-flood rainfall), indicated that maximum rainfall amounts of more than 100 mm occurred mostly over the Metropolitan area. This study found that such storms are most common in the summer months in the Sydney Metropolitan area, during the afternoon or evening hours. On the basis of thesefindings, some of the physical-environmental elements of the urban area appear to be important factors in affecting the amount of thunderstorm rainfall. SEVEN Thunderstorm Rainfall and Physiographic Parameters 121

It is more likely that urban areas can affect incoming solar radiation changing albedo rates and heating processes. This happens because the materials used in the City environment, such as paved surfaces and the multi-faceted nature of the rough urban surface, not only increase the absorption of heat energy, but also increase heat storage. The results of investigators such as McGrath (1971) and Kemp and Armstrong (1972) indicate that generally there is a considerable difference in temperatures between the City and the outlying rural areas in the Sydney region. The development of a 'heat island' may enhance vertical motion of air over the City and, as a result, enhance the subsequent convectional thunderstorm rainfalls. Although no measured data for Sydney's heat island exists, particularly for over a long time-span, experimental studies by Fitzpatrick and Armstrong (1973) and Kalma et al. (1973) confirmed that, in the Sydney region, there is a great difference between urban and non-urban areas in producing artificial energy. The high spatial variation in artificial heat generation during the day in summer may help the production of a heat island over the City. This effect can be increased by the high density of buildings within the City centre (CBD) and can create a heat island with greater cloudiness and higher rainfall.

It is also possible that increased suspended particles in Sydney's atmosphere which cause pollution, indirectly increase rainfall amounts during thunderstorms. Linacre and Edgar (1972) give evidence on the atmospheric pollution of Sydney which can be caused by urban development. There are different sources of pollutants emitted into Sydney's atmosphere. Industrial and commercial activities, including motor vehicles are important sources of particle emission (Carras and Johnson, 1982). It is likely that, under calm weather conditions, urban aerosols such as chemical materials (for example, nitrogen oxides and hydrocarbons) may act as nuclei or ice nuclei materials and therefore, help to induce cloud condensation in the atmosphere of Sydney. Despite intensive efforts devoted to the understanding of Sydney's atmospheric environment in the past (for example, Taylor, 1992), the role of pollutants upon thunderstorm rainfall remains uncertain. However, the present thesis shows that such a relationship may exist.

Moreover, the surface roughness of the CBD by way of tall buildings interspersed with roadways may modify the thunderstorm rainfall distribution to some degree. These structures contrast with the ground cover of surrounding rural areas, such as forests (TNP) and open rural areas (RUO), and they can produce local differences in Sydney's thunderstorm rainfall distribution patterns. Also, the aerodynamic roughness of Sydney's Metropolitan structure may enhance the development of thunderstorm activity. Generally, the peak in maximum thunderstorm rainfall over the central part of Sydney is evidence of the marked impact of the urban area upon rainfall processes. This mechanism is responsible for 17 per cent of variance in thunderstorm rainfall (Table 7.19). CHAPTER SEVEN Thunderstorm Rainfall and Physiographic Parameters 196

7.9 Summary and Conclusion

A "climatologically oriented" GIS was utilised to store, manipulate and display topo- climate data related to Sydney's physical environment. The GIS provided an effective way of displaying and manipulating the spatial data. Research presented in this chapter also investigated alternative ways in which the GIS can provide visual support to the analysis of rainfall distribution in relation to the physiographic parameters of the Sydney region. The rainfall data (averaged from the six biggest thunderstorm rainfall events) was, therefore, the fundamental skeleton over which other information, such as physiographic data, has been performed.

The GIS technique was then followed by a statistical procedure to verify correlations found between thunderstorm rainfall distribution and all independent variables. The results obtained can be closely linked to the major physiographic parameters of the Sydney region which were considered in this study. Analysing the available data, the following conclusions can be deduced.

1) Increases can be seen in thunderstorm rainfall when the distance from the coast decreases. Clearly, rainfall increases with proximity to the sea, but there are considerable variations in rainfall distribution along the coastal areas.

2) The urban area can affect or enhance the development of thunderstorms particularly over the Metropolitan area. As a result, it can increase thunderstorm rainfall amounts. The urban maximum results from increases in temperature, humidity, turbulence and the number of condensation and ice-nuclei substances. Such a physical environment can considerably influence the spatial distribution of thunderstorm rainfall amounts, hence, indicating the reality of the urban effects. The strongest effect of urbanisation can be seen over the CBD and the eastern part of the City near the coast.

3) Rainfall increases with elevation, but exposure to rain-bearing directions (east and west aspects) also seems to be important factors. Despite the effects of aspect, high- elevated areas such as the Blue Mountains, the Illawarra Plateau and less importantly the Hornsby Plateau, are seen to be the more significant topographic units affecting the distribution of rainfall patterns. These areas are thus found to be subject to the highest thunderstorm rainfall amounts.

As the percentages of variance from the step-wise regression method has indicated, there is still considerable unexplained variance (about 30 per cent) which suggests that new independent variables need to be incorporated in the models predicting the distribution of thunderstorm rainfall. In fact, the existence of this unexplained variance illustrates that Thunderstorm Rainfall and Physiographic Parameters 197 factors, other than those under consideration, may contribute to the thunderstorm rainfall distribution over the Sydney region.

To sum up, although synoptic weather patterns, such as those which were highlighted in the literature review, can introduce the relatively big and widespread thunderstorm rainfall events in the Sydney region, most often, as the results of the current chapter have indicated, there are more than one or two physical-environmental factors which simultaneously control the spatial distribution of the maximum thunderstorm rainfall amounts quite significantly. CHAPTER FIGHT Conclusions 128

CHAPTER 8

CONCLUSIONS

8.1 Introduction

This chapter summarises the findings and outcomes of this thesis. The results, given in detail in chapters 3 to 7, are presented briefly in the next section. The limitations of the data and techniques are outlined in section 3. In section 4 the relevant implications of the research are presented. Suggestions for future work are outlined in section 5.

8.2 Major Conclusions of the Thesis

Historically, the organisation of data concerning thunderstorm rainfall in time and space has been an important objective for many climatologists. During the last few decades, the Sydney region has experienced some of the heaviest rainfalls on record from thunderstorms. The consequentflooding ha s caused considerable property damage and the loss of human life (Riley et al., 1985; Colquhoun and Shepherd, 1985; Colls, 1991). The structure and causes of Sydney's thunderstorms has been well studied (Colquhoun, 1972 and 1994; Speer and Geerts, 1994). Severe thunderstorms can be generated by active fronts, troughs or squall-lines (Morgan, 1979b), by local disturbances controlled by climatic-environmental factors (Linacre and Hobbs, 1977), or by supercells (Dickins, 1994). Often thunderstorms originate in response to the daily heating of hill slopes (Gentilli, 1971). The mountain ranges in the west of the Sydney area can set off thunderstorms in potentially unstable airflows, and these thunderstorms can then drift over the adjacent lowlands and coast (Foreman and Rigby, 1990). Local convection in the Sydney region as a result of surface heating and within warm, humid and unstable air­ masses can generate thunderstorms that produce light precipitation (Mitchell and Griffiths, 1993; Batt, 1994). These thunderstorms are common and comprise nearly 95 per cent of all thunderstorms in the region. On average they generally produce less than 11 mm of rainfall. Although these kinds of thunderstorms are normally small in size, they can contain vigorous parcels of rising and descending air occasionally accompanied by intense rainfalls (Morgan, 1979a). In coastal areas the presence of the Tasman Sea has a great influence on the occurrence of precipitation, because it can furnish and sustain a plentiful supply of moisture. Sea-breeze circulation (Linacre, 1992) along with surface heating of land can enhance convective activity over coastal areas (Clarke, 1955 and 1960; Drake, 1982; Abbs CHAPTER EIGHT Conclusions 122

and Physick, 1992). Individual events linked to these factors and producing heavy rainfalls have been described in the thesis.

Thunderstorm rainfalls are thus neither spatially nor temporally uniform. The prime objective of this thesis was to examine this variability and relate it to the main climatic characteristics and physiography of the Sydney region. Four sets of results were obtained. First, the general behaviour of thunderstorm rainfall (frequency and amount) were characterised over time (at yearly, seasonal, monthly and diurnal levels) using measures of central tendency and dispersion. Results, presented in Chapter 3, indicate that thunderstorms have marked diurnal and seasonal variations, and are most frequent in the spring (October to November) and summer (January to March) during the late afternoon and early evening. Thunderstorms are most frequent over the west of the region and least frequent over the lowland interior. Stations which are located near the ocean receive more thunderstorm rainfall than those located inland, even near the Blue Mountains. Second more detailed associations exist between thunderstorm rainfall and climatic factors such as air and sea temperatures, and air humidity. Results presented in Chapter 4 indicate that there are casual relationships between these climatic variables and thunderstorms, particularly for coastal stations. This association becomes weaker as one moves inland. Specifically, the amount of thunderstorm rainfall is affected by sea-surface temperature, the effects of unequal heating of land surfaces and the availability of moisture in the atmosphere. Third, the patterns of spatial variation and distribution of thunderstorm rainfall during the thundery months of the year (October to March) were examined. Results are presented in Chapter 6 using data from 191 stations in the Sydney region, for the 34-year period 1960 to 1993. The probability distribution of thunderstorm rainfall amounts was shown to be described well using the gamma distribution. This technique provided two measures (beta and alpha values) which described the patterns of thunderstorm rainfall in the Sydney region. In addition, a GIS method was used to characterise the spatial pattern of thunderstorm rainfall over the Sydney region. The distribution of mean thunderstorm rainfall in the Sydney region, reflects topographic, coastal and urban effects. Thunderstorm rainfall increases with proximity to the ocean, in the vicinity of elevated topography over the Illawarra Plateau and Blue Mountains, and over built-up metropolitan areas, especially the CBD and eastern suburbs of the city. In the latter case, rainfall may be enhanced by urban heating, increased surface roughness and air pollution. These urban areas are more subject to flash flooding. Finally, because the physical environment affects the spatial distribution of thunderstorm rainfall, more detailed analysis was undertaken using a "climatologically oriented" GIS in conjunction with a stepwise regression technique. These results are presented in Chapter 7. While synoptic conditions initiate thunderstorm weather systems, physiographic parameters considerably influence the spatial distribution of the resulting rainfall amounts. CHAPTER EIGHT Conclusions 200

8.3 Limitations of the Study

8.3 1 Limitations of Data Used

During the research considerable time was spent in collecting data from different relevant sources and then summarising these. The main limitations of these data can be summarised as follows:

1) The distribution of the stations recording thunder reflects the distribution of major population clusters, older suburbs, dams, post offices and railway stations. This sampling network is spatially uneven and forms only a first approximation of the distribution of any rainfall. In addition, some records were sporadic over time. To overcome this limitation, only the best and longest records were initially chosen to study the generalised distribution of thunderstorm rainfall over time.

2) The number of daily thunderstorm observations for each particular station was not the same, because some stations reported every three hours and some only once or twice a day. To overcome this problem, thunderstorm occurrence was studied on a daily basis. If a station recorded at least one observation of thunder on a particular day, then all of the rainfall for that day was considered as being thunderstorm derived. All rainfall values used in this study should thus be considered maximum amounts.

3) Some rainfall stations did not have a complete record for the 34 years under consideration. Therefore, the period of study may not be represented adequately for some parts of the region. While no temporal constraints appear to be defined in the literature for rainfall records (Alexander, 1945; Longley, 1952 and 1974), in this study, only records of ten years or more were utilised.

4) Some problems were encountered with missing data. Generally, both thunder- recording and rainfall stations with extensive missing records have been removed from the data base. No attempt was made to compensate for missing values in the remaining records. There is, therefore, a possibility of error being introduced into the data set because of missing data.

5) The sea-surface temperature data were recorded weekly. Therefore, it was not possible to calculate and analyse the associations between thunderstorm data and sea surface temperature on a daily basis.

6) Generally, the north-east and south-west areas of the Sydney region suffer from a poor coverage of rain-gauges, leading to an incomplete picture of thunderstorm rainfall. CHAPTER EIGHT Conclusions 201

This uneven distribution of rainfall stations, particularly in mountainous areas, together with the equally problematic issue of missing data, should be satisfactorily dealt with in future research in order to understand accurately the distribution of thunderstorm rainfall. One of the best ways to overcome this problem is to increase the density of the rain-gauge network in the Greater Sydney Region.

8.3.2 Limitations of Techniques Applied

GIS was found to be a useful tool for data analysis and display of climatic variables. GIS has many advantages. For example, it has the capability of combining large amounts of spatial data in varied formats. Furthermore, GIS was very helpful in extracting the various study area characteristics, derived from topographic and landuse maps. While the simple functions of GIS provided the means for handling massive datafiles, GI S also allowed multivariate analyses of the rainfall data and associated physiographic variables of both a spatial and non-spatial nature. However, there are some disadvantages in using the SPANS GIS technique for climatic purposes:

1) The major limitation to GIS, in climatic modelling, is the current inability of the SPANS system to incorporate temporal change. The application of dynamic GIS in this field requires specialised analytical tools. The author hopes to carry out further study on this topic.

2) The raster-based GIS is limited by the fact that the minimum resolution of a raster map has to be decided when the map isfirst created . This resolution has to be set small enough in order not to lose spatial information; however this sometimes leads to very large grids where interpolation of variables may occur without much underlying data control. Currently the interpolation functions in SPANS GIS are primitive. A broader range of interpolation algorithms is required to overcome this problem. For example, a veryflexible TI N function is needed to create smoother isohyets.

3) Given that our future climatic data and our future modelling requirements will become more complex, it is necessary that the SPANS GIS can include new functions which will permit improved transfer of data, more efficient storage, and more flexibility and faster modelling capabilities. If these improvements are made, the SPANS GIS can be more widely used in research of climatic data.

8.4 Advantages and Implications of the Study

While researchers have examined many different aspects of thunderstorm activity in the Sydney region over the last four decades, three points have not been thoroughly considered. These are: CHAPTER EIGHT Conclusions 202

1) the general characteristics of the thunderstorm rainfall distribution over time; 2) the causes of variation in thunderstorm rainfalls across the region; and 3) the spatial variability and distribution of thunderstorm rainfall over time.

These aspects of thunderstorm rainfall climatology of the Sydney region were examined in the present thesis.

8.4.1 Advantages of the Study

The results of the above mentioned aims are presented in chapters 3 to 7. Thunderstorm activity can be viewed not only in terms of climatic variables, but also in terms of physiographic parameters which control the variation and distribution of thunderstorm rainfall. Three important techniques used in defining these relationships may have wider applicability in climate studies:

1) First, and of great importance, the gamma method offers summarised mathematical information about the variability of rainfall amounts from thunderstorms. This provides a convenient means of estimating the probabilities of receiving rainfall based on point observations, showing consistent spatial patterns. These patterns can be realistically compared with the real rainfall data which can then be linked to the mechanisms creating and controlling thunderstorms, such as synoptic systems and physiographic parameters. The spatial distribution of alpha and beta values help in broad- scale environmental planning and in establishing climatic regions where further detailed analyses, such as time series, can be performed.

2) Second, GIS computer technology offers an excellent means of analysing multivariate climatic-physiographic relationships. GIS can play a useful role in the analysis of the spatial distribution of rainfalls from thunderstorms. GIS also offers modelling capability of data from up to 19 map-layers. The application of GIS to the modelling of thunderstorm rainfall potential based upon physiographic features is one of the important outcomes of this thesis (see Chapter 5). Many climatologists (for example, Sajecki, 1991; Brignall et al., 1991) believe that the SPANS GIS system provides the logical framework for complex analysis and mapping of such climatic data.

3) Finally, the stepwise multiple regression technique is a useful technique for defining the relative importance of important climatic and physiographic factors influencing thunderstorm rainfall. This procedure was used twice in the study. The most important advantage of this technique is that it can be used to obtain the maximum degree of explanation of a dependent variable from a composite of independent variables. CHAPTER EIGHT Conclusions 201

8.4.2 Implications of the Study

Information obtained about the distribution of thunderstorm rainfall in the Sydney region in this thesis has some important management consequences as follows:

First, the results of this study explain, simply, the times when thunderstorm rainfall should occur in different zones of the Sydney region. This was highlighted in Chapter 3 where the temporal distribution of thunderstorm rainfall was discussed. Such information can be used by the State Emergency Service to narrow down the time of day when a response to heavy precipitation during thunderstorms is most likely to be required.

Second, spatial fluctuations in thunderstorms impact differently on human activity and development across the Sydney region. While isohyet maps contain very comprehensive information about the distribution of rainfall over the region, the gamma estimators (beta and alpha values) facilitate easy reading of the displayed probability values of thunderstorm rainfall at any point in the study area. Urbanisation is progressively increasing the amount of run-off during short rainfall events and leading toflash flooding. Areas where this is most common can be targeted by planners and engineers using these gamma maps and their associated Z-Score values.

Third, results from the study should make planners aware of the increased risk of heavy rain from thunderstorms induced by heavy industrial, residential and commercial development. The State Planning Authority of NSW (1970 and 1994) believes that within 25 years, more than 70 per cent of the people of NSW will live in the Greater Sydney Region. This will only exacerbate the chance and impact of heavy thunderstorm rainfall. It is imperative that detailed data on the urban heat island be acquired in order to understand how intense urbanisation affects the rainfall process. Perhaps a project like METROMEX (Changnon et al., 1971) should be conducted in the Sydney region or the Metropolitan Air Quality Study (MAQS) extended in scope to cover the flooding risk from intense thunderstorms.

Fourth, the spatial variations in urban rainfall are more important than those in rural areas, simply because more people will be affected. Adequate information on thunderstorm rainfall within Sydney can be of major importance in the planning and design of drainage systems, outdoor social and recreational activities, intra-city transport systems, watershed protection, andfinally flood prevention. It is hoped that the recognition of climatological landuse types defined in the thesis can be of assistance in this planning. Urban climatologists should work closely with the landuse planners, zoning authorities, architects, and hydrologists to ensure that heat islands are reduced as much as possible; CHAPTER EIGHT Conclusions 20A

that vegetative areas are interspersed throughout the Metropolitan area; that air pollution is reduced at its source, particularly over the CBD and nearby coast; and that hydrologic problems are reduced by increasing absorptive areas, providing ample storage in streams and lakes, and by constructing more channels to handle the rate of run-off. This planning is needed particularly in those areas which have been shown by this study to have a high potential for heavy amounts of thunderstorm rainfall and subsequentflash flooding.

8.5 Suggestions for Future Studies

This thesis is not inclusive. Important areas of further research should include the following:

1) the regional development of synoptic weather maps for the Greater Sydney Region that can be linked to thunderstorm development. This would allow rainfall predictions to be made in advance of the progress of a thunderstorm;

2) the application of the gamma distribution to other types of synoptic patterns generating heavy rainfall in the Sydney region. The present study did not include all rainfalls produced from east coast lows which are a major factor in generating heavy rainfall in the Sydney region (Bryant, 1991);

3) the techniques used in the study should be applied to a wider area of the New South Wales coast where other heavy thunderstorm rainfall events are know to occur; and

4) sophisticated time series models should be applied to the data to better characterise temporal change in thunderstorm rainfalls. The application of advanced models, such as the Auto-Regressive Integrated Moving Averages (ARTMA) technique, may help to construct better thunderstorm rainfall prediction models over time.

8.6 Concluding Remarks

Although the problems caused by thunderstorms are many and varied and the solutions do not seem simple, this current thesis has been built upon solid research provided over the last 25 years by meteorologists and climatologists (for example Hobbs, 1971, 1972, 1995; Colquhoun et al., 1985; Bryant, 1991; Williams, 1991; Linacre, 1992; Griffiths et al., 1993; Matthews, 1993; Colquhoun, 1994; Laudet et al., 1994; Speer and Geerts, 1994; Batt et al., 1995; and Matthews and Geerts, 1995). It is hoped that information and models contained within this thesis can also contribute to this understanding of thunderstorm rainfall. In terms of the spatial variation and distribution of thunderstorm rainfalls, most specifically from the analyses of the available data, the following conclusions can also be deduced: CHAPTER EIGHT Conclusions 201

1) The distribution of thunderstorms over the Sydney region varies from year-to- year and reflects the overall impacts of synoptic patterns and physiographic parameters. In general, the probability of receiving heavy precipitation (gamma estimators) from thunderstorms is greatest in the east of the Sydney region, particularly over the Metropolitan area, in spring and summer. This rainfall decreases towards the western suburbs and then increases again over the Blue Mountains. The decrease is most marked over the Southern Tableland.

2) The proximity to the Tasman Sea is the most important physiographic parameters controlling the spatial patterns in rainfall from thunderstorms. Generally, areas near the coast play a major part in the 'turning-on' of thunderstorms producing heavy rainfalls.

3) Spatial rainfall variations are associated with thermodynamic and/or kinematic characteristics of the landuse patterns. Intense urbanisation results in the heaviest rainfalls.

4) Thunderstorm rainfall intensity is also dependent on aspect and elevation. Generally areas of higher relief receive more rainfall.

The information obtained here can be used in many areas such as urban planning, design of rain-gauge network, flash flood control programs and emergency response management. Hopefully the study will spur other research into the identification and explanation of thunderstorm rainfall patterns along the east coast of Australia. REFERENCES 2M

REFERENCES

Abbott, P.F., and Tabony, R.C. 1985. The estimation of humidity parameters. Meteorological Magazine, 114, 49-56. Abbs, D.J. and Physick, W.L. 1992. Sea-breeze observations and modelling: a review. Australian Meteorological Magazine, 41, 7-19. Alberti, M. 1991. Suitability analysis and environmental impact modelling with GIS. Proceedings of the Urban and Regional Information Systems Conference, San Francisco 11-15 August 1991, 110-123, USA. Alexander, W.H. 1945. Hydrometeorological report No. 5, part 2. Hydrometeorological Section Office of Hydrological Director, Weather Bureau, USA. Alford, P. 1994. Thunderstorm classification. 4th Severe Thunderstorm Conference (8-12 August 1994), Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Allan, R.J. 1985. The Australian summer monson, teleconnections, and flooding in the Lake Eyre basin. South Australian Geographical Papers No. 2, Royal Geographical Society of Australia. Allan, R.J. 1988. El Nino Southern Oscillations influnces in the Australian region. Progress in Physical Geography 12, 313-348. Allan, R.J. 1989. ENSO and climatic fluctuations in Australia. Proceedings of Third International Conference on Southern Hemisphere Meteorology and Oceanography, Buenos Aires, 371-372. Amanatidis, G.T., Housiadas, C. and Bartzis, J.G. 1991. Spatial distribution of rainfall in the Greater Athens Area. Meteorological Magazine, 120, 41-50. Andersson, T., Andersson, M., Jacobsson, C. and Nilsson, S. 1989. Thermodynamic indices for forecasting thunderstorms in Southern Sweden. Meteorological Magazine, 118, 141-146. Armstrong, J.G. and Colquhoun, J.R. 1976. Intense rainfalls from the thunderstonns over the Sydney Metropolitan and Illawarra districts - 10 and 11 March 1975. Technical Report NO. 19, Bureau of Meteorology, Sydney, Australia. Aronoff, S. 1989. Geographic information systems: a management perspective. WDL Publications, Ottawa, Canada. Aspinall, R. and Miller, D. 1990. Mixing climate change models with remotely-sensed data using raster based GIS. In M.G. Coulson (ed.). Proceedings of the 16th Annual Conference of the Remote Sensing Society. University of Swansea, UK. Astling, E.G. 1984. On the relationship between diumal mesoscale circulations and precipitation in a mountain valley. Journal of Climate and Applied Meteorology, 23, 1635-1644. Atkinson, B.W. 1968. A preliminary examination of the possible effect of London's urban area on the distribution of thunder rainfall 1951-60. Transactions. Institute of British Geographers, 44, 97-118. REFERENCES 201

Atkinson, B.W. 1969. A further examination of the urban maximum of thunder rainfall in London, 1951-60. Transactions. Institute of British Geographers, 48, 97-120. Atkinson, B.W. 1971. The effect of an urban area on the precipitation from a moving thunderstorm. Journal ofApplied Meteorology, 10, 47-55. Atkinson, B.W. 1975. The mechanical effect of an urban area on convective precipitation. Occasional Paper 3, Dept. Of Geography, Queen Mary College, University of London. Atkinson, B.W. 1977. Urban effects on precipitation: an investigation of London's influence on the severe storm in August 1975. Occasional paper No. 8, Department of Geography, Queen Mary College, University of London, UK. Atkinson, B.W. 1981. Meso-scale atmospheric circulations. Academic Press, INC. London, UK. Atkinson, B.W. 1983. Rainshadow - the dynamical factor. Weather, 38, 264-269. Atkinson, B.W. and Smithson, P.A. 1974. Rain pulses in a warm sector. Weather, 29, 44-53. Auer, A.H. Jr. 1978. Correlation of land use and cover with meteorological anomalies. Journal of Applied Meteorology, 17, 636-643. AUSLIG, Australian Surveying and Land Information Group. 1993. Sydney, Australia. Bader, M.J. and Roach, W.T. 1977. Orographic rainfall in warm sectors of depressions. Quarterly Journal. Royal Meteorological Society, 103, 269-280. Bahr, V.J., Kemp, R.L. and Kurzeme, D. 1973. Report on a study of the severe thunderstorms over Sydney Metropolitan area on 21st August, 1971. Special Report, No. 3. Bureau of Meteorology, Australia. Baines, P.G. 1990. What's interesting and different about Australian meteorology? Australian Meteorological Magazine, 38, 123-141. Balchin, W.G. and Pye, N. 1948. Local rainfall variations in Bath and the surrounding district. Quarterly Journal. Royal Meteorological Society, 74, 361-378. Balogun, E.E. 1981. Seasonal and spatial variations in thunderstorm activity in Nigeria. Weather, 36, 192-197. Banta, R.M. 1984. Daytime boundary-layer evolution over mountainous terrain, part 1: observation of the dry circulation. Monthly Weather Review 112, 340-356. Barkley, H. 1934. Thunderstorms in Australia. Bulletin No. 19, Bureau of Meteorology, Melbourne, Australia. Barnes, F.A. 1960. The intense thunder rains of 1 July 1952 in the northern Midlands. E. Midld Geogr., 14, 11-26. Barnes, S.L. and Newton, CW. 1986. Thunderstorms in the synoptic setting, In E. Kessler (ed.) thunderstorms dynamics and morphology. University of Oklahoma Press, 75-112, USA. Bartzokas, A. and Metaxas, D.A. 1994. Spatial and temporal sea-surface temperature covariances in the Mediterranean. International Journal of Climatology, 14, 202-213. Batt, K. 1994. The Manyana severe thunderstorm event of 8 November 1991. Meteorological Note 203, Bureau of Meteorology, Sydney, Australia. REFERENCES 208.

Batt, K., Hobbs, J. and Carradd, G. 1995. Climatology of severe local storms in New South Wales. Workshop on Atmospheric Hazards. Process, Awareness and Response. University of Queensland, September 20-22 1995. Benjamin, S.G. 1983. Some effects of surface heating and topography on the regional severe storm environment. Ph.D. thesis, the Pennsylvania State University, the Graduate School, Department of Meteorology. USA. Bergeron, T. 1965. On the low-level redistribution of atmospheric water caused by orography. Proceedings of the International Conference on Cloud Physics, May 24 - June 1, 1965, Tokyo and Sapporo, Supplement, 96-100, Japan. Berndtsson, R. 1989. Topographical and coastal influence on spatial precipitation patterns in Tunisia. International Journal of Climatology, 9, 357-369. Berndtsson, R. and Niemczynowicz, J. 1988. Spatial and temporal scales in rainfall analysis - some aspects and future prespectives. Journal ofHydrology, 100, 293-313. Bernhardsen, T. 1992. Geographic information systems. Arendel, Norway. Berry, J.K. 1986. Fundamental operations in computer-assisted map analysis. International Journal of Geographical Information Systems, 2, 119-136. Berry, J.L. and Marble, D.F. 1968. Spatial analysis; a reader in statistical geography. Prentice- Hall, INC., Englewood Cliffs, New Jersey., USA. Berry, L.K. 1993. Beyond mapping: concepts, algorithms, and issues in GIS. GIS World Books, Inc. USA. Beswick, A., Brook, K., Carter, J., Danaher, T., Duncalfe, F., Flood, N., Hugman, P., Peacock, A. and Young, R. 1993. Visualisation of environmental models: a case study and demonstration. In S.J. Riley, C. Devonport, P.W. Waggitt and B. Fitzpatrick (eds.). Proceedings of the North Australian Remote Sensing and GIS Forum, 9-11 August 1993, Darwin, Australia. Bjerknes, J. 1963. Changes of climate. UNESCO, Geneva. Blake, D. 1933. Storm types and resultant precipitation in the San Diago area. Monthly Weather Review, 61, 223-225. Blong, R.J. 1991. Natural hazards in the Australian region: implications for insurance and reinsurance. In N.R. Britton and J. Olive (eds.). Natural and technological hazards: implications for the insurance industry. Proceedings of a seminar sponsored by Sterling Offices Ltd, University of New England, 51-70, Australia. Bonell, M. and Sumner, G.N. 1992. Autumn and winter daily precipitation areas in Wales, 1982- 1983 to 1986-1987. International Journal of Climatology, 12, 77-102. Bonham-Carter, G.F. 1993. Mineral potential mapping with GIS. Notes for a short course held at Department of Geography, University of Wollongong, Australia. Bonham-Carter, G.F. 1994. Geographic information systems for geoscientists: modelling with GIS. Pergamon, Canada. REFERENCES 201

Bonham-Carter, G.F., Agterberg, F.P. and Wright, D.F. 1988. Integration of geological datasets for gold exploration in Nova Scotia Photogrammetric Engineering and Remote Sensing, 54, 1586-1592. Bradshaw, M. and Weaver, R. 1993. Physical geography; an introduction to earth environments. Moshy, Sydney, Australia. Bridges, T.C. and Haan, C.T. 1972. Reliability of precipitation probabilities estimated from the gamma distribution. Monthly Weather Review, 100, 607-611. Bridgman, H.A. 1984. The spatial distribution of rainfall in the Hunter region. Conference on Australian Rainfall Variability, 6-8 August 1984. Ridgetops-Arkaroola, Australian Academy of Science, Bureau of Meteorology, Australia. Bridgman, H.A. 1990. Global air pollution: problems for the 1990s. Belhaven Press, London. Brignall, P., Parry, M. and Porter, J. 1991. GIS and the impacts of climate change. Awareness, 5(6), 32-34. Brooks, C.E.P. 1925. The distribution of thunderstorms over the globe. Geophysical Mem., Meteorological Office (London) No. 24, 145-164, UK. Brown, N.J. and Norris, D.A. 1988. Early applications of GIS at the Institute of Terrestrial Ecology. International Journal of Geographical Information Systems, 2, 153-160. Browning, K.A. 1962. Cellular structure of convective storms, Meteorological Magazine, 91, 341-350. Browning, K.A. 1980. Structure mechanism and prediction of orographically enhanced rain in Britain. WMO-ICSU, Orographic effects on planetary flows, GARP No. 23, 85-114. Browning, K.A. and Hill, F.F. 1981. Orographic rain. Weather, 36, 326-329. Browning, K.A., Pardoe, CW. and Hill, F.F. 1974. The nature of orographic rain at wintertime coldfronts. Quarterly Journal. Royal Meteorological Society, 101, 333-352. Bryant, E.A. 1982. Local climatic processes in the Illawarra. Wollongong Studies in Geography, No. 11. Geography Department, Wollongong University, Australia. Bryant, E.A. 1983a. Regional sealevel, Southern Oscillation and beach change, NSW, Australia. Nature, 305,213-216. Bryant, E.A. 1985a. Rainfall and beach erosion relationships, Stanwell Park, Australia, 1895- 1980: Worldwide implications for coastal erosion. Zeitschrift fur Geomorphologie. Supplementband, 57,51-65. Bryant, E.A. 1988. Storminess and high tide beach change, Stanwell Park, Australia, 1943- 1978. Marine Geology, 79, 171-187. Bryant, E.A. 1991. Natural hazards. Cambridge University Press. UK. Bryceson, K. and Bryant, M. 1993. The GIS / rainfall connection. The Australian Geographic Information Systems Applications Journal, 4, 32-35. Bufill, M.C. 1989. Effects of urbanisation on floods. Ph.D thesis, Department of Civil and Mining Engineering, Wollongong University, Australia. REFERENCES 210.

Bureau of Meteorology. 1965 to 1991. Monthly Weather Review (various months), NSW, Australia. Bureau of Meteorology. 1975. Monthly Weather Review (March, 1975). NSW, Australia. Bureau of Meteorology. 1976. Monthly Weather Review (November, 1976). NSW, Australia. Bureau of Meteorology. 1979. Climatic survey - Sydney. Australian Government Publishing Service, Canberra. Bureau of Meteorology. 1981. Monthly Weather Review, Western Australia, Perth, Australia. Bureau of Meteorology. 1984. Monthly Weather Review (February, 1984). NSW, Australia. Bureau of Meteorology. 1985. A report on the flash floods in the Sydney Metropolitan area over the period 5 to 9 November, 1984, Bureau of Meteorology, Australia. Bureau of Meteorology. 1987a. A report on the heavy rainfall and flood in the Sydney Metropolitan and nearby areas over the period 4-7 August 1986. Bureau of Meteorology, Sydney, Australia. Bureau of Meteorology. 1987b. Monthly Weather Review (October, 1987). NSW, Australia. Bureau of Meteorology. 1988. Monthly Weather Review (December, 1988). NSW, Australia. Bureau of Meteorology. 1989. Climate of Australia. Department of Meteorology, Administrative Services, Australia. Bureau of Meteorology. 1990. Monthly Weather Review (February, 1990). NSW, Australia. Bureau of Meteorology. 1991a. Climatic survey, Sydney and NSW. Australian Government Publishing Service, Canberra. Bureau of Meteorology. 1991b. Monthly Weather Review (January, 1991). NSW, Australia. Bureau of Meteorology. 1993a. Severe storms: facts, warnings and protection. Emergency management, Bureau of Meteorology, Australia. Bureau of Meteorology. 1993b. The Sydney hailstorm of 18 March 1990. Phenomena Report, Bureau of Meteorology, Australia. Bureau of Meteorology. 1995. The storm spotters handbook. Government Publishing Service, Canberra. Bureau of Meteorology. 1988a. Climatic atlas of Australia. Department of Adminstrative Services, Australia. Burrough, P.A. 1989. Principles of geographical information systems for land resources assessment. Clarendon Prees, Oxford, UK. Buttenfield, B.P. 1987. Automating the identification of cartographic lines. The American Cartographer, 14, 7-20. Byers, H.R. and Braham, R.R. 1949. The thunderstorm: report of the thunderstorm project. U.S. Government Printing Office. Washington DC, USA. Byers, H.R. and Rodehurst, H.R. 1948. Causes of thunderstorms of the Florida Peninsula. Journal of Meteorology, 5, 275-280. REFERENCES 21L

Calkins, H.W. and Tomlinson, R.F. 1977. Geographical information systems: methods and requirements for land use planning. IGU Commission on Geographical Data Sensing and Processing (RALI), USA. Campbell, A. 1906. Sonora storms and Sonora clouds of California. Monthly Weather Review, XXXJV, 464-465. Carpenter, F.A. 1913. The climate and weather of San Diego. California, San Diego Chamber Commerce, USA. Carras, J.N. and Johnson, G.M. 1982. The urban atmosphere - Sydney, a case study. CSIRO, Australia. Castro, A., Sanchez, J. L. and Fraile, R. 1992. Statistical comparison of the properties of thunderstorms in different areas around the Ebro-Valley (Spain). Atmospheric Research, 28, 237-257. Chacon, R.E. and Fernandez, W. 1985. Temporal and spatial rainfall variability in the mountainous region of Reventazon River Basin, Costa Rica. Journal of Climatology, 5, 175-188. Chandler, T.J. 1965. The climate ofLondon. Hutchinson, UK. Changnon, D. 1994. Regional and temporal variations in heavy precipitation in South Carolina. International Journal of Climatology, 14, 165-177. Changnon, S.A. Jr. 1968. The La Porte weather anomaly - fact orfiction? Bulletin. American Meteorological Society, 49,4-11. Changnon, S.A. Jr. 1969. Recent studies of urban effects on precipitation in the United States. Bulletin. American Meteorological Society, 50, 411-421. Changnon, S.A. Jr. 1973. Secular trends in thunderstorm frequencies. Eighth Conference on Severe Local Storms, October 15-17, 1973, Denver, Colorado, USA. Changnon, S.A. Jr. 1978. Urban effects on severe local storms at St. Louis. Journal of Applied Meteorology, 17, 578-586. Changnon, S.A. Jr. and Huff, F.A. 1973. Enhancement of severe weather by the St. Louis urban- industrial complex. Eighth Conference on Severe Local Storms, October 15-17, 1973, Denver, (122-129), Colorado, USA. Changnon, S.A. Jr. and Huff, F.A. 1986. The urban- related nocturnal rainfall anomaly at St. hams. Journal of Climate and Applied Meteorology, 25, 1985-1994. Changnon, S.A. Jr., Huff, F.A. and Semonin, R.G. 1971. METROMEX: an investigation of inadvertent weather modification. Bulletin. American Meteorological Society, 62, 958-968. Chen, C, Chan, W. and Deng Z. 1991. A study of a mountain-generated precipitation system in Northern Taiwan during TAMEXIOP 8. Monthly Weather Review, 119, 2574-2606. Cheong, H.F. and Tay, J.H. 1982. Statistical analysis of tropical thunderstorm intensities. In V.P. Singh (ed) Statistical analysis of rainfall and runoff. Water Resources Publications, Colorado, USA. REFERENCES 211

Chidley, T.R.E. and Keys, K.M. 1970. A rapid method of computing areal rainfall. Journal of Hydrology, 12, 15-24. Chidley, T.R.E. and Keys, K.M. 1972. Comment on calculation of mean areal depth of precipitation. By J. A. Akin. Journal ofHydrology, 15, 169-170. Chuan, G.K. and Lockwood, J.G. 1974. An assessment of topographical controls on the distribution of rainfall in the Central Pennines. Meteorological Magazine, 103, 275-287. Ciaramaglia, F., Calisti, C, Lungo, A. D., Nunzio, D. D. and Redat, C 1992. Forestry: a SPANS application for climatological analysis and productivity estimation, 31-35. First TYDAC European SPANS Users Conference 1992. Clarke, R.H. 1955. Some observations and comments on the sea breeze. Australian Meteorological Magazine, 11, 47-68. Clarke, R.H. 1960. The rain-producing potential of cold and sea breeze fronts. Seminar on Rain, Paper II/I, Bureau of Meteorology, Australia. Cliff, A., Hagget, P., Ord, J.K., Bassett, K.A., and Davies, R.B. 1975. Elements of spatial structure: a quantitative approach. Cambridge Unversity Press, London. Cliff, A.D. and Ord, J.K. 1981. Spatial processes models and applications. Pion Ltd. London, UK. Cohen, D.D., Martin, J.W., Bailey, G.M., Crisp, P.T., Bryant, E., Rothwell, R, Banks, J. and Hyde, R. 1994. A twelve-month survey of lead in fine airborne particles in the major population areas of NSW. Clean Air, 28, 79-88. Colls, K. 1991. Climatology of major flood events in the Sydney Metropolitan area. Bureau of Meteorology, NSW Regional Office, Sydney, Australia. Colls, K. and Whitaker, R. 1990. The Australian weather book. Child and Associates Publishing, Sydney, Australia. Colquhoun, J.R. 1972. An investigation of the severe thunderstorm at Richmond, NSW on the 25 December 1971. Meteorological Note, 56, Bureau of Meteorology, Sydney, Australia. Colquhoun, J.R. 1987. A decision tree method of forecasting thunderstorms, severe thunderstorms and tornadoes. American Meteorological Society, 2, 337-345. Colquhoun, J.R. and Shepherd, D.J. 1985. A report on the flash floods in the Sydney Metropolitan area over the period 5 to 9 November 1984. Bureau of Meteorology, Sydney, Australia. Colquhoun, J.R., Shepherd, D.J., Coulman, CE., Smith, RK. and Mclnnes, K 1985. The Southerly Burster of South Eastern Australia: an orographically forced cold front. Monthly Weather Review, 113, 2090-2107. Colquhoun. J. R. 1994. Severe thunderstorm forecasting in NSW - thefirst six years. 4th Severe Thunderstorm Conference (8-12 August 1994) Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Cook, B.G. 1978. The structural and algorithmic basis of a geographic data base. Hardware papers on GIS: First International Advanced Study Symposium on Topological Data REFERENCES 211

Structures for GIS. In G. Dutton (ed.). Laboratory for computer graphics and spatial analysis, Harvard University, USA. Cornish, E.A., Hill, G.W. and Evans, M.J. 1961. Inter-station correlations of rainfall in southern Australia. Technical paper 10, CSIRO, Australia. Court, A. 1960. Thunderstorm frequency in Northern California. Bulletin. American Meteorological Society, 41, 406-409 Cox, P.F. 1983. Precipitation regimes of the Illawarra coast and adjacent highlands. Wollongong Studies in Geography, No. 12, Department of Geography, University of Wollongong, Australia. Craig, D. 1980. Two examples of the effects of topography on rainfall distribution patterns. Weather, 55,301-307. Cressie, N. 1991. Statistics for spatial data. John Wiley, New York, USA. Critchfield, H.J. 1983. General climatology (4th ed.). Prentice-Hall, USA. Critchfield, H.J. 1987. Sources of climatic data. In J.E. Oliver and R.W. Fairbridge (eds.). Encyclopedia of Climatology. Van Nostrand Reinhold, 272-276, USA. Crowder, B. 1995. The wonders of the weather. Australian Government Publishing Service, Canberra, Australia. Cuff, D.J. and Mattson, M.T. 1982. Thematic maps, their design and production. Methuen, USA. Dangermond, J. 1986. The software toolbox approach to meeting the user's needs for GIS analysis. GIS Workshop, 1986, USA. Dangermond, J. 1989. The organisational impact of GIS technology. ARC News Summer, 25-26. Dangermond, J. and Smith, L.K. 1988. GIS and the revolution in cartography: the nature of the role played by a commercial organisation. American Cartographer, 15, 301-310. Dargie, S. 1994. The Gatton Pulse severe thunderstorm of 5th December 1993. 4th Severe Thunderstorm Conference (8-12 August 1994) Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Dasarathy, B.V. 1991. Nearest neighbor (NN) Norms: NN pattern classification techniques. IEEE Computer Society Press, Washington, U.S.A. Davey, C. 1976. The metropolis and its natural environment: Sydney. In J.H. Holmes (ed.) Man and the environment: regional perspectives, Longman Cheshire, 133-157, Melbourne, Australia. Davidson, A.R. 1981. Flooding in Wollongong 1945-1977: a survey. Occasional Paper 30, Faculty of Military Studies, Department of Geography, University of NSW, Sydney, Australia. Davis, F. 1991. Environmental analysis using integrated GIS and remotely sensed data: some research needs and priorities. Photogrammetric Engineering and Remote Sensing, 57, 689- 698. Davis, J.C. 1973. Statistics and data analysis in geology. Wiley, New York, USA. REFERENCES

Department of Planning. 1995. Cities for the 21st Century - integrated urban management for Sydney, Newcastle, the Central Coast and Wollongong. NSW Government, Sydney, Australia. Dickins, J. 1994. South Australian supercells - a composite hodograph. 4th Severe Thunderstorm Conference (8-12 August 1994) Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Dickinson, H.J. and Calkins, H.W. 1988. The economic evalusion of implementing a GIS. International Journal of Geographical Information Systems, 2, 307-327. DoE, Department of the Environment. 1987. Handling geographic information. Report of the Committee of Enquiry chaired by Lord Chorley. HMSO, London, UK. Doswell, CM. and Brooks, H.E. 1993. Supercell thunderstorms. Weather, 48, 209-210. Drake, V.A. 1982. Insects in the sea breeze front at Canberra: a radar study. Weather, 37, 134. Duckstein, L., Fogel, M.M. and Thames, J.L. 1973. Elevation effects on rainfall: a stochastic model. Journal ofHydrology, 18 , 21-35. Duff, G.A. Eamus, D. 1992. Soil-climate-biotic interaction in Northern Australia Savannas- The GIS potential. Proceedings of the GIS and Environment Rehabilitation Workshop, Australia. E-RMS, Environmental Resource Mapping System. 1989. User's Manual, version 2.1, National Parks and Wildlife Service of NSW, Sydney, Australia. Eagle, R.J. and Geary, M.G. 1985. Sydney storms November 1984 - hydrological aspects- Public Workers Department Report No. 85014, 1-31, Australia. Easterbrook, CC and Rogers, CW. 1974. Case studies of coastal convective storms as observed byDoppler radar. CALSPAN report, CK-5077-M-2, USA. Easterling, D.R. 1989. Regionalization of thunderstorm rainfall in the Contiguous United States. International Journal of Climatology, 9, 567-579. Easterling, D.R. 1990. Persistent patterns of thunderstorm activity in the Central United States. Journal of Climate, 3, 1380-1389. Easterling, D.R. 1991. Climatological patterns of thunderstorm activity in South-Eastern USA. International Journal of Climatology, 11, 213 -221. Easterling, D.R. and Robinson, P.J. 1988. Synoptic-scale variability in the probability of precipitation from thunderstorms in the United States. Physical Geography, 9, 385-396. Ebdon, D. 1985. Statistics in geography: a practical approach (2nd edn.). Basil Blackwell, Oxford, UK. Egger, G. 1991. Flash flood warnings - public communication issues. Paper delivered at a flood warning workshop, Australian Counter Disaster College (now Australian Emergency Management Institute), 20-25 October 1991, Australia. Eklundh, L. and Pilesjo, P. 1990. Regionalization and spatial estimation of Ethiopian mean annual rainfall. International Journal of Climatology, 10, 473-494. REFERENCES 211

Elliott, J. 1994. Flash flood warning. 4th Severe Thunderstorm Conference (8-12 August 1994) Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Estoque, M.A. 1962. The sea breeze as a function of the prevailing synoptic situation. Journal of Atmospheric Science, 19, 244-250. Fairbridge, R.W. 1967. The encyclopedia of atmospheric sciences and astrogeology (vol. 2). Reinhold Publishing Corporutium, New York, USA. Fandry, CB. and Leslie, L.M. 1984. A two-layer quasi-geostrophic model of summer trough formation in the Australian easterlies. Journal of Atmospheric Science, 41, 807-818. Ferrier, S. and Smith, A.P. 1990. Using GISs for biological survey design, analysis and extrapolation. Australian Biologist, 3, 105-116. Findlater, J. 1963. Some aerial explorations of coastal airflow. Meteorological Magazine, 92, 231-243. Fitzpatrick, E.A. and Armstrong, J. 1973. Structure and terrain of Sydney area. In H.A. Nix (ed.). The city as a life system. Proceedings Ecological Society of Australia, 7, 1-26. Flowerdew, R. and Banting, D. 1989. Evaluating the potential role of GIS for a market analysis company. North West Regional Research Laboratory, Research Report 2. NWRRL, Lancaster, UK. Fogel, M.M. and Hyun, K.K. 1990. Simulating spatially varied thunderstorm rainfall. Proceedings of the International Symposium Hydraulics / Hydrology of Arid Lands, 1990, Notional Conference on Hydraulic engineering. San Diego. USA. Foreman, I.K., and Rigby, E.H. 1990. Flooding and flood mitigation in the Illawarra Region. 30th Annual Flood Mitigation Conference: Wollongong 15-18, May 1990, Technical paper No. 7, Australia. Fuquay, D.M. 1962. Mountain thunderstorms and forestfires. Weatherwise, 15, 140-141. Gentilli, J. 1968. Sun, climate and life. Jacaranda Press, Brisbane, Australia. Gentilli, J. 1971. Climates of Australia and New Zealand. World Survey of Climatology (vol. 13). Elsevier Publishing Company, Amsterdam. Gentry, R.C. and Moore, P.L. 1954. Relation of local and general wind interaction near the sea coast to time and location of air-masses showers. Journal of Meteorology, 11, 507-511. Golde, R.H. 1977. Lightning: physics of lightning (vol. 1). Academic Press, London. Golding, B.W. and Leslie, L.M. 1993. The impact of resolution and formulation on model simulations of an east coast low. Australian Meteorological Magazine, 42, 105-116. Goldreich, Y., and Manes, A. 1979. Urban effects on precipitation patterns in the Greater Tel- Aviv area. Arch. Met. Geoph. Biokl, Ser. B, 27, 213-224. Goodchild, M.F. and Kemp, K.K. 1990. Introduction to GIS. NCGIA Core Curriculum National Centre for Geographic Information and Analysis, University of California, Santa Barbara, USA. Grace, W., Watson, A., Tepper, G. and Manson, R. 1989. A climatology of severe thunderstorms in South Australia. Technical Report No. 62, Bureau of Meteorology, Adelaide, Australia, REFERENCES 216

Griffiths, D.J., Colquhoun, J.R, Batt, K.L. and Casinader, T.R. 1993. Severe thunderstorms in the NSW: climatology and means of assessing the impact of climate change. Climate Change, 25, 369-388. Griffiths, G.A. and Saveney, M.J. 1983. Distribution of mean annual precipitation across some steepland regions of New Zealand. New Zealand Journal of Science, 26, 197-209. Hales, J.E. 1978. The Kansas City flash flood of 12 September 1977. Bulletin. American Meteorological Society 59, 706-710. Hane, CE. 1978. Scavenging of urban pollutions by thunderstorm rainfall: numerical experimentation. Journal of Applied Meteorology, 17, 699-710. Harnack, R.P. and Landsberg, H.E. 1975. Selected cases of convective precipitation caused by the metropolitan area of Washington DC. Journal of Applied Meteorology, 14, 1050-1079. Hastenrath, S. 1984. Interannual variability and annual cycle: mechanisms of circulation and climate in the tropical Atlantic sector, Monthly Weather Review, 112, 1097-1107. Hauser, O. 1974. Some problems in the use of stepwise regression techniques in geographical research. Canadian Geographer, 18, 148-158. Head, L.M., O'Neill, A.L., Marthick, J.K. and Fullagar, R.L.K. 1992. A comparison of aboriginal and pastoralfires i n the north-west Northern Territory. In I. Moffatt and A. Webb (eds.). Conservation and Development Issues in Northern Australia. Australian National University, Darwin. Henry, J.A., Dicks, S.E. and Marotz, G.A. 1985. Urban and rural humidity distributions: relationships to surface materials and land use. Journal of Climatology, 5, 53-62. Hicks, R.A. 1984. An example of mesocale convective complex development in Melbourne, Australia, 15 November 1982. Meteorological Note 154, Bureau of Meteorology, Australia. Hill, F.F., Browning, K.A. and Bader, M.L. 1981. Radar and raingauge observations of orographic rain over South Wales. Quarterly Journal. Royal Meteorological Society, 107, 643-670. Hirst, A. and Linacre, E.T. 1978. Associations between coastal sea-surface temperatures, onshore winds, and rainfalls in the Sydney Area. Search, 9, 325-327. Hobbs, J. 1971. Rainfall regimes of Northeastern NSW. Australian Meteorological Magazine, 19,91-116. Hobbs, J. 1972. An appraisal of rainfall trends in Northeastern NSW. Australian Geographical Studies, 10, 42-60. Hobbs, J. 1995. Storm of data on big weather questions in project using 450 research assistants. University of New England, Campus Review, September 14-20, 1995, Armidale, Australia. Hobbs, J. and Littlejohns, S. 1991. A case study of severe storm patterns: implications of mesozonation for hazard insurance. In Natural and Technological Hazards: Implications for the Insurance Industry (71-95). Sterling Offices (Australia) Ltd. and Centre Disaster Management, University of New England, Armidale, Australia. REFERENCES 211

Holland, G.J., Lynch, A.H. and Leslie, L.M. 1987. Australian East- Coast cyclones, part 1: synoptic overview and case study. Monthly Weather Review, 115, 3024-3036. Holroyd, E.W. 1982. Some observations on mountain-generated cumulonimbus rainfall on the northern Great Plains. Journal of Applied Meteorology, 21, 560-565. Holton, J.R. 1992. An introduction to dynamic meteorology (3rd edn.). Academic Press, New York, USA. Hopkins, L.C. and Holland, G.J. 1994. Australian east-coast cyclones and heavy-rain days: 1958- 1992. Submitted to Journal of Climate. Houghton, D.D. 1985. Handbook of applied meteorology. John Wiley and Sons, New York, USA. Houze, Jr. R.A. 1993. Cloud dynamics. Academic Press, New York. Huff, F.A. 1975. Urban effects on the distribution of heavy convective rainfall. Water Resources Research, 11, 889-896. Huff, F.A. and Changnon S.A. Jr. 1973. Precipitation modification by major urban areas. Bulletin. American Meteorological Society, 54, 1220-1232. Huff, F.A. and Changnon, S.A. Jr. 1972. Climatological assessment of urban effects on precipitation at St. Louis. Journal of Applied Meteorology, 11, 823-842. Huff, F.A. and Changnon, S.A. Jr. 1986. Potential urban effects on precipitation in the winter and transition seasons at St. Louis, Missouri. Journal of Climate and Applied Meteorology, 25, 1887-1986. Huff, F.A. and Vogel, J.L. 1978. Urban, topographic and diurnal effects on rainfall in the St. Louis region. Journal ofApplied Meteorology, 17, 565-577. Hutchinson, M.F. and Bischof, R.J. 1983. A new method for estimating the spatial distribution of mean seasonal and annual rainfall applied to the Hunter Valley, New South Wales. Aust. Met. Mag., 31, 179-184. Hutchinson, P. 1970. A contribution to the problem of spacing raingauges in rugged terrain. Journal ofHydrology, 12, 1-14. Ibbs, T.J. and Stevens, A. 1989. Quadtree storage of vector data. International Journal of Geographical Information Systems, 2, 43-56. Ichiye, T. and Paterson L.R. 1963. The anomalous rainfall of the 1957-58 winter in the equatorial central pacific arid area. Journal of Applied Meteorology Society Japan, Series 41, 172- 182. Intera TYDAC. 1992a. SPANS GIS Reference Manual. Intera TYDAC Technologies Inc., Ottawa, Canada. Intera TYDAC 1992b. SPANS GIS Learning System. Intera TYDAC Technologies Inc., Ottawa, Canada. Intera TYDAC. 1993. SPANS GIS Reference Manual. Intera TYDAC Technologies Inc., Ottawa, Canada. REFERENCES 218.

James, J.N. 1992. A preliminary study of mesoscale convective complexes over the Mid-Latitudes of Eastern Australia. Technical Report No. 66, Bureau of Meteorology, Melbourne, Australia. Jessup, N.J. 1985. An analysis of thunderstorm activity in southeast Tasmania on 11 November 1984. Meteorological Note, 165, Bureau of Meteorology, Regional Office, Hobart, Australia. Jessup, N.J. and Hughes, P.F. 1991. An analysis of thunderstorm activity over Tasmania on Monday 9 December 1985. Meteorological Note, 197, Bureau of Meteorology, Tasmanian Regional Office, Australia. IMP (Version 2). 1989. Software for statistical visualization on the Apple Macintosh. SAS Institute Inc. USA. Johnson, D.L. and Worobec, A. 1988. Spatial and temporal computer analysis of insects and weather: Grasshoppers and rainfall in Alberta. TYDAC, Technologies Inc. Ottawa, Canada. Johnson, R.W., Blong, R.J. and Ryan, CJ. 1995. Natural hazards: their potential in the Pacific Southwest. Australian Geological Survey Organisation, Australia. Johnston, CA., Detenbeck, N.E., Bonde, J.P. and Nicmi, G.J. 1988. GIS for cumulative impact assessment. Photogrammetric Engineering and Remote Sensing 54, 1609-1615. Johnston, K.M. 1987. Natural resource modelling in the GIS environment. Photogrammetric Engineering and Remote Sensing 53, 1411-1415. Joy, CS. 1991b. Natural disaster costs in Australia; Climate Change Impacts and Adaptation Workshop, Macquarie University 13-15 May, 1991, Sydney, Australia. Joy, CS. 1993. The cost of natural disasters in Australia. Climate Change, 25, 335-351. Kalma, J.D. 1974. An advective boundary-layer model applied to Sydney, Australia. Boundary- Layer Meteorology, 6, 351-361. Kalma, J.D. and Byrne, G.E. 1976. Energy use and the urban environment: some implications for planning. Proceedings of WMO Symposium on Meteorology Related to Urban Regional Landuse Planning. Geneva: WMO No. 444, 237-258. Kalma, J.D., Aston, A.R. and Millington, R.J. 1973. Energy use in the Sydney area. In H.A. Nix (ed.). The city as a life system. Proceedings Ecological Society of Australia, 7, 125-141, Australia. Karl, T.R. 1993. Missing pieces of the puzzle, research and exploration. Publication of the National Geographic Society, 9, 234-249. Kelly, J. 1994. Cartography, map-making, and GIS. The Society of University Cartographers (SUC) Bulletin, 28, 23-26. Kemp, R.A. and Armstrong, J.G. 1972. An investigation of temperature trends at Sydney (Observatory Hill) over all years of record. Working paper, Bureau of Meteorology, Melbourne, Australia. REFERENCES 211

Kessler, E. (ed.). 1986. Thunderstorms: a social, scientific and technological documentary (vol. 3). Instruments and techniques for thunderstorm observation and analysis (2nd ed.), University of Oklahoma Press, USA. Kessler, E. 1983. Thunderstorm morphology and dynamics (2nd ed.). University of Oklahoma Press, USA. Khemani, L.T. and Murty, R.B. 1973. Rainfall variations in an urban industrial region. Journal of Applied Meteorology, 12, 187-194. Kingwell, J. 1982. Some examples of the use of the Fawbush-Miller technique in the forecasting of extreme thunder squalls, Bureau of Meteorology, Perth, Australia. Klitch, M.A., Weaver, J.F., Kelly, F.P. and Vonder, T.H. 1985. Convective cloud climatologies constructed from satellite imagery. Monthly Weather Review, 113, 326-337. Knaap, W.G. 1992. The vector to raster conversion, (mis) use in GIS. International Journal of Geographical Information Systems, 6, 159-170. Konrad, CE. and Meentemeyer, V. 1994. Lower tropospheric warm air advection patterns associated with heavy rainfall over the Appalachian Region. Professional Geographer, 46, 143-155. Koshkariov, A.V., Tikunov, V.S. and Trofimov, A.M. 1989. The current state and the main trends in the development of GIS in the USSR. International Journal of Geographical Information Systems, 3, 257-272. Kuo, J.T. and Orville, H.D. 1973. A radar climatology of summertime convective clouds in the Black Hills. Journal of Applied Meteorology, 12, 359-368. Kurz, M. 1993. Severe thunderstorms over Western Germany - a case-study of the weather situation on 20 August 1992. Meteorological Magazine, 122, 177-188. L'hermitte, R. 1974. Kinematics of Seabreeze storms. Geophysical Research Letters, 1, 123-125. Landsberg, H.E. 1962. City air - better or worse, air over cities. Public Health Service, Technical Report A62.5, Cincinnati, 1-22, USA. Landsberg, H.E. 1981. The urban climate. Academic Press, London, UK. Laudet, T., King, P. and Colquhoun, J.R. 1994. A preliminary climatology of lightning in NSW. Proceedings of Fourth Conference on Severe Thunderstorms. Bureau of Meteorology, Australia. Lavery, B., Kariko, A., and Nicholls, N. 1992. A historical rainfall data set for Australia. Aust. Met. Mag., 40, 33-39. Leary, C.A., and Houze, E.N. 1987. The life cycle and internal structure of a mesoscale convective complex. Monthly Weather Review, 115, 1503-1527. Lee, D.O. 1984. Urban climate. Progress in Physical Geography, 8, 1-31. Lee, T.J., Pielke, R.A., Kittel, T.G.F., and Weaver, J.F. 1991. Atmospheric modeling and its spatial representation of land surface characteristics. Proc, First International Conference/Workshop on Integrating Geographic Information Systems and Environmental Modeling, Invited contribution, September, 1991, Boulder, Colorado, USA. REFERENCES

Leighton, R.M. and Spark, E. 1995. Relationship between synoptic climatology and pollution events in Sydney. Technical Report No. 68, Bureau of Meteorology, Sydney, Australia. Leslie, L.M., Holland, G.J. and Lynch, A.H. 1987. Australian east-coast cyclones. Part 2: Numerical modelling study. Monthly Weather Review, 115, 3037- 3053. Liasat, M.C and Ramis, C 1989. Synoptic, mesoscale and satellite study of the floods in Catalonina region during October 1987. Technical Note 298, WMO, Geneva. Lilly, D.K. 1986. The structure, energetics and propagation of rotating convective storms: helicity and storm stabilization. Journal of Atmospheric Science, 43, 126-140. Lilly, D.K. 1990. Numerical prediction of thunderstorms - Has its time come?. Quarterly Journal. Royal Meteorological Society 116, 779-798. Linacre, E.T. 1970. Problems of urban and industrial air pollution. Proceeding of Ecological Society, 5, 30-37. Linacre, E.T. 1992. Climate data and resources: a reference and guide. Routledge, London. Linacre, E.T. and Barrero, J.A. 1974. Surveys of surface winds in the Sydney region. Proceedings of International Geographical Union Regional Conference, and 8th New Zealand Geographical Conference, Palmerston North, Geographical Society, 247-260, New Zealand. Linacre, E.T. and Edgar, C. 1972. Climatology of Sydney in relation to air pollution. Proceeding 1972 International clean Air Conference Clean Air Society of Australia and New Zealand. Linacre, E.T. and Hobbs, J. 1977. The Australian climatic environment. John Wiley and Sons, Brisbane, Australia. Longley, R.W. 1952. Thunderstorm data for Canada. Technical Circular 2159, Meteorological Division, Department of Transport, Canada. Longley, R.W. 1974. Spatial variation of precipitation over the Canadian Prairies. Monthly Weather Review, 102, 307-312. Lough, J.M. 1992. Variations of sea-surface temperatures off North - Eastern Australia and associations with rainfall in Queensland: 1956-1987. International Journal of Climatology, 12, 765-782. Ludlam, F.H. 1962. A case study of thunderstorms on an occasion of small wind shear. Technical (Sci.) Note No. 11, Department of Meteorology, Imperial College, London. Ludlam, F.H. 1963. Severe local storms: a review. American Meteorological Society, 27, 1-31. Lutgens, F.K. and Tarbuck, E.J. 1982. The atmosphere: an introduction to meteorology. Prentice-Hall INC., Englewood Cliffs, New Jersey, USA. Lyons, T.J. 1977. Mesoscale flow analysis. Archiv fuer Meteorologie Geophysik, und Bioklimatologie. Serie B, 25, 233-249. Maddox R.A. 1980. Mesoscale convective complexes. Bulletin. American Meteorological Society, 61, 1374-1387. Maguire, D.J. 1989. Computers in geography. Longman Scientific and Technical, UK. REFERENCES 221

Maguire, D.J. 1991. An overview and definition of GIS. In D.J. Maguire, M.F. Goodchild and D.W. Rhind (eds.). Geographical information systems, principles and application. Longman, London. Maguire, D.L. and Dangermond, J. 1991. The functionality of GIS. In D.J. Maguire, M.F. Goodchild and D.W. Rhind (eds.). Geographical information systems - principles and application. Longman, London. Maine, R. 1962. Meteorological conditions associated with thunderstorms. Bureau of Meteorology, Melbourne, Australia. Marthick, J.K. 1995. Modelling of ground parrot (Pezoporus Wallicas) habitat in barren grounds nature reserve and Budderoo National Park. Hons thesis, Department of Geography, School of Geoscience, University of Wollongong, Australia. Martin, F.C 1985. Using a GIS for forest land management. Photogrammetric Engineering and Remote Sensing, 51, 1753-1759. Martin, J.J. 1982 Organization of geographical data with quadtrees and least squares approximation. Proceeding Symposium on Pattern Recognition and Image Processing, Las Vegas, Nevada, 458-465, IEEE Computer Society, USA. Mass, C. 1982. The topographically forced diurnal circulations of western Washington State and their influence on precipitation. Monthly Weather Review, 110, 170-183. Mathews, J.H. 1982. The sea-breeze - forecasting aspects. Aust. Met. Mag., 30, 205-209. MATLAB (Matrix Laboratory programming language) for Windows. 1994. Version 4.2b, Mathworks, INC, USA. Matthews, C. 1993. Distribution of thunderstorms and thunderstorm tracks in the Sydney Metropolitan area. Hons thesis, Macquarie University, Sydney. Matthews, C. and Geerts, B. 1995. Characteristic thunderstorm distribution in the Sydney area. Australian Meteorological Magazine, 44, 127-138. Max, N, Crawfis, R. and Williams, D. 1993. Visualisation for climatic modelling. IEEE Computer Graphics and Applications, 13 (4), 34-40. McCutchan, M.H. and Fox, D.G. 1986. Effect of elevation and aspect on wind, temperature and humidity. Journal ofClimate and Applied Meteorology, 25, 1996-2013. McGrath, CA. 1971. Observations of an urban heat island. Newcastle University, Geography Studies Journal, 1, 36-40. Mclnnes, K.L., Leslie, L.M. and McBride, J.L. 1992. Simulation of Australian east coast cut-off lows. Journal of Climatology, 12, 783-794. Merva, G.E., Strommen, N.D. and Kidder, E.H. 1976. Rainfall variations as influenced by wind and topography. Journal of Applied Meteorology, 15, 728-732. Michener, W.K. 1991. GIS assessment of large-scale ecological disturbances (hurricane Hugo). GIS/LIS, Atlanta, GA, ASPRS/ACSM, 343-355, USA. Mitchell, E. and Griffiths, D. 1993. Report on the Sydney hailstorm ofMarch 1990. Phenomenon Report, Bureau of Meteorology, Australia. 221.

Mooley, D.A., and Crutcher, H.L. 1968. An application of the gamma distribution function to Indian rainfall. ESSA Technical Report EDS 5, Department of Commerce, Environmental Data Service, Silver Spring, USA. Moore, A.W., Cook, B.G. and Lynch, L.G. 1981. Information systems for soil and related data. Proceeding 2nd Australian Meeting of the International Soil Science Society Working Group on Soil Information Systems, 19-21, February 1980. PUDOC, Wageningen, Canberra, Australia. Moran, J.M. and Morgan, M.D. 1991. Meteorology: the atmosphere and the science of weather. Macmillan Publishing Company, New York, USA. Morgan, P.A. 1979a. The Sydney severe thunderstorms of 10 November, 1976. Meteorological Note 96, Bureau of Meteorology, Sydney, Australia. Morgan, P.A. 1979b. The Sydney squall line of 21 January 1977. Meteorological Note 102, Bureau of Meteorology, Sydney, Australia. Mortimore, K.O. 1990. Thunderstorm climatological research in Great Britain and Ireland: a progress report and aims future study. Weather, 45, 21-26. Moss, J.M. 1965. Meteorological aspects of air pollution in the Sydney Region. Proceeding 1965 International Clean Air Conference Clean Air Society of Australia and New Zealand. Mottram, J. 1967. An investigation of a rapid and intense shear line. Meteorological Note 11, Bureau of Meteorology, Australia. Musk, L.F. 1988. Weather systems. Cambridge University Press, New York, USA. Nanson, G.C. and Hean, D. 1985. The West Dapto flood of February 1984: rainfall characteristics and channel changes. Australian Geographer, 16, 249-258. Neiburger, M., Edinger, J.G. and Bonner, W.D. 1982. Understanding our atmospheric environment (2nd ed.). Freeman and Company, USA. Nicholls, N. 1984. The Southern Oscillation, sea surface temperature, and interannual fluctuations in Australian tropical cyclone activity. J. Climatol, 4, 661-670. Nicholls, N. 1995. Lonh-term climate monitoring and extreme events. Climate Change, 31, 231- 245. Nicholls, N, and Kariko, A. 1993. East Australian rainfall events: Interannual variations, trends, and relationships with the Southern Oscillation. J. Clim., 6, 1141-1152. Norusis, M.J. 1994. SPSS advanced statistics 6.1. Chicago 111. SPSS. USA. O'Mahoney, G. 1961. Time series analysis of Australian rainfall data. Study No. 14, Bureau of Meteorology, Australia. O'Neill, A.L., Head. L.M. and Marthick, J.K. 1993. Integrating remote sensing and spatial analysis techniques to compare Aboriginal and pastoralfire patterns in the East Kimberley, Australia. Applied Geography, 13, 67-85. O'Neill, A.L., Moffatt, I. and Waggitt, P. 1992. Summary. Proceedings of the GIS and Environmental Rehabilitation Workshop. The Northern Territory University, Australia. 221.

Ohring, G. 1972. Application of stepwise multiple regression techniques to inversion of nimbus "IRIS" observations. Monthly Weather Review, 100, 336-344. Oke, T.R. 1977. The climate of urban areas. In Yearbook of Science and Future. Encyclopaedia Britannica, 116-131, Chicago, USA. Oke, T.R. 1979. Review of urban climatology. WMO publication, Technical Note 169, WMO, Geneva. Oladipo, E.O. and Mornu, M.E. 1985. Characteristics of thunderstorms in Zaria, Nigeria. Weather, 40,3X6-222. Oliver, J. 1986. Natural hazards. In D.N. Jeans (ed.) Australia: a geography. Sydney University Press, 283-314, Sydney. Oliver, J.E. and Fairbridge, R.W. 1987. The encyclopedia of climatology. Van Nostrand Reinhold Company, New York, USA. Osborn, H.B. 1982. Quantifiable differences between airmass and frontal-convective thunderstorm rainfall in the Southwestern United States. In V.P. Singh (ed.) Statistical analysis of rainfall and runoff, 21-32. Water Resources Publications, Colorado, USA. Paine, J., McLean, H. and Mayhew, S. 1988. Environmental geography (book one). Macmillan, Melbourne, Australia. Parry, M. 1956. An "urban rainstorm" in the Reading area. Weather, 11, 41-48. Passarelli, R.E. and Boehme, H. 1983. The orographic modulation of pre-warm-front precipitation in Southern New England. Monthly Weather Review, 111, 1062-1070. Pedgley, D.E. 1970. Heavy rainfalls over Snowdonia. Weather, 25, 340-350. Periago, M.C, Lana, X., Serra, C and Fernandez Mills, G. 1991. Precipitation regionalization: an application using a meteorological network in Catalonia (NE Spain). International Journal of Climatology, 11, 529-543. Pittock, A.B. 1975. Climate change and the pattern of variation in Australian rainfall. Search, 6, 498-503. Prezerakos, N.G. 1989. An investigation into the conditions in which air-mass thunderstorms occur at Athens. Meteorological Magazine, 118, 31-36. Prichard, R.J. 1990. A rare summer thunderstorm. Weather, 44, 164-168. Priestley, C.H.B. 1964. Rainfall / sea-surface temperature associations on the NSW coast. Australian Meteorological Magazine, 47, 15-25. Priestley, C.H.B. 1970. Interaction between oceans and atmosphere. Symposium on Marine Sciences, Canberra. (Reports of the Australian Academic Society, No, 12), 25-57, Australia. Priestley, C.H.B. and Troup, A.J. 1966. Droughts and wet periods and their association with sea- surface temperature. Australian Journal of Science, 29, 56-57. Prisley, S.P. 1986. Commercial GIS's for natural resources managements: what a manager needs to know. GIS Workshop April 1-4, 1986- Atlanta, Georgia, USA. 22A-

Ramage, CS. 1972. Sea surface temperature and local weather. Monthly Weather Review, 105, 540-544. Raper, J. 1993. Geographical information systems. Progress in Physical Geography, 17, 493-502. Rasuly, A.A. 1991. An introduction to GIS. Sepehr - Journal of National Geographic Organisation, 3, 50-55, Tehran, Iran. Rasuly, A.A. 1993. A review of different applications of GIS. Sepehr - Journal of National Geographic Organisation, 7, 42-51, Tehran, Iran. Reeder, M.J. and Smith, R.K. 1992. Australian spring and summer coldfronts. Australian Meteorological Magazine, 41, 101-124. Reid, I. 1973. The influence of slope aspect on precipitation receipt. Weather, 28, 490-494. Rhind, D.W. 1988. A GIS research agenda. International Journal of Geographical Information Systems, 2, 23-28. Richardson, CW. 1982. A comparison of three distributions for the generation of daily rainfall amounts. In V.P. Singh (ed). Statistical analysis of rainfall and runoff (67-7'8). Water Resources Publications, Colorado, USA. Riddle, A.M. 1991. Spill model for the protection of the fourth estuary. In D.G. Farmer and M.J. Rycroft (eds.). Computer modelling in the environment science. Clarendon Press, 49-64, Oxford, UK. Riley, S.J. 1980. Aspects of the flood record at Windsor, Proceedings, Institute of Australian Geographers Conference, Bathurst, Australia. Riley, S.J., Luscombe, G.B. and Williams, A.J. 1985. Urban stormwater design: lessons from the 8 November 1984 Sydney storm. Australian Geographer, 17, 40-48. Robinson, P.J. and Easterling, D.R. 1988. The frequency distribution of thunderstorm durations. Journal of Applied Meteorology, 27, 77-82. Rochford, D.J. 1977. Monitoring of coastal sea temperatures around Australia. Search, 8, 167- 169. Rodewald, M. 1963. Changes of climate, UNESCO, Geneva. Ryan, CJ. 1992. Dynamical classification of Australian thunderstorms. Meteorological Study No. 40, Bureau of Meteorology, Sydney, Australia. Ryan, CJ. 1993. Costs and benefits of tropical cyclone, severe thunderstorms and bushfires in Australia. Climate Change, 25, 353-367. Sajecki, P.J.F. 1991. A report on the use of SPANS to display LTM elements. Canadian Climate Centre, Canada. Salter, M. and de Cale, S. 1921. The rainfall of the British Isles. University of London Press. UK. Samet, H. 1989. Applications of spatial data structure. Addison-Wesley Publishing Company, Ontario, Canada. Sargent, G.P. 1980. Computation of vapour pressure, dew-point and relative humidity from dry- and wet-bulb temperatures. Meteorological Magazine, 109, 238-246. REFERENCES 225

Schermenhom, V.P. 1967. Relations between topography and annual precipitation in Western Oregon and Washington. Water Resources Research, 3, 707-711. Schwerdtfeger, P. 1982. Meteorological measurement relevant in planning urban environment. In Flinders Institute for atmospheric and marine sciences, research report No. 3, Australia. Sekiguti, T. and Tamiya, H. 1970. Precipitation climatology of Japanese city area, in urban climatology. Technical Note No. 108, WMO, Geneva. Shanahan, B.W. 1968. Local heavy rainfall - Sydney area. Meteorological Note 23, Bureau of Meteorology, Australia. Sharma, T.C 1987. Modelling extreme rain sequences in Zambia. Journal ofHydrology, 93, 101- 111. Sharon, D. 1983. The linear organization of localized storms in the summer rainfall zone of South Africa. Monthly Weather Review, Ul, 529-538. Sharon, D. and Kutiel, H. 1986. The distribution of rainfall intensity in Israel, its regional and seasonal variations and its climatological evaluation. Journal of Climatology, 6, 277-291. Shaw, E.M. 1962. An analysis of the origins of precipitation in Northern England, 1956 - 1960. Quarterly Journal. Royal Meteorological Society, 88, 539-547. Shaw, G., Wheeler, D. 1985. Statistical techniques in geographical analysis. John Wiley and Sons, New York, USA. Shenton, L.R. and Bowman, K.O. 1970. Remarks on Thorn's estimators for the gamma distribution. Monthly Weather Review, 98, 154-160. Shepherd, D.J. and Colquhoun, J.R. 1985. Meteorological aspects of an extraordinary flash flood event near Dapto, NSW. Australian Meteorological Magazine, 33, 87-102. Simmons, M. 1986. Crop yield estimation: South-Western Ethiopia, TYDAC Technologies INC., Canada. Simpson, J. 1972. Use of the gamma distribution in single-cloud rainfall analysis. Monthly Weather Review, 100, 309-312. Simpson, J.E. 1964. Sea breezefronts in hemisphere. Weather, 19, 208-219. Simpson, J.E. and Mansfied, D.A. and Milford, J.R. 1977. Inland penetration of sea breeze fronts. Quarterly Journal. Royal Meteorological Society, 103, 47-76. Sims, P.C 1981. Variations of rainfall catch in a small upland valley. Weather, 36, 197-201. Skidmore, A.K. 1989. A Comparison of techniques for calculating gradient and aspect from a girded digital elevation model. International Journal of Geographical Information Systems, 3, 323-334. Skidmore, A.K. 1990. Terrain position as mapped from a grided digital elevation model. International Journal of Geographical Information Systems, 4, 33-49. Smith R.B. 1982. A differential advection model of orographic rain. Monthly Weather Review, 110, 306-309.

Smith, D.I. 1993. Greenhouse climate change and flood damages, the implications. Climate Change, 25,319-333. REFERENCES 22L

Smith, K.B. 1975. Principles of applied climatology, McGraw Hill, London. UK. Smith, R.B. 1979. The influence of mountains on the atmosphere. Advances in Geophysics, Academic Press, 21, 87-230. Smith, R.B. 1985. Comment on 'interaction of low-level flow with the Western Ghat Mountains and offshore convection in the summer monsoon'. Monthly Weather Review, 113, 2176- 2177. Smith, R.B. 1989. Mechanisms of orographic precipitation. Meteorological Magazine, 118, 85- 88. Spark, E. and Casinader, T. 1995. The 21 January 1991 Sydney severe thunderstorm. Bureau of Meteorology, Sydney, Australia. Speer, M.S. and Geerts, B. 1994. A synoptic/mesoalpha-scale climatology of flash floods in the Sydney Metropolitan area. Australian Meteorological Magazine, 43, 87-103. Speer, M.S., and Leslie, L.M. 1994. Predicability of meso-6 convection using a high resolution NWP model. 4th Severe Thunderstorm Conference (8-12 August 1994) Australian Emergency Management Institute, Mount Macedon, Victoria, Australia. Spreen, W.C. 1947. A determination of the effect of topography upon precipitation. Transactions. American Geophysical Union, 28, 285-290. State Pollution Control Commission of NSW. 1974. Air Quality measurements in NSW. Annual Review, Sydney, Australia. State Pollution Control Commission of NSW. 1975. Quarterly report of air quality measurements in NSW. Sydney, Australia. Stol, PH. TH. 1972. The relative efficiency of the density of rain-gage networks. Journal of Hydrology, 15, 193-208. Storr, D. and Ferguson, H.L. 1983. The distribution of precipitation in some mountainous Canadian watersheds. Proceeding Geilo. Symposium (vol. 2), 243-263, WMO, Geneva. Streten, N.A. 1981. Southern Hemispheric sea-surface temperature variability and apparent associations with Australia rainfall. Journal of Geophysical Research, 86, 485-497. Streten, N.A. 1983. Extreme distributions of Australian annual rainfall in relation to sea-surface temperature. Journal of Climatology 3, 143-153. Strobl, J. 1992. Climatological modelling in Alpine regions. First TYDAC European SPANS Users Conference 1992, 47-54. Sumner, G.N. 1983a. Seasonal change in the distribution of rainfall over the Great Dividing Range: general trends. Australian Meteorological Magazine, 31, 121-130. Sumner, G.N. 1983b. The spatial organization of daily rainfall in Eastern NSW. Journal of Climatology, 3, 361-374. Sumner, G.N. 1984. The impact of wind circulation on the incidence and nature of rainstorms over Dar es Salaam, Tanzania. Journal of Climate, 4, 35-52. Sumner, G.N. 1988. Precipitation: process and analysis. John Willey and Sons, New York, USA. 222-.

Sumner, G.N. and Bonell, M. 1990. The spatial organization of daily rainfall in Wales during autumn-winter. Weather, 45, 174-183. Sumner, G.N., Ramis, C and Cuijarro, J. A. 1993. The spatial organization of daily rainfall over Mallorca, SPAIN. International Journal of Climatology, 13, 89-109. Suppiah, R. 1992. The Australian summer monsoon: a review. Progress in Physical Geography, 76,3.283-318. Surface Mapping System (Surfer). 1994. Version 5.01, Golden Software Inc., USA. Tabony, R.C 1980. Urban effects on trends of annual and seasonal rainfall in the London area. Meteorological Magazine, 109, 189-202. Tapp, R.G. and Skinner, CJ. 1990. Urban climatology in Australia: a brief survey of some recent studies. Meteorological Note 193, Bureau of Meteorology, Australia. Tapper, N. and Hurry, L. 1993. Australia's weather patterns: an introductory guide. Dellasta Pty. Ltd, Victoria, Australia. Taylor, C. 1992. Smog Moves West as Sydney grows. Ecos, 70, 17-22. Tepper, M. 1950. A proposed mechanism of squall-lines: the pressure jump line. Journal of Meteorology, 7,21-29. The State Planning Authority of NSW. (1970 and 1994). Sydney Region: Outline plans, Sydney, Australia. Theakstone, W.H. and Harrison, C. 1971. The analysis of geographical data. Heinemann Educational Books, London. Thom, H.CS. 1958. A note on the Gamma Distribution. Monthly Weather Review, 86, 117-122. Thom, H.CS. 1968. Approximate convolution of the gamma and mixed gamma distributions. Monthly Weather Review, 96, 883-886. Tomlinson, R.F. 1987. Current and potential uses of GIS: the North American experience. International Journal of Geographical Information Systems, 3, 303-218. Tomlinson, R.F., Calkins, H.W. and Marble, D.F. 1976. Computer handling of geographical data. UNESCO, Geneva. Treloar, A.B.A. 1991. The assessment and interpretation of severe thunderstorm damage in the Melbourne Metropolitan area on 22 December 1990. Meteorological Note, 196, Bureau of Meteorology, Melbourne, Australia. Tubbs, A.M. 1972. Summer thunderstorms over Southern California. Monthly Weather Review, 100, 799-807. Tucker, D.F. 1993. Diumal precipitation variations in south-central New Mexico. Monthly Weather Review, 121, 1979-1991. Tversky, A. 1983. Nearest neighbour analysis and point processes: applications to multidimensional scaling. JournalMathmatic Psychology, 27, 235-250. U.S. Weather Bureau. 1947. Thunderstorm rainfall. Hydrometeorological report 5, Washington, DC, USA. Unwin, D.J. 1981. Introductory spatial analysis. Methuen, New York, USA. REFERENCES 228

Urban Planning Centre. 1993. Department of Planning, Cartographic Branch, Sydney, Australia. Vogel, J.L. and Huff, F.A. 1978. Relation between the St. Louis urban precipitation anomaly and synoptic weather factors. Journal ofApplied Meteorology, 17, 1141-1152. Wallace, J.M. 1975. Diumal variation in precipitation and thunderstorm frequency over the Conterminous United States. Monthly Weather Review, 103, 406-419. Wallace, J.M. and Hobbs, P.V. 1977. Atmospheric Science: an introductory survey. Academic Press. New York, USA. Watson, D.F. 1992. Contouring: a guide to the analysis and display of spatial data. Pergamon Press, New York, USA. Webster, R. and Oliver, M.A. 1990. Statistical methods in soil and land resource survey. Oxford University Press, New York, USA. Webster, T. 1990. SPANS in geology. TYDAC Technologies Inc. Nepean, Ontario, Canada. Webster, T. 1992. GIS capabilities for geoscience applications. TYDAC Technologies Inc. Nepean, Ontario, Canada. Weeks, B. 1992. Flooding in the Illawarra: an historical perspective. BSc thesis, Department of Geography, University of Wollongong, Australia. Weibel, R. and Heller, M. 1991. Digital terrain modelling. In D.J. Maguire, M.F. Goodchild and D.W. Rhind (eds.). Geographic information systems: principles and applications. Longman, UK. Weisner, C.T. 1970. Hydrometeorology. Chapman and Hall, Ltd., London, UK. Wheeler, D. 1990. Modelling long-term rainfall patterns in North-East England. Meteorological Magazine, 119, 68-74. Whetton, P.H. 1988. A synoptic climatological analysis of rainfall variability in Southeastern Australia. Journal of Climatology, 8, 155-177. Whetton, P.H. 1990. Relationships between monthly anomalies of Australian region sea-surface temperature and Victorian rainfall. Aust. Met. Mag., 38, 31-41. White, D. 1985. Exceptionally heavy spring rains in Sydney. Weather, 40 , 210-212. Willett, H.C and Sanders, F. 1959. Descriptive meteorology (2nd ed.). Academic Press, New York, USA. Williams, A. 1984. Spatial and temporal distribution of the rainfall. In S. Riley and C Luscombe and A. Williams (eds.). Urban Flooding Conference. 16-44. Macquarie University, March 1985, Sydney, Australia. Williams, D.T. 1948. A surface microstudy of squall line thunderstorms. Monthly Weather Review, 76, 239-246. Williams, M. 1991. Regional forecasting (course notes). Bureau of Meteorology, Sydney, Australia. Wilson, K.J. and Ryan, B.F. 1987. The subsynoptic scale environment of mesocale convective weather systems at genesis over Southesthern Australia. Beitraege zur Physik der Atmosphaere, 60, 210-228. REFERENCES 221

Winkler, J.A. 1988. Climatological characteristics of summer-time extreme rainstorms in Minnesota. Annals of the Association ofAmerican Geographers, 78, 57-73. Wood, R.A. 1985. A dangerous family: the thunderstorm and its offspring. Weatherwise, 38, 131- 151. World Meteorological Organization (WMO). 1988. Manual on codes (vol.1). No. 306. WMO, Geneva. World Meteorological Organization. 1975. Technical regulations, Vol. l,No. 49 Geneva. Zar, J.H. 1984. Biostatistical analysis (2ed). Prentice-Hall of Australia PTY, Sydney. A List of Computer Programs 210.

APPENDIX A

List of Computer Programs

A.l Computer Program Number 1

To find the common thunderstorm-days in the Sydney region the Computer Program Number 1 was used (see Chapter 4). This program was written using a C1"4" programming language.

#include long lines (FILE *file); main() long int thun[1000][12]; int nil, nl2, nl3, nl4, nl5, n!6, nl7, nl8; int nl9, nllO, nlll, nll2, il, i2, i3, i4, i5, i6, i7, i8, i9; int ilO, ill, il2, i, j, k, m, n; long int xl, x2, x3, x4, x5, x6, x7, x8, x9, xlO, xll, xl2; FILE *fpl, *fp2, *fp3, *fp4, *fp5, *fp6, *fp7, *fp8, *fp9; FILE*fplO, *fpll, *fpl2; fpl =fopen("a:thul.txt", V); fp2 = fopen("a:thu2.txt", "r"); fp3 =fopen(,,a:thu3.txt", "r"); fp4 = fopen("a:thu4.txt", V); fp5 = fopen("a:thu5.txt", "r"); fp6 = fopen("a:thu6.txt", "r"); fp7 = fopen("a:thu7.txt", "r"); Q)8 = fopen("a:thu8.txt", "r"); fp9 = fopen("a:thu9.txt", "r"); fplO= fopen("a:thul0.txt", "r"); fpll= fopen("a:thull.txt", "r"); fpl2= fopen("a:thul2.txt", "r"); nll = lines(Q)l); n!2 = lines(fi)2); nl3 = lines(Q)3); nl4 = lines(^)4); nl5 = lines(fp5); nl6 = lines(lp6); nl7 = lines(fp7); nl8 = lines(lp8); nl9 = lines(fp9); nllO= lines(fpl0); nlll=lines(fpll); nll2= lines(fpl2); rewind(fpl) rewind(jfp2) rewind(fp3) rewind(Q)4) rewind(jQ)5) rewind(fp6) rewind(fp7) APPENDIX A List of Computer Programs 211

rewind(fp8); rewind(fp9); rewind(fplO); rewind(fpll); rewind(fpl2); k=0; for(il=0; iKnll; ++il) { fscanf(fpl,"%ld\n",&xl); thun[il][0]=xl; } for (il=nll; iKIOOO; ++il) wun[il][0]=0; for(i2=0; i2

fscanf(fp8,',%ld\n",&x8); thun[i8][7]=x8; } for (i8=nl8; i8<1000; ++i8) thun[i8][7]=0; for(i9=0; i9 for (ill=nlll; ilKlOOO; ++ill) thun[ill][10]=0; for(il2=0; il2

A.2 Computer Program Number 2 (C1^)

Due to the large volume of thunderstorm rainfall data recorded on a daily basis, it was very difficult and time consuming to extract the specified data sets. This was due to the data-set filesfrom mor e than 350 rainfall stations located in the Sydney region operated by the Bureau of Meteorology during the last 34 years (1960 to 1993). There were also more than 100 stations which have been covered by the Sydney Water. These data sets were in different formats, so it was impossible to extract the data manually, and the possibility of errors were very high. To overcome these difficulties three computer programs were developed, each for a specific format, using three different computer programming languages as follows:

This computer program was written to extract the daily rainfall events from the Sydney Water data sets. Each rainfall station data was in ASCI format.

#include #include #include int lines (FILE *file); main(int argc, char *argv[]) { FILE *fpl, *fp2, *lp3; char c[500], cc; intij, k, kk, nil, nl2; float x, y, z, ml, m2, m3, m4, m5; if((fpl = fopen(*++argv, "r"))==NULL) { printf("File \"%s\" does not exist.\n", *argv); printf("The usage is: Y'Alil inputfile outputfile\"\n"); return 1; } else if ((fp3 = fopen(*++argv, "w")) == NULL) { printf("File \"%s\" does not exist.\n", *argv); printffThe usage is: V'Alil inputfile outputfileVW); return 1; } fp2 = fopen("match.txt", "r"); nil = lines (fpl); nl2 = lines (fp2); rewind (fpl); rewind (fp2); prinrf("nll=%d nl2=%d\n", nll,nl2); for (k=0; k rewind (fpl); } rewind (fpl); } /* */

/* This function counts the number of lines in a file */ int lines (FILE *file) { int c, nl; nl = 0; while ((c = getc(file)) != EOF) if(c = V) ++nl; return nl;

A.3 Computer Program Number 3 (Fortran 77)

This computer program was written to extract the daily rainfall events from the Sydney Water data sets. Each rainfall station data was not in ASCI format. This program is written in a standard Fortran which can be used on mainframes or PCs. The program has two inputfiles as follows: afile named "MATCH"; and afile named "RAINFALL".

Every record of the MATCH consists of a date on which the value of rainfall is required. The RAINFALL contains dates and their corresponding values of rainfall for several years. First the program reads a data from the MATCHfile an d then searches that date in the RAINFALL. Afterfinding the date from the RAINFALL, the value of rainfall which corresponds to this date is also read and both the date and value of rainfall will be written in the outputfile calle d OUTPUT. This procedure is repeated for all records of the MATCH file. dimension iday (400), month (400), isal (400) character*5 AAA open (unit =5,file= 'match.txt' ,status= 'old') open (unit = 6,file = 'match.out', status = 'unknown') open (uint = 7 .file = '566020', status = old') APPENDIX A List of Computer Programs 211

open (uint = 8 ,file = 'final', status = unknown) do 1100 i = 1,383 read (5, *) iday (i), month (i), isal (i) isal (i) = isal (i) + 1900 1100 continue do 12 i = 1, 383 c print *, 'i = ',i do 120 m = 1, 10 c print *, 'm= *,m if(m. eq. 1) then nn=3 do 10 i i = 1, nn read (7,*) c print *, 'i i- ,i i, m 10 continue end if read (7,11) year 11 format (127X, i4) print *, *year=', year do 151 k= 1,4 read (7,*) 151 continue if (isal (i). eq. year) then do 121 j= 1,31 124 read (7, 122) irooz 122 format (9x, i3) if (irooz.eq.iday (i)) go to 123 go to 124 123 if (month (i) .eq. 1) then backspace (unit = 7) read (7, 13) AAA 13 format (18x, A5) write (8, *) iday (i), month (i), isal (i),' \AAA go to 1020 end if if (month (i). eq. 2) then backspace (unit = 7) read (7, 14) AAA 14 format (27x, A5) write (8, *) iday (i), month (i), isal (i),' ', AAA go to 1020 end if if (month (i). eq. 3) then backspace (unit = 7) read (7, 15) AAA 15 format (36x, A5) write (8, *) iday (i) month (i), isal (i),' ', AAA go to 1020 end if if (month (i) .eq. 10) then backspace (unit = 7) read (7, 16) AAA 16 format (lOOx, A5) write (8, *) iday (i), month (i), isal (i),' ', AAA go to 1020 end if if(month(i).eq. 11) then APPENDIX A List of Computer Programs 216.

backspace (unit = 7) read (7, 17) AAA 17 format (108x,A5) write (8, *) iday (i), month (i), isal (i),' ' AAA go to 1020 end if if (month (i) .eq. 12) then backspace (unit = 7) read (7, 18) AAA 18 format (117X,A5) write (8, *) iday (i), month (i), isal (i),' ';AAA go to 1020 end if 121 continue else dol25j = l, 58 read (7, *) 125 continue end if 120 continue 1020 rewind (unit = 7) 12 continue stop end

A.4 Computer Program Number 4 (Quick Basic)

This computer program was written to extract the daily rainfall events from the Sydney Bureau of Meteorology data sets which are in a long ASCI format, in each rainfall station.

CLS 100 REM 110 REM ** initialisation **** 120 DIM nm(12), P(12, 31), PM(12), PMAX(12), NOP(12), dy(500), mon(500), year(500) 130 Y$ = "#######.#": X$ = " ##": V$ = "####.#": W$ = " ##" 140 Z$ =" M ## #### ###.#" 150 DATA 31,28,31 FOR i = 1 TO 3 READ nm(i) NEXTi DATA 31,30,31 FOR i = 10 TO 12 READ nm(i) NEXTi INPUT "ENTER INPUT MATCH FILE NAME:"; 12$ OPEN 12$ FOR INPUT AS #3 WHILE NOT EOF(3) K = K+1 nmatch = nmatch + 1 INPUT #3, daym, monm, yearm dy(K) = daym mon(K) = monm year(K) = yearm + 1900 APPENDIX A List of Computer Programs 211

' PRINT dy(k), mon(k), year(k) WEND 190 INPUT "ENTER INPUT FILE NAME:"; i$ 200 OPEN i$ FOR INPUT AS #1 INPUT "ENTER OUTPUT FILE NAME:"; 0$ OPEN 0$ FOR OUTPUT AS #2 otyp= 1 330 WHILE NOT EOF(l) 340 INPUT #1,A$ 365 SW=1 370 year = 1000 + VAL(MID$(A$, 1, 6)) 380 IF year MOD 4 = 0 THEN nm(2) = 29 ELSE nm(2) = 28 390 mon = VAL(MID$(A$, 7, 6)) 440 MAXP = -1 450 FOR i = 1 TO nm(mon) '460 P$ = MID$(A$, 13 +1 * 8, 8) P$ = MID$(A$, 5 + i * 8, 8) 470 IFP$ = " " THEN P(mon, i) =-1 ELSE P(mon, i) = VAL(P$) •480 PCD$ = MLD$(A$, 24 +1 * 6, 1) PRINT 490 IF P(mon, i) > MAXP THEN MAXP = P(mon, i) 500 NEXTi 510 PMAX(mon) = MAXP 520 IF otyp = 1 THEN GOSUB 770 ' printing the results in a file 560 WEND 570 CLOSE 580 END 590 REM ** SETTING UNAVAILABLE DATA ** 600 FOR i = mon+1 TO 12 610 PM(i) = -l:NOP(i) = -l:PMAX(i) = -l 620 FORJ=lT0 31 630 P(i, J) = -1 640 NEXT J 650 NEXTi •670 RETURN 770 REM ** PRINTING OUTPUT FILE *** 780 FOR i = 1 TO nm(mon) FOR c = 1 TO nmatch IF year(c) o year THEN GOTO 999 IF mon(c) o mon THEN GOTO 999 IF dy(c) = i THEN PRINT #2, USING Z$; i; mon; year; P(mon, i) 'c = nmatch PRINT" year = "; year ELSE ENDIF 999 NEXT c 666 PRINT 800 NEXTi PRINT " it is working "; c 810 RETURN A.5 Computer Program Number 5

This computer program was written in the C1"4" computer programming language environment. The program calculates the horizontal distance between a rainfall station and the average coastline in the study area. To run the program, three files should be entered after the name of the program. One file contains the data of the stations; a second file for coastline data, and a thirdfile fo r the output.

#include "include.h" #include "function, h" #include "func.c" main(int argc, char *argv[]) { int nil, nl2, i, ii; char sn[50]; float x, y, xl, yl, x2, y2, x3, y3, x4, y4, gl, g2, dist; float dxl, dyl, dx2, dy2, m, ml, m2, h; FILE *fl, *f2, *f3; printf ("\n"); printf ("\n"); printf ("This program calculate the distance between the horizontal^"); printf ("line from a station and the coastal line.\n"); printf ("\n"); printf ("\n"); printf ("To run the program threefiles shoul d be entered after the \n"); printf ("name of the program. File contains the data of the stations,\n"); printf ("file contains the coastline data, and output file.\n"); printf ("\n"); printf ("\n"); if ((fl = fopen (*++argv, "r")) = NULL) { printf ("Program can't openfile % s as input.\n", *argv); exit (0); } if ((f2 = fopen (*++argv, "r")) == NULL) { printf ("Program can't openfile %s as input.\n", *argv); exit (0); } if ((f3 = fopen (*++argv, "w")) = NULL) { printf ("Program can't openfile %s as output.\n", *argv); exit (0); } printf ("argc = %d\n", argc); printf ("\n"); printf ("\n"); nil = lines (fl); nl2 = lines (f2); rewind (fl); rewind (f2); for (i=0; i

= fscanf (fl, "%s %f %f %f %f %f\n", sn, &xl, &yl, &gl, &g2, &h); APPENDIX A List of Computer Programs 211

x2 = xl+100; y2 = yi; fscanf (12, "%f %f\n",&x3 , &y3); for (ii=0; ii=x3 && x<=x4 && y>=y3 && y<=y4 || x<=x3 && x>=x4 && y>=y3 && y<=y4 || x>=x3 && x<=x4 && y<=y3 && y>=y4 || x<=x3 && x>=x4 && y<=y3 && y>=y4) && y<=yl) { dist = distance (xl, yl, x, y); printf ("The distance between Station %s and the coastal line is %.3fm\n", sn, dist); fprintf (f3, "The distance between Station %s and the coastal line is %.3fm\n", sn, dist); ii = nl2-l; rewind (f2); } else { x3 =x4; y3=y4; } } } rewind (f2); } fclose (fl); fclose (f2); fclose (0); return (0); } ********************************** rstorm

APPENDIX B

Thunderstorm Rainfall Data

B.l Common thunderstorm-days in the Sydney region between 12 thunder-recording stations

* The numbers in the table are representative of the thunder-recording stations.

1 = Richmond

2 = Katoomba

3 = Parramatta

4 = Prospect Dam

5 = Sydney Regional Office

6 = Liverpool

7 = Bankstown

8 = Sydney Airport

9 = Lucas Heights

10 = Camden Airport

11 = Wollongong University

12 = Bowral

Date is represented as Year, Month and Day. Thunderstorm Rainfall Data

Table 4.2 Common thunderstorm-days in Sydney region.

Row Date Row Date Row Date 1 19600103 5* 8 531 19731010 1 1061 19850402 1 12 2 19600129 1 5 8 532 19731011 1 1062 19850403 1 3 5 6 7 8 3 19600209 1 5 533 19731101 1 9 1063 19850404 4 10 4 19600214 5 8 534 19731102 1 5 7 8 9 1064 19850406 12 5 19600307 1 5 535 19731104 1 3 8 1065 19850407 8 6 19600314 1 536 19731111 1 7 8 1066 19850408 1 5 7 12 7 19600322 1 537 19731120 1 8 1067 19850409 1 7 8 19600402 5 538 19731121 7 9 1068 19850425 4 9 19600420 1 539 19731126 8 1069 19850426 1 4 10 19600421 5 540 19731129 5 7 8 10 1070 19850501 1 11 19600518 5 541 19731206 1 4 8 1071 19850506 5 8 12 19600519 5 8 542 19731215 1 1072 19850507 1 7 11 13 19600520 5 543 19731227 1 5 7 8 1073 19850520 5 7 8 11 14 19600916 5 544 19731228 5 7 1074 19850606 1 5 8 15 19600917 5 8 545 19740213 1 3 1075 19850620 5 8 16 19600924 5 8 546 19740214 10 1076 19850707 5 8 17 19601004 5 8 547 19740218 1 5 8 11 12 1077 19850901 1 10 11 18 19601024 1 5 8 548 19740219 9 1078 19850911 1 2 3 4 8 10 19 19601025 8 549 19740323 1 4 5 7 8 1079 19850912 4 11 20 19601026 8 550 19740328 1 1080 19850913 1 5 7 8 21 19601030 8 551 19740411 1 5 1081 19850925 4 11 22 19601031 5 8 552 19740425 7 1082 19851007 12 23 19601103 5 ! 1 553 19740601 5 8 1083 19851011 8 12 24 19601118 8 554 19740616 5 7 8 9 1084 19851012 1 25 19601121 5 555 19740901 1 3 4 6 7 8 10 1085 19851016 1 3 7 12 26 19601124 1 5 ! 556 19740902 9 1086 19851017 1 5 7 8 12 27 19601126 5 8 557 19740924 1 5 8 10 11 11 1087 19851018 12 28 19601203 1 5 8 558 19740930 1 1088 19851019 12 29 19601212 1 559 19741017 10 1089 19851023 1 5 7 8 10 30 19601214 5 8 560 19741022 5 8 1090 19851106 8 31 19601215 1 8 561 19741023 5 9 1091 19851107 7 8 10 11 12 32 19601216 8 562 19741028 1 1092 19851108 1 5 7 8 10 33 19610101 5 8 563 19741029 1 4 5 7 8 10 1093 19851109 1 7 34 19610112 1 564 19741030 1 9 10 1094 19851115 1 35 19610113 1 5 8 i 565 19741031 1 1095 1985112S 1 8 12 36 19610131 1 566 19741101 1 4 7 8 9 10 1096 19851126 5 8 12 37 19610205 5 567 19741102 11 1097 19851127 1 5 7 8 10 12 38 19610206 1 568 19741126 1 1098 19851128 1 7 10 39 19610207 1 5 8 569 19741207 10 1099 19851130 1 12 40 19610208 1 5 8 570 19741217 8 1100 19851201 1 41 19610209 5 571 19750102 1 5 8 1101 19851208 1 12 42 19610223 1 8 i 572 19750103 1 4 7 10 1102 19851209 8 10 43 19610228 I i 573 19750104 5 8 1103 19851210 3 6 7 8 10 44 19610315 1 574 19750108 1 5 7 8 10 1104 19851211 5 8 45 19610405 1 575 19750109 9 10 12 1105 19851214 1 7 8 10 46 19610413 5 8 576 19750110 9 1106 19851216 1 5 8 10 47 19610503 5 8 577 19750117 1 2 1107 19851217 1 7 48 19610604 5 8 578 19750125 1 8 10 1108 19851223 5 7 8 10 12 49 19610608 8 1 579 19750126 10 1109 19851224 7 50 19610821 8 580 19750227 1 5 10 1110 19860104 1 5 8 10 51 19610823 8 581 19750228 5 8 9 1111 19860105 1 7 10 10 12 52 19610824 8 582 19750310 5 7 8 1112 19860108 1 53 19610826 8 583 19750311 1 4 5 6 7 8 9 1113 19860115 5 8 10 12 54 19610829 8 584 19750313 1 10 1114 19860116 7 8 11 55 19610905 8 585 19750315 1 5 8 12 1115 19860117 8 11 56 19611011 1 5 8 586 19750328 7 9 10 1116 19860122 1 7 57 19611023 1 587 19750329 1 2 4 5 8 9 12 1117 19860130 5 8 58 19611101 1 5 8 588 19750330 7 9 10 1118 19860131 1 7 11 59 19611102 1 5 589 19750414 1 1119 19860204 2 5 8 10 11 12 60 19611103 1 8 590 19750415 1 1120 19860205 5 7 8 10 61 19611104 1 591 19750416 5 7 1121 19860206 7 62 19611107 5 592 19750419 1 10 12 1122 19860209 8 63 19611108 5 593 19750501 1 1123 19860222 5 8 64 19611114 1 5 8 594 19750621 1 2 5 9 1124 19860223 7 65 19611115 1 5 595 19750622 9 1125 19860225 2 66 19611120 1 5 8 596 19750623 9 1126 19860309 5 8 67 19611121 1 597 19750810 1 1127 19860318 2 68 19611123 1 598 19750923 1 1128 19860326 5 8 APPENDIX R Thunderstorm Rainfall Data

Table 4.2 cont....

Row Date Row Date Row Date 69 19611124 5 599 19750928 1 1129 19860327 4 11 70 19611125 5 8 600 19750930 1 8 1130 19860412 5 7 8 10 71 19611130 1 5 5 8 601 19751002 1 5 8 1131 19860413 4 72 19611201 8 602 19751010 1 5 8 10 1132 19860417 10 73 19611215 8 603 19751011 9 1133 19860418 4 10 74 19611217 5 8 604 19751020 1 4 5 7 8 10 1134 19860430 8 75 19611218 5 8 605 19751024 1 5 1135 19860804 1 4 7 8 76 19611231 1 606 19751031 8 9 10 1136 19860805 4 7 8 77 19620104 1 607 19751104 4 7 8 9 1137 19860925 1 5 7 8 78 19620109 1 5 8 608 19751109 1 8 12 1138 19861001 1 79 19620110 1 5 9 609 19751116 1 3 4 7 10 1139 19861003 5 7 8 10 80 19620111 9 610 19751123 8 1140 19861004 3 81 19620118 1 611 19751124 4 1141 19861019 1 82 19620129 8 612 19760109 1 1142 19861028 8 83 19620131 1 5 613 19760205 1 1143 19861111 10 84 19620205 1 614 19760206 5 1144 19861117 5 8 85 19620208 5 8 615 19760219 1 2 3 4 5 7 8 9 10 1145 19861118 5 7 8 12 86 19620209 1 616 19760220 4 6 7 1146 19861119 7 87 19620214 8 617 19760225 7 1147 19861126 1 5 8 8 88 19620215 9 618 19760226 2 1148 19861127 5 7 8 89 19620220 1 5 8 619 19760228 1 3 5 8 9 10 1149 19861205 12 90 19620302 1 8 620 19760327 1 5 8 1150 19861215 7 8 10 91 19620308 8 621 19760328 4 9 1151 19861216 1 92 19620318 5 8 622 19760405 5 7 9 1152 19861231 12 93 19620319 8 623 19760418 1 1153 19870101 10 94 19620320 5 624 19760615 9 1154 19870103 1 10 95 19620429 5 625 19760617 3 7 8 1155 19870104 1 5 8 10 12 96 19620522 1 626 19760701 5 8 1156 19870202 1 97 19620806 1 627 19760813 5 1157 19870210 1 2 5 8 10 12 98 19620915 5 628 19760823 5 1158 19870211 1 7 99 19620927 5 8 629 19760827 8 1159 19870215 10 100 19621104 5 630 19760828 10 1160 19870221 1 5 101 19621111 5 631 19760917 1 10 1161 19870319 1 2 11 102 19621128 1 i 632 19760918 4 1162 19870320 1 5 7 8 10 11 103 19621203 1 633 19760920 5 8 1163 19870326 1 2 104 19621204 5 634 19760921 1 2 3 4 5 8 1164 19870327 1 8 105 19621207 1 5 8 I 635 19760922 4 1165 19870328 5 7 8 106 19621208 5 ! 636 19761001 8 9 10 1166 19870329 1 7 11 107 19621214 5 8 ! 637 19761006 5 7 8 1167 19870409 1 108 19621218 1 5 8 ! 638 19761007 10 1168 19870514 8 109 19621219 5 8 I 639 19761014 1 5 7 8 9 10 1169 19870515 7 110 19630102 1 8 640 19761015 7 9 1170 19870722 5 7 8 111 19630103 1 5 6 8 9 641 19761017 3 1171 19870723 7 112 19630127 6 8 i 642 19761018 5 8 1172 19870728 1 1 113 19630129 1 8 643 19761030 3 4 5 7 9 10 1173 19870810 1 5 8 114 19630201 1 6 644 19761102 1 5 8 1174 19870920 10 115 19630204 5 8 645 19761103 4 5 9 1175 19871016 1 5 7 8 10 11 12 116 19630218 1 1 i 646 19761104 1 5 8 9 1176 19871017 2 7 117 19630310 8 1 ! 647 19761110 5 8 1177 19871019 1 2 10 118 19630314 1 648 19761111 1 2 9 11 1178 19871023 1 5 8 12 i i 119 19630409 1 649 19761112 1 2 3 5 7 8 10 1179 19871024 5 6 8 10 11 12 120 19630420 5 8 I 650 19761114 1 1180 19871025 1 5 7 8 121 19630421 8 | 651 19761115 1 5 7 8 11 1181 19871108 2 122 19630424 5 652 19761117 1 5 8 1182 19871109 2 8 123 19630517 1 5 653 19761118 8 9 1183 19871110 2 6 7 10 11 124 19630603 9 654 19761120 1 4 5 6 7 8 9 11 1184 19871111 6 7 125 19630604 5 655 19761121 1 2 5 7 8 9 12 1185 19871115 1 12 126 19630625 5 656 19761122 1 5 7 8 9 1186 19871116 1 2 5 12 127 19630626 5 8 657 19761123 9 1187 19871117 6 7 128 19630710 5 8 658 19761203 7 8 1188 19871118 2 129 19630713 5 659 19761210 5 8 1189 19871120 1 5 7 8 12 130 19630814 5 660 19761215 1 5 8 12 1190 19871201 8 131 19630822 1 661 19761216 4 -7 9 11 1191 19871205 12 132 19630823 1 5 662 19761217 1 5 6 7 8 9 10 11 1192 19871210 5 8 133 19630825 5 663 19770101 1 5 8 1193 19871211 8 10 134 19630829 5 8 664 19770102 9 1194 19871216 2 12 135 19630916 1 665 19770110 5 1195 19871220 1 2 7 8 12 136 19630923 1 5 6 8 666 19770113 1 10 12 1196 19871221 1 l\ APPENDIX R Thunderstorm Rainfall Data 243

Table 4.2 cont....

—r Row Date Row Date Row Date 137 19630930 1 5 6 8 667 19770114 10 1197 19871228 10 138 19631012 5 6 8 668 19770115 9 10 1198 19880101 1 2 5 8 139 19631025 1 669 19770116 1 5 8 12 1199 19880102 1 2 7 10 140 19631029 1 670 19770117 7 9 1200 19880108 2 2 5 8 141 19631030 1 671 19770121 1 5 8 1201 19880109 1 2 7 10 142 19631031 1 5 6 8 9 672 19770122 7 10 1202 19880120 2 143 19631111 1 673 19770123 1 1203 19880121 1 2 3 5 6 7 8 10 11 144 19631120 5 6 8 9 674 19770124 11 1204 19880122 2 7 10 145 19631122 1 5 6 8 9 675 19770128 12 1205 19880123 1 2 5 6 7 8 10 11 146 19631129 9 676 19770129 1 1206 19880124 10 11 147 19631211 6 677 19770202 9 11 1207 19880131 1 2 10 148 19631213 8 678 19770204 1 5 7 8 9 12 1208 19880207 1 2 149 19631215 1 5 679 19770205 12 1209 19880213 7 150 19631216 1 5 6 8 9 680 19770206 9 1210 19880214 6 151 19631217 8 681 19770218 1 5 8 1211 19880228 1 10 152 19631223 5 8 682 19770219 9 11 1212 19880229 10 153 19631227 1 5 6 8 683 19770223 1 1213 19880301 2 154 19640124 1 8 684 19770225 1 2 12 1214 19880304 1 2 155 19640126 1 685 19770228 1 2 12 1215 19880324 2 6 7 10 156 19640207 8 686 19770304 1 1216 19880325 10 157 19640209 5 8 687 19770305 1 2 5 8 1217 19880429 5 158 19640215 5 8 688 19770306 9 1218 19880430 1 2 5 7 8 159 19640216 1 8 | 689 19770320 8 1219 19880521 8 160 19640223 8 690 19770321 1 1220 19880527 10 161 19640301 5 8 ! 691 19770407 1 5 7 8 9 11 12 1221 19880528 5 8 162 19640609 1 5 8 1 692 19770528 5 6 7 8 1222 19880615 5 163 19640610 5 8 693 19770608 1 5 8 1223 19880827 5 6 7 8 10 164 19640702 5 8 694 19770806 1 5 7 8 9 10 11 1224 19880828 10 165 19640824 5 695 19770812 1 1225 19880917 1 5 7 8 10 166 19640827 5 I 696 19770927 1 2 3 4 5 6 7 8 9 10 11 12 1226 19880919 5 8 167 19640828 1 697 19770928 7 1227 19880920 6 7 168 19640829 5 ! 698 19771019 1 5 1228 19880927 1 5 6 7 8 11 169 19641008 1 5 8 1 699 19771029 5 8 12 1229 19880928 1 5 6 7 8 10 1 170 19641023 5 700 19771101 1 3 7 1230 19881103 5 171 19641029 5 8 1 701 19771104 1 1231 19881110 1 1 172 19641103 1 5 8 702 19771114 1 3 4 5 8 10 11 1232 19881117 1 173 19641109 6 703 19771115 7 9 1233 19881120 1 2 5 8 174 19641119 1 5 8 9 704 19771117 1 8 1234 19881121 1 2 3 5 6 7 8 10 11 175 19641209 8 ! 705 19771118 1 2 3 4 5 7 8 11 1235 19881122 5 6 7 8 176 19641223 1 5 706 19771119 9 1236 19881123 1 5 7 8 177 19641226 1 5 6 8 1 707 19771120 5 8 1237 19881124 3 178 19641229 5 8 708 19771130 1 2 3 5 8 10 11 1238 19881125 5 8 179 19650111 5 1 709 19771201 1 4 10 1239 19881126 1 7 8 10 180 19650124 1 5 8 ! 710 19771202 8 9 1240 19881127 1 2 3 5 6 7 8 11 181 19650130 1 5 8 1 711 19771210 1 11 1241 19881203 1 182 19650216 5 8 9 712 19771211 4 1242 19881209 1 2 5 8 183 19650217 1 6 | 713 19771214 5 1243 19881210 1 2 3 5 6 7 8 10 11 184 19650218 1 714 19771215 5 9 1244 19881211 1 2 5 6 7 8 10 11 185 19650219 5 8 715 19771221 8 1245 19881216 2 186 19650222 5 8 716 19771225 1 5 8 1246 19881220 1 2 187 19650410 5 8 9 717 19771226 7 1247 19881222 2 188 19650622 1 5 8 718 19780101 1 12 1248 19881223 2 189 19650623 5 8 719 19780102 1 8 12 1249 19881226 6 7 190 19650718 1 5 720 19780104 4 1250 19890102 10 191 19650802 5 8 721 19780115 1 5 8 12 1251 19890104 2 192 19650909 1 8 722 19780116 1 5 7 8 9 11 1252 19890105 1 2 6 10 193 19650913 1 723 19780118 1 5 7 8 9 1253 19890106 5 194 19651024 1 5 6 8 9 724 19780123 1 2 5 1254 19890110 2 195 19651025 1 725 19780124 1 5 8 10 11 1255 19890117 8 196 19651027 1 5 8 726 19780125 7 1256 19890118 1 2 3 5 7 8 10 197 19651103 5 727 19780127 1 2 3 5 8 11 12 1257 19890119 1 3 7 198 19651124 1 5 8 72S 19780211 3 5 8 11 1258 19890207 2 199 19651202 1 9 729 19780219 1 1259 19890219 2 5 200 19651203 1 9 730 19780221 8 10 1260 19890220 1 2 5 7 8 201 19651214 5 8 731 19780222 4 9 1261 19890226 6 7 202 19651230 5 8 9 732 19780225 1 3 1262 19890305 2 203 19651231 1 8 733 19780227 1 5 8 1263 19890309 1 5 8 204 19660116 1 5 8 734 19780228 4 7 1264 19890310 3 6 7 APPENDIX R Thunderstorm Rainfall Data 244

Table 4.2 cont....

—r Row Date Row Date Row Date 205 19660117 4 5 8 735 19780303 5 7 8 11 1265 19890311 2 7 206 19660128 1 5 8 736 19780308 12 1266 19890312 1 2 207 19660129 1 8 737 19780313 2 1267 19890313 7 8 208 19660130 1 4 738 19780321 1 11 1268 19890331 7 209 19660204 1 739 19780322 1 1269 19890406 2 210 19660205 4 8 740 19780323 1 10 17 1270 19890412 1 5 8 211 19660214 4 8 741 19780327 1 4 5 8 11 1271 19890413 6 7 212 19660215 4 5 8 742 19780401 9 1272 19890421 5 8 213 19660216 5 743 19780410 1 12 1273 19890422 5 6 8 214 19660309 1 5 8 12 744 19780518 11 1274 19890423 7 215 19660310 1 4 745 19780602 1 1275 19890426 11 216 19660311 1 5 8 746 19780831 5 7 8 1276 19890504 1 217 19660318 5 9 747 19780912 8 1277 19890505 11 218 19660321 1 8 12 748 19780918 1 5 8 1278 19890506 4 219 19660322 4 9 749 19780919 5 1279 19890507 8 220 19660323 1 5 8 750 19780922 1 2 5 8 10 12 1280 19890623 5 221 19660324 4 5 8 9 751 19781004 5 8 1281 19890624 1 5 7 8 11 222 19660414 1 12 752 19781006 8 1282 19890816 1 2 223 19660415 9 753 19781018 1 1283 19890817 5 7 8 224 19660519 1 754 19781031 1 5 8 10 1284 19890820 1 3 5 7 8 225 19660521 4 5 755 19781101 9 1285 19890823 1 226 19660610 5 8 756 19781107 1 7 9 1286 19890826 2 227 19660817 1 757 19781111 1 1287 19890926 1 228 19660831 1 5 8 9 1 758 19781113 1 2 7 8 9 1288 19891004 2 229 19660916 5 j 759 19781117 8 1289 19891023 1 2 1 230 19660921 4 5 8 760 19781126 1 2 12 1290 19891025 6 231 19661002 12 1 761 19781128 8 1291 19891105 1 i 232 19661003 1 4 5 8 9 i 762 19781129 8 1292 19891106 1 2 233 19661006 1 1 763 19781130 9 1293 19891107 3 11 234 19661016 1 5 8 764 19781203 1 5 8 1294 19891112 1 2 5 235 19661017 1 5 6 8 9 12l 765 19781212 1 2 5 8 10 12 1295 19891116 3 5 7 8 236 19661018 9 766 19781213 4 9 10 11 1296 19891117 5 8 237 19661019 1 8 9 767 19781214 5 1297 19891118 3 6 7 i 238 19661020 9 768 19781215 7 9 1298 19891202 2 i i 239 19661026 1 8 769 19781217 1 5 8 10 1299 19891203 1 240 19661027 1 9 ;' 770 19781218 4 9 10 1300 19891205 1 2 3 5 7 8 11 241 19661109 1 5 8 771 19781222 1 1301 19891209 6 7 242 19661110 9 i 772 19781225 12 1302 19891210 1 2 5 11 243 19661111 8 '! 773 19790102 5 8 1303 19891211 1 2 5 8 244 19661121 1 | : ' 774 19790103 4 7 9 1304 19891212 3 6 7 245 19661123 1 1 ! 775 19790107 1 1305 19891214 1 3 5 8 246 19661124 5 8 776 19790108 8 1306 19891219 2 247 19661205 1 5 6 8 8 9 777 19790211 1 5 8 1307 19891220 3 5 11 248 19661214 1 5 9 | 778 19790226 5 1308 19891221 6 249 19661222 1 ! ' i 779 19790302 1 4 5 7 1309 19900101 5 8 250 19661223 1 780 19790303 1 5 7 8 9 1310 19900102 1 6 7 251 19661226 1 781 19790314 1 5 8 1311 19900106 1 2 3 5 6 7 8 252 19661229 1 782 19790315 1 5 8 1312 19900107 1 2 3 5 6 7 8 253 19661230 8 12 783 19790321 1 12 1313 19900108 1 2 3 254 19661231 1 784 19790324 11 1314 19900112 2 255 19670102 1 5 6 8 9 785 19790402 5 7 8 9 10 11 1315 19900113 1 2 5 8 256 19670111 1 786 19790403 1 1316 19900114 3 6 257 19670116 9 ! 787 19790415 12 1317 19900119 1 258 19670129 1 9 788 19790416 1 1318 19900120 1 2 5 8 259 19670208 1 789 19790418 9 1319 19900121 1 3 6 7 260 19670213 1 5 6 8 790 19790510 5 8 1320 19900205 1 261 19670225 5 8 791 19790719 1 5 8 1321 19900206 2 262 19670226 5 8 792 19790720 4 9 10 1322 19900207 1 2 3 5 7 8 263 19670306 9 793 19790726 8 10 1323 19900208 1 2 3 7 264 19670504 12 794 19790819 8 1324 19900209 1 2 3 5 6 7 8 11 265 19670505 12 795 19790911 5 1325 19900210 1 2 3 5 6 7 8 266 19670806 5 796 19790919 5 1326 19900211 3 6 7 19670817 5 8 267 797 19790920 11 1327 19900217 1 2 5 8 — 268 19670905 5 8 798 19791004 10 1328 19900218 3 6 7 11 269 19671011 5 799 19791006 5 7 8 1329 19900222 2 270 19671013 1 5 8 800 19791011 1 5 8 9 11 1330 19900223 6 7 8 271 19671018 1 5 ; 801 19791016 5 1331 19900224 1 3 5 7 11 19671028 8 9 272 1 1 802 19791022 1 2] 1332 19900225 11 APPENDIX R Thunderstorm Rainfall Dat>

Table 4.2 cont....

Row Date — Row Date Row Date 273 19671029 9 803 19791023 1 5 8 12 1333 19900305 1 2 5 8 11 274 19671104 5 8 804 19791024 9 10 11 1334 19900306 1 3 5 6 7 8 11 275 19671105 9 805 19791028 8 10 1335 19900307 5 6 7 8 11 276 19671108 1 5 8 806 19791104 1 1336 19900316 3 6 7 277 19671119 8 807 19791105 1 2 5 8 10 11 1337 19900318 1 2 5 8 278 19671120 9 808 19791106 7 8 10 11 1338 19900319 6 7 279 19671130 1 809 19791111 5 1339 19900403 8 280 19671214 8 810 19791112 1 3 4 7 8 1340 19900413 11 281 19671215 9 811 19791115 1 5 8 1341 19900416 8 282 19671229 1 5 6 8 9 812 19791116 4 7 9 10 1342 19900701 11 283 19680104 1 813 19791119 9 1343 19900719 11 284 19680117 3 8 9 814 19791120 1 5 7 9 10 1344 19900724 1 2 285 19680122 1 815 19791123 1 5 8 1345 19900725 1 286 19680206 1 5 8 816 19791124 1 4 5 6 7 8 9 10 11 1346 19900801 11 287 19680207 1 3 8 9 817 19791126 5 6 7 8 9 10 11 1347 19900802 1 2 3 4 5 7 8 11 288 19680304 1 5 8 9 12 818 19791204 1 8 1348 19900815 1 6 8 289 19680312 1 819 19791205 1 5 8 10 1349 19900901 1 290 19680316 1 820 19791206 7 1350 19900913 1 291 19680318 5 821 19791208 1 5 8 9 10 1351 19900914 5 8 292 19680319 5 8 822 19791221 1 8 9 1352 19900915 1 293 19680320 8 823 19791231 1 1353 19901002 3 5 294 19680323 5 8 824 19800106 1 2 1354 19901011 4 5 7 8 12 295 19680324 1 5 8 825 19800110 1 3 4 5 6 8 9 10 12 1355 19901012 7 296 19680325 1 5 8 826 19800111 9 1356 19901015 2 297 19680412 8 827 19800112 1 5 7 8 1357 19901019 7 298 19680415 8 828 19800113 1 5 8 1358 19901021 2 299 19680416 1 829 19800131 1 1359 19901101 5 300 19680612 8 830 19800201 1 2 5 6 8 10 11 12 1360 19901103 5 8 12 301 19680722 5 8 831 19800202 4 7 9 11 12 1361 19901104 1 4 7 8 11 302 19680815 8 9 832 19800205 1 12 1362 19901109 1 2 5 12 303 19680820 5 833 19800206 7 1363 19901110 4 7 11 304 19680913 1 834 19800301 2 1364 1990U15 2 5 8 12 305 19680916 5 835 19800416 2 1365 19901116 1 2 3 4 306 19681019 4 836 19800417 5 8 1366 19901129 1 2 4 5 8 12 307 19681104 1 837 19800430 8 1367 19901130 2 3 7 308 19681110 3 838 19800501 1 5 8 9 11 1368 19901201 1 309 19681111 3 8 839 19800528 8 1369 19901203 1 2 4 5 8 12 310 19681120 5 I 840 19800529 1 5 7 8 9 1370 19901204 5 8 311 19681209 1 3 4 5 8 841 19800609 1 1371 19901208 1 2 3 5 8 312 19681210 5 842 19800610 7 1372 19901209 3 7 313 19681225 1 3 4 5 6 8 843 19800824 10 1373 19901210 1 5 12 314 19681226 4 844 19800826 8 1374 19901211 4 10 315 19681227 1 845 19800918 2 1375 19901220 1 316 19681228 1 846 19801002 1 1376 19901221 7 317 19690102 1 5 8 847 19801012 1 3 5 8 9 1377 19901227 12 318 19690103 1 6 8 848 19801013 4 1378 19901231 2 12 319 19690114 4 5 8 849 19801018 2 1379 19910101 2 320 19690122 1 850 19801019 2 1380 19910109 2 321 19690205 8 851 19801020 1 2 3 4 5 8 9 10 1381 19910110 1 2 5 8 10 12 12 322 19690206 1 i 852 19801021 7 1382 19910U1 2 3 4 7 10 11 12 323 19690207 1 4 6 7 9l 853 19801028 7 1383 19910112 1 2 4 11 12 324 19690223 1 5 8 854 19801107 1 1384 19910113 3 325 19690224 9 855 19801109 1 7 8 1385 19910115 2 326 19690225 5 8 856 19801110 7 1386 19910U6 1 3 327 19690314 1 857 19801203 1 1387 19910118 1 2 8 10 12 328 19690319 1 858 19801204 1 12 1388 19910119 7 10 329 19690327 8 859 19801216 5 8 1389 19910120 1 2 3 5 7 8 10 11 12 330 19690328 1 9 860 19801229 1 3 5 6 7 8 10 1390 19910121 1 2 3 4 5 7 8 10 11 12 331 19690329 12 861 19801230 10 1391 19910122 3 4 6 7 10 332 19690330 1 862 19810103 1 2 1392 19910125 12 333 19690331 1 5 8 863 19810106 1 2 12 1393 19910126 1 2 3 4 5 7 8 11 12 334 19690401 9 864 19810107 1 7 1394 19910127 3 4 335 19690415 1 865 19810112 7 1395 19910205 2 8 12 336 19690501 1 866 19810121 1 2 4 5 8 10 12 1396 19910206 1 2 3 4 8 11 12 337 19690515 5 867 19810122 4 7 10 11 1397 19910207 1 3 4 6 8 338 19690609 5 8 868 19810125 1 2 8 10 12 1398 19910215 2 5 8 12 339 19690610 1 5 8 9 869 19810126 2 9 10 1399 19910216 1 2 5 7 8 11 340 19690715 8 870 19810128 1 12 1400 19910222 2 12 APPENDIX R Thunderstorm Rainfall Data 246

Table 4.2 cont...

Row Date Row Date Row Date 341 19690721 5 871 19810205 2 5 8 10 12 1401 19910223 2 10 12 342 19690901 8 I 872 19810206 2 4 7 10 1402 19910310 1 2 343 19690908 7 873 19810210 1 2 8 10 1403 19910311 1 2 5 7 8 11 344 19690918 1 3 5 8 874 19810211 4 1404 19910312 1 7 10 345 19690922 1 875 19810212 8 1405 19910320 1 346 19690929 1 876 19810219 11 1406 19910411 12 347 19690930 5 877 19810302 1 5 8 12 1407 19910412 1 4 5 7 8 10 11 348 19691004 1 3 7 878 19810303 4 7 9 10 11 1408 19910426 1 2 3 5 7 8 349 19691015 1 3 4 8 879 19810310 11 1409 19910509 5 8 350 19691021 1 5 880 19810405 4 1410 19910515 8 351 19691022 1 881 19810406 1 2 4 5 6 7 8 10 1411 19910522 2 352 19691026 1 5 882 19810407 4 9 10 1412 19910530 5 8 353 19691029 1 5 883 19810424 8 1413 19910609 2 354 19691030 1 4 884 19810504 7 11 1414 19910612 11 355 19691101 1 5 885 19810827 5 8 1415 19910703 8 356 19691106 1 5 7 8 886 19811011 1 2 12 1416 19910823 12 357 19691108 1 887 19811012 1 1417 19910926 1 2 3 4 5 7 8 12 358 19691110 1 4 8 888 19811015 1 9 1418 19910927 3 4 7 359 19691114 1 889 19811021 1 5 8 1419 19911005 2 360 19691118 1 6 9 890 19811029 5 8 1420 19911007 3 4 361 19691119 1 5 8 1 891 19811030 1 2 9 1421 19911024 2 362 19691128 5 7 8 892 19811104 1 8 1422 19911025 1 5 5 8 363 19691211 1 4 5 6 7 8 12 893 19811105 7 1423 19911026 3 10 364 19691212 1 5 8 12 894 19811113 10 12 1424 19911031 1 11 12 365 19691222 1 3 5 7 8 895 19811114 4 10 1425 19911116 1 2 5 8 11 12 366 19691230 8 896 19811115 1 5 8 1426 19911117 7 367 19700101 1 5 7 8 9 12 897 19811116 4 9 1427 19911126 5 8 12 368 19700102 9 898 19811121 1 2 1428 19911127 1 2 12 369 19700104 5 899 19811128 5 8 1429 19911130 3 370 19700110 1 5 8 900 19811205 1 1430 19911203 8 12 371 19700111 1 5 7 8 901 19811212 1 5 8 1431 19911204 1 2 3 4 5 7 8 10 12 372 19700118 1 3 4 5 7 8 902 19811213 1 4 1432 19911210 12 373 19700119 1 903 19811219 4 5 8 8 9 12 1433 19911211 1 3 374 19700121 1 904 19811220 7 1434 19911215 1 375 19700125 1 5 905 19811223 1 1435 19911221 1 2 5 7 8 376 19700126 1 4 5 6 7 8 12 906 19811224 2 1436 19911222 1 5 7 8 12 377 19700210 5 i 907 19811225 5 7 8 1437 19911224 2 378 19700212 1 4 5 8 12 1 908 19811226 1 1438 19911227 5 8 12 379 19700215 1 j 909 19811229 1 7 9 1439 19911228 1 3 5 7 8 380 19700216 1 7 i 910 19811230 5 7 1440 19920101 1 12 381 19700226 1 ! 911 19820101 8 1441 19920103 2 5 8 382 19700227 1 912 19820102 1 1442 19920104 1 3 5 7 383 19700228 5 8 9 913 19820107 1 1443 19920105 5 7 8 384 19700306 1 5 8 914 19820113 1 1444 19920106 7 385 19700307 5 915 19820117 1 2 1445 19920109 1 2 5 12 386 19700316 5 8 916 19820124 2 1446 19920110 2 12 387 19700318 12 917 19820130 1 8 1447 19920121 12 388 19700319 1 12 i 918 19820131 1 2 3 4 5 8 12 1448 19920122 2 7 12 389 19700320 5 8 I 919 19820226 1 2 12 1449 19920123 1 2 5 8 12 390 19700425 8 920 19820227 10 1450 19920124 3 4 7 391 19700528 1 921 19820321 3 7 1451 19920131 2 392 19700603 5 8 i 922 19820324 1 1452 19920201 2 393 19700621 1 5 8 1 923 19820417 1 7 1453 19920202 2 12 394 19700802 5 8 1 924 19820425 12 1454 19920203 5 8 395 19700831 1 925 19820426 11 1455 19920204 1 2 3 4 5 8 10 12 396 19700901 8 926 19820612 2 1456 19920205 3 4 397 19700902 1 927 19820709 8 1457 19920206 12 398 19700909 7 928 19820816 7 8 1458 19920211 1 3 4 8 12 — 399 19700923 1 5 929 19820927 2 5 8 1459 19920212 1 2 3 5 8 11 400 19700928 7 930 19820930 5 8 1460 19920213 1 3 4 6 401 19701019 1 5 931 19821008 5 8 1461 19920222 1 402 19701024 5 8 932 19821016 1 8 11 1462 19920223 1 2 3 4 7 11 12 403 19701106 1 5 8 933 19821203 1 3 5 6 8 9 10 1463 19920224 1 404 19701108 1 4 5 8 934 19821207 5 1464 19920303 2 405 19701111 5 8 935 19821214 1 2 10 1465 19920304 2 406 19701112 9 936 19821215 5 8 10 11 12 1466 19920305 11 12 407 19701115 3 8 9 12 937 19821216 4 1467 19920322 11 408 19701123 1 938 19821230 1 10 1468 19920328 1 Thunderstorm Rainfall Data 247

Table 4.2 cont....

Row Date Row Date Row Date 409 19701124 1 3 4 5 7 8 9 939 19821231 1 10 1469 19920426 2 410 19701201 1 940 19830102 8 1470 19920427 8 411 19701205 7 941 19830103 7 1471 19920428 4 5 7 8 10 12 412 19701211 1 12 942 19830109 1 3 5 8 1472 19920429 10 413 19701212 1 9 943 19830121 5 7 8 10 1473 19920510 4 414 19701214 1 8 944 19830122 2 7 10 1474 19920518 2 415 19701215 9 945 19830126 1 2 5 8 12 1475 19920626 1 5 6 7 8 416 19701218 1 5 8 946 19830201 10 1476 19920628 8 417 19701219 9 947 19830202 1 1477 19920804 1 5 8 418 19701222 1 948 19830203 1 12 1478 19920824 5 419 19701223 1 8 949 19830204 7 , 1479 19920828 1 5 8 420 19701228 3 950 19830207 1 1480 19920919 2 421 19701229 1 5 8 951 19830209 1 8 1481 19921015 1 5 6 8 422 19710103 1 4 5 7 8 9 952 19830210 1 5 8 10 1482 19921016 1 3 4 5 6 8 11 12 423 19710104 1 8 9 953 19830211 7 1483 19921018 1 3 4 8 424 19710113 1 954 19830212 1 10 1484 19921020 8 425 19710115 3 7 955 19830213 2 7 12 1485 19921031 2 426 19710116 1 5 956 19830222 1 2 1486 19921101 1 2 5 8 427 19710117 4 957 19830223 7 1487 19921102 1 2 4 428 19710118 1 958 19830225 1 1488 19921104 1 5 8 11 12 429 19710126 1 959 19830226 1 1489 19921105 4 10 430 19710128 1 3 4 7 12 960 19830305 1 1490 19921106 3 431 19710210 1 3 8 961 19830306 4 10 1491 19921109 5 8 12 432 19710315 5 8 962 19830314 1 1492 19921110 4 433 19710322 1 3 4 7 963 19830315 1 2 7 8 1493 19921111 8 434 19710323 1 964 19830730 10 1494 19921116 12 435 19710413 5 8 965 19830903 5 8 1495 19921117 10 436 19710520 1 966 19830904 4 5 7 1496 19921119 1 2 437 19710521 1 967 19830906 7 1497 19921120 1 438 19710821 1 5 6 7 8 968 19830908 1 5 1498 19921121 1 2 5 8 12 439 19710916 8 969 19830909 7 1499 19921122 1 10 440 19711031 3 970 19830915 1 1500 19921124 1 2 4 5 6 8 12 441 19711107 1 3 4 5 7 8 9 971 19830929 1 5 8 1501 19921125 4 442 19711109 1 972 19830930 7 1502 19921129 1 443 19711113 1 4 5 8 9 973 19831003 1 2 5 8 12 1503 19921130 4 444 19711114 1 3 4 974 19831004 7 10 1504 19921204 12 445 19711128 1 975 19831015 1 7 1505 19921205 1 3 5 6 8 12 446 19711129 5 8 976 19831019 1 2 3 5 5 7 8 10 1506 19921206 2 3 5 5 8 12 447 19711202 12 977 19831020 7 10 1507 19921213 2 5 8 12 448 19711206 1 3 5 9 978 19831024 5 7 8 1508 19921214 1 2 3 449 19711214 1 3 4 5 8 9 12 979 19831127 1 8 10 1509 19921221 2 8 12 450 19711225 4 5 980 19831130 1 5 8 1510 19921222 1 2 4 8 10 11 12 451 19711226 1 4 5 6 8 981 19831201 4 5 7 1511 19921223 1 2 3 4 8 12 452 19711227 5 982 19831207 5 8 12 1512 19921224 1 2 4 8 11 12 453 19720105 1 10 983 19831208 4 7 1513 19921225 1 2 4 8 11 12 454 19720118 1 5 8 984 19831210 1 2 12 1514 19921226 1 4 455 19720122 1 5 8 985 19831211 4 1515 19930102 2 456 19720123 1 3 986 19831212 4 1516 19930104 1 2 3 4 5 8 10 11 12 457 19720126 5 987 19831213 1 5 8 1517 19930106 1 5 8 11 458 19720127 1 988 19831214 1 4 7 1518 19930107 1 3 4 11 12 459 19720203 1 3 8 9 10 989 19840108 1 5 7 8 1519 19930111 1 460 19720214 1 990 19840109 1 7 8 11 12 1520 19930115 2 461 19720217 1 4 5 7 8 9 12 991 19840110 7 10 1521 19930116 1 12 462 19720218 1 5 8 992 19840113 1 1522 19930118 1 3 12 463 19720219 1 993 19840121 4 5 7 10 1523 19930119 1 4 464 19720220 5 8 9 10 994 19840204 5 8 1524 19930120 12 465 19720221 5 8 995 19840205 1 4 5 7 8 1525 19930121 1 8 12 466 19720303 1 5 8 996 19840206 4 7 11 1526 19930122 4 467 19720306 5 8 997 19840207 1 2 5 8 12 1527 19930124 1 2 5 8 12 468 19720307 1 5 7 8 998 19840208 1 4 7 1528 19930125 1 3 4 469 19720422 1 5 7 8 9 10 999 19840214 1 2 5 7 10 12 1529 19930202 1 2 12 470 19720605 8 1000 19840215 5 8 12 1530 19930203 4 12 471 19720621 3 4 11 11 1001 19840216 4 12 1531 19930204 1 8 472 19720623 3 11 1002 19840217 2 1532 19930205 4 11 12 473 19720826 1 1003 19840218 7 1533 19930209 2 12 474 19720829 1 8 1004 19840221 10 12 1534 19930210 1 475 19721004 1 1005 19840319 1 7 10 11 1535 19930212 1 476 19721016 1 5 1006 19840325 1 1536 19930216 2 12 APPENDIX R Thunderstorm Rainfall Data 248

Table 4.2 cont....

Row Date Row Date Row Date 477 19721020 1 11 1007 19840530 5 8 1537 19930217 1 3 4 5 6 6 8 478 19721024 1 1008 19840620 5 7 8 1538 19930218 1 479 19721025 1 12 1009 19840806 4 7 1539 19930220 12 480 19721026 10 1010 19840810 5 7 8 1540 19930221 2 481 19721029 1 4 1011 19840901 1 5 8 1541 19930226 1 482 19721101 5 7 1012 19840902 1 4 7 1542 19930307 2 5 6 8 12 483 19721104 1 5 8 1013 19840904 1 5 8 1543 19930308 1 3 4 5 8 10 484 19721106 1 5 8 1014 19840905 4 5 7 8 1544 19930309 1 2 3 5 8 11 12 485 19721109 1 5 8 1015 19840908 1 2 5 8 1545 19930320 1 5 8 11 486 19721110 1 11 1016 19840909 7 1546 19930321 3 4 487 19721111 1 1017 19840914 10 1547 19930324 2 12 488 19721116 1 1018 19840915 4 10 1548 19930325 2 12 489 19721125 3 7 1019 19840921 11 1549 19930326 1 2 4 8 11 12 490 19721207 1 5 7 8 9 1020 19840928 11 1550 19930327 1 3 4 491 19721208 9 1021 19841013 1 8 1551 19930328 1 12 492 19721215 1 5 5 8 11 1022 19841014 1 4 7 1552 19930329 1 3 8 493 19721216 1 5 8 10 11 1023 19841025 1 1553 19930405 1 494 19721221 1 5 7 10 12 1024 19841026 11 1554 19930406 3 495 19721222 1 10 1025 19841028 1 2 4 5 8 1555 19930428 1 5 8 496 19730109 1 5 8 9 10 1026 19841029 4 7 1556 19930429 5 8 497 19730113 10 1027 19841103 1 2 5 8 12 1557 19930510 1 2 498 19730125 1 12 1028 19841104 1 7 1558 19930523 1 3 4 5 8 11 499 19730126 1 1029 19841105 5 8 1559 19930804 3 500 19730127 10 1030 19841106 11 1560 19930810 1 501 19730129 10 1031 19841107 1 8 1561 19930825 1 2 5 8 11 12 502 19730131 5 8 1032 19841108 1 3 5 6 7 8 10 11 1562 19930826 3 503 19730201 1 4 5 6 7 8 9 10 1033 19841109 5 7 8 11 1563 19930913 2 5 8 ll| 504 19730202 1 1034 19841111 1 5 7 8 10 12 1564 19930914 8 505 19730203 1 5 7 8 9 10 1035 19841112 1 7 8 11 1565 19930919 2 506 19730204 1 1036 19841113 1 5 7 8 11 1566 19930920 1 507 19730205 10 1037 19841115 11 1567 19931004 12 508 19730217 7 1038 19841211 1 2 5 8 1568 19931018 1 12 509 19730221 1 8 1039 19841212 4 7 1569 19931023 1 12 510 19730222 1 1040 19841215 7 1570 19931024 ll 511 19730223 7 8 10 11 1041 19841221 11 1571 19931025 5 512 19730226 1 1 1042 19841225 5 8 11 1572 19931102 2 3 5 8 12 513 19730227 1 i 1043 19841226 1 4 7 10 1573 19931113 1 8 514 19730228 1 12 1044 19841230 7 1574 19931114 1 2 8 515 19730301 10 1045 19850102 5 8 1575 19931117 2 5 516 19730304 1 9 1046 19850103 7 8 1576 19931118 1 2 3 8 12 517 19730312 1 1047 19850109 11 1577 19931119 2 5 6 8 11 12 518 19730406 12 i 1048 19850116 1 5 8 1578 19931120 1 3 8 519 19730407 1 5 8 12 I 1 1049 19850117 5 1579 19931124 12 520 19730408 10 | i 1050 19850122 8 1580 19931204 1 2 3 8 11 521 19730409 5 8 11 1051 19850123 11 1581 19931212 12 522 19730430 1 1052 19850128 1 3 7 1582 19931213 2 523 19730501 1 5 7 8 9 10 11 1053 19850129 10 1583 19931214 1 8 524 19730502 11 1054 19850207 I 7 1584 19931226 1 2 8 12 525 19730710 7 1055 19850208 1 3 5 6 7 8 526 19730811 8 1056 19850222 2 12 527 19730825 4 51 1057 19850319 2 528 19730912 1 5 8! 1058 19850320 1 529 19730913 5 101 1059 19850324 1 5 8 12 530 19731004 5 RI 1060 19850325 7 11 APPENDIX B Thunderstorm Rainfall Data 249

B.2 Monthly Thunderstorm Rainfall Data at Richmond

Table 5.1 Monthly thunderstorm rainfall frequency at Richmond station.

Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 1960 1 3 1 0 0 0 0 0 1 1 3 11 61 2 1 1 0 0 0 0 0 2 7 0 16 62 5 0 0 1 0 0 1 0 0 0 2 12 63 2 1 1 1 0 0 2 3 3 1 2 18 64 1 0 0 0 1 0 1 0 0 2 2 8 65 0 0 0 0 1 1 0 2 2 1 2 10 66 3 5 0 0 0 0 2 0 3 1 5 20 67 1 0 0 0 0 0 0 0 2 1 0 5 68 2 4 1 0 0 0 0 1 0 0 3 13 69 3 4 1 1 1 0 0 3 6 6 3 31 70 7 2 0 1 1 0 0 2 1 4 7 29 71 2 2 0 2 0 0 1 0 0 4 3 15 72 5 2 1 0 0 0 2 0 4 6 4 29 73 2 2 1 1 0 0 0 1 2 4 1 22 74 0 1 1 0 0 0 0 2 3 2 0 10 75 5 4 3 1 1 0 1 3 2 1 0 22 76 1 1 0 0 0 0 0 1 0 7 2 13 77 5 2 1 0 1 0 2 1 0 6 2 25 78 8 4 1 0 1 0 0 2 2 3 4 26 79 0 3 1 0 0 1 0 1 2 5 3 16 80 4 0 0 2 0 0 0 0 1 2 1 12 81 5 1 1 0 0 0 0 0 4 2 3 17 82 6 1 1 0 0 0 0 0 1 0 4 14 83 1 6 2 0 0 0 0 0 2 3 2 2 18 84 3 4 2 0 0 0 0 0 4 3 7 2 25 85 1 2 0 5 2 1 0 0 2 4 6 5 28 86 3 0 0 0 0 0 0 1 1 1 1 0 7 87 2 3 5 0 0 0 1 1 0 4 3 2 21 88 5 1 1 1 0 0 0 0 3 0 6 4 21 89 2 1 0 1 0 1 0 2 1 1 1 5 15 90 6 7 3 0 0 0 2 2 3 0 4 5 32 Total 93 73 56 23 12 9 5 18 38 57 96 81 561 Average 2.74 2.15 1.65 0.68 0.35 0.26 0.15 0.53 1.12 1.68 2.82 2.38 16.50

Table 5.2 Monthly thunderstorm rainfall (in mm) at Richmond station.

Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total

1960 2.5 24.6 25 2.5 0 0 0 0 0 0.8 44.1 18.5 118 61 9.5 8.6 5 9 0 0 0 0 0 8.2 59.7 0 100 62 21 47.3 0 0 4.8 0 0 0.5 0 0 0 12.5 86.1 63 33.3 19 0.3 6.4 22 0 0 9 15.6 10.9 0.5 8 125 64 17.3 2.5 0 0 0 257 0 4.8 0 0 59.2 23.5 133 65 0.5 0.5 0 0 0 7.6 93.2 0 37 26.2 1.6 1.1 167.7 66 12.7 0.8 12.3 0 0 0 0 1.8 0 7.4 23.4 31.2 89.6 67 0.3 17.3 0 0 0 0 0 0 0 22.3 18.5 0 58.4 68 5.7 5 52.2 1.8 0 0 0 0 11 0 0 21 96.7 69 7.5 5.1 126 15.2 1 8.6 . 0 0 12.5 73.5 57 4.7 311.1 70 92.6 18.3 19 0 3 3.6 0 0 16.5 2.5 70 38.3 263.8 71 11.5 12.7 13 0 29.7 0 0 20.3 0 0 20.4 21.5 129.1 72 60.2 19.6 10 1.5 0 0 0 8 0 32 21 9 161.3 73 7.8 46 5 26.7 9 0 0 0 10.2 8.1 21.4 38.1 172.3 74 0 1.2 8 27.4 0 0 0 0 2.6 32.6 3.8 0 75.6 75 22.2 25 34.2 20.4 8.8 11 0 7.4 23 22.8 32.2 0 207 76 8.6 18.2 25.8 0 0 0 0 0 1.6 0 32.6 18 104.8 77 44.2 61.2 8.2 4 0 5.8 0 6 9.2 0 28.8 25.2 192.6 78 43.6 0.8 26.8 0.2 0 27 0 0 6.7 12 38 30.8 185.9 79 0 0 20.2 0.8 0 0 27 0 4.2 3.2 19 13 87.4 80 3.7 22.2 0 0 14.4 0 0 0 0 11.6 13.8 4.3 70 81 34.2 7.4 23.8 5.8 0 0 0 0 0 20.8 3.2 39.8 135 82 58 11.6 5 17 0 0 0 0 0 2 0 17.8 111.4 83 3.8 51.4 4.4 0 0 0 0 0 24.6 15.8 10 5.4 115.4 84 9.8 36.2 3 0 0 0 0 0 7.8 4.6 52.4 12.2 126 85 3.6 24.5 0 33.8 3 47 0 0 12.6 19.4 81.2 28.4 253.5 86 28.6 0 0 0 0 0 0 2 0.4 7 25.2 0 63.2 87 4.6 16 26.8 0 0 0 3.2 12 0 59.3 3.6 12 137.5 88 46.8 7.4 1.2 19 0 0 0 0 20.6 0 26.3 64.5 185.8 89 14.2 12 0 6.4 0 0.8 0 7.6 8 16.8 6 63.7 135.5 90 40.2 79.8 48.2 0 0 0 18 75 64.5 0 31 13 369.7 Total 648.5 602.2 503.4 197.9 957 137.1 141.4 154.4 288.6 419.8 803.9 575.5 4568 Average 19.1 17.7 14.8 5.8 2.8 4.0 4.2 4.5 8.5 12.3 23.6 16.9 147.3 APPENDIX B Thunderstorm Rainfall Data

B.3 Monthly Thunderstorm Rainfall Data at Sydney Regional Office

Table 5.3 Monthly thunderstorm frequency at Sydney Regional Office station.

Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 1960 2 2 0 2 3 0 0 0 2 3 2 2 18 61 2 3 0 1 1 1 0 0 0 1 9 2 20 62 3 2 3 1 0 0 0 0 1 0 0 6 16 63 1 1 0 2 1 3 2 3 2 2 1 4 22 64 0 1 1 0 0 2 1 3 0 3 2 2 15 65 3 3 0 1 0 2 1 1 0 2 1 2 16 66 3 2 4 0 1 1 0 1 2 3 1 0 18 67 1 3 0 0 0 0 0 2 1 1 2 1 11 68 0 1 6 0 0 0 1 1 0 0 0 1 10 69 2 2 1 0 1 2 1 0 2 3 3 3 20 70 6 2 4 0 0 2 0 1 1 2 4 2 24 71 2 0 1 1 0 0 0 1 0 0 3 4 12 72 3 4 3 1 0 0 0 0 0 1 4 2 18 73 1 3 0 2 1 0 0 1 2 1 3 1 15 74 0 1 1 1 0 2 0 0 1 3 0 0 9 75 2 2 4 1 0 1 0 0 0 4 1 0 15 76 0 2 1 1 0 0 1 2 2 4 9 2 24 77 4 2 1 1 1 1 0 1 1 2 4 2 20 78 5 1 2 0 0 0 0 1 1 2 0 4 16 79 1 1 4 1 0 0 0 0 1 4 7 2 21 80 3 0 0 0 2 0 0 0 0 2 0 1 8 81 1 1 1 1 0 0 0 1 0 1 2 3 11 82 1 0 0 0 0 0 0 0 2 1 0 2 6 83 3 1 0 0 0 0 0 0 3 3 1 2 13 84 1 5 0 0 0 1 0 1 4 1 6 2 21 85 2 1 1 2 2 1 1 0 0 2 3 3 18 86 3 3 1 1 0 0 0 0 1 1 4 0 14 87 1 2 2 0 0 0 1 0 0 4 2 1 13 88 4 0 0 2 1 1 0 1 4 0 6 3 22 89 2 2 1 2 0 1 0 2 0 0 2 4 16 90 4 5 3 0 0 0 0 1 1 2 5 2 23 Total 66 58 45 24 14 21 9 24 34 58 87 65 505 Average 1.9 1.7 1.3 0.7 0.4 ! 0.6 0.3 0.7 1.0 1.7 2.6 1.9 14.9

Table 5.4 Monthly thunderstorm rainfall (in mm) at Sydney Regional Office station.

Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total

1960 8.7 18 0 7.2 38 0 0 0 7.3 12 21.8 22.3 135.3 61 5.6 12.1 0 2 2.5 8.1 0 0 0 10.7 133.8 41 215.8 62 8.4 38.4 33.2 0.5 0 0 0 0 2.5 0 0 29 112 63 18.3 4.8 0 5.8 5.8 53.8 16.8 53.8 7.9 13 3.6 47.5 231.1 64 0 12.2 14 0 0 71 4.6 6 0 10 61.5 16.5 195.8 65 21.4 3.2 0 2.3 0 41.6 15 0.8 0 19 2.3 3.8 109.4 66 10.4 15.8 52.1 0 20.1 4.6 0 1.6 4 21.2 3.3 0 133.1 67 0.3 53 0 0 0 0 0 37.6 41.4 2 89.4 5.8 229.5 68 0 10.4 45 0 0 0 8.4 2.3 0 0 0 4.6 70.7 69 29.2 17.5 33 0 1.8 82 3 0 18.3 10 11.5 20.5 226.8 70 69.2 29.5 88 0 0 14.7 0 5.8 2 2.3 68.6 29.7 309.8 71 8.2 0 25 1 0 0 0 112.7 0 0 18 20 184.9 72 23.7 47 31.8 4.3 0 0 0 0 0 28.7 8.7 14.8 159 73 7.4 143 0 43 29.5 0 0 6.4 7.4 3.6 24.3 6.6 271.2 74 0 16.4 14.2 7.5 0 26.8 0 0 5.4 17 0 0 87.3 75 3.6 12 240 7 0 17.6 0 0 0 41.6 0.7 0 322.5 76 0 65.5 6 1.8 0 0 17.6 3.4 25.4 40 90.4 23 273.1 77 28.2 10.2 18.4 5.4 9.2 0.6 0 6.8 12.1 1.5 15 12.6 120 78 25.7 2.2 17.2 0 0 0 0 18.2 5.1 6 0 38.6 113 79 7.8 0.4 16.8 7.4 0 0 0 0 6 27.3 49.1 0.4 115.2 80 21.5 0 0 0 64 0 0 0 0 3.6 0 23.6 112.7 81 9.2 23.6 25.8 14.2 0 0 0 1.4 0 8.8 3.2 23.6 109.8 82 5.6 0 0 0 0 0 0 0 27.8 2.2 0 10.4 46 83 20.7 8.6 0 0 0 0 0 0 15.7 53.5 3.6 22.2 124.3 84 30.7 54 0 0 0 21.7 0 1.8 21 2.6 334 21 486.8 85 5 15.3 9.4 12.4 10 16.4 8 0 0 13.6 10.3 16 116.4 86 33 23 5 0.6 0 0 0 0 0.2 6.6 8.6 0 77 87 30.6 9.3 11 0 0 0 4.6 0 0 128.8 8.6 3 195.9 88 43.3 0 0 75.5 4 2 0 3.4 26.4 0 51 69 274.6 89 50 7 36 32.2 0 4 0 35 0 0 12 40.3 216.5 90 28.4 85.4 48.4 0 0 0 0 4 34 1 45 4.4 250.6 Total 554.1 737.8 770.3 230.1 184.9 364.9 78 301 269.9 486.6 1078 570.2 5626 Average 16.3 21.7 22.7 6.8 5.4 10.7 2.3 8.9 7.9 14.3 31.7 16.8 181.5 APPENDIX B Thunderstorm Rainfall Data

B.4 Monthly Thunderstorm Rainfall Data at Sydney Airport

Table 5.5 Monthly thunderstorm rainfall frequency at Sydney Airport station.

Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 1960 2 1 0 0 1 0 0 0 1 5 0 4 14 61 2 2 0 1 1 1 0 5 1 1 4 2 20 62 1 2 4 0 0 0 0 0 0 0 0 4 11 63 2 1 1 1 0 1 1 1 2 2 2 5 19 64 0 4 1 0 2 1 0 0 0 2 2 1 13 65 2 2 0 1 0 2 0 1 1 2 1 3 15 66 3 3 4 0 0 1 0 1 1 4 2 1 20 67 0 3 0 0 0 0 0 1 1 3 3 0 11 68 1 2 5 0 0 1 1 1 0 0 1 1 13 69 1 2 1 0 0 2 1 0 2 1 4 3 17 70 5 1 3 0 0 2 0 1 1 1 5 4 23 71 0 2 1 1 0 0 0 1 1 0 3 2 11 72 2 5 3 1 0 1 0 1 0 0 3 1 17 73 1 3 0 2 1 0 0 1 1 1 6 1 17 74 0 1 1 0 1 2 0 0 2 2 1 1 11 75 4 1 4 0 0 0 0 0 1 4 2 0 16 76 0 2 1 0 0 1 1 1 2 4 9 4 25 77 3 2 2 1 1 1 0 1 2 1 4 3 21 78 5 3 2 0 0 0 0 1 3 3 1 3 21 79 1 1 2 1 1 0 2 1 0 3 7 1 20 80 3 [___!_ 0 2 3 0 0 1 0 2 1 2 15 81 3 3 1 2 0 0 0 1 0 2 3 3 18 82 2 0 0 0 0 0 0 1 2 1 0 2 8 83 3 1 1 0 0 0 0 0 2 3 2 2 14 84 2 4 0 0 0 1 0 1 3 1 8 2 22 85 3 1 1 1 2 1 1 0 1 3 5 4 23 86 5 4 2 2 0 0 0 2 1 2 4 1 23 87 1 1 3 0 1 0 1 1 0 4 2 3 17 88 4 0 0 1 1 0 0 1 4 0 5 3 19 89 2 1 2 3 1 1 0 1 0 0 2 2 15 90 5 4 4 2 0 0 0 2 1 1 4 1 24 Total 68 63 49 22 16 19 8 29 36 58 96 69 533 Average 2.00 1.85 1.44 0.65 0.47 0.56 0.24 0.85 1.06 1.71 2.82 2.03 15.68

Table 5.6 Monthly thunderstorm rainfall (in mm) at Sydney Airport station.

Year Jan j Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total

1960 30 9 0 0 2 0 0 0 3.6 15.7 0 55.5 115.8 61 5 2 0 15.5 0.5 0.3 0 13.2 2.8 11.5 67.5 10 128.3 62 1.3 46.5 4.1 0 0 0 0 0 0 0 0 30 81.9 63 26 1.8 2.5 2.5 0 10.2 2.5 33.5 17 13 1.5 91.5 202 64 0 5.7 8.5 0 0 59.2 2.8 0 0 7 17.3 8.4 108.9 65 15 1 0 2 0 33 0 0.8 0.5 15 1.5 23 91.8 66 12.5 15.5 102 0 0 1.3 0 7 1 7.8 5 9.5 161.6 67 0 92.5 0 0 0 0 0 24 35 4 54 0 209.5 68 2 19 26 0 0 6.5 10 7 0 0 0.3 6 76.8 69 16.5 12.5 22 0 0 26 0.5 0 4 0.3 14.2 29 125 70 24 26 85 0 0 7.5 0 8 28 1.3 95.5 38.6 313.9 71 0 44 10 49 0 0 0 48.5 15.5 0 21 28.5 216.5 72 16.5 56.5 51 10.5 0 1.3 0 6.6 0 0 12.2 10 164.6 73 6.6 30.5 0 54 16 0 0 16 4 6.6 23 2.3 159 74 0 13.2 17 0 19 13 0 0 20.5 13 0.4 5 101.1 75 6.2 42 202 0 0 0 0 0 16 22.6 14.5 0 303.3 76 0 23 4.6 0 0 29.2 35 3.6 13 67 98 20 293.4 77 19.5 14 14.2 10 3.6 0.2 0 4.6 18 0.8 15.4 4.6 104.9 78 47.2 0.8 35 0 0 0 0 24.5 4 56 1.4 31.5 200.4 79 5.4 10 10 3.6 0.2 0 12 1 0 25.6 51.5 1.2 120.5 80 18.2 40 0 5.5 18.5 0 0 0.4 0 17 33 14 146.6 81 13.5 33.2 1 24 0 0 0 0.2 0 3.6 20 59.2 154.7 82 18.5 0 0 0 0 0 0 15.5 21.6 1.8 0 6 63.4 83 9.2 1.2 9.6 0 0 0 0 0 19 54.2 21.5 28.5 143.2 84 15.3 27 0 0 0 15 0 4 15.5 2.8 265.5 46.5 391.6 85 5.2 3.4 6.6 18.2 20 17 0.4 0 0.4 13.6 62 17.6 164.4 86 44 30.4 2.5 10 0 0 0 111 0.4 23.5 24 8 253.8 87 11.5 18 41 0 1 0 9 2 0 109 9.5 24.5 225.5 88 36.5 0 0 70 3 0 0 6 19 0 65.5 112.5 312.5 89 12 0 26.6 74 10 7.2 0 11.1 0 0 11 33 184.9 90 19 69.3 32.2 17 0 0 0 10 41 11 11 3 213.5 Total 436.6 688 713.4 365.8 93.8 226.9 72.2 358.5 299.8 503.7 1017 757.4 5533 Average 12.8 20.2 21.0 10.8 2.8 6.7 2.1 10.5 8.8 14.8 29.9 22.3 173.4 APPENDIX B Thunderstorm Rainfall Data 252

B.5 List of Rainfall Stations

All stations were sorted by latitudes and longitudes (in decimal) from north­ west to south-east of the Sydney region. * The Sydney Water stations are identified by a prefix of the number 5.

Table 6.2 List of stations and the periods from which data were used. Station Elevation No. of Row Number Name of Stations Latitude0 Longitude0; in m Period Events 63057 Mount Wilson 33.51 150.37 1027 !1960-78 100 63184 [ Blaxland Ridge 33.51 150.90 177 1962-79 109 63246 Mount Wilson 33.52 150.37 1005 1969-86 133 63118 Bilpin (Fern Grov.) 33.52 150.50 610 1960-93 282 63013 Berambing 33.53 150.43 792 1960-93 104 63043 Kurrajong Heights 33.53 150.63 495 1960-93 227 63042 Kurrajong P.Of. 33.55 150.67 152 1960-91 184 -AAJW UUUAJJUUUUU . ,A**J^UUIAJJUUUUUUU«JJUUMU' HII 67090 Arcadia 33.58 151.07 205 1964-93 202 66143 Kuring-Gai Chase 33.58 151.30 170 1969-91 140 63056 Mount Victoria 33.60 150.27 1064 1960-90 159 11 63248 Grose Wold 33.60 150.68 61 1969-93 233 12 67033 Richmond AMO/MO 33.60 150.78 19 1960-93 271 1960-93

66128 Palm Beach (G.C) 151.32 1965-89 17 67021 Richmond (H.A.C.) 33.62 150.75 20 1960-93 268 18 67073 Maralya Boudary RD. 33.62 150.90 49 1963-93 263 19 63009 Blackheathp P.Of. 33.63 150.28 1065 1960-93 198 ivwvvvwevvvn rr *VWWVWWW*VWVIrt Y**1 * 567100; Riverstone 33.65 150.83 25 1984-93 120 21 66119 Mount Kuring-Gai 33.65 151.13 215 1964-93 242 22 67002 Castlereagh 33.67 150.68 15 1965-93 216 «*wu M***WWiV^^WI 23 66183 Ingleside Walter Ave. 33.67 151.27 160 1984-93 130 24 66045 Newport B.C. 33.67 151.32 1960-93 213 25 563059 Katoomba 33.68 150.30 950 1984-93 122 26 67086 Dural (Old Northern RD) 33.68 151.03 216 1973-93 181 27 566051 "hWamewoo d 33.68 151.30 15 1982-93 135 28 63227 Wentworth Falls 33.70 150.37 900 1967-93 176 29 563070 Linden W.F.Dam 33.70 150.48 520 i1984-93 123 30 63077 Springwood P.Of. 33.70 150.57 ;366 1960-93 240 • rrrr>rr*rrT+rr** 31 567076 Castle Hill 33.70 150.98 65 1984-93 123 32 66028 Hornsby Police STN 33.70 151.10 181 1960-93 110 33 63044 Lawson P.Of. 33.71 150.43 1715 1960-93 193 34 63039 Katoomba Composite 33.72 150.30 1030 1960-93 190 51 563067 Wentworth Falls 33.72 150.38 823 1982-93 101 Leura P.Of™ 36 63045 33.72 150.43 1975 1960-93 186 37 66063 Wahroonga Reservoir 33.72 151.12 205 1960-93 188 WWAMMMArtMNMrVVV - *?_ f VAWAVAWAWVAW ^M^^M^'J^^^^V^W^WVt ^VJWIWVMMI^AVWVriJl^^.NW^AV^A^WMVA^V W 38 567087 ST. Marys S.T.P. 33.73 150.77 20 1984-93 129 67076 Quarkers Hill 33.73 150.88 33 1966-92 184 66158 Turramurra (kiss. PT.RD) 33.73 151.13 160 1960-93 301 66157 Pymble(Canisius College) 33.73 1151.15 1165 1960-93 281 66044 Cromer 33.73 151.27 10 1960-93 172 66182 Frenchs Forest 33.74 151.23 1960-93 ...•JJUUIIIU'WT' rwowvwwwv»»l»TW> irvvwwvwwwwvww 155 277 63230 Blaxland western Highway 33.75 150.60 234 1968-80 100 APPENDIX B Thunderstorm Rainfall Data 253

Table 6.2 cont....

r*'""v*'"v'~wwww,iMWV 45 67018 Penrrith 33.75 150.68 25 1960-93 256 46 67089 West Pnnant Hills 33.75 151.04 120 1960-93 187 47 67098 Pennant Hills West 33.75 151.05 168 1960-93 233 48 66118 Frenchs Forest (F. Av) 33.76 151.23 150 1964-82 129 49 63185 Glenbrook B.C. 33.77 150.62 183 1963-93 229 50 67067 Emu Plains Gough ST 33.77 150.65 31 1960-93 274 51 67024 ST Marys B.C 33.77 150.77 35 1960-84 127 52 67059 Blacktown Kidare RD 33.77 150.88 58 1963-93 256 53 67080 Winston Hills 33.77 151.00 75 1968-93 215 54 566040 West Epping 33.77 151.05 100 1980-93 153 55 66020 Epping (Chester St) 33.77 151.08 92 1960-93 273 56 66156 Marsfield (Macq.Uni.) 33.77 151.12 55 1970-92 210 57 66120 Gordon B.C. 33.77 151.15 96 1960-93 274 58 67026 Seven Hills Exp.Farm 33.78 150.93 55 1960-90 224 59 66032 West Lindfield 33.78 151.15 60 1960-92 248 60 66056 Roseville B.C. 33.78 151.18 116 1960-79 106 61 66145 Seaforth (Castle Circuit) 33.78 151.23 85 1968-92 208 | 62 66153 Manly Vale(ManDam) 33.78 151.25 20 1968-93 147 63 566025 Manly Dam 33.78 151.24 21 1963-93 201 64 66089 North Manly B.C. 33.78 151.27 5 1961-87 113 65 67084 Orchard Hill 33.80 150.72 93 1970-92 188 66 66124 Parramatta North 33.80 151.02 60 1965-93 273 | 67 66087 Eastwood B.C. 33.80 151.08 78 1960-93 150 68 66081 North Ryde Stround St. 33.80 151.13 70 1960-79 101 69 566017 Chastswood 33.80 151.18 92 1963-93 243 70 66010 Chatswood Council 33.80 151.20 96 1960-93 214 71 66167 Northbridge B.C. 33.80 151.22 35 1980-92 128 72 66002 Balgowlah 33.80 151.25 70 1960-89 217 73 567092 South Prospect 33.82 150.90 65 1975-93 153 74 67019 Prospect Dam 33.82 150.92 61 1960-93 283 75 67032 Westmead Austral Av. 33.82 150.98 26 1960-92 268 76 66130 Northbridge 33.82 151.22 80 1960-80 127 77 66042 Mosman (Bapaume RD) 33.82 151.24 70 1960-93 280 78 66138 Manly Army North Head 33.82 151.30 85 1968-88 111 79 67070 Merrylands Wellsford ST 33.83 150.98 45 1968-93 254 80 66085 Auburn/ Granville Composite 33.83 151.02 8 1960-93 242 81 66134 Granville Shell Refinery 33.83 151.03 3 1960-93 197 82 66131 Riverview Observatory 33.83 151.17 23 1960-78 103 83 66075 Waverton B.C. 33.83 151.20 21 1960-93 258 84 66166 Cremorne Grasmere RD 33.83 151.22 61 1963-89 185 85 566027 Mosman 33.83 151.23 85 1960-93 164 86 66082 Concord West Plaster Mills 33.84 151.08 5 1961-82 121 87 567079 Guildford 33.85 150.97 50 1972-93 190 88 66013 Concord G.C. 33.85 151.10 15 '1960-93 245 89 66163 rWatson s Bay 33.85 151.28 25 1968-93 138 90 67029 Wallacia 33.87 150.63 50 1960-93 267 91 67068 Badgerys Creek RES STN 33.87 150.73 65 1960-93 230 92 567077 Fairfield 33.87 150.95 5 1981-93 138 93 67008 Guildford 33.87 150.98 31 1960-77 105 > „..«....— 94 566022 Home Bush 33.87 151.08 10 1969-93 152 95 66017 Fivedock Council DEP 33.87 151.12 6 1960-93 224 96 66062 Sydney Regional Office 33.87 151.20 42 ,1960-93 287 97 66006 Sydney Botanic Gardens 33.87 151.22 15 1960-93 257 B Thunderstorm Rainfall Data

Table 6.2 cont.... 98 66098 Rose Bay R/S G. C. 33.87 151.27 6 1979-93 136 99 566038 Vaucluse B.C. 33.88 151.27 75 1980-93 159 100 563046 Mcmahons Loo 33.88 150.38 655 1982-93 119 101 568045 Wairagamba 33.88 150.58 180 1982-93 133 102 567093 St.Johns Park 33.88 150.88 .3 5 1980-93 144 103 566050 Villawood 33.88 150.98 30 1980-93 143 104 66164 Rookwood 33.88 151.05 41 1974-93 142 105 66070 Strathfield G.C. 33.88 ]151.0 7 21 1960-93 191 106 66000 Ashfield B.C. 33.88 151.13 25 1960-93 259 107 566032 Paddington 33.88 151.22 45 1961-87 100 108 66005 Bondi B.C. 33.88 151.27 15 1960-82 120 109 66050 Potts Hill (pumping st.) 33.90 151.03 55 1960-93 267 110 566020 Enfield 33.90 151.08 10 1983-93 128 111 66160 Centennial Park 33.90 151.23 38 1960-93 225 112 66073 Randwick Racecourse 33.91 |151.2 3 25 1960-93 196 113 67035 Liverpool Council 33.92 150.92 21 1962-93 272 114 66025 Warwick Farm 33.92 ]150.9 3 5 1960-90 189 115 566049 Liverpool 33.91 !150.9 3 5 1960-93 258 116 66003 Bankstown (Condell Park) 33.92 ;151.0 2 10 1960-79 127 | 117 566026 Marrickville 33.92 151.15 5 1980-93 145 118 66052 Randwick B.C. 33.92 151.24 75 1960-93 248 119 67036 Austral Eighth Ave 33.93 150.82 60 1964-89 158 120 66137 Bankstown AMO 33.93 1 150.98 9 1968-92 223 121 566033 Padstow 33.93 151.02 20 1981-93 109 122 66076 Wiley Park 33.93 ^ 151.07 45 1960-87 146 123 66037 Sydney Airport 33.93 151.17 6 1960-93 288 124 66171 MoorebankN.B. 33.95 150.95 22 1968-80 100 125 66054 Revesby 33.95 151.00 15 1960-93 164 126 66004 Bexley G.C. 33.95 151.10 10 1960-93 208 127 67015 Bringelly (Maryland) 33.97 150.72 122 1960-91 196 128 67009 Glenfield Composite Macquarie 33.97 150.90 23 1960-83 132 129 66148 Peakhust Golf Course 33.97 151.05 ,3 9 1969-87 114 130 566047 Mortdale B.C. 33.97 151.07 40 1978-93 140 131 567078 Glenfield 33.98 150.90 15 1984-93 125 132 66181 Oatley (Woronora Parade) 33.98 151.08 42 1982-93 149 133 66069 Hurstville Grove 33.98 151.10 5 1960-81 140 134 66051 Little Bay (Coast G.C.) 33.98 151.25 22 1962-93 216 135 66058 Sans Souci 34.00 151.13 9 1960-93 290 136 66072 Kurnell(A.6.R) 34.02 151.22 3 11960-93 284 137 68007 Camden (Brownlow Hill) 34.03 150.65 61 1960-93 269 138 566018 Cronulla S.T.P. 34.03 151.15 10 1972-93 200 139 66086 Cronulla W.P.C.P. 34.03 151.17 |10_ ;1960-93 234 140 68192 Camden j^irport 34.05 150.68 70 !1960-93 i 171 141 66078 Lucas Hts (A.A.E.) 34.05 150.98 140 1960-93 183 142 566056 Yarrawarrah 34.05 151.02 50 1983-92 112 143 66040 Miranda Blackwood ST 34.05 151.10 40 1960-93 274 144 66014 Cronulla South B.C. 34.05 151.15 30 1960-93 266 145 563037 Barragorang 34.07 150.40 180 !1983-93 132 146 568138 Oakdale 34.07 150.43 410 1984-93 129 147 568130 West Camden 34.07 150.67 75 j 1983-93 110 148 66090 Engadine 34.07 151.02 170 j 1962-93 128 149 66001 Audley National Park Bottom 34.07 151.05 23 f1960-79 100 150 68081 Campbelltown S.C. 34.08 150.52 75 1960-92 171 MWWWMJWM1MIMI r„..nn^nflJiJuvll APPENDIX B Thunderstorm Rainfall Data

Table 6.2 cont.. 151 66176 Audley Royal Notional Park 34.08 151.05 120 1979-93 156 152 66116 Bundeena Composite 34.09 151.15 45 1964-78 101 153 566052 Woronora Dam 34.12 150.93 180 1983-93 126 154 563036 Yerranderie 34.13 150.30 298 1982-93 139 155 68013 Menangle (JMAI) 34.13 150.73 80 1960-92 ; 229 156 68052 Picton Composite 34.18 150.62 171 1960-92 215 157 568072 Cobbong 34.18 150.85 280 .1984-93 128 158 568069 Reverces 34.18 150.92 305 1983-92 108 159 568051 Oakdale 34.20 150.50 410 1982-93 148 160 68001 Appin (Bulli Road) 34.20 150.78 230 1960-93 1 135 161 68028 'Helensburgh P.Of. 34.20 150.98 150 1960-93 182 162 568038 Wollondilly River 34.23 150.32 200 1982-93 133 j 163 68024 Darker Forest (Kintyre) 34.23 150.92 370 1960-93 276 | 164 568139 Buxton P.O. 34.25 150.53 390 1966-93 245 165 568048 Cataract Dam(w.B.S) 34.27 150.80 340 1982-93 133 166 68016 Cataract Dam 34.27 150.81 340 1960-93 279 167 568065 Letle Box Tower 34.27 150.87 449 1982-93 146 168 568004 Cordeaux Air.St. 34.28 150.72 380 1984-93 122 169 568050 Hill Top 34.30 150.42 580 1981-93 137 170 568060 Ironbark 34.30 150.67 '30 0 1983-92 103 171 568047 Nepean Dam 34.33 150.60 390 1982-93 118 172 568049 Cordeaux Quart 34.33 150.75 335 1984-93 100 173 568067 Beth Salem 34.33 150.85 }36 6 1982-93 142 174 568046 Avon Dam 34.35 150.63 j39 0 1982-93 135 I 175 568097 Mount Keira 34.37 150.82 430 1982-93 138 176 68030 Mittagong 34.39 150.30 j73 5 1960-93 245 177 568061 Browns Roard 34.40 150.70 442 1984-93 133 178 568068 Upper Cordeaux 34.40 150.77 >s 330 1982-93 141 179 68188 Wollongong Uni. 34.40 150.88 •3 0 1960-92 224 180 568099 Leicester Park 34.43 150.38 670 1960-93 220 181 568118 Wollongong STP 34.43 150.90 5 1980-93 155 182 68044 Mittagong Pool 34.45 150.47 625 1960-93 187 | 183 568136 Wollongong 34.45 150.90 15 1975-93 148 184 68033 Mittagong (K.O) 34.47 150.50 625 1960-93 244 185 568058 Hambridge 34.47 150.63 491 1982-93 141 186 568071 Upper Avon 34.47 150.73 330 1982-93 154 187 68023 Dapto West 34.47 150.77 42 1960-87 143 188 68102 Bowral (P.D) 34.48 150.40 690 1960-93 237 189 568054 Mittagong Ma. Cr. 34.48 150.52 570 1981-93 126 190 68053 Port Kembla S.ST. 34.48 150.90 11 1960-92 138 191 68022 Dapto B.C. 34.49 150.78 10 1960-93 235 APPENDIX C Synoptic Weather Charts 256

APPENDIX C Synoptic Weather Charts

This appendix contains the 6 sets of different synoptic charts. During the domination of these weather systems the biggest thunderstorm rainfall events (for each thundery mounts, October to March 1975 to 1993) occurred in the Sydney region (see Chapter 6 for more details). Synoptic charts were taken from the Monthly Weather Review of the New South Wales (Bureau of Meteorology). All attempts have been made to maintain the clarity and detail of information on the synoptic charts.

Synoptic charts 6.1 from 23th to 25th October 1987. APPENDIX C Synoptic Weather Charts 257

Synoptic charts 6.2 from 5th to 11th November 1984.

-~"jl,"__**H006* ^»: i

lo)" ~ ^—' / "JOB 'oo* APPENDIX C Synoptic Weather Charts 258

Synoptic charts 6.3 from 9th to 11th December 1988. APPENDIX C Synoptic Weather Charts 259

Synoptic charts 6.4 from 19th to 22th January 1991.

—'—^?—=—f-ita* ,,0T—~+"~~ X^ ^X+Xi \ A T — r^ + I 1012 \ / *?°jj—=»-r~-' / ' wi2 \ V r\J 1 + \ V \iv Jk '• + \ o 1012 s \ \ -A F"^^ jXT~ r0W\ J> ' L ki + \ 1 ^^^-W20

T?K ' + 18) -10 W-\- Jr -U^H - 1020 + 1oW \

T* *' m \ •r" -w^xVi /^ V +i^ t ,AT K f + \ / ,JE™^ ' J r—Ksr\ x. V"^

1004*" + \\ A (v \JL *J N-J012

^—-+^'°^\ \ \ ni + f b-1020^_ / t^ion V. + 10VT ihi T " APPENDIX C Synoptic Weather Charts 260

Synoptic charts 6.5 from 7th to 11th February 1990. APPENDIX C Synoptic. Weather Charts 261

Synoptic charts 6.6 from 10th to 11th March 1975. APPENDIX D Data Used for GIS and Statistical Models 262

APPENDIX D Data Used for GIS and Statistical Models

Geographical Location of Rainfall Stations and their Attributes

Landuse classes

CBD = Central Business District URT = Urban-Residential (treed) IND = Industrial areas RUS = Rural / Semi-Urban URB = Urban-Residential (barren) RUO = Rural / Open areas TNP = Treed (National / Urban parks)

* Indicates the location of a rainfall station located in one of the four sub-topographic regions A, B, C and D. For stations numbers and their names see Appendix B,Table 6.2.

Table 7.2 Geographical location of rainfall stations and their attributes and average rainfall of the 6 biggest monthly thunderstorm rainfall events (October to March - 1975 to 1993). APPENDIX D Data Used for GIS and Statistical Models 263

Table 7.2 cont.... APPENDIX D Data Used for GIS and Statistical Models 264

Table 7.2 cont. APPENDIX D Data Used for GIS and Statistical Models 265

Table 7.2 cont. Equations Used in SPANS GIS

APPENDIX E

Equations

Table 7.6 Equations which were written in SPANS GIS environment.

E Reclass Reclassification of Rainfall Map

: This equation reclassifies the thunderstorm rainfall map having more than 120 mm rainfall

A = { 1 if class (biggrain)>3, 0};

: where 'biggrain' is the name of the rainfall map (average of the biggest thunderstorm rainfall events

: where 1 and 3 are the number of classes on thunderstorm rainfall map

Result (A)

:The resulted map is the study area's map showing the areas having >120 rainfall amount

E Overlay Overlaying of the Reclassified Map

: This equation overlays (imposes) the reclassified rainfall map upon the physiographic maps

A = {class (input map) if class ('reclass map') = 1, 0 or o 0, 0};

: where input map is the name of each physiographic map in the equation. This equation was applied for all of the physiographic maps such as the proximity, elevation, aspect and landuse maps of the

Sydney region.

Result (A)

:The result of the 'GIS overlaying modelling technique' is shown in Figure 7.5 (a-d).