The Relationship of Flooding in Australian Dryland Rivers to Synoptic Weather Patterns, El Nino Southern Oscillation, Sea Surfac

University of Wollongong Thesis Collections University of Wollongong Thesis Collection

University of Wollongong Year 

The relationship of ﬂooding in Australian dryland rivers to synoptic weather patterns, El Nino southern oscillation, sea surface temperatures and rainfall distribution Graham D. Grootemaat University of Wollongong

Grootemaat, Graham D,The relationship of ﬂooding in Australian dryland rivers to synoptic weather patterns, El Nino southern oscillation, sea surface temperatures and rainfall distribution, PhD thesis, School of Earth and Environmental Sciences, University of Wollongong, 2008. http://ro.uow.edu.au/theses/262

This paper is posted at Research Online. http://ro.uow.edu.au/theses/262

The Relationship of Flooding in Australian Dryland Rivers to Synoptic Weather Patterns, El Nino Southern Oscillation, Sea Surface Temperatures and Rainfall Distribution.

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

DOCTOR OF PHILOSOPHY

from

The University of Wollongong

Graham D. Grootemaat BSc (Physical Geography) (Hons.) (University of Wollongong)

School of Earth and Environmental Sciences

The information in this thesis is entirely the result of investigations conducted by the author, unless otherwise acknowledge, and has not been submitted in part, or otherwise, for any other degree or qualification.

Graham Grootemaat

12th November 2008. Abstract

This thesis classifies and describes the synoptic weather patterns resulting in floods, investigates the spatial properties of dryland rainfall, and identifies the links between flooding and flow patterns with the Southern Oscillation Index (SOI) and Sea Surface Temperatures (SSTs) in Australian dryland rivers catchments.

Flooding of dryland rivers in the monsoon dominated northern areas, such as in Timor Sea and Gulf of Carpentaria drainage divisions, is exclusively the result of tropical trough/lows, deep tropical lows (monsoon depressions) and tropical cyclones. The Timor Sea division, however, has a larger proportion of floods caused by tropical cyclones (32% of the total) compared to the Gulf of Carpentaria division (7% of the total). In the latter division 80% of the floods are the result of tropical trough/lows and deep tropical lows whereas in the Timor Sea division only 48% of floods are the result of these two synoptic classes. They also cause a high proportion of floods in the Indian Ocean division together with the Gascoyne- Murchison-Greenough region of the central west of Western Australia. Tropical cyclones tend to produce the largest floods in all these divisions.

In the more centrally located regions such as the Lake Eyre division, the northeast Murray- Darling Rivers, the Bulloo, Paroo and Warrego Rivers, and the Greenough, Murchison and Gascoyne Rivers, there are many more types of flood-producing weather patterns compared to the monsoon dominated drainage divisions further north. Flooding can result from at least six different synoptic classes individually, including tropical trough/lows, deep tropical lows, tropical cyclones, frontal systems and easterly dips (continental and offshore). Twenty-nine percent of floods result from combined synoptic classes with 15 different combinations involving every synoptic class. Tropical trough/lows and easterly dips (continental) are dominant. Easterly dips (continental) are significant flood producing weather patterns in these centrally located regions and tend to occur in the winter months in association with favourable upper atmosphere conditions and strengthening high-pressure systems.

In southern areas, such as in the southwest coastal division of Western Australia and the southern Murray-Darling basin, frontal systems and cut-off lows are the dominant flood-

i producing weather patterns. However, in the latter a larger proportion of floods result from cut-off lows (18%) relative to frontal systems (8%), whilst in the southwest coastal division frontal systems produce a larger proportion of floods (29%) relative to cut-off lows (26%). Cut-off lows were also found to result in a significant number of floods in the Greenough, Murchison and Gascoyne Rivers in central Western Australia. These can often combine with northwest cloudbands in this area, the largest flood on record for the Gascoyne River being the result of such a combination.

In contrast to many other dryland regions of the world, rainfall in dryland Australia (both flooding and storm rainfall totals) is relatively widespread and uniform rather than localized and convective. However, rainfall in the arid Todd River region around Alice Springs is slightly less widespread than rainfall in the semi-arid Thomson River region, particularly for smaller storm-rainfall events. It appears that the smaller rainfall events in more arid areas of Australia are the result of more convective, smaller scale weather patterns that lead to more discrete, localized rainfall totals. In comparison, larger rainfall events that produce flooding are the result of synoptic scale weather patterns such as tropical trough/lows, deep tropical lows or ex-tropical cyclones, and tend to result in relatively widespread and uniform rainfall across the region.

Only north and northeast Australia show a clear, consistent and predictable relationship between the magnitude of partial series flood events and the Southern Oscillation Index (SOI) and Sea Surface Temperature (SST) indices, with correlation coefficients modest but reaching 0.40-0.50. The SST1 anomaly (The SST1 anomaly represents the sea surface temperatures in the eastern equatorial Pacific Ocean) and the SOI appear to be the most reliable indicators. For seasonal correlations in north and northeast Australia, spring, summer and autumn monthly flows are more highly correlated with the SOI and SSTs, whereas in southern Australia winter and spring monthly flows are more highly correlated. This demonstrates that the specific synoptic weather patterns that cause high seasonal monthly flows are related to the SOI and SSTs. In Western Australia autumn and winter seasonal flows are strongly negatively correlated with Indian Ocean SSTs. These correlations are strongest in the northwest and are related to the Indian Ocean dipole.

Over the last 300Ka it has been inferred that the monsoon played a pivotal role during warmer interglacial periods, generating greater runoff throughout much of dryland

ii Australia. Under climate change, where temperatures would be higher, it thought that the monsoon, with its characteristic wet and dry seasons, would affect areas further to the south. Monsoon lows and ex-tropical cyclones would more frequently affect southern areas, whilst mid-latitude weather patterns, such as frontal systems and cut-off lows, would only seriously affect the very southern areas. Further to this the more central dryland regions, such as the northern Murray-Darling Rivers, that currently receive flooding through a large range of weather patterns, would probably become dominated by tropical systems.

iii Acknowledgements

I wish to acknowledge the following people and organizations for the integral part they have played in this thesis.

Gerald Nanson – for his academic guidance, advice and for his patience with me.

Ivars Reinfelds – for his academic guidance, advice and support.

Robert Dahni from the Bureau of Meteorology - for supplying Mapbrowser 1.1 which was used for analysing synoptic weather patterns. This program was invaluable.

Peter Baddiley from the Bureau of Meteorology – for his valuable assistance and supply of streamflow data.

Gordan McKay from the Bureau of Meteorology – also for valuable assistance and supply of Streamflow data.

Francis Chiew and Wasyl Drosdowsky – for their valuable advice regarding SST analysis.

Australian Survey and Land Information Group (AUSLIG) – for supply various digital data layers of Australia.

The Departments of Land and Water Conservation (NSW), Sustainability and Environment (VIC), Natural Resources and Mines (QLD), Lands Planning and Environment (NT), Western Australia Rivers Commission (WA) – for the supply of Streamflow data for their respective state areas.

Family and Friends – I acknowledge the support of family and friends in the completion of this thesis.

iv Contents

Abstract ……………………………………………………………………………………..i

Acknowledgements……………………………………………………………………….. iv

Contents……………………………………………………………………………………. v

List of Figures …………………………………………………………………….……… ix

List of Tables ..…………………………………………………………………………... xii

Chapter 1. Introduction and Regional Setting………………………………………. 1

1.1 Introduction and Aims ……….……………………………….………... 1 1.2 Specific Objectives …………………….………………………………..2 1.3 Thesis Outline ..……………………..……….………………….……… 5 1.4 Australia’s Physiography ………………………………………………..6 1.5 Climatic and meteorological processes driving Australia’s rainfall patterns …………..………………………………………………...10 1.6 Conclusion…………………………………………………………….. 15

Chapter 2. The Study Rivers and associated Drainage Divisions…….……………. 16

2.1 Australia’s dryland rivers: hydrology and climatology ………………. 16 2.1.1 The Timor Sea drainage division ……………………………17 2.1.2 The Gulf of Carpentaria drainage division …………………. 22 2.1.3 Lake Eyre drainage division ……………………….……….. 26 2.1.4 Bulloo-Bancannia drainage division ……………………….. 31 2.1.5 The Murray-Darling drainage division …………………….. 34 2.1.6 Southwest coastal division …………………………………. 39 2.1.7 Indian Ocean drainage division …………………………….. 43 2.2 Conclusion ……………………………………………………………. 49

v Chapter 3. Review of Relevant Literature …………………………………………. 52

3.1 Introduction and Objectives……………………………………………52 3.2 Uniqueness, diversity and character of Australia’s dryland rivers …………………………………………………………………………….. 53 3.3 Synoptic weather patterns and flooding worldwide ………………….. 56 3.4 Synoptic weather patterns and flooding in Australia ……….…60 3.5 El Nino southern oscillation: flooding rainfall and stream flow in Australia ………………………………………………………………….. 65 3.6 Conclusion ……………………………………………………………. 71

Chapter 4. Synoptic Weather Patterns and Major Flooding of Australia’s Dryland Rivers …………………………………………………………... 73

4.1 Introduction and Objectives ………………………………………….. 73 4.2 Methods and data sources ………….………………………………….. 73 4.3 The synoptic weather types: Definition and description ……….…….. 76 4.3.1 Tropical troughs/lows ………………………………………. 76 4.3.2 Deep tropical lows ………………………………………….. 78 4.3.3 Tropical cyclones and associated rainfall depressions ……... 80 4.3.4 Northwest cloud bands ……………………………………... 83 4.3.5 Frontal systems ……………………………………………... 84 4.3.6 Cut-off lows ………………………………………….……... 86 4.3.7 Easterly dip and east coast low ……………………………... 89

4.4 Flood producing synoptic weather patterns: class, seasonality and flood magnitude ……………………………………………………………93 4.4.1 The Timor Sea division …………………………………….. 93 4.4.2 The Gulf of Carpentaria division …………………………… 97 4.4.3 The Lake Eyre division …………………………………….100 4.4.4 Bulloo-Bancannia division, Paroo and Warrego Rivers ..…. 104 4.4.5 The northeast Murray Darling drainage sub-division .……..108 4.4.6 Southern Murray-Darling sub-division …………………… 112

vi 4.4.7 The southwest coastal division ……………………………. 116 4.4.8 Indian Ocean division: Greenough, Murchison and Gascoyne Rivers ………………………………………………… 120 4.4.9 The Indian Ocean: Pilbara Rivers …………………………. 124 4.5 Discussion …………………………………………………………... 127 4.5.1 Synoptic weather patterns responsible for frequency of flooding ………………………………………………………. 127 4.5.2 Flood magnitude by synoptic type ………………………... 133 4.6 Conclusion …………………………………………………………... 135

Chapter 5 The Spatial Properties of Rainfall Events in Dryland Australia ……... 138

5.1 Introduction and Objectives ………………………………….……... 138 5.2 Catchment Description ……………………………………………… 139 5.3 Methods ……………………………………………………………... 139 5.4 Results ………………………………………………………………. 142 5.4.1 Thomson River catchment…………………………………. 142 5.4.2 Todd River catchment ……………………………………...143 5.5 Discussion …………………………………………………………... 147

Chapter 6 Sea Surface Temperatures and Australia’s Dryland Rivers: Links to Flooding and flow patterns ……………………………………………. 151

6.1 Introduction …………………………………………………………. 151 6.2 Methods ……………………………………………………………... 154 6.3 Results: Correlations with indices …………………………………... 158 6.3.1 Avoca River ………………………………………….……. 158 6.3.2 Macquarie River …………………………………………... 159 6.3.3 Warrego River …………………………………………….. 160 6.3.4 Cooper Creek ……………………………………………… 160 6.3.5 Thomson River ……………………………………………. 161 6.3.6 Flinders River ……………………………………………... 162 6.3.7 Fitzroy River ………………………………………………. 163 6.3.8 Ashburton River …………………………………………... 164

vii 6.3.9 Gascoyne River …………………………………………… 165 6.3.10 Murchison River ………………………………………… 165 6.3.11 Murray River (Western Australia) ………………………. 166 6.4 Flooding and the SOI ………………………………………….……. 167 6.4.1 Avoca River ………………………………………………. 167 6.4.2 Thomson River …………………………………………… 168 6.4.3 Flinders River ……………………………………………... 169 6.4.4 Fitzroy River ………………………………………………. 170 6.4.5 Gascoyne River …………………………………………… 171 6.4.6 Murray River (Western Australia) ………………………… 172 6.5 Discussion …………………………………………………………... 173 6.6 Conclusion …………………………………………………………... 178

Chapter 7 Summary of Flooding Synoptic Weather Patterns, Climatology and Spatial Rainfall Patterns for Australia Dryland Rivers .……….. 179

7.1 Introduction …………………………………………………………. 179 7.2 Synoptic weather patterns leading to flooding ……………………… 180 7.3 Synoptic weather patterns; past and future climate change ………… 182 7.4 Rainfall properties in arid to semi-arid regions …………………….. 186 7.5 The role of ENSO and SSTs in the flooding of Australian arid to semi-arid rivers ………………………………………………….. 187 7.6 Conclusion ………………………………………………………….. 188

References ……………………………………………………………… 190

viii

List of Figures Figure 1.1 Australia’s dryland rivers and surrounding seas and oceans ……………… 3 Figure 1.2 Australian average annual rainfall ……………………………………..…. 4 Figure 1.3 Average annual evaporation for Australia ………………………………… 4 Figure 1.4 Australia’s physiography ………………………………………………….. 9 Figure 1.5 Sub-tropical high pressure ridge …………………………………………. 13 Figure 1.6 The monsoon trough ………………………………………………….…...14 Figure 1.7 Pilbara and Cloncurry heat-lows ………………………………………… 15 Figure 2.1 Australian drainage divisions ……………………………………………. 18 Figure 2.2 Timor Sea drainage division ………………………………………….….. 19 Figure 2.3 Mean monthly rainfall totals for Katherine ……………………………… 20 Figure 2.4 Monthly temperatures for Katherine ………………………………….…. 20 Figure 2.5 Mean monthly flows Ord River ………………………………………….. 22 Figure 2.6 Gulf of Carpentaria drainage division …………………………………… 23 Figure 2.7 Mean monthly rainfall totals for Normanton …………………………….. 24 Figure 2.8 Monthly Temperatures Normanton ……………………………………… 25 Figure 2.9 Mean monthly flows Flinders River ……………………………………... 26 Figure 2.10 Lake Eyre drainage division ……………………………………………... 27 Figure 2.11 Mean monthly rainfall totals for Birdsville ………………………….…... 28 Figure 2.12 Monthly temperatures for Birdsville ………………………………….…. 29 Figure 2.13 Mean monthly flows for Cooper Creek …….………………………….…. 31 Figure 2.14 Bulloo-Bancannia drainage division …………………………………….. 32 Figure 2.15 Mean monthly rainfall totals for Adavale ……………………………….. 33 Figure 2.16 Monthly temperatures for Adavale ……………………………………… 33 Figure 2.17 Mean monthly flows for the Bulloo River …………………………….… 34 Figure 2.19 Mean monthly rainfall totals for Charleville ………………………….…. 36 Figure 2.18 Murray-darling drainage division ……………………………………….. 37 Figure 2.20 Mean monthly rainfall totals for Euroa ……………………………….…. 37 Figure 2.21 Mean monthly flows for the Paroo River ………………………………... 38 Figure 2.22 Mean monthly flows for the Avoca River …………………………….…. 39 Figure 2.23 The southwest coastal drainage division ………………………………… 40 Figure 2.24 Mean monthly rainfall totals for Perth …..………………………………. 41 Figure 2.25 Monthly temperatures for Perth ……………………………………….…. 42 Figure 2.26 Mean monthly flows for the Avon River ………………………………... 43 Figure 2.27 Indian Ocean drainage division ………………………………………..… 45 Figure 2.28 Mean monthly rainfall totals for Onslow ..………………………………. 45 Figure 2.29 Monthly temperatures for Onslow ……………………………………….. 46 Figure 2.30 Mean annual discharge for selected rivers in the Indian Ocean drainage division ………………………………………… ..47 Figure 2.31 Mean monthly flows for the De Gray River ..………………………….… 48 Figure 2.32 Mean monthly flows for the Murchison River …………………………... 49 Figure 3.1 Diversity in hydrological input, output, throughput, and channel characteristics within the arid zone ……………………………………… 55 Figure 3.2 The synoptic weather patterns for the 5th February 1967 ………………... 63 Figure 3.3 The areas that are most consistently affected by El Nino ……………….. 66 Figure 3.4 The 4 regions that Ropesewski and Halpert (1987) ……………………... 68 Figure 3.5 Rainfall sites from Nichols and Kariko (1993) ………………………….. 69 Figure 4.1 Tropical trough/low-pressure system 1/3/1994 ………………………….. 79

ix Figure 4.2 500hPa upper synoptic chart 1/3/1994 …………………………………... 79 Figure 4.3 Deep tropical low over Central Australia on the 13/01/1984 ………….… 81 Figure 4.4 500hPa synoptic chart 13/01/1984 ………………………………………. 81 Figure 4.5 Tropical Cyclone Wylva situated over the Northern Territory 18/2/2001 ………………………………………………………………… 82 Figure 4.6 500hPa synoptic chart 18/02/2001 ………………………………………. 83 Figure 4.7 Synoptic weather pattern for 21/06/1980 associated with NW cloudband …………………………………………………………… 85 Figure 4.8 Outgoing longwave radiation image 21/06/1980 ……………………….... 85 Figure 4.9 Synoptic weather pattern on the 17/071995 ……………………………... 87 Figure 4.10 500hPa synoptic chart 19/07/1995 ………………………………………. 88 Figure 4.11 Synoptic weather pattern on 30/07/1986 ……………………………….... 88 Figure 4.12 500hPa synoptic chart 30/07/1986 ………………………………………. 89 Figure 4.13 An easterly dip (continental) over eastern Australia on 18/04/1990 …….. 91 Figure 4.14 An easterly dip (offshore) off the NSW/Queensland coastline on 31/07/1990 ……………………………………………………………. 91 Figure 4.15 A well-developed east coast low on 27/07/1984 ……………………….... 92 Figure 4.16 The 500hPa synoptic chart for 27/07/1984 ……………………………… 92 Figure 4.17 The synoptic weather patterns and flood events for the Timor Sea division rivers ……………………………………………. 95 Figure 4.18 Flooding seasonality for the Timor Sea division rivers ………………….. 94 Figure 4.19 Flood Magnitude and Synoptic Class Fitzroy River at Me No Savvy …... 96 Figure 4.20 Flood magnitude and synoptic class for the East Banes River …………... 96 Figure 4.21 The synoptic weather patterns and flood events for the Gulf of Carpentaria division rivers …………………………………………………………….. 98 Figure 4.22 Seasonality of flood events for Gulf of Carpentaria division rivers ……... 97 Figure 4.23 Flood magnitude and synoptic class for the Gregory River at Gregory Downs ……………………………………………………….…. 99 Figure 4.24 Flood magnitude and synoptic class Cloncurry River at Cloncurry ……. 100 Figure 4.25 The Synoptic Weather Patterns and Flood Events for the Lake Eyre division rivers ……………………………………………….. 103 Figure 4.26 The seasonality of flood events for Lake Eyre Basin Rivers …………... 101 Figure 4.27 Flood Magnitude and Synoptic Class for Cooper Creek Currareva ……. 102 Figure 4.28 Flood Magnitude and Synoptic Type for the Todd River at Anzac Oval …………………………………………………………... 104 Figure 4.29 Synoptic weather patterns and flooding in the Bulloo, Paroo and Warrego Rivers …………………………………………………….. 106 Figure 4.30 Flood seasonality in the Bulloo, Paroo and Warrego Rivers ………….... 105 Figure 4.31 Flood magnitude and synoptic type for the Bulloo River at Autumnvale ………………………………………………………….. 107 Figure 4.32 Flood magnitude and synoptic type for the Warrego River at Wyandra .. 108 Figure 4.33 Synoptic weather patterns and flood events for the northeast Murray-Darling Rivers …………………………………………………. 110 Figure 4.34 The flood seasonality of Northeast Murray-Darling Rivers ……………. 111 Figure 4.35 Flood magnitude and synoptic type for the Maranoa River ……………. 111 Figure 4.36 Flood magnitude and synoptic type for the Bogan River ………………. 112 Figure 4.37 The synoptic weather patterns and flood events for the rivers of the southern Murray-Darling rivers …………………………………………. 114 Figure 4.38 Flood seasonality of the Southern Murray-Darling Rivers …………….. 113 Figure 4.39 Flood magnitude and synoptic class for the Avoca River ……………… 115

x Figure 4.40 Flood magnitude and synoptic class for the Broken River ……………... 116 Figure 4.41 Synoptic weather patterns and flood events for southwest coastal rivers of Western Australia ………………………………………………………. 118 Figure 4.42 Seasonality of flood events for the rivers of southwest Western Australia ………………………………………………………. 117 Figure 4.43 Flood magnitude and synoptic class for the Lort River (Fairfield) …….. 119 Figure 4.44 Flood Magnitude and synoptic class for the Murray River (Western Australia) ……………………………………………………... 120 Figure 4.45 Synoptic weather patterns and flood events for the Greenough, Murchison and Gascoyne Rivers of Western Australia ……………………………... 121 Figure 4.46 Seasonality of flood events for the Greenough, Murchison and Gascoyne Rivers of Western Australia ………………………………….. 122 Figure 4.47 Flood magnitude and synoptic class for the Greenough River …………. 123 Figure 4.48 Flood magnitude and synoptic class for the Gascoyne River …………... 123 Figure 4.49 Synoptic weather patterns and flooding of the De Gray, Fortescue and Ashburton Rivers ………………………………………………………... 125 Figure 4.50 Seasonality of flood events for the De-Gray, Fortescue and Ashburton Rivers ………………………………………………………... 126 Figure 4.51 Flood magnitude and synoptic class for the Ashburton River …………. 126 Figure 4.52 Flood magnitude and synoptic class for the De Gray River …………… 127 Figure 5.1 Rainfall stations in the Todd River and Coopers Creek Catchment areas ……………………………………………………………………... 141 Figure 5.2 Average monthly rainfall totals for Alice Springs and Longreach ……... 141 Figure 5.3 Linear regression fitted to correlations coefficient values for rainfall stations in the Thomson River Catchment ……………………………… 144 Figure 5.4 Correlation coefficient values in the Thomson River catchment for storm rainfall totals ……………………………………………………………. 144 Figure 5.5 Linear regression fitted to correlations coefficient values for rainfall Stations in the Todd River catchment …………………………………... 146 Figure 5.6 Correlation coefficient values in the Todd River catchment for storm rainfall totals ……………………………………………………………. 146 Figure 6.1 Map of the Bureau of Meteorology SST1 and SST2 sea surface temperature anomalies ………………………………………………….. 156 Figure 6.2 Indian Ocean SST index averaging area ……………………………….. 157 Figure 6.3 Selected Australian dryland rivers for ENSO and SST correlations …… 157 Figure 6.4 The largest flood magnitude and the percentage of floods per SOI class for the Avoca River at Quambatook ……………………………………. 168 Figure 6.5 The largest flood magnitude and the percentage of floods per SOI class for the Thomson River at Longreach ……………………. 169 Figure 6.6 The largest flood magnitude and the percentage of floods per SOI class for the Flinders River at walkers Bend …………………………………. 170 Figure 6.7 The largest flood magnitude and the percentage of floods per SOI class for the Fitzroy River at Philips Range ………………………………….. 171 Figure 6.8 The largest flood magnitude and the percentage of floods per SOI class for the Gascoyne River at Fishy Pool …………………………………… 172 Figure 6.9 The largest flood magnitude and the percentage of floods per SOI class for the Murray River at Baden Powel Station ………………………….. 173 Figure 6.10 Schematic diagram showing the connection between Indian Ocean climate variability and (a) dry, (b) wet years over southwest Western Australia ………………………………………………………. 177

List of Tables

Table 2.1 Gauge statistics for selected rivers in the Timor Sea drainage division …. 21 Table 2.2 Gauge statistics for selected Gulf of Carpentaria Rivers ………………… 25 Table 2.3 Gauge statistics for selected Lake Eyre Basin Rivers …………………… 30 Table 2.4 Gauge statistics for selected Murray-Darling Rivers ……………………. 38 Table 2.5 Individual site details for selected southwest coastal division Rivers …… 41 Table 2.6 Individual site details for Indian Ocean drainage division Rivers ………. 48 Table 3.1 Correlations of annual SOI with number, length, and intensity of rain events and annual rainfall totals for 5 east Australian sites ……………… 70 Table 4.1 Synoptic classes ………………………………………………………….. 76 Table 5.1 Regression statistics for the Thomson and Todd River catchments ……. 143 Table 5.2 Wet Season (October-April) correlations between individual rainfall stations for the Thomson River Catchment 1990-1995 …………………………. 145 Table 5.3 Dry Season (May-September) correlations between individual rainfall stations for the Thomson River Catchment 1990-1995 ………………… 145 Table 5.4 Wet season correlations between individual rainfall stations for storm rainfall totals in the Todd River catchment between 1990-1995 ………. 147 Table 5.5 Dry season correlations between individual rainfall stations for storm rainfall totals in the Todd River catchment between 1990-1995 ………. 147 Table 6.1 Correlation coefficient values between climatic indices and flow for the Avoca River at Quambatook …………………………………………… 158 Table 6.2 Correlation coefficient values between climatic indices and flow for the Macquarie River at Gongolgon ……………………………………. 159 Table 6.3 Correlation coefficient values between climatic indices and flow for the Warrego River at Auguthella ………………………………………. 160 Table 6.4 Correlation coefficient values between climatic indices and flow for Cooper Creek at Currareva ……………………………………………… 161 Table 6.5 Correlation coefficient values between climatic indices and flow for the Thomson River at Longreach …………………………………………… 162 Table 6.6 Correlation coefficient values between climatic indices and flow for the Flinders River at Walkers Bend ………………………………………… 163 Table 6.7 Correlation coefficient values between climatic indices and flow for the Fitzroy River at Philips Range ………………………………………….. 163 Table 6.8 Correlation coefficient values between climatic indices and flow for the Ashburton River at Capricorn Range ……………………………….. 164 Table 6.9 Correlation coefficient values between climatic indices and flow for the Gascoyne River at Nine Mile Bridge …………………………………… 165 Table 6.10 Correlation coefficient values between climatic indices and flow for the Murchison River at Emu Springs ……………………………………….. 166 Table 6.11 Correlation coefficient values between climatic indices and flow for the Murray River at Baden Powell ………………………………………….. 167

xii

Chapter 1

Introduction and Outline

1.1 Introduction and aims

It is well known that, with the exception of Antarctica, Australia is the driest continent, and is certainly temporally one of the most climatically variable. However, to this date there remains relatively little detailed information on meteorological conditions resulting in the flooding of Australia’s dryland river systems. This study aims to provide information about the synoptic weather patterns that produce flood events, the seasonality of these patterns, and the magnitude of the floods produced by them.

There are many studies demonstrating the correlation between Australian rainfall patterns, streamflow records and temperature with the El Nino Southern Oscillation (ENSO) phenomenon (Pittock 1975; McBride and Nichols 1983; Ropelewski and Halpert 1987, 1989; Nichols 1989; Stone and Auliciems 1992; Drosdowski 1993; Nichols and Kariko 1993; Simpson et al 1993; Allen et al 1996; Chiew et al 1998; Drosdowsky and Chambers 1998; Piechota et al 1998 and Whetton and Pittock 2001). There are, however, few studies that show a direct correlation between ENSO and flood magnitude. Although it is well known that ENSO and Sea Surface Temperature (SST) patterns are closely linked to rainfall patterns, there is not a good understanding of how these influence the timing or magnitude of flood events. As part of this study, the correlation between the Southern Oscillation Index (SOI), various SST indices from the tropical Pacific and the Indian Oceans, and floods from Australian dryland rivers are investigated.

Australia’s dryland rivers (Figure 1.1) occupy parts of the semi-arid and arid portions of Australia, which cover approximately 80% of the continental landmass. In this study, semi-arid is defined as those areas that receive less than 600mm but greater than 300mm average annual rainfall, as measured by the Bureau of Meteorology over a standard 30-year period from 1961 to 1991, and covers 30% of the Australian landmass. The arid zone is defined for this study as

those areas which receive less than or equal to 300mm average annual rainfall and constitutes a further 50% of the continent (Bureau of Meteorology 2003) (Figure 1.2). River systems in semi-arid and arid regions of Australia will be collectively referred to in this study as dryland rivers. Of course many dryland areas in Australia are traversed by long river systems partly affected by conditions in their headwaters that may not be reflective of areas where the average rainfall is less than 600mm. However, such rivers often exhibit characteristics prescribed by the dryland areas through which they pass and are, therefore, of interest in this study. The rivers selected include those sourced entirely in dryland areas (autochthonous) as well as those sourced, in part, upstream of strictly dryland reaches of the catchment (allochthonous). Furthermore the input of rainfall, resulting in runoff to rivers, on various parts of the continent is greatly affected by Australia’s highly spatially-variable evaporation (Figure 1.3).

Flooding of Australia’s dryland river systems results from particular types of synoptic weather patterns depositing massive amounts of rainfall over large areas. The broad aim of this study is to provide a greater understanding of synoptic weather patterns and meteorological processes that cause floods in Australia’s dryland rivers. Further to this, these weather patterns and processes are likely to be altered/shifted through climate change. It is therefore relevant to discuss how climate change could impact the future characteristics of dryland river flooding.

1.2 Specific objectives

The specific objectives of this thesis are to:

• Identify, document and describe the synoptic weather patterns and associated climatology that result in the flooding of Australia’s dryland rivers. • Document the seasonality and flood magnitudes of such events and their relationship to El Nino Southern Oscillation (ENSO) and Sea Surface Temperature (SST) indices. • Assess whether flooding rainfall, in dryland regions of Australia, is spatially heterogeneous or homogeneous.

Figure 1.1: Australia’s dryland rivers and surrounding seas and oceans

Figure 1.2: Australian average annual rainfall

Figure 1.3: Average annual evaporation for Australia.

1.3 Thesis outline

Chapter 2 identifies and describes the rivers studied as part of this research project. Detailed information is provided on the hydrology and climatology of the seven drainage divisions that contain the majority of Australia’s major (as well as minor) dryland river systems.

Chapter 3 places this study in the context of other research performed on synoptic weather- patterns, on the climatological drivers of intense rainfall within dryland areas, and on the spatial properties of rainfall within these areas.

Chapter 4 identifies, documents and describes the types of synoptic weather patterns that have resulted in major flood events within the dryland river systems of Australia. It begins by developing an original synoptic classification scheme specifically for use in this study. Each major and many minor dryland river basins across continental Australia are compared and contrasted in terms of the type of synoptic weather patterns that have resulted in major flood events. This information is evaluated and discussed in terms of its relevance to understanding of the climatological causes of major flooding in dryland Australia.

Chapter 5 provides an analysis of the flood-event climatology of Australia’s dryland rivers. Stream-gauge records from around Australia are related to the SOI and various tropical Pacific and Indian Ocean SST indices in an attempt to identify the climatological controls of flood- event magnitude. Where correlations are provided their statistical significance is tested. This chapter concludes with a discussion of where this analysis adds to knowledge of flood event climatology.

Chapter 6 examines the spatial distribution of individual rainfall events in Australian dryland areas in order assess whether Australian dryland rainfall is made up of spatially heterogeneous or homogeneous patterns.

Chapter 7 provides a discussion of the major findings in relation to Australian dryland hydrometeorology including aspects of past and possible future climate change.

1.4 Australia’s physiography

Australia is the world’s lowest continent with an average elevation of only 330m and spatially restricted highland areas (Figure 1.4), hence orographic effects are spatially restricted.

Australia is divided into 3 broad but distinct geomorphic provinces, namely: the Western Plateau or Shield Zone, the Interior Lowlands and the Eastern Uplands (Figure 1.4) (Kotwicki 1986; Warner 1988).

The Western Plateau occupies over half of the Australian landmass making up most of Western Australia, Northern Territory and Western South Australia (Figure 1.4). The Plateau is a surface with relatively low relief, the highest points rarely exceed 1200m with the bulk of the land surface having an elevation of 450-600m (Heathcote 1994). Towards the eastern boundary of the Plateau lie two parallel east-west trending mountain ranges, the more northerly being the MacDonnell Ranges with the town of Alice Springs at their heart, and to the south the Musgrave Ranges (Heathcote 1994). These consist of narrow steep sided ranges rising to 1000-1500m with narrow deep gorges and sandy watercourses.

North of Adelaide, lie the Flinders Ranges of South Australia. These mark the eastern most edge of the Plateau, with the highest peaks rising to 1100m and receiving a dusting of snow during the winter months.

The Kimberly Plateau, in the north of Western Australia, rises to 1000m and is made up of highly dissected sandstones with numerous watercourses carving their way through the strata. The other major area of relief is the Hammersly Plateau, which lies on the western boundary near the Pilbara region. It has similar physiography to the Kimberly but is less dissected and has peaks rising to over 1200m.

The Pilbara region is situated in the northwest of Western Australia, south of the Kimberly region and north of the Hammersly Plateau, and extends from the Indian Ocean to the Northern Territory border. The region is comprised of three distinct physiographic provinces. There is a

distinct coastal plain, extensive inland ranges (including the Hammersly Plateau) and an inland arid desert region.

Also of note are the Swan coastal lowlands in the southwest corner of Western Australia around the city of Perth. This area is situated on the western edge of the Plateau and consists of faulted down blocks. Many rivers drain from the Darling Ranges before flowing across the costal lowlands to the Indian Ocean. A similar but larger area of costal lowlands exists to the north around Carnarvon. The Stirling Ranges provide some dramatic topography north of Albany at the bottom of Western Australia, with Bluff Knoll rising to 1096m.

Overall though, the Western Plateau consists of areas of low relief, slopes are gentle to flat and where there is significant relief it is of limited extent. Sand sheets and dunes cover almost half of the plateau and are derived from the underlying geology. They make up the large areas of desert that predominate on the plateau (Heathcote 1994).

The Interior Lowlands stretch from the Gulf of Carpentaria in the north to the mouth of Murray River in the south (Figure 1.4), a distance of over 1900km. It consists of three major drainage divisions: the Lake Eyre basin in the center, the Murray Darling basin in the south and the Gulf of Carpentaria drainage division in the north.

As with the Western Plateau the majority of the zone is of low relief but it has significant areas of relief around the margins. In the east, the highlands of Queensland and NSW make up the north-south trending Great Dividing Range that broadly separates the lowlands zone from the eastern highlands. In the east of the zone the topography slopes gently up to the Great Dividing Range. This is particularly evident in the south inland from Canberra on the Riverine Plains and Southwest Slopes. The MacDonnell and Musgrave Ranges rise to over 1400m and provide local dramatic relief in the west, as do the Flinders Ranges in the south, which rise to over 1100m.

Probably the major physical feature of the Interior Lowlands is the thick fluvial deposits laid down by the numerous rivers that traverse this zone. This is particularly evident in the

‘Channel Country’ in southwestern Queensland and in the Murray-Darling basin. All the major river systems in the Murray-Darling and Lake Eyre drainage divisions, including the 3700km Murray-Darling River system, Australia’s longest, drain from the eastern uplands where rainfall is more plentiful and the gradients are steeper. The Murray-Darling River eventually reaches the sea in the southeast corner of South Australia.

The rivers of the Lake Eyre basin terminate at Lake Eyre, a 9,500km2 lake that reaches 10-15m below sea level and fills only infrequently as a result of large floods.

The Eastern Uplands form an unbroken 150 to 500km wide belt of uplands stretching from Cape York to Tasmania (CSIRO 1970). Mountains within the Great Dividing Range are highest in the south within the Snowy Mountains, Victorian Highlands and Tasmania. In the Victorian Alps several peaks rise beyond 1800m and are snow covered during the winter months. Tasmania’s highest peak, Mount Ossa, reaches over 1600m and forms part of the southern end of the Eastern Uplands. The Queensland section of the Eastern Uplands is wider but lower and only in the Atherton Tablelands are there large areas over 650m in elevation, with some peaks reaching 1500m. Around Cape York, plateaus reach over 600m in elevation. Despite the relatively high mountain peaks within the Eastern Uplands it predominantly consists of elevated plateaus rather than mountain peaks. Many streams drain eastwards from the mountains down steep short valleys, then across the coastal plain to the sea.

Australia’s landscape is summed up well by Twidale and Campbell (1993). Firstly, it is a continent of low elevation, with an average altitude of just 330m, with Mt Kosciusko being the highest point at just 2228m. Secondly, Australia is a predominantly planar landscape. Plains of remarkably flatness and huge extent occupy much of Australia’s interior. Thirdly, Australia is the driest continent (excluding Antarctica) with one-third being desert with an average annual rainfall of 250mm or less and 80% being arid or semi-arid with a rainfall of less than 600mm average annual rainfall.

Figure 1.4: Australia’s physiography

1.5 Climatic and meteorological processes driving Australia’s rainfall patterns

Australia is a large island continent in the southern hemisphere with a diverse range of climatic environments, from the hot tropical north with its dramatic monsoon driven climate, to the large expanses of the arid interior and the more temperate regions in the south. Seasonal changes can be extreme with temperatures ranging from 50°C in central Australia during summer to well below freezing in upland areas of southeastern Australia and Tasmania during the wintertime.

Australia’s shape, size and position make it the world’s continent most completely affected by the large high-pressure systems that travel globally from west to east (Gentilli 1972). This means that Australia is predominantly a dry warm country. However, proximity to the surrounding oceans is important as are the various airflows that pass from these oceans over the landmass and how these airflows vary seasonally (Figures 1.5 and 1.6).

The average annual rainfall map of Australia (Figure 1.2) clearly shows the dry interior with higher rainfall areas increasingly concentrically outwards towards the coastal fringe. There is orographic enhancement associated particularly with the Eastern Uplands and a northern enhancement associated with the tropical monsoon. The driest location on the continent is near Lake Eyre and receives an average of 125mm per year, but often much less. The wettest locations are around Cape Tribulation and Innisfail, North Queensland, with up to an average of 8000mm per year in the area of Bellenden Ker, near Innisfail (Bureau of Meteorology 2000). This is a result of moisture-laden southeasterly trade winds being forced to rise over coastal mountains. Areas of southwestern Tasmania also receive very high rainfall totals, from the southern Westerlies, averaging over 4000mm in the wettest locations. Parts of Tasmania and southwestern Western Australia receive some of the most reliable rainfall due to the relentless strength of the moisture-laden roaring forties that affect these regions. The vast majority of Australia has very high potential evaporation with interior regions recording between 2400 and 4000mm per annum (Figure 1.3). This, combined with low rainfall, produces a high moisture deficit causing surface runoff to be very low, with rivers seasonal and

often ephemeral. The average annual discharge to the sea, per unit area, of Australia’s rivers being by far the lowest of any of the continents.

Apart from its low value, the single most defining factor in Australia’s climate is the variability of rainfall, both temporally and spatially. Rainfall variability increases dramatically with decreasing average annual totals such that the interior areas not only receive the lowest totals but also have the most variable values. Also, the nature of Australian arid-zone rainfall means that when rainfall does arrive it is often patchy in spatial extent. Temporal variability in particular is linked to the ENSO phenomenon that is a response to conditions in the tropical Pacific Ocean and its overlying atmosphere. This means that Australia experiences periods of drought followed by very wet periods and flooding in a somewhat erratic fashion but broadly cyclical every 2-10 years (Allan et al 1996).

The movements of large-scale belts of high and low pressure dominate the weather and climate of Australia (Bureau of Meteorology 2000). There are three main belts that affect Australia’s weather patterns (Figures 1.5 and 1.6).

• The subtropical high-pressure ridge: this is a zone of slowly subsiding, drying air which generally produces little or no rainfall. It migrates north during winter and south during summer. • The monsoon trough: this is a summer feature and is the predominant source of rainfall over northern Australia. • The midlatitude westerlies: these contain embedded fronts and depressions that are often the main source of rainfall for southern Australia, particularly during the winter months.

The sub-tropical ridge

This is a zone where individual cells of high-pressure dominate, resulting in generally stable weather (Figure 1.5). The Australian weather map is frequently dominated by the presence of these large mobile cells of high-pressure slowly progressing across Australia from west to east.

The location of the ridge migrates north-south seasonally. During summer to early autumn it lies over southern Australia bringing fine weather to most areas in the southern states. During late autumn the ridge moves northward and intensifies, to be located over inland South Australia and New South Wales by winter (Bureau of Meteorology 2000). This allows weather systems, which are usually pushed well to the south by the ridge in summer, to influence the southern states in winter. At the same time the ridge directs the southeasterly trade winds onto the Queensland coast. When sufficient moisture is picked up over the Tasman and Coral Seas (Figure 1.1) coastal rainfall can result, especially where orographic enhancement occurs.

The monsoon trough

As the summer season approaches, large scale heating in the Australian tropics and sub-tropics results in the development of a low-pressure area that is represented in detail by two distinct and persistent troughs, namely the Queensland and west coast troughs (Figure 1.7). Intense heating occurs in areas where cloud generation is insufficient to prevent the heating, a scenario particularly effective in arid to semi-arid regions. The Queensland trough is located to the west of the Great Dividing Range at an average position of about 700km from the east coast. The west coast trough shows considerably more day-to-day variation in its position and but is generally located inland from the Western Australia coast (Sturman and Tapper 1996). Associated closed lows are frequently observed situated to the north of the west coast and Queensland troughs, the Pilbara and Cloncurry lows, respectively. These are quite shallow, extending only to about 1500m (850hPa level). Upper highs replace these lows at about 3000m (700hPa level) (Sturman and Tapper 1996).

With the steady development of the Pilbara and Cloncurry lows (Figure 1.7) and associated moist air being drawn in from the surrounding oceans off northwestern and northeastern Australia, the monsoon trough becomes established over northern Australia. The monsoon heralds the arrival of the summer rainy season although more correctly refers to the annual reversal of winds in northern Australia from southeasterlies during the dry season to northwesterlies during the rainy (Bureau of Meteorology 2000).

Figure 1.5: Sub-tropical high pressure ridge (Bureau of Meteorology 2003a)

The northern Australia wet season is defined as extending from October to April, but the average monsoon onset date (wind reversal) is December 24th with a standard deviation of 15 days (Sturman and Tapper 1996). The monsoon season proper last for 74 days on average but is also variable with a standard deviation of 15 days. The Australian monsoon is typically characterized by “breaks” of dry weather. Flooding often occurs when the monsoon is well developed and in contrast the summer rains can fail with a weak or erratic monsoon. Darwin has a median rainfall of 1530mm but varies from a decile one rainfall of 1120mm to a decile nine rainfall of 1900mm. During the monsoon season, rainfall is often in the form of thunderstorms, which are associated with the position of the heat lows. Heavy rainfall also frequently results from monsoon depressions and tropical cyclones and these at times can travel well south bringing heavy flooding rain to New South Wales, Victoria and the southern areas of Western Australia.

Figure 1.6: The monsoon trough (key as in Figure 1.5) (Bureau of Meteorology 2003a)

The mid-latitude westerlies

Located well to the south of the Australian continent is a belt of low-pressure known as the sub-Antarctic trough (Sturman and Tapper 1996). Between this trough and the sub-tropical ridge to the north is the zone of mid-latitude westerlies (Figure 1.5 and 1.6). This zone moves north during the winter season to affect southern Australia. Embedded fronts and depressions within this westerly flow are deepest and strongest during the winter months and result in the southern areas of Australia receiving high winter rainfalls with some areas, including western Tasmania, having highly changeable and showery weather throughout the year. Average annual rainfall totals increase south of the center of the continent, demonstrating the effect of the westerlies.

Figure 1.7: Pilbara (left) and Cloncurry (right) heat-lows associated with the Queensland and west coast troughs respectively (Source: Tapper and Hurry 1993)

1.6 Conclusion

Australia is a low, climatically variable continent with up to 80% of the landmass being defined as arid to semi-arid. Average annual rainfall decreases, and temporal and spatial variability increases, with distance inland. A large number of dryland rivers, many of which originate in relatively well-watered uplands, flow through these dryland areas. The large belt of globally migrating high-pressure systems provide Australia with a predominantly warm, dry climate. The monsoon does however bring significant rainfall to northern and sometimes central Australia, with monsoon related weather patterns also occasionally bringing significant rainfall to southern areas during the summer months. During the winter months the mid- latitude westerlies result in frontal weather patterns affecting all of southern Australia and even, at times penetrating well into the center.

15 Chapter 2

The Study Rivers and associated Drainage Divisions

2.1 Australia’s dryland rivers: hydrology and climatology

Despite the widespread aridity and associated extreme climatic variability that influences much of Australia, the continent’s dryland areas support a large number and diverse range of river systems (Figure 1.1). Many of these rivers are dry for years at a time, but when flow does occur, the landscape can be transformed by a wide range of flora and fauna taking advantage of favorable moisture conditions. Following heavy rainfalls enormous flows of up to 40km3, with instantaneous discharges reaching 30,000m3/s, can be generated in the larger rivers (Kotwicki and Isdale 1991).

Australia’s drainage patterns are the result of geologic, geomorphic and climatic factors. For example the Western Plateau drainage division, which occupies a huge area of Western Australia, Northern Territory and South Australia (Figure 1.4), has predominantly uncoordinated drainage with no major and only a few minor rivers. This almost riverless landscape exists because of the arid climate and unique geomorphic controls associated with surface sands and extensive areas of limestone producing high infiltration and gentle gradients that limit any substantial overland flow. In contrast, in most of the other drainage divisions after heavy prolonged rainfall (Figure 1.4) rivers flow over large distances for many months at a time (Knighton and Nanson 2001).

Defined by major topographic features and the main climatic zones Australia is divided into 12 major drainage divisions (Figure 2.1) that give broadly homogenous hydrologic regions (Australian Water Resources Council 1965). There is a narrow continuous peripheral zone of external coordinated drainage that includes the southeast coast, northeast coast, Gulf of Carpentaria, Timor Sea, Indian Ocean, southwest coastal and South Australian Gulf divisions. The Murray Darling division has a coordinated internal drainage network with an external outlet in South Australia for a small part of its discharge. The Lake Eyre division is essentially

16 a coordinated internal drainage network with the majority of the low inputs negated by high evaporation rates. In years with very high rainfall, flow may find its way to Lake Eyre and on rare occasions fill it. Both the Bulloo-Bancannia and the Western Plateau divisions are regions of uncoordinated or disconnected drainage that have, even in the case of the latter which extends to the coast, no outlet to the ocean.

In this chapter the seven drainage divisions that have major (and many minor) dryland rivers are discussed in some detail. These are the Timor Sea, the Gulf of Carpentaria, the Lake Eyre, the Bulloo-Bancannia, the Murray Darling, the southwest coastal and the Indian Ocean divisions.

2.1.1 The Timor Sea drainage division

The Timor Sea drainage division is situated in northern Australia occupying the entire ‘top end’ of the Northern Territory and Western Australia with an area of 547,000km2 (Figure 2.2). It includes some of Australia’s most rugged and spectacular landscapes including the dissected plateaus and deep gorges that make up the Kimberley in the west and Arnhem Land in the east (Figure 1.4).

The climate is best described as hot tropical with two distinct seasons: the ‘wet’ and the ‘dry’. The ‘wet’ technically extends from October to April but usually starts around late December (Bureau of Meteorology 2000). During the ‘wet’, or more correctly the monsoon season, northern Australia receives the majority of its rainfall. This results from the monsoon trough, together with monsoonal lows and tropical cyclones, all of which affect northern Australia at this time. During the winter months the sub tropical high-pressure system provides hot dry weather.

Average annual rainfall totals range from 1600mm along the northern coastline to 500mm in the south to (Figure 1.3). This drainage division does not as such contain any areas of truly arid country, although much of the southern parts are classified as semi-arid, with rainfall becoming

17 more erratic and unreliable with distance south. Figure 2.3 presents the monthly rainfall totals for Katherine, located 300km south of Darwin, demonstrating the highly season rainfall of the summer months.

Figure 2.1: Australian drainage divisions.

Temperatures across the Timor Sea drainage division are approximated by those recorded at Katherine. Due to the monsoon, with its associated cloud cover and humidity, maximum temperatures are recorded at the end of the dry season with October and November recording monthly means of 37.7°C and 37.5°C, respectively (Figure 2.4). January, typically the hottest month in Australia, records an average of only 35.1°C. In the middle of the dry season (the southern winter) northern Australia cools slightly with June and July recording average maxima of around 30°C and minima of around 13°C. Temperatures are less extreme closer to the coast where, for example, Darwin’s mean maximum temperature is 33.1°C for both

18 October and November. Darwin has an average overnight July minimum of 19.3°C, which is warmer relative to sites further inland to the south.

Figure 2.2: Timor Sea drainage division.

There are a number of major river systems that drain the top end of Australia to the Timor Sea. The Fitzroy, Ord and Victoria Rivers are selected for this study. Table 2.1 shows that mean annual flows are relatively high and reliable for this region. For example, the Fitzroy River at Dimond Gorge has a mean annual discharge of 2,543,921 Ml from only 17,560km2 whereas the De Gray River in the Indian Ocean division has a mean annual discharge of 1,404,703Ml from a catchment area of 50,190km2 (Table 2.6).

The Ord River, with a total catchment area of 50,000km2, is situated in the east Kimberly region of Western Australia and extends into northwestern Northern Territory. The Ord River

19 is 560km long and eventually drains into Cambridge Gulf near Wyndham. Rainfall ranges from 450mm in the south of the catchment to 780mm in the north, with most of this falling between November and April with the heaviest falls between January and February.

Mean Monthly Rainfall for Katherine

300

250

200

150

100

50 Monthly Rainfall (mm) 0

l n b r r y n u g p ct v c a e a p a u J u e o e J F M A M J A S O N D

Figure 2.3: Mean monthly rainfall totals for Katherine based on Bureau of Meteorology data from 1942-2000. Average annual total: 1060mm.

Monthly Temperatures for Katherine 50 Mean Max 45 40 Mean Min 35 Highest 30 Max 25 Lowest Min Deg C 20 15 10 5 0

l n r y n u ct a eb pr u J ug ep ov ec J F Ma A Ma J A S O N D

Figure 2.4: Monthly temperatures for Katherine based on Bureau of Meteorology data from 1942-2000.

20 Figure 2.5 demonstrates the effect of the monsoon on mean monthly river flow in the Ord River at Ord River Homestead with a catchment area of 19,600km2. The majority of river flow occurs between December and April, with February providing the highest mean monthly flow. The Ord River also has a more reliable flow with a coefficient of variation (defined as the standard deviation divided by the mean annual flow) of 0.70 on the annual flow volume, quite low compared to other semi-arid to arid rivers in Australia.

In summary the Timor Sea drainage division being situated in the north of Australia has a hot tropical climate with the ‘wet’ from October to April and the ‘dry’ for the remainder of the year. The rivers of the top end of Australia therefore have a flow regime that is highly seasonal with the majority of river flow being between December and April.

River Gauge Catchment Mean Annual Coefficient of Period of 2 Area (km ) Discharge Variation CV Record (Ml) Fitzroy Dimond 16,800 2,543,921 0.82 1970-2001 Gorge

Fitzroy Philips 5,020 729,950 0.82 1966-2002 Range Fitzroy Me No 7,800 818,633 0.88 1965-2002 Savvy Ord Ord River 19,600 1,551,098 0.70 1962-1996 Homestead Ord Bedford 550 54,213 0.82 1967-2002 Downs Ord Mistake Creek 7,700 530,745 0.84 1970-2002 Homestead Victoria Coolibah 44,900 3,184,000 0.69 1952-2002 Homestead Banes River Victoria 2,342 406,814 0.86 1964-2002 Highway Table 2.1: Gauge statistics for selected rivers in the Timor Sea drainage division Data source: Water and Rivers Commission, Western Australia.

21 Mean Monthly Flows Ord River

900000 800000 700000 600000 500000 400000 300000 Monthly Flow (Ml) 200000 100000 0

n b n l t a e pr u Ju ug ep c J F Mar A May J A S O Nov Dec

Figure 2.5: Mean monthly flows Ord River Data source: Water and Rivers Commission, Western Australia. Period of record used for analysis is 1962-1996

2.1.2 The Gulf of Carpentaria drainage division

This division, with a total area of 638,000km2, is also located in the far north of Australia. It makes up a large section of northeastern Northern Territory and northern Queensland, including most of Cape York Peninsula (Figure 2.6).

The hot tropical climate is very similar to that of the Timor Sea division in that it is dominated by two distinct seasons: the ‘wet’ and the ‘dry’. On average, much of northern Australia receives more than 80% of its rainfall from December to March (Sturman and Tapper 1996). Mean monthly rainfall for Normanton supports this with 810mm of the annual average of 920mm falling from December to March (Figure 2.7). Average annual rainfall totals range from 450mm around Mt Isa in the far south of the division, to 1200mm at the southern edge of the Gulf of Carpentaria, and up to 2000mm at the top of Cape York Peninsula. In the northeast corner of the Northern Territory, average annual rainfall totals reach 1400mm (Figure 1.3).

Temperatures across the division vary depending on proximity to the coastline and latitude. Places further south and inland have more variation in temperature, both diurnally and

22 seasonally. For example, inland at Mt Isa mean maximum temperatures range from 37°C in December to 24.7°C in July whereas on the coast at Normanton mean maximum temperatures range from 36.8°C in November to 29.1°C in July (Figure 2.6 and 2.8). Mean minimum temperatures for Mt Isa in July are 8.9°C with Normanton recording a mean minimum temperature of 15.2°C.

Figure 2.6: Gulf of Carpentaria drainage division.

There are a number of rivers that drain from the drier interior of the division across a vast open coastal plain to the Gulf of Carpentaria. Those included as part of this study are the Gregory (tributary to the Nicholson River), Leichardt, Flinders and its tributary the Cloncurry (Figure 2.6). The Nicholson River is located in northwest Queensland and has a catchment area of 53,200km2. The headwaters rise in the Barkly Tablelands in the Northern Territory. The major tributary of the Nicholson River, the Gregory River, also rises on the Barkly Tablelands about

23 250km southeast of the Nicholson. The Gregory finally joins the Nicholson near Burketown before flowing across the coastal plain and into the Gulf of Carpentaria.

The Leichardt River has a total catchment area of 33,000km2 (Table 2.2) and is located to the east of the Nicholson River in northwest Queensland where its headwaters rise in the Selwyn Ranges 40km east of Mt Isa. The Leichardt River flows roughly northwards before being joined by several tributaries (Gunpowder Creek and Fiery Creek), flowing to the Gulf 30km northeast of Burketown.

The Flinders River differs from the Nicholson and Leichardt Rivers in that it rises in the Great Dividing Range to the east. It flows initially westwards and then northwards over the vast savannah country before entering the Gulf. The Cloncurry River, one of the Flinders’ main tributaries, rises on the same plateau as the Leichardt River before flowing northwards to join the Finders River.

Mean Monthly Rainfall Normanton

300

250

200

150

100

50 Monthly Rainfall (mm) 0

n b n l t a e pr u Ju ug ep c J F Mar A May J A S O Nov Dec

Figure 2.7: Mean monthly rainfall totals for Normanton based on Bureau of Meteorology data from 1872-2000. Average annual total: 920mm.

24 Monthly Temperatures Normanton

Mean Max 50 Mean Min 45 Highest Max 40 Lowest Min 35 30 25

Deg C 20 15 10 5 0

ay Jul ov ec Jan Feb Mar Apr M Jun Aug Sep Oct N D

Figure 2.8: Monthly Temperatures Normanton based on Bureau of Meteorology data from 1872-2000.

The mean annual flow of these rivers is relatively high because of the large amount of rainfall that falls over a relatively short time during the wet (Table 2.2). The flow regime for all three rivers is highly seasonal due to the wet, with normally no flow during the dry season. The mean monthly flows for the Flinders River at Walkers Bend, with a catchment area of 107,500km2, show that flows are concentrated from January to March with minor flows occurring mostly from April to December (Figure 2.9).

River Gauge Catchment Mean Annual Coefficient of Period of Area Discharge Variation CV Record (km2) (Ml) Flinders Walkers 107,500 3,910,957 1.27 1970-2001 Bend Cloncurry Cloncurry 5,975 367,249 1.11 1967-1994 Gregory Gregory 12,690 664,375 1.10 1969-2002 River Downs Leichardt Floraville 22,270 728,414 1.31 1969-2002

Table 2.2: Gauge statistics for selected Gulf of Carpentaria Rivers. Data source: Department of Natural Resources and Mines, Queensland.

25 Mean Monthly Flows Flinders River

1800000 1600000 1400000 1200000 1000000 800000 600000 400000 Monthly Flow (Ml) 200000 0

n b r y a e a a Jul J F M Apr M Jun Aug Sep Oct Nov Dec

Figure 2.9: Mean monthly flow Flinders River, northwest Queensland Data source: Natural Resources and Mines, Queensland. Period of record used for analysis is 1971-2001.

In summary, rivers that flow to the Gulf of Carpentaria are highly seasonal, the monsoon related rainy season in the summer months sharply contrasting with the winter dry season. They flow from areas of low rainfall into areas of higher rainfall and are highly variable in flow volume from year to year.

2.1.3 Lake Eyre drainage division

The Lake Eyre drainage division, with an area of 1,170,000km2, is the second largest division in Australia and one of the largest arid inland drainage systems in the world (Figure 2.10). In Australia only the Western Plateau is larger. In terms of a mostly coordinated network, the Lake Eyre division is clearly the largest o the two, as much of the Western Plateau is an un-coordinated network. The division is bounded in the west by the Western Plateau division, in the east by the Bulloo-Bancannia and Murray-Darling divisions, in the northeast by the Northeast coast division and in the north by the Gulf of Carpentaria division (Figure 2.1).

Figure 2.10: Lake Eyre drainage division.

The climate is semi-arid to arid with hot summers and mild to cool winters, the latter particularly in the southern parts. The northern parts are strongly affected by monsoonal rainfall between December and April. Eastern and central parts are affected by both monsoon- related weather systems and temperate or extra tropical weather systems, particularly those such as easterly dips originating from the east coast of Australia.

Average annual rainfall ranges from less that 150mm around Lake Eyre in the south to around 650mm in the far northeast corner, mostly over the warmer months and particularly in the north. For example, Camooweal in the far north of the division receives 290mm of its 390mm average annual rainfall between December and April whereas Maree in the far south receives just 70mm of its 160mm average total for the year over the same period. Figure 2.11 presents the mean monthly rainfall totals for Birdsville located in the center of the division.

27 Mean Monthly Rainfall Birdsville

35 30 25 20 15 10 5 Monthly Rainfall (mm) 0

ct v c Jul Jan Feb Mar Apr May Jun Aug Sep O No De

Figure 2.11: Mean monthly rainfall totals for Birdsville based on Bureau of Meteorology data from 1892-2000. Average annual total: 170mm.

Despite latitude ranging from 19° to 32° south, temperatures do not vary greatly as the Lake Eyre division is strongly continental. The December and January mean maximum temperatures for Camooweal are 38.1°C and 37.4°C with the mean minimum temperature for July being 8.7°C. Maree has equivalent maximum temperatures of 36°C and 37.5°C, with a minimum of 4.8°C. Alice Springs on the western edge of the division, and located in the MacDonnell Ranges at an altitude of 600m, receives occasional frosts during the winter months and has recorded an extreme winter minimum of –7.5°C. Figure 2.12 provides the mean monthly minimum and maximum temperatures, together with the highest recorded maximum and lowest recorded minimum temperatures for Birdsville.

The Lake Eyre basin contains a surprising density of rivers despite the widespread aridity that exists over much of the division. The north and northeast is dominated by three large river systems; Cooper Creek, the Diamantina River and the Georgina River (Figure 2.10). Their catchments make up a large area of western Queensland and the lowland areas of them are referred to as the Channel Country. Morphologically very similar, they are made up of numerous shallow anastomosing channels with very wide flood plains of low gradient (Knighton and Nanson 2001). Their headwaters are in higher rainfall areas but they flow through semi-arid to arid country before terminating in Lake Eyre. They are ephemeral, especially in their lower reaches where they only flow occasionally, about once each decade

28 (Kotwicki 1986). Their middle reaches flow for longer and consist of large perennial waterholes. Despite their erratic nature, they are the main source of flow for the filling of Lake Eyre during infrequent major floods.

Monthly Temperatures Birdsville

Mean Max 60 Mean Min 50 Highest 40 Max Lowest Min 30 20 Deg C 10 0 -10

n ct eb Jul ep Ja F Mar Apr May Jun Aug S O Nov Dec

Figure 2.12: Monthly temperatures for Birdsville based on Bureau of Meteorology data from 1872-2000.

The Georgina River system drains an area of approximately 210,000km2 with the major tributaries being the Sandover River to the west, the Rankin River in the north and the Hamilton River to the east (Figure 2.10).

The Rankin and Sandover Rivers rise in the northern Territory on the Barkly Tablelands and MacDonnell Ranges, respectively (Figure 1.4). The Hamilton River drains the Selwyn Ranges to the southeast of Mt Isa before joining the Georgina River.

The Diamantina River, with a catchment area of 119,000km2 at Birdsville (Figure 2.10), rises in the Swords Range to the east of the Selwyn Ranges in southwest Queensland. The river initially flows in a north and easterly direction before swinging around to the southwest.

At Currareva near Windorah (Figure 2.10), Cooper Creek with a catchment area of 150,000km2, forms from its two main tributaries, the Thomson and Barcoo Rivers. These both rise on the western side of the Great Dividing Range in eastern Queensland before draining in a

29 southwest direction, with large flood flows reaching Lake Eyre after several months. Cooper Creek, Diamantina River and Georgina River have high coefficients of variation demonstrating the degree of stream flow variability within the Channel County Rivers (Table 2.3).

River Gauge Drainage Mean Annual Coefficient of Period of Area Discharge (Ml) Variation CV record (km2) Cooper Currareva 150,220 3,115,066 1.35 1939-1988 Creek Thomson Longreach 57,587 1,281,919 1.73 1968-2002

Barcoo Blackall 8,782 67,812 1.36 1968-2002

Diamantina Birdsville 119,000 1,044,800 1.67 1949-1988

Georgina Roxborough 118,398 1,041,627 1.65 1967-1988 Downs Georgina Glengyle 203,315 1,512,945 1.71 1965-1990

Todd Anzac Oval 445 12,300 1.49 1952-2000

Table 2.3: Gauge statistics for selected Lake Eyre Basin Rivers. Data source: Department of Natural Resources and Mines, Queensland.

The mean monthly flows for Cooper Creek at Currareva show that the majority of the flow occurs from January to April, with an additional small peak of flow in June (Figure 2.13). The latter peak is associated with rainfall from temperate weather systems during the winter months.

The rivers in the western parts of the arid zone of Lake Eyre division drain from lower rainfall areas and are truly ephemeral, with flow only occurring after heavy rain. Flow onset is usually rapid and short lived, reflecting both the nature of rainfall within the desert zone and the fact that little vegetation cover exists to retard overland flow into stream channels. The main ephemeral rivers in the western Lake Eyre division include the Todd River which flows through Alice Springs, the Fink River to the south of the Todd River and the Hay River to the northeast. All flow to the southeast before terminating into the sands of the Simpson Desert,

30 and they are much smaller than the large anastomosing rivers of the Georgina, Cooper and Diamantina Rivers.

Mean Monthly Flow Cooper Currareva

1200000

1000000

800000

600000

400000

Monthly Flow (Ml) 200000

b r y v c Jul o e Jan Fe Ma Apr Ma Jun Aug Sep Oct N D

Figure 2.13: Mean monthly flows for Cooper Creek at Currareva. Data source: Department of Natural Resources and Mines, Queensland. Period of record used for analysis is 1939-1988.

In summary the Lake Eyre drainage division is Australia’s largest inland coordinated drainage division. Its rivers can be divided into two broad groups: the Cooper, Diamantina and Georgina Rivers of the Channel Country; and the smaller, more sandy western rivers that are even more ephemeral, and drain the geographical center of Australia. Rainfall over the division is predominantly associated with the northern ‘wet’ during the summer months, however significant rainfall can be associated with more temperate weather patterns during the cooler months of the year.

2.1.4 Bulloo-Bancannia drainage division

The Bulloo-Bancannia drainage division is situated in southwest Queensland and far northwest NSW. It consists of 100,570km2 of internal disconnected drainage (Figure 2.14). The division is situated between the Lake Eyre division to the west and the Murray-Darling division to the

31 east (Figure 2.1). Average annual rainfall ranges from 500mm in the north of the division to around 250mm in far western NSW.

Figure 2.14: Bulloo-Bancannia drainage division.

Mean monthly rainfall is highest from December to March with another slight rise in early winter, as demonstrated by the mean monthly rainfall for Adavale in the north of the division (Figure 2.14 and 2.15). Rainfall totals are seasonally associated with the northern wet and its associated low-pressure troughs. Autumn to winter totals are predominantly the result of northwest cloudbands, easterly dips and occasional northwards penetration of southern cold fronts or cut off lows.

Summer temperatures can be very hot with winters typically being mild to warm. Temperatures for Adavale range from a December mean maximum of 36°C to a July

32 maximum of 20°C (Figure 2.16). The mean minimum overnight temperature for July is 5.1°C with occasional frosts during winter.

Mean Monthly Rainfall Adavale

70 60 50 40 30 20 10 Monthly rainfall (mm) rainfall Monthly 0

b r r y g t e a a Jul u Jan F M Ap M Jun A Sep Oc Nov Dec

Figure 2.15: Mean monthly rainfall totals for Adavale, Southwest Queensland, based on Bureau of Meteorology data from 1885-1985. Annual rainfall: 390mm.

Monthly Temperaturs for Adavale

Mean Max 50 Mean Min 40 Highest Max 30 Lowest Min

20 Deg C 10

-10

v c Jul o e Jan Feb Mar Apr May Jun Aug Sep Oct N D

Figure 2.16: Monthly temperatures for Adavale based on Bureau of Meteorology data from 1872-2000.

The division only has one main drainage system, the Bulloo River, with a catchment area of 55,000km2. Its tributaries rise in the Gowan Range to the north of Adavale then flow south through the towns of Quilpie and Thargomindah before terminating in sandy country near the

33 Queensland-New South Wales border. Like many of Australia’s dryland rivers, the Bulloo flows only after heavy rain, which in this case occurs predominantly in the headwaters. The Bulloo River at Quilpie has a catchment area of 15,380 km2, a mean annual flow of 471,170Ml (based on the analysis of records from 1967-1992), and a coefficient of variation, (based on annual discharge) of 0.98. At Autumnvale, the Bulloo River has a catchment area of 26,755 km2, a mean annual flow of 586,651Ml and a coefficient of variation of 0.93 (based on the analysis of records 1966-2002). There are two periods of higher than average monthly flow; December to February and April to May (Figure 2.17).

In summary the Bulloo-Bancannia drainage division is a small area of internal disconnected streams with only one moderately large river, the Bulloo, flowing from the higher rainfall regions in the north before being absorbed in the desert sands to the south. Rainfall and the resultant river flow are in two distinct periods; December-February and April-May.

Mean Monthly Flows Bulloo River

160000 140000 120000 100000 80000 60000 40000 Monthly Flow (Ml) 20000 0

n b n l t a e pr u Ju ug ep c J F Mar A May J A S O Nov Dec

Figure 2.17: Mean monthly flow for the Bulloo River at Quilpie. Data source: Department of Natural Resources and Mines, Queensland. Period of record used for analysis is 1967-1992.

2.1.5 The Murray-Darling drainage division

The Murray-Darling drainage division extends across four states from Queensland in the north, through NSW, to Victoria and South Australia in the south (Figure 2.18). Among the drainage

34 divisions with coordinated drainage, the Murray-Darling division with an area of 1,062,530km2 is second in size to the Lake Eyre division with which it differs in that it has an outlet to the sea. However, its total discharge is very low relative to the drainage area due to large evaporation losses and, more recently, irrigation and water storage.

Climate is highly variable across the division from its headwaters in the seasonally snow-clad and well-watered Snowy Mountains in the southeast to the parched arid environments in the north and west where summer temperatures are extreme and rainfall infrequent. The majority of the division is semi-arid to arid with only a thin strip of higher rainfall areas along the southern and eastern margins. Average annual rainfalls range from 2000mm in the Snowy Mountains and Victorian Highlands to less than 250mm west of the Darling River, with rainfall totals generally decreasing with distance inland. Rainfall is concentrated in the summer months in the north of the division and in the winter and springtime in the south. The highest mean monthly rainfall totals for Charleville, southwest Queensland, are from January to March due to the Queensland trough triggering summertime thunderstorms (Figure 2.19). In Euroa, Central Victoria, rainfall totals are highest from June to August due largely to frontal systems and cut-off lows (Figure 2.20). However, most parts of the Murray Darling division are affected at one time or another by a wide variety of weather patterns, even including tropical cyclones (as rainfall depressions) and monsoon depressions that move well south, and northwest cloudbands that move across the continent to bring heavy rainfall over southeastern Australia.

The Murray-Darling division contains an extraordinary diversity of river types from the high energy boulder strewn channels of the upland reaches through to the slow flowing meandering channels in the lowland reaches. The majority of the rivers begin on the Great Dividing Range before flowing westwards. In the north, all the rivers (from the Paroo River in the southwest of Queensland across to the Bogan River in Central NSW) will, after heavy rain, eventually flow into the Darling River. In the south, all the rivers (from the Lachlan River south to the Victorian Rivers) reach the Murray River, with the Darling River joining the Murray in far western NSW near Wentworth.

35 Mean Monthly Rainfall Charleville Queensland

Monthly rainfall (mm) 10

b t e ar ay ov ec Jan Apr Jun Jul Oc F M M Aug Sep N D Figure 2.19: Mean monthly rainfall totals for Charleville, southwest Queensland based on Bureau of Meteorology data from 1942-2000. Average annual total: 490mm.

Clearly the biggest rivers, in terms of mean annual discharge, are the Murray and Murrumbidgee Rivers in the south of the division (Table 2.4). These have their headwaters in the high rainfall areas of the Great Dividing Range, and with relatively reliable flow regimes they contrast with other small discharge rivers within the division. By way of contrast, the Bogan River has a catchment area of 27,970km2 and a highly variable flow with a mean annual discharge of 329,090Ml compared to the slightly larger catchment of the Murrumbidgee River with an area of 34,200km2 at Naranderra, which has a discharge of 3,838,900Ml about 11 times greater. Average annual rainfall totals in the headwater of the Bogan are only about 700mm compared to the headwaters of the Murrumbidgee with up to 1600mm. The seasonal shift from summer-autumn dominated rainfall in the north of the division to winter dominated rainfall in the south is demonstrated in the mean monthly flows for the Paroo River (Figure 2.21) and the Avoca River (Figure 2.22), respectively.

Figure 2.18: Murray-darling drainage division.

Mean Monthly Rainfall for Euroa-Victoria

90 80 70 60 50 40 30 20

Monthly Rainfall (mm) 10 0

ct v c Jul Jan Feb Mar Apr May Jun Aug Sep O No De

Figure 2.20: Mean monthly rainfall totals for Euroa, Central Victoria based on Bureau of Meteorology data from 1883-1997. Average annual total: 655mm.

37 River Gauge Drainage Mean annual Coefficient of Period of Area Discharge Variation Record 2 (km ) (Ml) (CV) Paroo Caiwarro 23,570 564,722 0.91 1967-2002 Warrego Wyandra 42,865 532,115 1.15 1967-2002 Condamine Maranoa 19,490 127,841 1.41 1969-2002 Barwon Mungundi 44,070 689,560 0.92 1973-2000 Namoi Bugilbone 31,100 600,359 1.19 1971-2000 Bogan Gongolgon 27,970 329,090 1.97 1948-2000 Lachlan Condoblin 25,200 954,734 0.89 1973-2000 Murrumbidgee Naranderra 34,200 3,838,908 0.53 1973-2000 Murray Corowa 18,800 5,018,471 0.42 1894-2002 Darling Wilcannia 569,800 2,481,677 0.87 1886-2001 Avoca Quambatook 4,740 50,064 0.84 1973-2001 Broken Rices Weir 7,723 102,302 0.43 1975-1997 Table 2.4: Gauge statistics for selected Murray-Darling Rivers. Data source: Department of Land and water Conservation, NSW and Department of Natural Resources, Victoria.

Mean Monthly Flows Paroo Caiwarro

140000 120000 100000 80000 60000 40000

Monthly Flow (Ml) 20000 0

l t n b r y n u a e pr u J ug ep ov ec J F Ma A Ma J A S Oc N D

Figure 2.21: Mean monthly flow for the Paroo River at Caiwaro. Data source: Department of Natural Resources and Mines, Queensland. Period of record used for the analysis is 1967-2002.

38 Mean Monthly Flows Avoca River at Quambatook

12000

10000

8000

6000

4000

Monthly Flow (Ml) 2000

n b r r y e g p t v c a e a p a n ly u e c o e J A u u O F M M J J A S N D

Figure 2.22: Mean monthly flow for the Avoca River at Quambatook. Data source: Department Sustainability and Environment, Victoria Period of record used for the analysis is 1973-2001.

2.1.6 Southwest coastal division

This drainage division is 314,000km2 in size and is situated in the southwest corner of Western Australia and extends approximately from Esperance in the southeast to Leeman in the north. It includes Perth, the capital city of Western Australia (Figure 2.23). Average annual rainfall totals within this region range from 1600mm in the far south to 300mm in the central and northern sections, with rainfall totals dropping very rapidly with distance inland and more slowly to the north (Figure 1.2). The climate of southwestern Western Australia is defined dominantly by winter rainfall derived from cold fronts and cut off low-pressure systems, although occasionally tropical air does move far enough south to bring rainfall during the warmer months. The climate statistics for Perth provide a good idea of the broad climatic patterns that affect this region of Australia. The seasonality of rainfall does however decrease slightly with distance inland. Figure 2.24 shows that the majority of the rainfall of Perth is received between May and September, with June receiving the highest monthly average of 172mm. Rainfall over the summer months is very low with December and January only receiving averages of 11 and 8mm, respectively.

39 Temperatures vary over the division depending on proximity to the coastline, altitude and latitude. Temperatures on average become more extreme in the inland, with higher summer and lower winter temperatures. The Stirling Ranges, around Albany in the extreme south, occasionally receive snowfalls in the winter months but temperatures increase northwards. Figure 2.25 shows the mean maximum and minimum monthly temperatures, and the highest and lowest monthly temperatures for Perth Airport. The highest mean daily maximum occurs in February with 31.8°C and the lowest mean daily maximum occurs in July with 17.7°C. The mean minimum for February is 17.4°C with July being decidedly colder with 8.1°C.

Figure 2.23: The southwest coastal drainage division.

The southwest coastal division contains numerous rivers draining from the Darling Fault Scarp that runs north-south along the coast of Western Australia. Several of these are large, extend well inland from the dryland areas and run generally westward to the sea. The rivers assessed

40 as part of this study from this division include the Avon, Murray, Blackwood, Pallinup and Lort Rivers (Figure1.1). Table 2.5 presents the details of their catchment areas and mean runoff values.

River Gauge Catchment Mean Coefficient Period of Area (km2) Annual of Record Flow (Ml) Variation CV Avon Northam 99,610 147,895 0.78 1977-2003

Murray Badden 6,687 256,283 0.65 1952-2003 Powell Blackwood Manywaters 9,354 23,370 0.80 1982-2002

Pallinup Bull Crossing 4,897 26,509 1.25 1973-2003

Lort Fairfield 3,069 5,829 1.49 1973-2003

Table 2.5: Individual site details for selected southwest coastal division Rivers. Data source: Waters and Rivers Commission, Western Australia.

Mean Monthly Rainfall Totals Perth Airport

200 180 160 140 120 100 80 60 40

Monthly rainfall (mm) 20 0

n b n l a ar pr ay u Ju ug ep ct ov ec J Fe M A M J A S O N D

Figure 2.24: Mean monthly rainfall totals for Perth Airport based on Bureau of Meteorology data from 1944-2000. Average annual total: 795mm.

41 Monthly Temperatures Perth Airport

Mean Max 50 Mean Min 40 Highest Max Lowest Min 30

20 Deg C 10

-10

ct ar ay Jul Jan Feb M Apr M Jun Aug Sep O Nov Dec

Figure 2.25: Monthly temperatures for Perth Airport based on Bureau of Meteorology data from 1944-2000.

The Avon River flows from arid areas in the west of this drainage division through the higher rainfall areas associated with the coastal escarpment and eventually into the Swan River, which flows through the city of Perth to the Indian Ocean. With a total catchment area of 120,000km2 the Avon and Swan Rivers make up the largest river system in the southwest coastal drainage division. The average annual rainfall over the Avon catchment ranges from 800mm in the far west in the proximity of the 300m high Darling Fault Scarp, to around 250mm in the eastern parts. The majority of this catchment sits atop a dissected plateau and is flat apart from minor relief in the western parts associated with the escarpment. Inter-connected playa lakes make up the driest regions in the east. At Northam the Avon River has a catchment area of 99,610km2. Historic mean monthly river flows are highly seasonal with 70% of the total occurring during the winter months of June, July and August (Figure 2.26) and low or no flow during the summer months.

In summary, the southwest corner of Western Australia has a temperate climate, with temperatures on average neither excessively hot nor extremely cold relative to the more northern or central regions. Rainfall is concentrated in the winter months and is relatively reliable, although rainfall does become rapidly less reliable with distance inland as average rainfall totals become less. Flows reflect the highly seasonal rainfall regime in that the majority occur during the winter months.

42 Mean Monthly Flow-Avon River Northam

50000 45000 40000 35000 30000 25000 20000 15000 10000 Monthly Flow (Ml) 5000 0

b r t c ay c ov Jan Apr O Fe Ma M June July Aug Sep N De

Figure 2.26: Mean monthly flows for the Avon River at Northam. Data source: Waters and Rivers Commission, Western Australia. Period of record used for analysis is 1977-2003.

2.1.7 Indian Ocean drainage division

The Indian Ocean drainage division covers an area of 518,000km2 and extends from approximately Geraldton in the south to Port Headland in the north (Figure 2.27). It occupies all of northwest Western Australia and includes spectacular coastal cliffs and dissected inland plateaus of the Hammersly and Pilbara regions (Warner 1986).

The climate is best described as semi-arid to arid with hot summers and mild winters (Bureau of Meteorology 1998). Rainfall occurs in two fairly distinct seasons (Figure 2.28); January to March associated with the northern monsoon and it’s related low-pressure troughs and tropical cyclones, and May to June commonly due to frontal penetration northwards along with northwest cloudbands. Frontal systems, however, have more of an effect on the southern parts of the division. Predominantly though, the division is under the influence of the sub tropical ridge of high pressure resulting in dry stable weather. Average annual rainfalls range from 450mm in the far south to under 200mm in the center (Figure 1.2). The north has higher totals than the center, with around 400mm, however, the vast majority of the division receives less than or equal to 300mm and as such is classified as arid. In contrast to the eastern states of

43 Australia where average annual rainfalls increase with proximity to the coastline this aridity extends right to the sea. As is normal for areas of low precipitation, rainfall is highly variable. Many locations receive very low rainfall, some for several consecutive years, only to be followed by heavy rainfalls from a tropical cyclone crossing the coastline in summer or a northwest cloudband in the winter months.

Temperatures are highest in January and February when inland averages generally exceed 37°C with temperatures towards the mid 40s not uncommon in the northern inland regions (Bureau of Meteorology 1998). Mean maximum temperatures are significantly less for coastal areas due to the reliable sea breeze. Onslow, situated on the coast approximately 400km southwest of Port Headland, demonstrates this effect with a mean maximum of 35.8°C in January (Figure 2.29). Maximum temperatures are also less for more southern areas. For example at Geraldton on the coast the mean maximum temperature for January and February is 31.7°C and 32.6°C, respectively. Average winter maximum temperatures for July range from 18°C in the far south to over 26°C in the northern inland regions. Overnight minimums for July average over 10°C in coastal and more northern parts to less than 6°C in the southern inland areas where temperatures below freezing are occasionally recorded.

Despite the widespread aridity of the Indian Ocean drainage division, many large rivers drain its vast areas. Although they are predominantly dry or contain only small flows, they do at times flow with extraordinary force and volume. The large rivers of this division all drain to the Indian Ocean and include the Greenough and Murchison Rivers in the far south, the Gascoyne River in the central parts, and the Ashburton, Fortescue and De Gray Rivers in the north.

Figure 2.27: Indian Ocean drainage division.

Mean Monthly Rainfall-Onslow

10 Monthly Rainfall (mm) 0

n b n l t a e pr u Ju ug ep c J F Mar A May J A S O Nov Dec

Figure 2.28: Monthly rainfall for Onslow-Western Australia based on Bureau of Meteorology data from 1886-2000. Average annual total: 275mm

45 The Pilbara Rivers (De Gray, Ashburton, and Fortescue) drain from the Hammersly Plateau, and associated upland areas in the northern inland of the Pilbara region (Figure 1.4). The Gascoyne River sits mostly within the Carnarvon Sedimentary Basin, which has low relief and open drainage apart from in the far northeast where the southern edge of the highly dissected Hammersly Plateau provides high relief and a well-developed drainage pattern. The Murchison River drains part of the Yilgarn Plateau (Figure 1.4) that consists of mainly granitic sandy plains and ridges made up of metamorphic rocks. The lower catchment areas of the Murchison River consist of dissected plateaus and hills with extensive sand plains (Jennings and Mabbutt 1986).

Figure 2.30 shows the mean annual discharge of the above-mentioned rivers at specific gauging stations. Mean annual discharge is obviously governed by factors such as physiography, geology, vegetation and climate. Despite having a relatively small catchment area, the De Gray River at Coolenar Pool has the largest mean annual discharge. The Gascoyne River at Nine Mile Bridge has the second largest mean annual discharge but from a significantly larger catchment area. The Murchison River in the far south of the drainage division has the largest catchment area but the second smallest mean annual flow.

Monthly Temperatures Onslow

60 Mean Max 50 Mean Min Highest Max 40 Lowest Min 30 Deg C 20

ct v c Jul Jan Feb Mar Apr May Jun Aug Sep O No De

Figure 2.29: Monthly temperatures for Onslow Western Australia based on Bureau of Meteorology data from 1886-2000.

46 Table 2.6 presents the hydrological characteristics, including the coefficient of variation, for each of the major rivers, with the most variable rivers being the Murchison, Ashburton and Gascoyne Rivers. This is in contrast to the rivers in the southwest of Western Australia where coefficients of variation are generally lower, in line with more reliable rainfall.

Mean Annual Flow

1400000 90000

1200000 80000 70000 1000000 60000 800000 50000 600000 40000 30000 400000 20000 200000

10000 Catchment Area (Km) Mean Annual Flow (Ml) 0 0

n h o ue ray c ison Megalitres G s h e burt te D h r Catchment Area Gascoyne As Fo Murc Greenoug Figure 2.30: Mean annual discharge for selected rivers in the Indian Ocean drainage division of Western Australia. Data Source: Water and Rivers Commission, Western Australia

Figure 2.31 presents the mean monthly flows for the De Grey River in the Pilbara region. It can be seen that bulk of river flow occurs in the summer to early autumn months of February to April. December and January have small mean monthly flows with the reminder of the year having no or low monthly flow. This clearly shows that flows are associated with tropical rainfall systems.

The mean monthly flows for the Murchison River located in the south of the Indian Ocean drainage division also show a peak from February to April along with another peak in the winter months from June to August (Figure 2.32). This latter peak is the result of cold fronts and cut-off lows bringing rain from further to the south. This contrasts the entirely summer dominated regime in the De Gray River further north (Figure 2.31).

47 Mean Monthly Flows De Gray River

600000

500000

400000

300000

200000

Monthly flow (Ml) 100000

l n b n ct a ar pr ay u Ju ug ep ov ec J Fe M A M J A S O N D

Figure 2.31: Mean monthly flows for the De Gray River. Data Source: Water and Rivers Commission, Western Australia. Period of record used for analysis is 1974-2002.

River and Gauge Catchment Mean Annual Co-efficient Period of Station Area (km2) Discharge of Record Name (ML) Variation (CV) De Gray Coolenar 50,190 1,404,073 1.16 1974- Pool 2002 Gascoyne Nine Mile 73,690 726,225 1.25 1957- Bridge 2003 Gascoyne Fishy Pool 71,212 728,449 1.08 1964- 2003 Ashburton Capricorn 43,510 339,331 1.42 1968- Range 2003 Ashburton Nanutarra 71,387 958,198 0.93 1972- 2202 Fortescue Bilano 49,020 281,132 0.92 1987- 2003 Fortescue Newman 2,822 47,470 1.28 1980- 2002 Murchison Emu 87,460 137,627 1.48 1967- Springs 2002 Greenough Karlanew 11,490 19,223 1.23 1971- Peak 2003 Greenough Pindarring 5,668 5,068 1.97 1976- Rocks 2002 Table 2.6: Individual site details for Indian Ocean drainage division Rivers. Data source: Water and Rivers Commission, Western Australia.

48 Mean Monthly Flows Murchison River

60000

50000

40000

30000

20000

Monthly flow (Ml) 10000

n b n l a ar pr ay u Ju ug ep ct ov ec J Fe M A M J A S O N D

Figure 2.32: Mean monthly flows for the Murchison River. Data Source: Water and Rivers Commission, Western Australia. Period of record used for analysis is 1967-2002.

In summary, the Indian Ocean drainage division occupies all of northwest Western Australia which has some of Australia’s most arid and climatically variable locations. The climate is basically hot and dry with mild to warm winters. Rainfall is highest in the extreme south and northern regions, with inland areas the driest. The rivers that drain this division are ephemeral, only flowing after heavy rain. River flow, once initiated, can be of a volume and force quite unexpected from such an arid landscape.

2.2 Conclusion

Australia is a large predominantly dry continent with a diverse array of dryland river systems. Many of these flow from higher rainfall regions into terminal lakes or desert sands while others flow from dry headwaters through higher rainfall areas to the sea.

The Timor Sea drainage division occupies the northwest of Australia and has a hot tropical climate with the ‘wet’ extending from October to April and the ‘dry’ for the remainder of the year. There are a number of major dryland rivers that drain the top end of Australia to the

49 Timor Sea, including the Fitzroy, Ord and Victoria Rivers. These have a relatively reliable, highly seasonal flow regime associated with the ‘wet’.

The Gulf of Carpentaria division in northern Australia has a very similar climate to the more westward Timor Sea division, with ‘wet’ and ‘dry’ seasons dominating. The Nicholson, Leichhardt, Flinders and Cloncurry Rivers all drain from inland ranges or tablelands onto the coastal plain and to the Gulf of Carpentaria. Mean annual discharges are large in response to the areas monsoon-related rainy season.

The Lake Eyre division is the largest coordinated drainage division in Australia and one of the largest inland drainage systems in the world. Eastern and northern parts have higher average annual rainfall due to temperate or extra tropical weather systems from the Pacific Ocean and monsoon related weather patterns, respectively. The rivers of the division can be broadly divided into two groups: those of the Channel Country rivers including the Cooper, Diamantina and Georgina; and the smaller more sandy western rivers that are more ephemeral and drain from the geographic center of Australia.

The Bulloo-Bancannia division consists of a small internal disconnected drainage system located between the Lake Eyre division to the west and the Murray-Darling to the east. It contains only one moderately large river that flows from higher rainfall regions in the north before being absorbed in the desert sands to the south.

The Murray-Darling division is only slightly smaller than the Lake Eyre division with which it differs by way of having an outlet to the sea. The climate varies considerably, from the seasonally snow-clad and well-watered Snowy Mountains to the parched arid desert environments in the north and west where summer temperatures are extreme and rainfall infrequent. The rivers are as diverse as the climatic environments, from high-energy boulder strewn channels of the upland reaches to the slow meandering channels in the lowland reaches.

The southwest coastal division of Western Australia has a temperate climate with rainfall concentrated in the winter months from frontal and cut-off low-pressure systems. Average

50 annual rainfalls range from 200mm in the inland parts to 1600mm in the far south. It contains several dryland rivers with many of these draining from the Darling Fault Scarp that runs north-south along the coast of Western Australia. Flows reflect the highly seasonal rainfall regime in that the majority occur during the winter months.

The Indian Ocean division extends from Geraldton to Port Headland and occupies much of the central coastal and inland regions of Western Australia. The climate is semi-arid to arid with hot summers and mild winters. Rainfalls generally occur from January to March associated with the northern monsoon, and between May and June as a result of northwest cloudbands, frontal systems and cut-off lows. The rivers that drain this region are ephemeral in nature and include the Greenough and Murchison Rivers in the south, the Gascoyne River in the central parts, and the Ashburton, Fortescue and De Grey Rivers in the north.

51 Chapter 3

Review of Relevant Literature

3.1 Introduction and objectives

Studies of synoptic situations in relation to flooding and rainfall across dryland Australia are relatively uncommon. There are a number of studies performed on Australian dryland river flow variability (Finlayson and McMahon 1988; McMahon et al. 1992; Knighton and Nanson 1994, 2000) and the resultant ecology (Kingsford and Norman 2002; McMahon and Finlayson 2003; Puckridge et al. 1998; Puckridge et al. 1999; Watts 1999; Young 1999). In addition there have been studies on specific flood events, such as that by Williams (1970), which investigated the central Australian stream floods of February-March 1967, and studies into the flooding and hydrology of Lake Eyre (Kotwicki 1986; Kotwicki and Isdale 1991). There has been a study focusing on significant rainfall events (Bureau of Meteorology 1995) and another looking into the temporal distribution of large and extreme rainfall bursts, both of these for southeastern Australia (Bureau of Meteorology 1998). However, these mainly focused on humid-zone rainfall and did not investigate the controlling synoptic situations or the spatial pattern of rainfall. Cordery and Fraser (2000) completed the only study to date into rainfall distribution in the Australian arid zone, but were restricted to a relatively small area of western New South Wales.

This study takes a continental approach to identifying the synoptic weather patterns that result in flooding of Australian dryland river systems of various size. While no such study has previously been performed for Australia, studies into synoptic circulation patterns and the resultant heavy rainfall and catastrophic flooding have been undertaken in the United States (Hirschboeck 1987; Keim 1996; Konrad 1997).

52 3.2 Uniqueness, diversity and character of Australia’s dryland rivers

The Australian landmass supports a large number and diverse range of dryland river systems. Although many are ephemeral and are dry more often than not, their flow regime is of importance to many agricultural areas and regional centers. Despite the fact that dryland river basins cover the majority of the Australian landmass they are much less studied compared to their more humid counterparts located around the coastline.

Dryland rainfall is highly variable with long dry periods separated by large but infrequent rainfall events that can generate extreme floods. It is argued that greater temporal variability often accompanies higher spatial variability in the form of localized convective thunderstorms (Knighton and Nanson 1997). However, the only study of spatial rainfall variability performed in dryland Australia showed that rainfall totals for western NSW were spatially uniform for annual, monthly and storm rainfall (Cordery and Fraser 2000). This appears to be in contrast to many other arid zones of the world, including Tanzania (Sharon 1974), Namibia (Sharon 1981) and Arizona (Osbourne et al. 1979) where rainfall has been found to be more localized and convective in nature.

The lack of rainfall in dryland areas means that vegetation is generally sparse and this leads to rapid runoff and a rapid rise in stream-flow hydrographs. Flood peaks tend to have a more prominent peak and recede more rapidly than rivers in more humid zones (Knighton and Nanson 1997). However, much of the previous research has been in small basins. With increasing drainage area there is a reduction in the steepness of the rising limb of the hydrograph and a broadening of the time base of the flow event.

Flood magnitude frequency curves tend to be much steeper for dryland rivers demonstrating that floods in intervening years are very small whilst others years are characterized by extreme events. In the United States the 12 largest rainfall-runoff floods were found to have all occurred in arid-semi-arid regions (Costa 1987).

53 Graf (1988) documents 4 types of flood events in dryland rivers: flash floods; single peak floods; multiple peak events and seasonal floods. Knighton and Nanson (1997) state that these 4 types are somewhat scale dependent, with flash floods tending to occur in smaller catchments where convective storms can involve the entire contributing area, and seasonal floods being produced in wetter, more mountainous regions where rainfall tends to have much greater seasonal reliability. In this continuum of flood flow occupancy, from ephemeral to perennial, there is some correlation with rainfall producing mechanisms. Flash floods result from convective storms and perennial flows from more seasonal rainfall (or snow melt). Between the two end members there is a wide range of conditions with tropical and more frontal weather patterns tending to result in single and multi-peak flood events (Knighton and Nanson 1997). Figure 3.1 demonstrates the diversity in hydrological input, throughput, output and channel characteristics.

Knighton and Nanson (2001) use event analysis as a vehicle for explaining dryland hydrological variability of the Channel Country rivers within the Lake Eyre basin, identifying three types of flood event: single; multiple and compound. These are similar to the types proposed by Graf (1988) in that single peak events correspond to his flash floods, and multiple and compound events to his multiple-peak floods. Single events have a shorter duration, a smaller magnitude and a more rapid time to peak, with the same hydrograph form as the flash floods of Graf (1988) but much longer in duration. Another critical difference between Graf’s (1988) flash flood and Knighton and Nanson’s (2001) single event is that the latter occurs in both large and small contributing basin. Multiple events have multiple peaks that progressively rise towards a well-defined maximum. Compound events are multi-peaked with no progressive rise and fall in the peaks around the maximum discharge. Interestingly, multiple events behave more like single events at small runoff volumes and with small basin areas, and more like compound events as the flood volume or basin area increases.

Figure 3.1: Diversity in hydrological input, output, throughput, and channel characteristics within the arid zone. Reproduced from Knighton and Nanson (1997)

Like South African rivers, those in Australia have much higher coefficients of variation than those on other continents (Finlayson and McMahon 1988). The coefficient of variation, defined as the standard deviation divided by the mean annual flow, is regarded as one of the most useful measures of hydrologic variability (McMahon 1979). In Australia the areas where the coefficient of variation is highest correspond to low rainfall areas in general.

Using 23 different measures of hydrological variability, Puckridge (1999) found that the Diamantina, Cooper and Paroo Rivers are in the top 6 of the most variable rivers in the world with Cooper Creek being ranked first. All the major ephemeral rivers of Australia lie within the Lake Eyre Basin (Kotwicki and Isdale 1991). They include the Georgina and Diamantina Rivers, and Cooper Creek.

55 Doupe and Pettit (2002) investigated the ecological perspectives of regulation and water allocation on the Ord River, Western Australia. They stated that the natural flows of the Ord River are highly variable despite the Ord River being dominantly affected by the Australian monsoon. On a worldwide scale the Ord River has a high coefficient of variation value (0.70 for the Ord River at Ord River Homestead) with only the truly arid rivers, such as Cooper Creek in the Lake Eyre basin, having higher.

3.3 Synoptic weather patterns and flooding worldwide

Hirschboeck (1987) analysed 21 catastrophic flood events within United States. She found it useful to separate flash floods in smaller drainage basins from more regional flooding that occurred in larger regional size basins or across several large basins. It was noted that extreme regional floods over broad areas were best explained by examining macro-scale anomalous behavior such as an unusual location or out of season occurrence of an other wise typical weather pattern, an uncommon combination of several common atmospheric processes, an extremely uncommon upper atmosphere weather pattern, or an unusual persistence of a weather pattern in both space and time. Specific macro-scale anomalous weather patterns include blocking of either a high or low-pressure system.

East coast low-pressure systems along the eastern coastline of Australia can result in serious flooding in both coastal and inland river systems. They have been found to be associated with a strong high-pressure system that may stall or block over the Tasman Sea, exacerbating the duration of the storm by directing onshore southeast to southerly winds on to the coast for several days (Bryant 1991).

In a four-fold classification scheme based upon already existing flash flood forecasting schemes for the United States, Maddox and Chappell (1978) and Maddox et al. (1979) identified synoptic, frontal, mesohigh and western type events. Maddox et al. (1980) subsequently divided the western type events into four.

56 Synoptic events tend to form with an intense synoptic-scale cyclone together with an almost stationary front at the surface and a major trough (often a cut-off low) at the 500hPa level (Hirschboeck 1987). This synoptic pattern is most common in spring, early summer and autumn in conjunction with the changeover from winter to summer circulation, and vice versa. Two maximum floods in Colorado resulted from this type of synoptic pattern (Costa 1987). In southern Australia, particularly around southwest coastal areas of Western Australia, a cut-off low-pressure system coupled with an intense 500hPa low can result in severe flooding. Similarly a 500hPa low in conjunction with a strong high-pressure system can induce an easterly trough system to develop over southeastern Australia directing an extremely moist and unstable easterly airflow onto the coastline.

Frontal events result from a stationary or slow moving frontal system at the surface that is oriented west to east. A broad ridge and a weak short-wave trough often dominate the 500hPa chart. The July 1965 flood event in northwest Missouri was the result of frontal weather patterns (Hirschboeck 1987). During the Australian winter, frontal weather patterns that are often associated with an intense low-pressure system, located well south of the mainland, bring substantial rainfall to southern Australia. During the warmer months of the year such frontal weather systems can at times interact with tropical air masses, form a low-pressure trough, and bring widespread rainfall across NSW and Victoria.

Mesohigh events are the result of extreme instability and convection along a cold air outflow boundary. This outflow boundary is usually the result of thunderstorm activity in the late afternoon or evening. The cold air associated with the rainy area of an intense thunderstorm moves ahead of the storm acting like a small cold front and generating a localized high- pressure system. This cold air that is falling through the small high-pressure system then forces warm moist air to rise over the outflow boundary resulting in heavy precipitation. Mesohigh features associated with thunderstorm activity were key factors in the 1977 Johnstown floods in the United States.

Western type events refer to a range of regional synoptic patterns over the Rocky Mountains (Bryant 1991). They are associated with zonal or meridonal circulation that are abnormal air

57 flow patterns within general air circulation at the macro scale. Western type events are associated with the summer rainfall in the southwest of the United States whereas elsewhere they appear linked to weather patterns over the North Pacific Ocean.

Texas has recorded some of the largest floods and rainfall totals both within the United States and around the world (Baker 1977; Caracena and Fritsch 1983; Costa 1987; Hirschboeck 1987). Bryant (1991) states that in the winter months when deep troughs and frontal systems move across the state they can interact with warm, moist air from the Gulf of Mexico. This creates extreme instability, which is often further orographically enhanced by the Balcones Escarpment. Along with synoptic conditions, tropical storms and hurricanes can also affect the area. In the southeast coastal areas of Australia orographic enhancement of warm moist air from the Tasman Sea during late summer to early autumn can result in some spectacularly high 24 to 48 hour rainfalls. Such was the case in the West Dapto NSW floods of February 1984 when the Illawarra Escarpment orographically enhanced extremely moist, unstable easterly airflow to produce a 48-hour rainfall total of 800mm. This was the largest rainfall total in temperate Australia (Nanson and Hean 1984).

In southeastern United States (which includes Texas), Keim (1996) found that heavy rainfall events were produced by three distinct synoptic weather types: frontal systems; tropical disturbances and air mass storms. Frontal systems were found to be the most dominant heavy rainfall generating mechanism, especially in the north away from the coastline. Tropical disturbances and air mass storms were more prominent (although still the minor causative mechanism) near the coastline.

In South Africa during February and March 1988, two of the largest flood events on record affected the Orange River within a three-week period (Jury et al. 1993). The first was the result of a well-developed tropical-temperate trough, with the second flood in March resulting from a cut-off low-pressure system or a west coast trough. High rainfall in the semi-arid to arid interior of South Africa is frequently related to cut-off lows during early summer, and to intense thunderstorms within organized large scale circulations late in summer (January to March). The late summer rainfall systems are termed tropical-temperate troughs due to the

58 interaction between tropical and mid-latitude weather patterns (Jury et al. 1993), in this case between a tropical low and a mid-latitude depression. Cloud bands associated with these weather patterns are usually west to east oriented and hence are easily identified on satellite imagery. Lindesay and Jury (1991) have noted that a key feature for the development of tropical-temperate troughs is the inland penetration of a tropical wave disturbance over sub- tropical Africa embedded in an easterly air flow from the southwest Indian Ocean. Australia is similar to South Africa in that interaction between tropical and mid-latitude weather patterns is common. Northwest cloudbands form when warm tropical air is forced to rise over much cooler air associated with the northwards penetration of a cold front. These cloudbands can stretch over 5,000km from northwest to southeast Australia and bring heavy rainfall and flooding, particularly about the Gascoyne region of Western Australia.

Komuscu and Seyfullah (1998) state that flash floods have become a hazardous phenomenon along the Mediterranean Coasts of Turkey in recent years. In analyzing the meteorological factors responsible for the Izmir flash flood of November 1995 the authors adopted the methods used by Maddox et al. (1979). At the mesoscale level they identified pronounced low level advection, positive vorticity and strong upper level divergence. A surface low together with a frontal system was situated in the Aegean Sea, which was enhancing the advection of warm moist air along a low level jet stream and producing a squall line (line of thunderstorms). Rainfall intensity was further increased through orographic enhancement. The combined meteorological features of this flood event in Turkey were remarkably similar to the synoptic type events as classified by Maddox et al. (1979) for the United States. In the central and southern regions of Western Australia frontal systems often combine with surface lows, most often cut-off lows, resulting in heavy rainfall with imbedded thunderstorms.

Sharon and Kutiel (1986) have analysed the distribution of rainfall intensities in Israel and described the controlling meteorological situations. They found that there is a relatively high frequency of high intensity rainfall along the coast and in the Rift Valley along with the Negev Desert. Among the mountainous regions in northern and central Israel, rainfall intensities were found to be lower, especially during winter and springtime. In these northern and central areas, where annual rainfall totals range up to 700mm, 40% was found to be due to frontal systems

59 whereas in the south, where average annual rainfall totals are as low as 100mm, frontal systems only account for 10% of the total. The majority of the rainfall in the south is the result of discrete cloud clusters associated with Benard Cells (Rosenfeld 1980), along with high intensity showers resulting from the low level Red Sea Trough extending from the Sudanese low. The Red Sea Trough mostly results in warm dry weather but can be associated with deep instability and the widespread occurrence of severe thunderstorms. Sharon and Kutiel (1986) concluded from their study, which compared the intensity of rainfall in humid regions with arid regions, that arid regions receive greater proportions of high intensity rainfall.

Dhar and Rakhecha (1979) analysed the incidence of heavy rainfall for Rajasthan, in the desert region of northwest India. Rajastha has an average annual rainfall of 310mm and the majority is received from the Indian monsoon between July and September. Dhar and Rakhecha (1979) state that the region is unique in that it receives heavy periods of rainfall in association with monsoon depressions and cyclonic disturbances. It is not unusual for specific localities to receive their entire annual rainfall total in one day. Indeed point rainfalls of between 250- 500mm were associated with single severe rainstorms. The spatial distribution of rainfall in this area is not known objectively but if the majority of the rainfall results from monsoon depressions and rainfall depressions it could be assumed that rainfall totals are of a widespread nature rather than convective and localised. Northern Australia often receives the bulk of its annual rainfall as a result of monsoon related weather patterns. Such weather patterns can at times travel inland bringing widespread heavy rainfalls and flooding to arid central Australia.

3.4 Synoptic weather patterns and flooding in Australia

In one of the few Australian studies into rainfall and synoptic patterns associated with flooding Weeks (1992) identified six main groups of synoptic weather patterns that were responsible for flood events in the Illawarra region of NSW:

1. Tropical cyclones - as rainfall depressions that move south;

60 2. Zonal synoptic patterns: Where high or low pressure directs moist air on to the coastline; 3. Inland depressions: Low-pressure depressions that originate from continental; Australia. They can dip down into NSW and intensify over coastal areas; Occurrence is most common in autumn in relation to the northern monsoon; 4. Easterly lows: Low-pressure troughs moving south from Queensland, generally just off the coastline. These occur predominantly in the winter months; 5. Continental lows: Low pressure systems that move across the continent either from southern continental areas or from southern Queensland; 6. Miscellaneous: Two patterns that don’t correlate with any of the other defined groups.

Her work stands in contrast to case studies associated with specific flood events.

Shepherd and Colquhoun (1985) and Nanson and Hean (1985) described the meteorological aspects of a high magnitude flash flood that occurred on the 18th February 1984 near Dapto, in the Illawarra region of New South Wales. The rainfall totals recorded during this event set new Australian point records for durations in temperate regions from 8-17 hours (Shepherd and Colquhoun 1985). The synoptic situation consisted of a high-pressure system centered well to the southeast of Tasmania directing moist unstable air on to the coastline.

There have been a number of other studies into the rainfall patterns associated with various floods in the Illawarra region. Armstrong and Colquhoun (1976) described the intense rainfalls resulting from severe thunderstorms over the Sydney and Illawarra metropolitan area during the 10th and 11th of March 1975. The August 1998 storm resulted in some of the worst flooding recorded in the Illawarra. Evan and Bewick (2001), Mcilwain (1999), and Reinfelds and Nanson (2001) all describe various aspects in relation to the rainfall during the August 1998 rainfall event. Grootemaat (2000) described the spatial rainfall properties of eight high magnitude rainfall events that affected the Illawarra region ranging from the October 1959 rainfall event through to October 1999.

61 Crohamhurst, situated within the Stanley River catchment (a tributary of the Brisbane River) received rainfall totals of nearly 900mm over 24hrs as a result of a cyclone that crossed the coast north of Brisbane in early February 1893 (Brunt 1958). Flooding was disastrous in both the Mary and Brisbane River catchments, with hundreds of homes lost. The cyclone had a well-defined trough extending southwards from its center along the Queensland coastline. There was a strong high-pressure center in the southern Tasman Sea feeding moisture into the trough. This produced a great deal of convergence around the tip of the trough. With the slow southward movement of the whole system, torrential rainfalls resulted. During the next two weeks a further two cyclones crossed the coastline providing further flooding. Brunt (1958) suggested these cyclones were the result of an almost stationary upper trough with blocking systems to the southeast.

Five severe short duration rainfall events in non-tropical Australia were analysed by Pierrehumbert and Kennedy (1981) who found they were basically the result of thunderstorm cloudbursts that are reported frequently in Australia during the summer months. The Woden Valley storm of January 1971 was the result of just such a thunderstorm. Rainfall totals were around 102mm over 1km2 and 99mm over 10km2. The thunderstorm remained stationary over several Canberra suburbs for about 90 minutes resulting in flooding of Woden Valley where several lives were lost. The synoptic situation consisted of a broad region of low pressure dipping southward from the northern monsoon. In the upper atmosphere over southeastern Australia there was a well-developed trough that together with the surface trough would have provided severe convective instability.

Williams (1970) documented the floods of February to March 1967 in central Australia, finding that the heavy rainfalls were basically the result of southward intrusion of monsoonal weather patterns (Figure 3.4). Tropical air penetrated southwards and convergence between it and an easterly airflow from a large high-pressure system resulted in widespread rainfall in early February. Northwesterly winds, associated with the monsoon trough, again penetrated well south during the third week of February, with a cold front undercutting the northwesterly winds from the south and causing further heavy rainfall. During early March ex-tropical

62 cyclone Gwen moved southeastwards as a rainfall depression, bringing torrential rainfall as it tracked across central Australia.

During the summer of 1974, simultaneous flooding of unprecedented magnitude occurred on all the major rivers of eastern Australia. The floods came on the back of a severe ENSO event that reached a maximum in Christmas 1972 (Bryant 1991). Following this, the turning on of Walker Circulation patterns brought two of the heaviest years of rainfall in eastern Australia. Bryant (1991) noted that the January 1974 floods represented, in terms of area, the largest natural disaster to occur in Australia to date, covering an area of over 3,800,000km2 from Alice Springs to the Pacific Ocean and from the Gulf of Carpentaria to the Murray River. During January, the active monsoon trough settled over northern Australia bringing torrential rainfall. This spawned the development of a series of deep tropical lows which traversed across the center of the continent, combining with Cyclone Wanda around the 25th of January, to provide ideal conditions for heavy, widespread and continuous rainfall over most of the eastern half of Australia.

Figure 3.2: The synoptic weather patterns for the 5th February 1967.

63 Wright (1989) analysed the synoptic climatology associated with winter rainfall in Victoria, identifying five distinct synoptic classes:

1. Interacting frontal; 2. Non-interacting frontal; 3. Post frontal; 4. Cold lows (cut-off lows); 5. Warm lows;

Interacting frontal systems were those where, at any atmospheric level, the cloudmass originated from tropical or sub-tropical regions before being retained in a frontal trough and moving south or southeast. The weather associated with these consisted of continuous rain well ahead of the front in northwesterly flow. These tended to bring the majority of rain to northern Victoria and less to other parts. Non-interacting frontal systems were those where no interaction with a tropical or sub tropical cloudmass occurred. Light rainfall or showers with the passage of the front characterize the weather associated with these. This synoptic type is common but only makes up a significant proportion of southern Victoria’s winter rainfall. Post- frontal rainfall occurred in the westerly air stream behind significant cold fronts and was found to be common in southern Victoria or in hilly areas where topographic enhancement occurred.

Cold lows, or cut-off lows, are characterized by a pool of cold air in the middle to upper atmosphere most frequently recognized through closed low pressure circulation at the 500hPa level. The surface circulation is most often cyclonic as well. Widespread rainfall often results from cold lows due to the upsliding of relatively warm air to the east of such a system. It was found that this type made up most of eastern Victoria’s winter rainfall total. Warm lows generally originate as heat lows over northern Australia that can extend or dip far enough south to affect northern Victoria, most frequently in the warmer months. The weather associated with warm lows is often characterized by heavy periods of rainfall and thunderstorms.

64 3.5 El Nino southern oscillation: flooding rainfall and stream flow in Australia

Worldwide the El Nino Southern Oscillation (ENSO) phenomenon is known to be one of the largest sources of natural climatic variability (Allen et al 1996). Allen et al (1996) suggests that climatic variability over much of the globe is influenced, on interannual time scales and, to varying degrees, by the ENSO phenomenon. The Bureau of Meteorology (2003) has also noted that the characteristic changes in the atmosphere that accompany those in the ocean result in altered weather patterns across the globe.

The ENSO phenomenon is the result of interactions between large-scale ocean and atmospheric circulation patterns (Chiew et al 1998). In detail, the term ‘El Nino’ refers to the presence of warm water along the coast of South America (eastern equatorial Pacific Ocean) whilst La Nina refers to the presence of cold-water upwelling in the same location. The ENSO phenomenon is part of a continuum where El Nino and La Nina events are simply the two opposite and extreme phases of the ocean-atmospheric system that has a tendency to cycle over a period of 2-10 years (Allen et al 1996). The Southern Oscillation Index (SOI) is a measure of the difference in atmospheric pressure between the Australian-Indonesian region (Darwin) and the eastern tropical Pacific Ocean (Tahiti). It is known to accurately reflect the underlying Sea Surface Temperature (SST) patterns. Positive SOI values mean a lower average air-pressure at Darwin than Tahiti, with airflow towards the Australian region, and a greater chance of above average rainfall and flooding (La Nina conditions) in many parts of Australia. Negative SOI values result in airflow towards the eastern Pacific region and a greater chance of below average rainfall and drought (El Nino conditions) in Australia.

Ropelewski and Halpert (1987) found that precipitation patterns could be directly related to the ENSO in the Pacific Ocean Basin, Australia, North America, South America, the Indian Sub- continent, Africa and one region in Central America. As a result of El Nino events parts of North America and Europe receive unusually hot weather, floods occur in South America and drought can dominate Australia and parts of Africa (Figure 3.5).

Figure 3.3: The areas that are most consistently affected by El Nino (Bureau of Meteorology 2003).

Whetton and Pittock (2001) state that Australia is one of the countries most affected by ENSO. Droughts occur during El Nino years and flooding rainfall most commonly occurs during La Nina phases (Pittock 1975 and Ropesewski and Halpert 1987). One of the most severe El Nino phases caused The Great World Drought of 1982-83 that affected the majority of Australia, Indonesia, India and South Africa (Bryant 1991). Many of Australia’s worst droughts have been followed by record floods, most often as a result of a sudden switch to La Nina conditions. Serious flooding over large areas followed the 1972-73 and 1982-83 droughts, which were both the result of severe El Nino conditions.

In detail for Australia, Ropesewski and Halpert (1987) found that four large regions showed ENSO related precipitation regimes: northern Australia; eastern Australia; southern Australia and Tasmania; and central Australia (Figure 3.6). The boundaries of the regions are based on phase differences in the ENSO harmonic vectors. For northern Australia, 22 out the 26 defined ENSO events were associated with dry conditions for September to March and with the disruption of the spring-summer monsoon. Ropelewski and Halpert (1989) found that high index phases of the Southern Oscillation (SO) were associated with enhanced precipitation, further demonstrating that tropical rainfall in Australia is intricately linked with the Southern Oscillation Index (SOI). For eastern Australia from February to January, 20 dry seasons out of 26 episodes were related to ENSO. The significance of the relationship in eastern Australia is

66 strengthened by the length of season over which the anomaly occurs, and by the area covered, often stretching from the sub-tropics through to mid-latitude regions (Figure 3.6). McBride and Nicholls (1983) found that the ENSO related precipitation signal is not well defined for areas east of the Great Dividing Range. In southern Australia and Tasmania, between May to October, eight of the 16 driest periods were ENSO years. Although the ENSO signal is not as strong in the far south, it is nevertheless significant. Ropesewski and Halpert (1987) state that it is significant that the central Australian region, which includes the Simpson and Great Western Deserts, shows a very consistent ENSO related rainfall relationship. They found that southwest Western Australia didn’t show a clearly defined ENSO precipitation regime. Significantly dryer periods did not occur during ENSO events.

Kane (1997) demonstrates the complexity of the response of the Australian climatic system to ENSO related phenomenon. Each year of the 120-year period from 1871-1990 was characterized as having an El Nino and/or Southern Oscillation Index minimum (SO), and/or warm or cold-water sea surface temperature anomalies in the equatorial eastern Pacific Ocean. In years where El Nino existed, the SO was at a minimum and SST warm water occurred between May and August, widespread drought resulted. When El Nino was present but the SO minimum and SST warm water was outside the period from May to August, droughts occurred in some parts with floods in others. Where cold SST water was present with no El Nino or SO minimum, widespread flooding generally resulted although some bizarre exceptions occurred with droughts instead of floods. Kane’s study shows that not all El Ninos are necessarily associated with periods of drought and that the relationships between ENSO, sea surface temperatures and the atmosphere are indeed complex.

Whetton and Pittock (2001) have analysed both the temporal and spatial evolution of the ENSO-rainfall relationship over Australia from 1889-1998, with an emphasis on multi-decadal changes. They performed 10-year correlations between rainfall and the SOI. They noted that during an ENSO warm event, anomalously low rainfall was received over northeastern Australia, leading to drought conditions. The significant finding of their study was that during periods of higher global mean temperatures the relationship between low rainfall and ENSO events weakened. This weakening of ENSO-rainfall relationship is first seen in northwestern

67 Australia before shifting to eastern and southeastern Australia and finally to northeastern Australia. The period from 1931-45 was a period where the interannual ENSO-rainfall relationship over northeastern Australia weakened, reversed and then rapidly recovered. It was thought that one of the reasons for this was that the SST anomalies in the tropical Pacific Ocean were weaker then compared with those averaged over all ENSO events.

Figure 3.4: The 4 regions that Ropesewski and Halpert (1987) found demonstrated ENSO related rainfall regimes. Arrow direction shows relationship season and whether correlation is positive or negative. EAU = Eastern Australia, NAU = Northern Australia, CAU = Central Australia, SWA = South Western Australia, SAT = Southern Australia and Tasmania.

In 1992 Stone and Auliciems found that little research had been undertaken relating ‘phases’ of the SOI with rainfall. Their work attempted to establish a method of identifying the life-cycle phases of the SOI and relating these to rainfall totals in eastern Australia. It was found that when the SOI was rising rapidly or was consistently positive, above median rainfall totals were recorded during autumn and spring. When the SOI was consistently negative, below average

68 rainfalls were recorded. As these SOI phases are actually related to rainfall amounts, in terms of probability, there is a direct application in using the index for agricultural management.

Nicholls and Kariko (1993) investigated the number, average length, and average intensity of rainfall events for five rainfall stations in eastern Australia (Figure 3.7). Average annual rainfall variations are also examined for significant trends over time. They found that the relationships with the SOI were all positive indicating that the SOI affects the number, length and intensity of rainfall events at all stations, and that it also influences annual totals. Correlations coefficients between 0.44 and 0.58 (significant at the 5% confidence level) were calculated for the annual totals (Table 3.1). Further significant correlations were found with the number of rainfall events and the intensity of rainfall during the events.

Figure 3.5: Rainfall sites from Nicholls and Kariko (1993)

69 Number Length Intensity Annual

Winton 0.35 0.13 0.27 0.44

Gatton-Lawes 0.33 0.12 0.24 0.47

Peak Hill 0.30 0.15 0.38 0.44

Canary Island 0.37 0.22 0.19 0.45

Carrick 0.19 0.37 0.41 0.58 Table 3.1: Correlations of annual SOI with number, length, and intensity of rain events and annual rainfall totals for 5 east Australian sites.

Chiew et al. (1998) presents an overview of the relationship between ENSO and rainfall, drought and stream flow in Australia. They found that the link between rainfall, streamflow and ENSO is statistically significant in most parts of the country. During the first half of the year, lag correlations were barely statistically significant. Lag correlations exist when the atmospheric response to sea surface temperatures lags behind by one or more months. The relationship was found to be strongest in the second half of the year, particularly over eastern parts of Australia where correlations between spring rainfall were greater than 0.4 and significant at the 5% level, with a lag of up to 3 months (lag 3). During summer only a small area of northeast Australia produced lag 1 correlations greater than 0.4 (p=0.05). For streamflow correlations with the SOI, the southeast coast/Murray region produced lag 2 correlation values of greater than 0.4 (p=0.05) during spring. The northeast and east coastal rivers had values greater than 0.4 (p=0.05) during the summer season for lag 1 and lag 3 streamflow-SOI correlations respectively. Chiew et al. (1998) concluded that ENSO indicators, such as the SOI, could be used with some success to predict spring runoff in southeast Australia and summer runoff in northeast and eastern Australia.

Puckeridge et al. (2000) looked at the hydrological persistence and the ecology of dryland rivers with particular emphasis on Cooper Creek. The 48-year hydrograph record for Currareva was analysed for the influence of ENSO. It was found that the effects of ENSO were apparent in the record for lagged correlations between discharge and the SOI. They found that flood events tended to be clustered around La Nina episodes. The highest statistically significant correlations (5% confidence level) were with a lag of 1 month with September recording the highest value of 0.52, with values of 0.45 and 0.43 for January and February. This indicates

70 that the linkage between SOI and discharge, for the Cooper Creek at Currareva, operates with a lag period of a month.

Overall, particularly over the majority of mainland Australia, the ENSO phenomenon has a more significant affect on rainfall patterns than other atmosphere-ocean modes. The Southern Annular Mode (SAM) is an atmosphere – ocean mode that influences Australian climatic patterns. The SAM is a ring of climate variability that encircles the South Pole and extends out to the latitudes of New Zealand and involves alternating changes in weather patterns and storm activity between the middle latitudes around New Zealand and southern Australia (40-50° S) (Renwick and Thompson 2006). Meneghini et al (2006) have shown the association between Australian rainfall and the SAM. They found that the SAM may have a greater inlfluence on inter-annual rainfall variability in areas affected by extra-tropical cyclones, cold fronts and westerlies (southern South Australia, Tasmania, southwest Western Australia, and Victoria) than areas affected by high-pressure systems and easterlies (Queensland and New South Wales). It appears that the SAM does not have a large influence on the monsoon or its related weather patterns such as tropical cyclones. Western Tasmania appears to be the main region where the SAM has a larger influence on seasonal rainfall variability than ENSO. Gillet et al (2006) have found that a positive phase of the SAM is associated with anomalously dry conditions in New Zealand and Tasmamia due to the southward shift of stormtrack. They also found that a positve phase results in increased easterly flow on to the eastern coast of Australia increasing rainfall in these areas. Hendon et al (2007) found a similar result. They suggest that 10-15% of weekly rainfall variability in the southwest and southeast of Australia during winter, and on the southeast coast during spring-summer, is explained by the SAM which is comparable to the variation explained by ENSO in these areas.

3.6 Conclusion

Flood events in dryland river basins throughout the world are generally the result of specific synoptic weather patterns. In tropical locations tropical depressions, cyclones (hurricanes in the northern hemisphere) and monsoon related troughs cause the majority of dryland river flood events. In mid-latitude more temperate locations, floods are generally the result of frontal

71 systems, cut-off lows and low-pressure troughs. However, many floods in temperate areas are also the result of tropical weather patterns that have traveled out of tropical zones. Near the margins of tropical areas floods can be the result of more temperate weather patterns. For example, a significant number of floods in the Lake Eyre basin of Australia are the result of easterly dip weather patterns. Much interaction between tropical and mid-latitude weather patterns occurs between these two zones and many flood events result from such interaction. The floods of 1974, that simultaneously flooded almost every river throughout the eastern half of Australia from Victoria to the Gulf of Carpentaria, were the result of tropical weather patterns that traveled down across much of Australia.

Australian dryland rivers are diverse and somewhat unique in their hydrological characteristics. Dryland rivers of Australia, together with those in South African, are amongst the most variable rivers in the world with extremely high coefficients of variation for annual discharges. Many such rivers are ephemeral, or have extremely low flow levels, and flow very infrequently. Dryland rainfall is highly variable with long dry periods separated by large but infrequent rainfall events. It has been assumed that high spatial variability accompanies high temporal variability. However there has only been one such study in Australia investigating this assumption and it has shown that annual, monthly and storm rainfall is remarkably uniform and not spatially diverse.

Australia is one of the countries most affected by ENSO. Generally, during El Nino conditions many areas receive below average rainfall and therefore below average stream flows and fewer and smaller flood events. In contrast to this situation, during La Nina conditions many areas receive well above average rainfall, higher than average stream flows and a higher number of larger flood events. The response of the Australian climate to ENSO is complex. There are El Nino years, or clusters of years, that are not associated with droughts and La Nina years that are not associated with flooding. However, there are other Atmosphere-ocean modes that influence Australian rainfall variability, such as the SAM, which influences areas in the south, southwest and southeast of Australia, and more dominantly in Tasmania.

72 Chapter 4

Synoptic Weather Patterns and Major Flooding of Australia’s Dryland Rivers

4.1 Introduction and objectives

Flooding of Australia’s dryland rivers is the result of particular synoptic patterns and processes interacting to produce heavy rainfall, generally over large areas. The aim of this chapter is to document and describe the type, seasonality, location and magnitude of the synoptic weather patterns that cause flooding.

4.2 Methods and data sources

Across Australia, a series of major dryland river basins were selected on the basis that they have an average annual rainfall of about 600mm or less, and stream gauges located in or strongly affected by a dryland contributing area. Of course many dryland areas in Australia are traversed by long river systems partly affected by conditions in their headwaters that may not be reflective of areas where the average rainfall is less than 600mm. However, such rivers often exhibit characteristics prescribed by the dryland areas through which they pass and are, therefore, of interest in this study. The rivers selected include those sourced entirely in dryland areas (autochthonous) as well as those sourced, in part, upstream of strictly dryland reaches of the catchment (allochthonous). Both types can have a profound impact on the hydrological and ecological conditions in dryland basins. On the basis of length of record and data quality, a total of 52 key stream gauges were selected from major river basins (Figure 1.1). The Departments of Land and Water Conservation, Sustainability and Environment, and Natural Resources and Mines supplied continuous stream gauge records for New South Wales, Victoria and Queensland, respectively, whilst the Departments of Lands Planning and Environment and the Western Australia Rivers Commission supplied records for the Northern Territory and Western Australia, respectively.

Flood events were selected from partial series data sets, where the number of (ranked- largest) independent flood events was set to equal the number of years of stream flow data. From this ranking only, floods that have an Annual Exceedence Probability (AEP) of 50% or less were analysed as part of this study. Only rated stream flow records were used for this analysis. Every effort was made to avoid selecting flow records on stream channels that have their flow regime altered by large water storage reservoirs. Where, due to a lack of alternative streams, this was unavoidable, only the floods larger in magnitude than the 20% AEP event (1:5 year flood) were selected. For this study, this is considered justifiable as the effect of the large water storage dam on the flow regime of regulated rivers, such as the Murrumbidgee, largely diminish with flood magnitudes greater than 20% AEP (Page 1988).

Flood peaks were determined to be independent of each other using a time minimum (Tmin) and a discharge minimum (Qmin). For this study these were set at 1440 minutes (Tmin) and 20 m3s-1 (Qmin), respectively. These parameters were found to be adequate in determining event independence in all the study streams apart from some highly regulated streams where the discharge maintained unnaturally high levels for extended periods due to water releases from large reservoirs. In these latter cases Tmin was retained at 1440 minutes but Qmin was increased to 100 m3s-1.

With the flood events selected, it was necessary to identify the dates when the flooding rains fell. This was achieved using the Commonwealth Bureau of Meteorology daily rainfall and weather stations across Australia, numbering over 16,000. These rainfall stations were overlain onto the Australian river basins data as supplied by the Australian Survey and Land Information Group (AUSLIG). Rainfall gauges were selected for each catchment using Arcview GIS. Rainfall records were then analyzed methodologically using the criterion of ≥5mm rainfall in 24hrs to identify the dates on which flooding rainfall fell. Daily read rainfall data was the only rainfall data that could be used for the identification of flooding rainfalls due the size of drainage basins and because pluviometer data is not widely available for dryland river basins in Australia. The standard reading time of 9.00am is used in all instances; however, some readings are taken over 2 or more days due to gauges only being read on weekdays. The delay between rainfall and river flow varied for each catchment depending on basin size,

channel form and topography. This delay or transmission time was found to be relatively easy to identify in dryland river catchments using a combination of rainfall and river flow records.

Synoptic situations for flood events were analysed using Map Browser 1.1, a computer-based application for browsing Australian weather maps from 1948. Mapbrowser 1.1 as produced by Robert Dahni from the Bureau of Meteorology, Australia is an interactive viewer of archived grids (e.g. National Centre for Environmental Prediction (NCEP) re-analysis datasets), tropical cyclones and synoptic types (e.g. Treloar and Stern 1993) over the Australian region. Flood events were selected from between 1948 and 2000 for this study.

The flooding rainfall dates have been classified based on the class of synoptic weather pattern(s) dominantly responsible for the rainfall. Mapbrowser 1.1 was used to view the NCEP reanalysis data set allowing the viewing of 6 hourly Mean Sea Level (MSL) atmospheric pressure, 500hPa height and outgoing longwave radiation (OLR). This allowed the identification and description of weather patterns that were responsible for the flooding rainfalls. The outgoing longwave radiation images, percentage cloud imagery from the National Oceanic and Atmospheric Administration (NOAA) Climate Diagnostics Center, and satellite imagery (GMS IR1) from the Bureau of Meteorology, all provided verification of synoptic weather patterns. Upper synoptic weather conditions are often crucial in determining the development and severity of a surface weather pattern. Favourable upper synoptic patterns allow moisture and heat to be transported to much higher levels, increasing instability and often the likelihood of rainfall and the total rainfall received. Therefore, the 500hPa charts were viewed to assess the role that the upper atmosphere played in specific synoptic weather-pattern situations.

Flooding rainfall events identified by date were classified into synoptic classes within each individual catchment and these were grouped together according to drainage division and gauge proximity. This allowed more accurate reporting, description, and comparison of the synoptic weather patterns that affect these specific hydrological regions, especially when some key river-gauge records contain a limited number of flood events. Flood magnitude, in peak flow rate (m3/s), was related to synoptic type to allow an assessment of which particular

synoptic pattern was responsible for each large flood event. Due to the synoptic scale at which weather patterns were observed it was often unavoidable that several catchments within the drainage division recorded floods from the same synoptic event. The two longest and most complete stream-flow records within each basin are presented to provide an understanding of how flood magnitude varies with synoptic type. For each stream-flow record, flood events are classified according to the month in which they occurred and are grouped into the same drainage divisions as the synoptic weather pattern analysis.

4.3 The synoptic weather types: Definition and description

The synoptic weather classes defined and used in this study are listed in Table 4.1. These classes were identified following a detailed literature review and a lengthy analysis of the synoptic weather patterns. Classes are specific to this study.

Class # Synoptic Class 1. Tropical lows/troughs 2. Deep tropical lows 3. Tropical cyclones 4. Northwest cloudbands 5. Frontal systems 6. Cut-off lows 7A Easterly dip (continental) 7B Easterly dip (offshore) 7C East coast low Table 4.1: Synoptic classes

4.3.1 Tropical troughs/lows

The meteorology of the Australia’s tropics and sub-tropics, particularly during the summer months, is dominated by two major semi-stationary low-pressure troughs in the trade-wind easterlies, one on each side of the continent. The Queensland trough is located west of the Great Dividing Range, approximately 700km from the coastline, and the west coast trough lies inland from the West Australian coastline (Sturman and Tapper 1996) (Figure 4.1). There are

two semi-permanent heat lows associated with the Queensland and west coast trough and these are known as the Cloncurry and Pilbara heat lows, respectively (Figure 1.7). The Queensland and west coast low-pressure troughs can become linked with the monsoon sheer line before developing into or promoting significant low-pressure trough development, generating extreme instability, heavy rainfall and flooding. This synoptic class has the following characteristics: • These troughs of low-pressure occur throughout the year over northern Australia but are more pronounced in the summer months when they are known to regularly dip south over eastern and western Australia (Adams 1986). • Tropical troughs/lows are distinct from deep tropical lows by retaining a visibly more open synoptic circulation pattern, retaining relatively higher central pressures (≥1000hPa), and being semi-stationary. • The Queensland trough is typically associated with a ridge of high-pressure near the coastline that directs a moist southeasterly air stream into the trough. This provides a distinct demarcation between relatively moist air to the east and much drier air to the west (Adams 1986). • The west coast trough exhibits considerably more day-to-day variability than the Queensland trough making temperature forecasting difficult. If the trough remains offshore for a number of days, temperatures of 40°C can be reached. However, if the trough is located onshore, sea breezes are likely to develop and penetrate well inland (Leslie and Skinner 1994). • A typical summertime Australian weather pattern consists of a broad area of low- pressure located over northern Australia with both the West Coast and Queensland Troughs in typical positions (Figure 4.1). • Surface troughs or low-pressure systems associated with the monsoon trough are generally shallow atmospheric features, often only extending to the 850hPa (1500m) level, and are replaced by upper level highs at the 700hPa level (3000m) (Sturman and Tapper 1996). However, it has been found that in many circumstances, upper troughs and/or lows, at the 500hPa level (5000m), overlie the surface troughs and low-pressure systems and lead to increased convective instability with enhanced rainfall (Figure 4.2). • Smaller scale low-pressure systems and thunderstorms, within these semi-stationary synoptic-scale low-pressure troughs, frequently result in relatively more localized

heavy rainfall and flooding at a catchment scale. These smaller scale lows can develop into tropical cyclones when they move over warm ocean waters such as the waters of the Gulf of Carpentaria or deep tropical lows over land.

4.3.2 Deep tropical lows

These are intense, often mobile, cyclonic depressions that have been described by a number of authors (Davidson and Holland 1987; Tapper and Hurry 1993). They predominantly develop and occur over land and have structures similar to that of the outer portions of tropical cyclones (McBride and Keenan 1982). They develop strong winds and can rapidly evolve into tropical cyclones over water (Foster and Lyons 1984). Their formation is always associated with, or embedded in, the monsoon trough in the summer months over northern Australia. This synoptic class has the following characteristics: • They often start out as a weak low or trough before deepening to form intense cyclonic depressions clearly identified on the weather charts. • In this study they are separated from tropical troughs/lows by attaining a central surface pressure of ≤1000hPa during some stage of their development and by being more mobile, often away from the monsoon trough. • Further to this deep tropical lows are often a standalone, more coherent weather pattern capable of sustaining very heavy rainfall, flooding and strong winds away from the monsoon sheer line. • Their presence normally dominates the northern (and at times the southern) Australian weather patterns. For example, in February 1974, a series of monsoon depressions brought record rainfall and flooding to many inland river systems, including those in southern Australia. These monsoon depressions moved away from the monsoon trough and developed a classic low-pressure pattern identified on weather charts.

Figure 4.1: Tropical trough/low-pressure system 1/3/1994

Figure 4.2: 500hPa upper synoptic chart 1/3/1994

• Deep tropical lows (often termed monsoon depressions) and tropical cyclones have similar atmospheric structures in that they are relatively warm cored at high atmospheric levels and relatively cold cored in the lower atmospheric levels. A study of the 500hPa charts demonstrates that monsoon depressions and tropical cyclones often have their closed circulation low-pressure counterparts at this level (Figure 4.4), with high-pressure at the 250hPa level aiding in air-mass evacuation. Sturman and Tapper (1996) state that with warm cored depressions, such as tropical cyclones and deep tropical lows, winds become anti-cyclonic in the middle to upper atmosphere. • The synoptic weather map for the 13th January 1984 has a deep tropical low positioned over central Australia (Figure 4.3). This low moved from northwest to the southeast of Australia over 7 days bringing widespread flooding to many inland locations. The synoptic map shows tight pressure gradients, low central pressures, and coherent structure even over land, which is characteristic of deep tropical lows.

4.3.3 Tropical cyclones and associated rainfall depressions

These are non-frontal relatively small-scale intense low-pressure systems that develop over warm ocean water. Pressure gradients within a tropical cyclone are very steep and wind speeds can reach 300km/h. McBride and Keenan (1982) have described the climatology of tropical cyclone formation in the Australian region. Evans and Allen (1992) have identified changes in monsoon and tropical cyclone activity around Australia in relation to El Nino/Southern Oscillation while Broadbridge and Hanstrum (1998) have described the relationship between tropical cyclones and the Southern Oscillation Index near Western Australia. Sturman and Tapper (1996) describe the development criteria for tropical cyclones and provide several case studies of tropical cyclones in the Australian region. • Tropical cyclones form over the warm waters (26-27°C) off the northern Australian coasts. • They typically have closed low-pressure circulation patterns along with very low central pressures, cyclonic winds and torrential rainfall.

Figure 4.3: Deep tropical low over Central Australia on the 13/01/1984

Figure 4.4: 500hPa synoptic chart 13/01/1984 showing well developed upper low

• They frequently decay to rainfall depressions over land where heavy rainfall, flooding and strong winds can continue for more than a week. • Rainfall depressions can bring flooding rainfalls as far south as southern NSW and Northern Victoria. • The cyclone season in Australia extends from December to May. • Upper synoptic patterns for the 500hPa level are similar to that of deep tropical lows with associated areas of closed low-pressure circulation immediately above the cyclone. • Tropical Cyclone Wylva was situated over the northern Territory on the 18th February 2001 before moving over central Western Australia (Figure 4.5). The 500hPa synoptic chart for cyclone Wylva shows an associated area of closed circulation low-pressure immediately above the cyclone (Figure 4.6).

Figure 4.5: Tropical Cyclone Wylva situated over the Northern Territory 18/2/2001

4.3.4 Northwest cloud bands

These consist of a large elongated band of middle and upper level cloud often stretching 5000km from Indonesia, across northwestern Australia through to southeastern Australia (Tapp and Barrel 1984; Tapper and Hurry 1993; Sturman and Tapper 1996). Eighty percent occur between April and September when the penetration of cold fronts or mid- latitude troughs towards the equator is most pronounced. • Their occurrence is associated with warm, moist tropical air moving poleward on the western limb of a high-pressure system affecting eastern and northern Australia. This air is forced to rise over colder, drier air in a mid-latitude trough or cold front that extends well into the tropics, typically to about 15° S. • Northwest cloudbands typically start to form when a deep cold front, with accompanying upper trough, affects Western Australia. • They have been known to bring heavy rainfall to northwestern, central and eastern Australia.

Figure 4.6: 500hPa synoptic chart 18/02/2001 associated with tropical cyclone Wylva

• Rainfall totals appear to be dependent upon sea surface temperatures and therefore not all northwest cloud bands result in substantial rainfall (Nichols 1989). • As a northwest cloudband forms in conjunction with a frontal system, cut-off low, or a mid-latitude trough, the upper atmospheric patterns are those associated with such surface patterns • Figure 4.7 shows the weather pattern responsible for the formation of a northwest cloudband associated with a cut-off low that is situated off the southwest coast of Western Australia. There was a strong high-pressure system centered in the eastern Bight drawing warm air down from northern Australia. An outgoing longwave radiation image for the 21/06/1980 at 00UTC shows a deep cloud layer that stretched from the northwest of Australia through into the Great Australian Bight (Figure 4.8).

4.3.5 Frontal systems

These are mid-latitude depressions that frequently develop in the westerly air stream south of the Australian mainland with their cold fronts affecting southern Australia. Wilson and Stern (1985) describe the synoptic and sub-synoptic aspects of frontal systems over southern Australia that result in the characteristic Australian summertime cool change. Wright (1989) provides a synoptic climatological classification of winter rainfall in Victoria. Frontal systems were found to be a major synoptic weather pattern leading to rainfall in Victoria. Smith et al. (1995) provide details of a study into cold fronts that penetrate well into central Australia during the northern dry season. Sturman and Tapper (1996) describe the climatology of frontal systems in the Australian region and how they differ from mid-latitude weather systems in the Northern Hemisphere.

Figure 4.7: Synoptic weather pattern for 21/06/1980 associated with NW cloudband

Figure 4.8: Outgoing longwave radiation image 21/06/1980 illustrating deep cloud cover

Frontal systems have the following characteristics: • Cold fronts result in significant temperature differences either side of the front and as such are known to produce intense thunderstorms and heavy though generally short- lived rainfall over southern and eastern Australia. • These systems often have a pre-frontal trough (easterly or westerly) producing a significant temperature change migrating ahead of the front, as was the case with the weather during the Ash Wednesday bushfires of 1983. • Pre-frontal westerly troughs have the potential to develop into significant weather systems (Hanstrum et al. 1990) bringing heavy rainfall. • They predominantly affect the weather of southern Australia, particularly in winter. • When significant rainfall is recorded, it is often the result of a cold front or mid-latitude depression interacting with a tropical weather air-mass to the north • Cold front penetration to latitudes of 15° S in winter is often observed, but little rainfall generally results at these latitudes. • Frontal systems have their upper atmospheric counterparts in the form of upper troughs of low-pressure as a result of the cold air associated with their passage (Figure 4.10). • A frontal system affected the southwest of Western Australia on the 19th of July 1995 brought strong winds and heavy rainfall with flooding (Figure 4.9).

4.3.6 Cut-off lows

These are synoptic-scale low-pressure systems that often extend into the middle to upper atmosphere. They typically become cut-off from a region of low pressure, further to the south (Sturman and Tapper 1996). They have closed circulation patterns both at the surface and 500hPa level occurring most commonly over the winter months. Leslie and Zhao (1999) describe the climatology of cut-off low-pressure systems as amongst the most important weather systems affecting southern Australia. Cut-off low-pressure systems have the following characteristics: • They commonly affect southern Australia, including southwest Western Australia but can also affect central parts of Western Australia.

• The preferred area of formation is off the southwest coast of Western Australia where they commonly form from deep troughs within the westerlies and can be accompanied by a cold front. They can also form in the Great Australian Bight before moving into Northern Victoria and southern NSW. • They can develop into an intense low-pressure system, most often in the Tasman Sea, with central pressures less than 990hPa. • Figure 4.11 shows a low situated over the southwest of Western Australia that was originally linked to the intense low-pressure system situated southeast of Tasmania. It can be seen that the high-pressure system slid underneath the low, essentially cutting it off. The 500hPa chart reveals a closed circulation low-pressure system that is associated with the surface low (Figure 4.12).

Figure 4.9: Synoptic weather pattern on the 17/071995 associated with a strong frontal system over Western Australia

Figure 4.10: 500hPa synoptic chart 19/07/1995

Figure 4.11: Synoptic weather pattern on 30/07/1986 associated with a cut-off low

4.3.7 Easterly dip and east coast low

This weather pattern consists of significant troughs and /or low-pressure systems that form in the easterly flow over eastern Australia, or the adjacent seaboard. The development and structure of easterly dips and east coast lows are closely related and are associated with the development of very heavy rainfall along the eastern seaboard of Australia (Sturman and Tapper 1996). Speer and Geerts (1994) provide a description of the two types of easterly dips: the onshore and offshore types. Holland et al. (1988) provides an overview and case study into Australian east coast cyclones. Leslie et al. (1987) undertook a study to try and improve the forecasting of rapidly developing and intense Australian east coast low-pressure systems. Hopkins and Holland (1997) provide a comprehensive climatology of Australian heavy-rain days and east coast cyclones from 1958-1992. Lynch (1987) documents the case study of an east coast low-pressure system in August 1986 that caused the then worst flood event in metropolitan Sydney.

Figure 4.12: 500hPa synoptic chart 30/07/1986 showing a deep upper low-pressure system

Easterly dips and east coast lows have the following characteristics: • Easterly dips form in the moist easterly flow associated with a strong high-pressure system often centered in the Tasman Sea/Pacific Ocean, in the Great Australian Bight or over continental areas of Australia. • Development is linked to influence of east coastal mountain range, and to a strong surface sea-surface temperature gradient associated with the warm Eastern Australian Current. • Their formation is almost exclusively associated with an upper level low-pressure system that is thought to often induce the surface low. • There is a general lack of tropical activity, in the form of a monsoon trough or associated heat lows, affecting the region receiving heavy rainfall from this synoptic type. • Easterly dips and east coast lows develop more frequently in winter due to higher land- sea temperature gradient. • Easterly dips can also occur in warmer months due to the position of high-pressure cells and the easterly winds they generate onto the Queensland/NSW coast. • Easterly dips are known to develop in three ways: I. Easterly dip (continental) develop over continental eastern Australia and can move further inland or offshore. The NSW floods of April 1990 resulted from this synoptic type (Figure 4.13). II. Offshore easterly dip-develops offshore and can move onshore or offshore (Figure 4.14). III. An east coast cyclone is an easterly dip that has developed into an intense cyclonic depression between 20°S and 40°S and within 500km of the coastline. An east coast cyclone must exhibit a component of movement parallel to the coastline and have a pressure gradient of at least 4 hPa/100km. During July 1984 an intense east coast cyclone moved along the NSW coastline (Figure 4.15). The well-developed 500hPa low-pressure system is shown in Figure 4.16.

Figure 4.13: An easterly dip (continental) over eastern Australia on 18/04/1990

Figure 4.14: An easterly dip (offshore) off the NSW/Queensland coastline on 31/07/1990

Figure 4.15: A well-developed east coast low on 27/07/1984

Figure 4.16: The 500hPa synoptic chart for 27/07/1984

4.4 Flood producing synoptic weather patterns: class, seasonality and flood magnitude

4.4.1 The Timor Sea division

For this study, rivers in the Timor Sea division include the Fitzroy, Ord and Victoria Rivers located in the northwest and far north of Australia (Figure 2.2). Tropical trough/lows (Class 1), deep tropical lows (Class 2) and tropical cyclones (Class 3), either on their own or in combination, have caused all of the flood events on these rivers (Figure 4.17). As mentioned previously, flood events for this analysis were those with an Annual Exceedance Probability (AEP) of 50% or less from the partial flood series. Out of a total of 149 such flood events, 32% were the result of tropical trough/lows with tropical cyclones resulting in 32%. Sixteen percent, a relatively low proportion, were caused by deep tropical lows. A combination of tropical trough/lows and tropical cyclones make up only 11% of the flood events. As discussed in the methods, broad scale synoptic weather patterns will cause simultaneous flooding of a number of large river systems within the drainage division.

The majority of the floods occurred between January and March with a small number between April and December (Figure 4.18). It is noteworthy that no flood events occurred over the cooler months from May to October.

Figures 4.19 and 4.20 show the three highest peak discharges for each synoptic class for the Fitzroy River at Me No Savvy and the East Banes River (tributary to the Victoria River) at Victoria Highway (Figure 2.2 and Table 2.1). The largest peak flood discharge for the Fitzroy River was recorded in February 1993, with a peak discharge of 1.19 m3s-1 km-2 (9,329 m3s-1). This flood was the result of a deep tropical low. Another large flood with a peak discharge 1.15 m3s-1 km-2 (8,979 m3s-1) occurred in February 1983 as a result of a tropical trough/low weather pattern. A tropical cyclone in January 1986 resulted in the third largest flood with a peak discharge of 0.70 m3s-1 km-2 (5,537 m3s-1). A number of smaller discharges were the result of tropical trough/lows, deep tropical lows or tropical cyclones.

The largest flood for the East Banes River, with a peak flood discharge of 0.60 m3s-1 km-2 (1,414 m3s-1), occurred in March 1964 and was the result of a tropical cyclone. In March 2001 a tropical cyclone caused a slightly smaller flood with a peak discharge 0.39 m3s-1 km-2 (919 m3s-1). A tropical trough/low in March 1976 caused the third largest flood on the East Banes River with a peak discharge of 0.38 m3s-1 km-2 (910 m3s-1). Tropical trough/lows and deep tropical lows caused a number of smaller but still significant flood magnitudes.

Timor Sea divsion 60

40 Number of Floods 30

Number of floods 10

g p v c pr uly u e ct Jan Feb Mar A May June J A S O No De Month

Figure 4.18: Flooding seasonality for the Timor Sea division rivers

Timor Sea Rivers

35 60 % By Synoptic Class Number of Flood Events 30 50 1. Tropical trough/low

25 2. Deep Tropical low 40 3. Tropical cyclone 20 4. Northwest cloudband 30 15 5. Frontal

Number of Flood Events 6. Cut of low 20 10 7. Easterly dip: Percentage By Synoptic Class Percentage By Synoptic A. Continental 10 B. Offshore 5 C. East coast low

0 0

1 2 3 4 5 6 2 3 3 7A 7B 7C 1& 1& 2& Synoptic Class

Figure 4.17: The synoptic weather patterns and flood events for the Timor Sea division rivers 95

Fitzroy River Me No Savvy Flood 1 Flood 2 1.4 Flood 3 1.2 1. Tropical trough/low 1 2. Deep Tropical low 3. Tropical cyclone

-2 0.8 4. Northw est cloudband

km 5. Frontal -1

s 6. Cut-off low 3 0.6 m 7. Easterly dip: A. Continental 0.4 B. Offshore C. East coast low 0.2

1 2 3 4 5 6 A B C 3 7 7 7 1& Synoptic Class

Figure 4.19: Flood Magnitude and Synoptic Class Fitzroy River at Me No Savvy

East Banes River

0.7 Flood 1 Flood 2 0.6 Flood 3

0.5 1. Tropical trough/low 2. Deep Tropical low -2 0.4 3. Tropical cyclone

km 4. Northw est cloudband -1 s 3 0.3 5. Frontal m 6. Cut-off low 0.2 7. Easterly dip: A. Continental 0.1 B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 3 7 7 7 1&3 2& Synoptic Class

Figure 4.20: Flood magnitude and synoptic class for the East Banes River (Victoria River catchment)

4.4.2 The Gulf of Carpentaria division

The major rivers selected for this study in the Gulf of Carpentaria drainage division include the Gregory (tributary to the Nicholson River), Leichardt, Flinders and Cloncurry Rivers (tributary of Flinders River) (Figure 2.6). Tropical trough/lows, deep tropical lows and tropical cyclones, either on their own or in combination, caused all of the 74 selected floods (Figure 4.21). Forty- nine percent of the total number of floods resulted from tropical trough/lows, 31% from deep tropical lows and 7% from tropical cyclones. The remaining 13% were the result of a combination of these synoptic classes.

The flood events are clustered around the warmer months of the year from November to March (Figure 4.22). February received the highest monthly total with 35% of the total of 74 floods. January received 31%, March 16%, December had a total of 15%, and November had just 2% (Figure 4.22).

Gulf of Carpentaria division v 30

15 Number of Floods

10 Number of Flood Events 5

n ly t a eb pr ay u ug ep c J F Mar A M June J A S O Nov Dec Month

Figure 4.22: Seasonality of flood events for Gulf of Carpentaria division rivers

Gulf Rivers

60 40 % By Synoptic Class Number of Flood Events

35 50

30 1. Tropical trough/low

40 2. Deep Tropical low 25 3. Tropical cyclone

30 20 4. Northwest cloudband

5. Frontal 15 20 6. Cut of low Number of Flood Events rrrrr 10 7. Easterly dip: Percentage By Synoptic Class lass Percentage By Synoptic 10 A. Continental 5 B. Offshore C. East coast low

0 0

1 2 3 4 5 6 A B C 7 7 7 &2 &3 &3 Synoptic Class 1 1 2

Figure 4.21: The synoptic weather patterns and flood events for the Gulf of Carpentaria division rivers

The two longest, most complete records for the Gulf of Carpentaria division are the Gregory River (tributary to Nicholson River) at Gregory Downs and the Cloncurry River (tributary to Flinders River) at Cloncurry (Figure 2.6 and Table 2.2). The largest flood on the Gregory River with a peak flood discharge of 0.28 m3s-1 km-2 (3,622 m3 s-1) occurred in December 2000 as a result of tropical trough/low weather pattern (Figure 4.23). A slightly smaller peak discharge of 0.26 m3s-1 km-2 (3,465 m3 s-1) occurred in March 1971 from a deep tropical low. The third largest flood on the Gregory River, with a peak discharge of 0.19 m3s-1 km-2 (2,554 m3 s-1), occurred in January 1971 as a result of a topical trough/low.

Gregory River Gregory Downs

0.3 Flood 1

0.25 Flood 2 Flood 3

0.2 1. Tropical trough/low

-2 2. Deep Tropical low

km 0.15 3. Tropical cyclone -1 s

3 4. Northw est cloudband

m 5. Frontal 0.1 6. Cut-off low 7. Easterly dip: 0.05 A. Continental B. Offshore C. East coast low 0

1 2 3 4 5 6 7A 7B 7C 2&3 Synoptic Class

Figure 4.23: Flood magnitude and synoptic class for the Gregory River at Gregory Downs

The largest flood on the Cloncurry River, with a peak discharge of 1.06 m3s-1 km-2 (6,393 m3 s-1), occurred in March 1997 from a deep tropical low (Figure 4.24). Another deep tropical low in March 1971 caused the second largest flood with a peak discharge of 0.57 m3s-1 km-2 (3,437 m3 s-1). In January 1991 a tropical trough/low caused a flood with a peak of 0.57 m3s-1 km-2 (3,425 m3 s-1). Several smaller peak discharges resulted from deep tropical lows and tropical trough/lows on their own and from a combination of synoptic classes 1&2, 1&3 and 2&3.

Cloncurry River Cloncurry

1.2 Flood 1 Flood 2 1 Flood 3

0.8 1. Tropical trough/low

-2 2. Deep Tropical low

km 0.6 3. Tropical cyclone -1 s

3 4. Northw est cloudband

m 5. Frontal 0.4 6. Cut-off low 7. Easterly dip: 0.2 A. Continental B. Offshore 0 C. East coast low

1 2 3 4 5 6 7A 7B 7C &2 &3 &3 1 1 2 Synoptic Class

Figure 4.24: Flood magnitude and synoptic class Cloncurry River at Cloncurry

4.4.3 The Lake Eyre division

The major rivers for this division include the Georgina, Diamantina and Coopers Creek (Figure 2.10). Smaller rivers include the Todd, the Rankin (tributary of the Georgina), and the Barcoo and Thomson Rivers (tributaries of Cooper Creek).

Tropical trough/lows, deep tropical lows and tropical cyclones have caused the majority of the 138 floods analysed for the Lake Eyre basin (Figure 4.25). Twenty six percent of all floods resulted from tropical trough/lows, with deep tropical lows and tropical cyclones making up 16% and 13%, respectively. A combination of these first three synoptic classes accounted for 31% of the remaining floods. Easterly dips-continental (Class 7B) account for 7% and a combination of tropical troughs/lows and northwest cloudbands (1&4), northwest cloudbands and easterly dips-continental (4&7A), and northwest cloudbands and frontal systems (4&5) account for just 3%.

100

Seventy percent of the 138 floods selected for the Lake Eyre division occurred in the summer months, with January having the highest total of 25% and February and March recording 23% and 22%, respectively (Figure 4.26). December and May both recorded 10% of the floods each, with April receiving 5%. It is interesting that August and September were the only months to record no floods, even June and July recording 2% and 1%, respectively.

Lake Eyre Division

40 35 Number of Floods 30 25 20 15 10 5 Number of Flood Events 0

n r y e y a eb pr n l ug ep ct ov ec J A u u O F Ma Ma J J A S N D Month

Figure 4.26: The seasonality of flood events for Lake Eyre Basin Rivers

The two longest and most complete stream-flow records for this study in the Lake Eyre division were Cooper Creek at Currareva and the Todd River at Anzac Oval (Figure 2.10 and Table 2.3). Cooper Creek is located in the eastern section of the Lake Eyre division with the Todd River running through the town of Alice Springs in the western most section of the division.

The largest flood magnitude for Coopers Creek at Currareva occurred in February 1974 and was caused by a combination of a deep tropical low and a tropical cyclone (synoptic Class 2&3) (Figure 4.27). This resulted in a peak discharge of 0.17 m3s-1 km-2 (25,467 m3 s-1), which is well above the second largest flood in March 1971 with a peak discharge of 0.07 m3s-1 km-2 (11,101 m3 s-1), as a result of a deep tropical low (Class 2). The third largest flood occurred in May 1955 as a result of an easterly dip (continental) (Class 7A) and had a peak discharge of

101

0.03 m3s-1 km-2 (5,138 m3 s-1). Tropical cyclones, easterly dip-continental, and a combination of tropical troughs/lows and deep tropical lows all resulted in moderate-sized floods. A number of smaller floods resulted from tropical trough/lows, and tropical trough/lows and tropical cyclones in combination.

Cooper Creek Currareva

0.18 Flood 1 0.16 Flood 2 0.14 Flood3 0.12 1. Tropical trough/low

-2 2. Deep Tropical low 0.1 3. Tropical cyclone km -1

s 4. Northw est cloudband

3 0.08

m 5. Frontal 0.06 6. Cut-off low 7. Easterly dip: 0.04 A. Continental 0.02 B. Offshore C. East coast low 0

1 2 3 4 5 6 3 7A 7B 7C &2 &3 & 1 1 2 Synoptic Class

Figure 4.27: Flood Magnitude and Synoptic Class for Cooper Creek Currareva

For the Todd River, deep tropical lows (Class 2) resulted in the two largest flood magnitudes with peak flood discharges of 2.70 m3s-1 km-2 (1,195 m3 s-1) in April 1988 and 2.03 m3s-1 km-2 (902 m3 s-1) in March 1983, respectively (Figure 4.28). In December 1973 a tropical trough/low (Class 1) caused the third largest flood with a peak discharge of 1.32 m3s-1 km-2 (587 m3 s-1). Smaller flood discharges resulted from tropical trough/lows and tropical cyclones (Class 1&3) and tropical trough/lows and deep tropical lows combined (Class 1&2).

102

Lake Eyre Division

30 50 Percentage by Synoptic Class 45 Number of Floods Events 25 40 1. Tropical trough/low

2. Deep Tropical low 35 20 3. Tropical cyclone 30 4. Northwest 15 25 cloudband

20 5. Frontal Number of Flood Events Percentage by Synoptic Class Percentage by Synoptic 10 6. Cut of low 15

7. Easterly dip: 10 5 A. Continental B. Offshore 5 C. East coast low

0 0

1 2 3 4 5 6 A B 2 3 3 4 5 7 7 C & & & & & 7 1 1 2 1 4 4&7A Synoptic Class

Figure 4.25: The Synoptic Weather Patterns and Flood Events for the Lake Eyre division rivers

103

Todd River at Anzac

3.0 Flood 1 Flood 2 2.5 Flood 3

2.0 1. Tropical trough/low

-2 2. Deep Tropical low 3. Tropical cyclone km

-1 1.5

s 4. Northw est cloudband 3

m 5. Frontal 1.0 6. Cut-off low 7. Easterly dip: 0.5 A. Continental B. Offshore 0.0 C. East coast low 2 3 4 A 1 2 3 4 5 6 7A 7B 7C & & & 7 1 1 1 4& Synoptic Class

Figure 4.28: Flood Magnitude and Synoptic Type for the Todd River at Anzac Oval

4.4.4 Bulloo-Bancannia division, Paroo and Warrego Rivers

The rivers of the Bulloo-Bancannia division (Bulloo River) (Figure 2.14) and those of northern inland Murray-Darling division (Paroo and Warrego Rivers) (Figure 2.18) were analysed as one group due to their morphological similarity and their adjoining localities. Tropical trough/lows (Class 1) make up 34% of the 90 flood events (Figure 4.29). Tropical cyclones (Class 3) and easterly dips-continental (Class 7A) were each responsible for 17% of floods. Deep tropical lows (Class 2) resulted in only 16% of the total. Combinations of synoptic classes produced the remaining floods. Synoptic Classes 1,2 and 3 in various combinations caused 11% of the floods while tropical trough/lows combined with northwest cloudbands (Class 4) caused 2%. Frontal systems (Class 5) and cut-off lows (Class 6) in combination resulted in 2% (Figure 4.29).

Fifty-seven percent of all the floods in the Bulloo, Paroo and Warrego Rivers were recorded in the months from December to February (Figure 4.30) and were due to tropical conditions. March and April combined recorded 21%, while May alone represented a slight rise, recording

104

16% of floods due to more temperate weather conditions. There were a small number of floods in the months of October and November in the buildup to the warmer months, and a few floods in July.

Bulloo, Paroo and Warrego River

15 Number of Floods

5 Number of Flood Events

ct Jan Apr uly O Feb Mar May June J Aug Sep Nov Dec Month

Figure 4.30: Flood seasonality in the Bulloo, Paroo and Warrego Rivers

The largest flood for the Bulloo River at Autumnvale (Figure 2.14), with a peak flood discharge of 0.06 m3s-1 km-2 (1,765 m3 s-1), occurred in January 1974 and was the result of a deep tropical low (Class 2) (Figure 4.31). A slightly smaller flood produced by a tropical trough/low (Class 1) occurred in February 1973 with a peak flood discharge of 0.055 m3s-1 km- 2 (1,621 m3 s-1). The third largest flood with a peak of 0.05 m3s-1 km-2 (1,455 m3 s-1) occurred in May 1989 as a result of an easterly dip (continental) (class 7A).

105

Bulloo, Paroo and Warrego Rivers

40 35 Percentage by Synoptic Class Number of Flood Events 35 30 1. Tropical trough/low

30 2. Deep Tropical low 25 3. Tropical cyclone 25 20 4. Northwest cloudband 20 5. Frontal 15 6. Cut of low 15

Number of Flood Events 7. Easterly dip: 10

Percentage By Synoptic Class Percentage By Synoptic 10 A. Continental B. Offshore C. East coast low 5 5

0 0

1 2 3 4 5 6 7A 7B 7C 1&2 1&3 2&3 1&4 5&6 Synoptic Class

Figure 4.29: Synoptic weather patterns and flooding in the Bulloo, Paroo and Warrego Rivers 106

The largest recorded flood in the Warrego River at Wyandra, which is located east of the Bulloo River (Figure 2.18 and Table 2.4), occurred in April 1990 as a result of an easterly dip (continental) with a peak discharge of 0.09 m3s-1 km-2 (3,975 m3 s-1) (Figure 4.32). In February 1997 a tropical trough/low (Class 1) synoptic pattern caused a large flood with a peak flood discharge of 0.08 m3s-1 km-2 (3,368 m3 s-1). The third largest flood, with a peak flood discharge of 0.07 m3s-1 km-2 (3,149 m3 s-1), was the result of a tropical cyclone (Class 3) and occurred in December 1971. Also, an easterly dip (continental) in May 1983 and a tropical trough/low in February 1973 resulted in two significant floods with peak flood discharges of 0.05 m3s-1 km-2 (2,148 m3 s-1) and 0.06 m3s-1 km-2 (2,442 m3 s-1), respectively.

Bulloo River at Autumnvale

0.07 Series1 0.06 Series2 Series3 0.05 1. Tropical trough/low

-2 0.04 2. Deep Tropical low 3. Tropical cyclone km -1

s 4. Northw est cloudband 3 0.03

m 5. Frontal 6. Cut-off low 0.02 7. Easterly dip: A. Continental 0.01 B. Offshore C. East coast low 0.00

1 2 3 4 5 6 3 7A 7B 7C & 1 Synoptic Class

Figure 4.31: Flood magnitude and synoptic type for the Bulloo River at Autumnvale

107

Warrego River at Wyandra

0.1 Flood 1 0.09 Flood 2 0.08 Flood 3

0.07 1. Tropical trough/low 0.06 2. Deep Tropical low

-2 3. Tropical cyclone 0.05

km 4. Northw est -1

s cloudband 3 0.04 m 5. Frontal 0.03 6. Cut-off low 0.02 7. Easterly dip: A. Continental 0.01 B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 3 3 4 6 7 7 7 & & & & 1 2 1 5 Synoptic Class

Figure 4.32: Flood magnitude and synoptic type for the Warrego River at Wyandra

4.4.5 The northeast Murray Darling drainage sub-division

This sub-division includes the Maranoa (tributary of the Balone River), Barwon, Gwydir, Namoi and the Bogan Rivers, making up the headwater streams of the Darling River (Figure 2.18). A large range of synoptic weather patterns caused the 85 floods recognised in this study for these river systems (Figure 4.33). Thirty-one percent were the result of tropical trough/lows (Class 1) while 12% were the result of tropical cyclones (Class 3) and easterly dip (continental) (Class 7A), respectively. Deep tropical lows (Class 2) caused 7%, easterly dip (offshore) 6% and frontal systems 4% (Figure 4.33). Because the sub-division straddles the summer and winter rainfall zones, a large range of combined synoptic weather patterns caused the remaining 29%. The most significant remaining floods were the product of tropical trough/lows and tropical cyclones (Classes 1&3), and northwest cloudbands and frontal systems (Classes 4&5).

108

Floods cluster around the warmer months from December to March with a successively smaller number of floods for each winter month from April to September, October recording no floods and November just two floods (Figure 4.34).

The largest flood by synoptic class for the Maranoa River (Figure 2.18 and Table 2.4) was produced by an easterly dip (continental) (Class 7A) in April with a peak flood discharge of 0.08 m3s-1 km-2 (1,603 m3 s-1) (Figure 4.35). The next largest flood in May 1983 was only half the size, and was the result of an easterly dip (continental) with a discharge of 0.04 m3s-1 km-2 (767 m3 s-1). The third largest flood, with a peak flood discharge of 0.03 m3s-1 km-2 (576 m3 s-1), resulted from a tropical cyclone in December 1971. Similar sized flood discharges were produced by tropical trough/lows, an easterly dip (offshore) and a combination of a frontal system and a cut-off low.

For the Bogan River (a tributary of the Macquarie) (Figure 2.18 and Table 2.4) the largest flood was produced by a deep tropical low in February 1955 with a peak flood discharge of 0.09 m3s-1 km-2 (2,546 m3 s-1) (Figure 4.36). This flood was almost twice as large as the second largest flood, with a peak flood discharge of 0.05 m3s-1 km-2 (1,379 m3 s-1), which occurred in November 1950 from a combination of a tropical trough/low and a deep tropical low. In April 1990 an easterly dip (continental) produced a flood of approximately the same size. A smaller magnitude flood resulted from a tropical cyclone whilst a large range of combination classes caused several smaller floods.

109

Northeast Murray-Darling Sub-Division

35 30

Percentage By Synoptic Class Number of Flood Events 1. Tropical trough/low 30 2. Deep Tropical low 25

3. Tropical cyclone 25 4. Northwest cloudband 20

5. Frontal 20

6. Cut-off low 15

15 7. Easterly dip: A. Continental B. Offshore

10 Number of Flood Events C. East coast low 10 Percentage By Synoptic Class Percentage By Synoptic

5 5

0 0

1 2 3 4 5 6 7A 7B 7C 7B 7C 1&2 1&3 1&4 2&4 4&5 4&6 5&6 1&7A 2&7A 5&7A 5&7B 6&7A 6&7B 7A& 7B& Syntopic Class

Figure 4.33: Synoptic weather patterns and flood events for the northeast Murray-Darling Rivers

110

Northeast Murray-Darling sub-division

18 16 14 12 10 8 Number of Floods 6

Number of Floods 4 2 0

b v c o e Jan Apr Oct Fe Mar May June July Aug Sep N D Month

Figure 4.34: The flood seasonality of Northeast Murray-Darling Rivers

Maranoa River

0.09 Flood 1 0.08 Flood 2 Flood 3 0.07

0.06 1. Tropical trough/low

-2 2. Deep Tropical low 0.05

km 3. Tropical cyclone -1

s 4. Northw est cloudband 3 0.04

m 5. Frontal 0.03 6. Cut-off low 7. Easterly dip: 0.02 A. Continental 0.01 B. Offshore C. East coast low 0

1 2 3 4 5 6 6 7A 7B 7C &3 & 1 5 Synoptic Class

Figure 4.35: Flood magnitude and synoptic type for the Maranoa River

111

Bogan River at Gongolgon

Flood 1 0.1 Flood 2 0.09 Flood 3 0.08 1. Tropical trough/low 0.07 2. Deep Tropical low

-2 0.06 3. Tropical cyclone 4. Northw est km 0.05 -1

s cloudband 3

m 0.04 5. Frontal 0.03 6. Cut-off low 7. Easterly dip: 0.02 A. Continental 0.01 B. Offshore C. East coast low 0

1 2 3 4 5 6 7A 5 7A 7B 7C 1&2 1&3 2&4 4& &7C 1& 5&7A5&7B 7B Synoptic Class

Figure 4.36: Flood magnitude and synoptic type for the Bogan River

4.4.6 Southern Murray-Darling sub-division

The rivers of the southern Murray-Darling sub-division include the Lachlan, Murrumbidgee, Avoca and Broken Rivers (all of which are tributaries to the Murray River) (Figure 2.18). This sub-division is strongly affected by winter and spring floods. The winter snow pack melting during the springtime contributes to flooding in those rivers that have their headwaters in above the snowline.

From a total of 38 floods, 18% were the result of cut-off lows (Class 6), 8% from frontal (Class 5), and 24% from cut-off lows and frontal patterns in combination (Classes 5&6) (Figure 4.37). Tropical trough/lows on their own, and tropical trough/lows and cut-off lows in combination (Class 1&6) both resulted in 11%. A small number of floods resulted from easterly dips (continental), easterly dip (offshore) and east coast lows on their own, and in various combinations with other synoptic weather patterns.

112

Flood frequency increases through the winter months before peaking in early to mid springtime (Figure 4.38). Out the total of 38 floods September recorded ten, October seven whilst June, July and August recorded four, five and four, respectively. Over the summer months both December and January recorded two floods with no floods occurring in February and November.

Flood Seasonaility of Southern Murray-Darling Rivers

10 Number of Floods 8

4 Number of Floods

b r y e e ar a Jan Ap Oct F M M Jun July Aug Sep Nov Dec Month

Figure 4.38: Flood seasonality of the Southern Murray-Darling Rivers (Lachlan, Murrumbidgee, Avoca, Broken Rivers)

The largest flood for the Avoca River (Figure 2.18 and Table 2.4) occurred in September 1983 from an easterly dip-continental (Class 7A) with a maximum daily discharge of 1.1 Mld-1 km-2 (5,090 Mld-1) (Figure 4.39). In September 1996 a cut-off low (Class 6) produced a flood with a maximum daily discharge of 0.90 Mld-1 km-2 (4,303 Mld-1). In June 1995 a flood of similar size, with a maximum daily discharge 0.89 Mld-1 km-2 (4,280 Mld-1), was produced by a combination of a frontal system and an easterly dip (continental) (Class 5&7A). There were a number of smaller magnitude floods produced by combined synoptic classes. These included cut-off lows and frontal systems (Class 5&6), and northwest cloudbands and cut-off lows in combination (Class 4&6).

113

Southern Murray-Darling Rivers Sub-Division

25 10 % By Synopitc Class Number of Flood Events 9

20 8 1. Tropical trough/low

7 2. Deep Tropical low

3. Tropical cyclone 15 6 4. Northwest cloudband 5 5. Frontal 10 4 6. Cut-off low 3 Number of Flood Events 7. Easterly dip: Percentage By Synoptic Class Percentage By Synoptic 5 2 A. Continental B. Offshore 1 C. East coast low

0 0

1 2 3 4 5 6 5 6 6 6 A A 7A 7B 7 7C 7 7C 1& 1& 4& 5& 5& A& 6& 7 Synoptic Class 5&

Figure 4.37: The synoptic weather patterns and flood events for the rivers of the southern Murray-Darling rivers 114

The largest flood for the Broken River (Figure 2.18 and Table 2.4), with a maximum daily discharge of 0.63 Mld-1 km-2 (2,096 Mld-1), occurred in December 1975 and was produced by a tropical trough/low (Figure 4.40). In September 1992 a flood with a maximum daily discharge of 0.58 Mld-1 km-2 (1,944 Mld-1) occurred as a result of a combination between a frontal system and a cut-off low. With a maximum daily discharge of 0.57 Mld-1 km-2 (1,904 Mld-1) the third largest flood occurred in September 1983 as the result of a cut-off low. A number of slightly smaller floods were the result of frontal systems and cut-off lows in combination, and a cut-off low on its own.

Avoca River at Quambatook

1.2 Flood 1 Flood 2 1 Flood 3

0.8 1. Tropical trough/low

-2 2. Deep Tropical low

km 3. Tropical cyclone

-1 0.6 4. Northw est cloudband

Mld 5. Frontal 0.4 6. Cut-off low 7. Easterly dip: 0.2 A. Continental B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 6 6 5 7 7 7 7A 4& 5& 1& & 5 Synoptic Class

Figure 4.39: Flood magnitude and synoptic class for the Avoca River

115

Broken River at Rices Weir

0.7 Flood 1 0.6 Flood 2 Flood 3 0.5 1. Tropical trough/low -2 2. Deep Tropical low 0.4 3. Tropical cyclone km

-1 4. Northwest cloudband 0.3 5. Frontal

Ml/d 6. Cut of low 0.2 7. Easterly dip: A. Continental 0.1 B. Offshore C. East coast low 0

1 2 3 4 5 6 6 6 7A 7B 7C 1& 5& Synoptic Class

Figure 4.40: Flood magnitude and synoptic class for the Broken River

4.4.7 The southwest coastal division

The main rivers of the southwest coastal division are the Avon, Murray, Blackwood, Pallinup and Lort Rivers all of which drain into the Southern Indian Ocean (Figure 2.23). This division is dominated by winter flooding. Frontal and cut-off low weather systems, either on their own or in combination, produced the majority of floods, 77% from a total of 69 flood events (Figure 4.41). Frontal weather systems (Class 5) contribute 29%, cut-off lows (Class 6) 26%, and in combination (Classes 5&6) 22%. A small number of flood events were the result of tropical trough/lows (Class 1), deep tropical lows (Class 2), tropical cyclones (Class 3), and from combinations of tropical trough/lows with either frontal systems (Classes 1&5) or cut-off lows (Classes 1&6).

The majority of the floods occurred over the winter months of June, July and August with monthly totals of 10, 26 and 11, respectively (Figure 4.42). January recorded a significant total of six flood events whilst December recorded no floods.

116

Southwest Coastal Divison

25 Number of Floods 20

5 Number of Flood Events 0

b t v c pr ug ep o e Jan A Oc Fe Mar May June July A S N D Month

Figure 4.42: Seasonality of flood events for the rivers of southwest Western Australia

The largest flood magnitude for the Lort River (Figure 2.23 and Table 2.5), with a peak flood discharge of 0.08 m3s-1 km-2 (229 m3 s-1), occurred in January 1999 and was the result of a deep tropical low (Figure 4.43). In November 1975 a cut-off low-pressure system produced a flood event with a peak flood discharge of 0.06 m3s-1 km-2 (169 m3 s-1). In June 1989 a cut-off low produced the third largest flood with a peak flood discharge of 0.05 m3s-1 km-2 (130 m3 s-1). The only other significant flood occurred in October 1992 and was the result of a tropical trough/low in combination with a frontal system, producing a peak flood discharge of 0.02 m3s- 1 km-2 (55 m3 s-1).

117

Southwest Coastal Division

35 25 % By Class Number of flood Events 30 20 1. Tropical trough/low 25 2. Deep Tropical low

15 3. Tropical cyclone 20 4. Northwest cloudband

15 5. Frontal 10 6. Cut-off low

10 Number of Flood Events

Percentage By Synoptic Class Percentage By Synoptic 7. Easterly dip: 5 A. Continental 5 B. Offshore C. East coast low

0 0

1 2 3 4 5 6 7A 7B 7C 1&5 1&6 5&6 Synoptic Class

Figure 4.41: Synoptic weather patterns and flood events for southwest coastal rivers of Western Australia 118

The largest flood on the Murray River (Figure 2.23 and Table 2.5) occurred in August 1964 with a peak flood discharge of 0.08 m3s-1 km-2 (553 m3 s-1) and was produced by a frontal system (Figure 4.44). The second largest flood, with a peak flood discharge of 0.07 m3s-1 km-2 (492 m3 s-1), occurred in July 1958 as a result of a cut-off low and a frontal system. A flood of similar magnitude occurred in August 1955 as a result of a frontal system, with a peak flood discharge of 0.06 m3s-1 km-2 (478 m3 s-1). A moderate sized flood, with a peak flood discharge of 0.05 m3s-1 km-2 (347 m3 s-1), occurred in January 1982 as a result of a tropical cyclone. Frontal systems and cut-off lows on their own, or in combination, produced a number of small to medium floods.

Lort River at Fairfield

0.090 Flood 1 0.080 Flood 2 0.070 Flood 3 0.060 1. Tropical trough/low

-2 2. Deep Tropical low 0.050 3. Tropical cyclone km -1

s 4. Northw est cloudband

3 0.040

m 5. Frontal 0.030 6. Cut-off low 0.020 7. Easterly dip: A. Continental 0.010 B. Offshore C. East coast low 0.000

1 2 3 4 5 6 6 7A 7B 7C &5 &6 & 1 5 1 Synoptic Class

Figure 4.43: Flood magnitude and synoptic class for the Lort River (Fairfield)

119

Murray River at Baden Powell

0.09 Flood 1 0.08 Flood 2 0.07 Flood 3

0.06 1. Tropical trough/low -2 0.05 2. Deep Tropical low

km 3. Tropical cyclone -1 s

3 0.04 4. Northw est cloudband

m 5. Frontal 0.03 6. Cut-off low 0.02 7. Easterly dip: A. Continental 0.01 B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 6 7 7 7 5& Synoptic Class

Figure 4.44: Flood Magnitude and synoptic class for the Murray River (Western Australia)

4.4.8 Indian Ocean division: Greenough, Murchison and Gascoyne Rivers

The Indian Ocean division (Figure 2.27) is extremely large, covering 518,000km2, and for the purposes of the study is divided into two regions. The division stretches over approximately 10° north-south and is such affected by differing synoptic weather patterns. The southern region extends from the Greenough through the Murchison to the Gascoyne. Tropical cyclones produced 29% of the total of 79 floods (Figure 4.45). Cut-off low-pressure systems produced 18%, while tropical trough/lows, northwest cloudbands, and deep tropical lows caused 14%, 11% and 10% of the floods respectively. Cut-off lows in combination with northwest cloudbands produced 9%, and frontal systems on their own produced 5%.

The seasonality of floods is clearly divided into two distinct periods throughout the year (Figure 4.46). The first is a summer mode from January to March, inclusive. The second is a winter mode from April through until the end of July. The months from August until the end of the year recorded few floods.

120

Indian Ocean Division: Greenough, Murchison and Gascoyne Rivers

35 30 % By Class Number of flood Events 30 25 1. Tropical trough/low

25 2. Deep Tropical low 20 3. Tropical cyclone 20 4. Northwest cloudband 15 15 5. Frontal

6. Cut-off low 10

10 Number of Flood Events 7. Easterly dip:

Percentage By Synoptic Class Percentage By Synoptic A. Continental 5 B. Offshore 5 C. East coast low

0 0

1 2 3 4 5 6 7A 7B 7C 4&5 4&6 5&6 Synoptic Class

Figure 4.45: Synoptic weather patterns and flood events for the Greenough, Murchison and Gascoyne Rivers of Western Australia

121

Indian Ocean Division: Greenough, Murchison and Gascoyne Rivers

18 16

14 Number of Floods 12 10 8 6 4 2 Number of Flood Events 0

n b t e pr ne ly ug ep c Ja A u O F Mar May Ju J A S Nov Dec Month

Figure 4.46: Seasonality of flood events for the Greenough, Murchison and Gascoyne Rivers of Western Australia

The largest flood on the Greenough River (Figure 2.27 and Table 2.6) occurred in March 1971, with a peak flood discharge of 0.03 m3s-1 km-2 (411 m3 s-1), as a result of a tropical cyclone (Figure 4.47). Another tropical cyclone in May 1988 produced the second largest flood with a peak flood discharge of 0.025 m3s-1 km-2 (320 m3 s-1). In January 1994 a cut-off low-pressure system produced the third largest flood with a peak flood discharge of 0.02 m3s-1 km-2 (237 m3 s-1). A number of smaller flood events were the result of a northwest cloudband combined with a frontal system, frontal systems, and cut-off low-pressure systems.

The largest flood magnitude for the Gascoyne River (Figure 2.27 and Table 2.6), with a peak flood discharge of 0.075 m3s-1 km-2 (5,509 m3 s-1), occurred in June 1980 as a result of a northwest cloudband in combination with a cut-off low (Figure 4.48). In February 1960 a deep tropical low produced a flood with a peak discharge of 0.071 m3s-1 km-2 (5,245 m3 s-1) while in February 1961 a tropical cyclone produce a similar sized flood with a peak discharge of 0.071 m3s-1 km-2 (5,227 m3 s-1). Several moderate sized floods were the result of tropical cyclones, cut-off low-pressure systems, and northwest cloudbands combined with cut-off low-pressure systems.

122

Greenough River at Karlenew Peak

Flood 1 0.035 Flood 2 Flood 3 0.03

0.025 1. Tropical trough/low 2. Deep Tropical low

-2 3. Tropical cyclone 0.02 4. Northwest cloudband km

-1 5. Frontal s

3 0.015 6. Cut-off low

m 7. Easterly dip: 0.01 A. Continental B. Offshore 0.005 C. East coast low

1 2 3 4 5 6 5 6 7A 7B 7C 4& 4& Synoptic Class

Figure 4.47: Flood magnitude and synoptic class for the Greenough River

Gascoyne River at Nine Mile Bridge

0.08 Flood 1 0.07 Flood 2 Flood 3 0.06 1. Tropical trough/low -2 0.05 2. Deep Tropical low 3. Tropical cyclone km 0.04 -1 4. Northwest cloudband s 3 5. Frontal

m 0.03 6. Cut-off low 0.02 7. Easterly dip: A. Continental 0.01 B. Offshore C. East coast low 0

1 2 3 4 5 6 7A 7B 7C 4&6 Synoptic Class

Figure 4.48: Flood magnitude and synoptic class for the Gascoyne River

123

4.4.9 The Indian Ocean: Pilbara Rivers

This region includes the De Gray, Fortescue and Ashburton Rivers from the northwest of western Australia (Figure 2.27). Tropical cyclones resulted in the highest number of floods with 43% of the total of 58 floods (Figure 4.49). Tropical trough/lows produced 28% and deep tropical lows 12%, while tropical trough/lows and tropical cyclones in combination accounted for 9%. Northwest cloudbands produced 3% on their own and 2% when in combination with both cut-off lows and tropical cyclones.

The majority of the floods in the Pilbara Region rivers occurred during the warmer months from January through until the end of March with February recording the highest monthly total of 22, which is 39% of the annual total (Figure 4.50). Monthly totals drop steadily, with autumn and winter months only recording a few floods and the period from September to the end of November receiving no floods.

The largest flood for the Ashburton River occurred in December 1975 and was produced by a tropical cyclone attaining a peak flood discharge of 0.11 m3s-1 km-2 (4,883 m3 s-1) Figure 4.51). In May 1971 a tropical trough/low-pressure system produced a flood with a peak discharge of 0.06 m3s-1 km-2 (2,645 m3 s-1). The third largest flood, with a peak discharge of 0.05 m3s-1 km-2 (2,156 m3 s-1), occurred in January 1987 and was produced by a tropical cyclone. A number of smaller floods resulted from tropical trough/lows, deep tropical lows, tropical cyclones, and tropical cyclones combined with northwest cloudbands.

The largest flood on the De Gray River, with a peak discharge of 0.18 m3s-1 km-2 (8,800 m3 s-1), was the result of a deep tropical low in March 1988 (Figure 4.52). In March 1987 and January 1983 tropical cyclones produced floods with peak discharges of 0.14 m3s-1 km-2 (6,843 m3 s-1) and 0.13 m3s-1 km-2 (6,223 m3 s-1), respectively. Tropical cyclones were responsible for another four moderate sized floods. A number of smaller floods resulted from tropical trough/lows, and deep tropical lows and tropical cyclones in combination.

124

India Ocean Division: Pilbarra Rivers

50 30 % By Class 45 Number of flood Events 25 40 1. Tropical trough/low

35 2. Deep Tropical low 20 30 3. Tropical cyclone

25 15 4. Northwest cloudband

5. Frontal 20 10 6. Cut-off low 15 Number of Flood Events

Percentage By Synoptic Class Percentage By Synoptic 7. Easterly dip: 10 A. Continental 5 B. Offshore 5 C. East coast low

0 0

1 2 3 4 5 6 4 7A 7B 7C 1&3 2&3 4&6 Synoptic Class 3 &

Figure 4.49: Synoptic weather patterns and flooding of the De Gray, Fortescue and Ashburton Rivers

125

Indian Ocean Division: Pilbara Rivers 25

Number of Floods 10

Number of Floods of Floods Number 5

b r y y t v c Jan Apr Oc Fe Ma Ma June Jul Aug Sep No De Month

Figure 4.50: Seasonality of flood events for the De-Gray, Fortescue and Ashburton Rivers

Ashburton River at Capricorn Range

0.12 Flood 1 Flood 2 0.1 Flood 3

0.08 1. Tropical trough/low

-2 2. Deep Tropical low 3. Tropical cyclone km 0.06 -1 4. Northw est cloudband s 3

m 5. Frontal 0.04 6. Cut-off low 7. Easterly dip: A. Continental 0.02 B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 3 7 7 7 1& 3&4 Synoptic Class

Figure 4.51: Flood magnitude and synoptic class for the Ashburton River

126

De Gray River Coolenar Pool

0.25 Flood 1 Flood 2 0.2 Flood 3

1. Tropical trough/low -2 0.15 2. Deep Tropical low

km 3. Tropical cyclone -1 s

3 4. Northw est cloudband m 0.1 5. Frontal 6. Cut-off low 7. Easterly dip: 0.05 A. Continental B. Offshore C. East coast low 0

1 2 3 4 5 6 A B C 3 7 7 7 2& Synoptic Class

Figure 4.52: Flood magnitude and synoptic class for the De Gray River

4.5 Discussion

4.5.1 Synoptic weather patterns responsible for frequency of flooding

The synoptic weather patterns that result in flood events vary widely around Australia, from the hot tropical north where floods are the result of monsoon-related synoptic weather systems to southern Australia where floods result from mid-latitude weather systems.

In detail, tropical trough/lows, deep tropical lows and tropical cyclones, either individually or in combination, cause all the floods that affect the dryland rivers of both the Timor Sea and Gulf of Carpentaria drainage divisions in the far north of Australia. These occur during the warmer months from December to April, with February recording the highest number of floods. In both divisions tropical trough/lows result in the highest number of flood events, although only by a small margin in the Timor Sea division. These tropical trough/lows are related to the active monsoon as the monsoon trough moves southward over northern Australia

127

bringing heavy convective rainfall. Deep tropical lows, which are elsewhere referred to as monsoon depressions or tropical lows (Tapper and Hurry 1993; Davidson and Holland 1987; Reinhard and Smith 1998), most often form over land in association with the monsoon trough within which they are frequently embedded. Most deep tropical lows tend to remain near the monsoon trough although it is not uncommon for them to drift slowly and coherently across the entire continent before decaying out to sea in the south. Although deep tropical lows have the same basic structure as a tropical cyclone, their ability to develop and intensify over land makes them significant flood producing weather patterns. The role of the monsoon, which in Australia tends to be erratic, is therefore the main driver of flood events over far northern Australia.

In terms of flood producing weather patterns, the main difference between the Timor Sea and Gulf of Carpentaria drainage divisions, is that in the former a much greater number of floods result from tropical cyclones, 23% compared to just 7%. This pattern is further exemplified by virtue of the number of floods in the Timor Sea division that result from a combination of synoptic weather patterns that involved tropical cyclones, frequently with tropical trough/lows. In the Timor Sea division, tropical cyclones combined with deep tropical lows and tropical trough/lows produced a further 17% of the floods while in the Gulf of Carpentaria division, this figure was only 9%. The high number of flood events as a result of tropical cyclones is not restricted to the Timor Sea division. The rivers of the Indian Ocean division further to the south in the Pilbara and Greenough-Gascoyne region also have a high proportion of floods resulting from tropical cyclones. Sturman and Tapper (1996) state that the areas of actual tropical cyclone generation are located in the Coral Sea off northeast Australia, in the western Gulf of Carpentaria and off the northwest coast, with equal numbers of cyclones generated in each region. It is apparent that tropical cyclones tend to be more efficient flood-producing synoptic weather patterns in the northwest of Australia relative to northeast and eastern parts of Australia. Relative to the Timor Sea division, the Gulf of Carpentaria division has a much higher proportion of floods resulting from tropical trough/lows and deep tropical lows. Eighty percent of the floods were the result of either tropical trough/lows or deep tropical lows. In fact, the Gulf of Carpentaria division has the highest percentage of both the tropical trough/lows and deep tropical lows than any other Australian drainage division containing

128

dryland rivers. The Lake Eyre drainage division also has a high number of floods resulting from tropical trough/lows (27% of the total) and deep tropical lows (17%) with a small number resulting from tropical cyclones (14%). Tropical cyclones make up a slightly higher proportion of floods than deep tropical lows in the Bulloo-Paroo-Warrego and Northeast Murray-Darling Rivers, although tropical trough/lows still result in the highest number of flood events by a large margin in both divisions.

In the Lake Eyre division, compared to the Timor Sea and Gulf of Carpentaria divisions further to the north, there is a larger range of flood-producing weather systems. Tropical weather patterns are still by far the dominant cause of floods, with tropical trough/lows, deep tropical lows and tropical cyclones making up 90% of the total. In the Lake Eyre division tropical trough/lows tends to come in two forms. The first is the southward movement of the monsoon trough, associated with an active monsoon circulation, into the most northern parts of the Lake Eyre division. This southward movement of the monsoon trough can also affect the very northern parts of the Pilbara region. Secondly, tropical trough/lows develop from the semi- permanent Queensland Trough and its associated heat low, the Cloncurry Low. The West Coast Trough, with the associated Pilbara Low, is analogous to the Queensland Trough in that tropical trough/lows that affect Western Australia are typically the result of this synoptic situation.

Afternoon thunderstorms frequently result from the deepening of the Queensland or West Coast Troughs as a result of intense surface heating during the day (Adams 1986). These thunderstorms can become well organized and long lasting under favorable conditions, providing intense rainfall over a sufficiently large area to generate a large flood. This is demonstrated in the Indian Ocean division where all the Pilbara Rivers (De Gray, Ashburton and Fortescue Rivers) and to a lesser extent Greenough, Murchison and Gascoyne Rivers have a significant number of floods resulting from tropical trough/lows, with many of these being southward extensions of the West Coast Trough. The Queensland Trough and its southward extension results in floods in the Lake Eyre division, Bulloo-Paroo-Warrego region and the Murray-Darling division, although to a lesser degree in the south of this division. More often than not, over inland Queensland and Northern New South Wales, troughs of low-pressure in

129

the upper atmosphere (500hPa level) promote further instability associated with these tropical trough/lows.

Easterly dips (continental) result in a number of floods within the Lake Eyre division, an interesting result considering that easterly dips are generally regarded as resulting in mostly coastal rainfall. The moisture laden easterly flow associated with an easterly dip is sourced well out in the Northern Tasman Sea or Southern Pacific Ocean before being transported large distances inland to bring floods over the Lake Eyre division. Upper troughs or low-pressure cells are found to accompany all the easterly dips (all types). These upper lows are thought to often induce and steer the surface system and appear to be related to and located just equator- wards of a strong high-pressure system, particularly if that high remains nearly stationary for a number of days. The strength of the high-pressure system, together with its stationary nature, appears to result in the upper atmosphere being cooled to the point that an upper low can form. Strong high-pressure systems are most common during autumn and winter and therefore easterly dips (all types) are most common during the colder months from April to June, with May recording the highest number of easterly dips. This seasonality of easterly dips is demonstrated in the Lake Eyre and Bulloo-Paroo-Warrego Rivers where a secondary peak of flood numbers occurred in May and to a lesser extent in April. It is well noted that although 90% of the floods in the Lake Eyre division result from monsoon related weather patterns, little attention is given to the fact that eight percent of floods, often very large floods such as April 1990, result from easterly dips that are not directly related to the monsoon.

Easterly dips (continental) are also a prominent flood producing weather pattern in the adjacent Bulloo, Paroo and Warrego Rivers and in the Northeast Murray-Darling Rivers, but are less significant in the Southern Murray-Darling Rivers. Seventeen percent of the floods in the Bulloo, Paroo and Warrego Rivers and 12% in northeast Murray-Darling division are a result of easterly dips (continental). This is the highest total percentage of floods, for all the drainage divisions, resulting from this synoptic class. Easterly dips (offshore) are only significant flood producing weather patterns in the Northeast Murray-Darling Rivers although it must be noted that the ability to decipher between easterly dip (continental) and (offshore) is at times

130

difficult. Easterly dip (east coast lows) resulted in a relatively insignificant 3% percent of total floods in the southern Murray-Darling Rivers.

Northwest cloudbands resulted in 12% of the floods in the Greenough, Murchison and Gascoyne Rivers of Western Australia. In this region, a further 11% of floods were the result of northwest cloudbands combining with cut-off lows that were frequently situated off the southern coast of Western Australia. Northwest cloudbands also combined with frontal systems in this region, something that is to be expected due to the fact that northwest cloudbands result in part from eqator-ward penetration by frontal systems. Tapp and Barrel (1984) stated that 80% of all northwest cloudbands occur between April and September. This study confirms that the vast majority of northwest cloudbands occur between May and August. This is well demonstrated in the Greenough, Murchison and Gascoyne Rivers, which recorded the highest number of northwest cloudbands, where the peak in flood events from May to July is in part due to these.

In the eastern regions of Australia only the Lake Eyre division recorded a flood resulting solely from a northwest cloudband, and that occurred in the month of July. However, northwest cloudbands in combination with other synoptic classes, such as tropical trough/lows and easterly dip-continental, caused a small number of floods in the Lake Eyre division, the Bulloo- Bancannia division, and the Paroo and Warrego Rivers. In the northeast and southern Murray- Darling divisions, northwest cloudbands combined with tropical trough/lows, frontal systems and cut-off lows to produce a small number of floods.

Frontal systems in combination with cut-off lows result in only 2% of floods in the Bulloo, Paroo and Warrego Rivers, and no floods in the Lake Eyre division. In the northeast Murray- Darling region frontal systems on their own result in 4% of floods, as where frontal systems combining with cut-off lows and easterly dips (continental) and (offshore) result in 6%. Also, the northeast Murray-Darling region recorded the largest range of synoptic weather patterns that cause floods, with 15 different combination synoptic classes.

131

In the Greenough, Murchison and Gascoyne Rivers, 18% of floods result from cut-off lows and 5% from frontal systems. This is significant, given that these rivers are situated approximately 600kms north of the southwest corner of Western Australia where frontal systems are frequent. This is not totally surprising; Leslie and Zhao (1999) stated that the most favorable location for formation of cut-off lows is in the southwestern region of Western Australia. As part of the climatic survey of the Gascoyne Region, the Bureau of Meteorology (1998) stated that at times a low can form further north than is normal and become cut-off in the easterly flow. The resultant rainfall is often significant, especially when interacting with tropical cloudbands. It was found that flooding in the Gascoyne region often results when cut-off lows interact with northwest cloudbands. Surface cut-off lows were always accompanied by an upper trough or low-pressure cell that appeared to steer the surface pattern. At times it was difficult to resolve whether the rainfall was dominantly the result of the surface or upper pattern. Once formed, cut-off lows tended to track east or southeastwards across the continent, generally remaining coherent before passing into the Tasman Sea off eastern Australia.

Cut-off lows, along with frontal systems, were found to be highly significant flood producing weather patterns in both the southern Murray-Darling Rivers and the rivers of the southwest coastal division. The southern Murray Darling had many more floods resulting from cut-off lows whereas the southwest coastal division had more floods resulting from frontal systems, but only by a small margin. Both regions had a large number of floods caused by a combination of frontal systems and cut-off lows. In these cases lows embedded in the westerly airflow associated with the frontal system would develop to the north or in front of the frontal system, before being ‘cut-off’ from the parent low to the south. In terms of seasonality frontal systems and cut-off lows occurred during the winter months of June, July and August, almost exclusively in the southwest coastal division of Western Australia. This is contrasted by the rivers of the southern Murray-Darling region, where they occurred most prominently in the months of September and October. This peak in flooding during springtime for the southern Murray-Darling Rivers is partially the result of snowmelt providing river systems with a high base flow.

132

4.5.2 Flood magnitude by synoptic type

The largest floods generally result from synoptic classes that feature frequently in that particular division. However, this was not always the case. For example, the largest flood in the Broken River catchment (southern Murray-Darling sub-division) was the result of a tropical trough/low, not a mid-latitude weather pattern that is more common this far south.

The tropical synoptic classes, associated with the monsoon, not surprisingly, resulted in all the large flood events in the far northern divisions along with many of the largest floods in other divisions. In fact, in every drainage division a number of the largest floods resulted from a tropical trough/low, a deep tropical low or a tropical cyclone, either on their own or in combination with other synoptic classes. There were a higher number of the largest floods caused by deep tropical lows and tropical trough/lows in the Gulf of Carpentaria division, Lake Eyre division and the Bulloo, Paroo and Warrego Rivers. In contrast, tropical cyclones cause the majority of the largest flood events in the Pilbara region, the Timor Sea division and in the Greenough, Murchison and Gascoyne Rivers. Central and northern dryland rivers of Western Australia, relative to other regions of Australia, receive a higher number of floods as a result of tropical cyclones, and these floods are often larger in magnitude.

Easterly dips (continental) caused the largest flood magnitudes in the Warrego River (Bulloo- Paroo-Warrego Region), the Condamine River in the Northeast Murray-Darling, and the Avoca River in the Southern Murray-Darling region. This reflects the effectiveness of this synoptic class in bringing heavy widespread rainfall southwards, sufficient to generate very large floods in dryland rivers. This class also occurred relatively frequently within each of these divisions. The fact that easterly dips (continental) are always associated with upper troughs or lows means that instability extends from the surface up through a deep layer of the atmosphere.

Despite the fact that in the Avoca River, in Victoria, the largest flood was a result of an easterly dip (continental), and in the Broken River the largest flood resulted from a tropical trough/low, both rivers had several large floods resulting from cut-off lows and frontal systems, as single events and in combination. Both rivers also had large floods resulting from various

133

combinations of tropical trough/lows with either frontal systems or cut-off lows. This, together with the fact that mid-latitude weather patterns didn’t necessarily result in the largest floods, demonstrates the importance of tropical weather patterns to flooding in the southern Murray- Darling Rivers.

The importance of tropical weather patterns to floods in southern catchments was also demonstrated in the Lort River in the southwest coastal division where the largest flood resulted from a deep tropical low. However, cut-off lows also resulted in several impressive floods. The Murray River, just south of Perth, was the only river to receive its largest flood event from a mid-latitude weather pattern, in this case a cut-off low and frontal system in combination, but as was the case in other southern regions, tropical weather patterns were also significant. A tropical cyclone also recorded a large flood magnitude in the Murray River, Western Australia.

The ability of tropical weather patterns, such as tropical trough/lows, deep tropical lows and tropical cyclones, to produce large floods in southern latitudes demonstrates the superior effectiveness of warm tropical air masses to generate instability compared to colder mid- latitude air masses. Although mid-latitude weather patterns such as cut-off lows and frontal systems, influence flood magnitudes in the Gascoyne River on the west coast, and in the Warrego River on the east coast, they are not as effective as tropical weather patterns in influencing flood magnitudes well out of the tropical region, such as in the southern Murray- Darling Rivers.

The Gascoyne River recorded its largest flood as result of a cut-off low in combination with a northwest cloudband. Northwest cloudbands on their own also produced some notably large floods. The Gascoyne region is the only area in Australia that appears to be vulnerable to large flood events as a result of northwest cloudbands. This appears to be particularly the case if the cloudbands, which are really tropically derived, combine with mid-latitude weather patterns such as cut-off lows. Sturman and Tapper (1996) state that heavy rainfall is often associated with northwest cloudbands over southern Australia, especially if combined with a cut-off low system. It does appear, however, that northwest cloudbands do not occur frequently enough

134

over southern Australian dryland rivers to produce large floods. This is contrasted with central and northern parts of Western Australia where the occurrence of cloudbands is frequent over the winter months, often generating heavy rainfall. Tapp and Barrell (1984) plotted the alignments and lengths of 15 cloudbands. From this it is apparent that the Gascoyne Region is affected by more cloudbands than other regions of Australia. They also found that cloudbands come straight off the Indian Ocean, providing abundant moisture that greatly influences rainfall. In southern Australia, unless the cloudband is associated with a secondary feature such as a cut-off low, it appears that there is no access to abundant moisture.

4.6 Conclusion

Floods in the far northern dryland rivers of Australia are the result of tropical trough/lows, deep tropical lows and tropical cyclones, either on their own or in combination. These synoptic patterns are associated with the wet season of the northern monsoon, which technically extends from December to March. In the Timor Sea division along with the Indian Ocean division, tropical cyclones tend to cause more flood events relative to those occurring in the dryland rivers of the Gulf of Carpentaria division, Lake Eyre division and the Bulloo-Paroo-Warrego region. The Gulf of Carpentaria division records the highest number of floods as a result of tropical trough/lows and deep tropical lows. From the Lake Eyre division through to the northeast Murray-Darling Rivers, tropical trough/lows and deep tropical lows result in a large number of floods. Along with causing a high number of the floods, tropical cyclones tend to produce the largest flood events in the Timor Sea and Indian Ocean divisions (including the Pilbara region and Greenough, Murchison and Gascoyne Rivers) of Western Australia. In the Gulf of Carpentaria division, Lake Eyre division, and the Bulloo, Paroo and Warrego Rivers, deep tropical lows and tropical trough/lows tend to cause the largest floods.

Mid-latitude weather patterns more dominantly influence drainage divisions in southerly latitudes, such as the southern Murray-Darling division and the Southwest Coastal division. Tropical weather patterns still cause floods in these divisions but the dominant flood producing

135

patterns are mid-latitude in origin. However, the largest floods in the south can also result from tropical systems.

Because they straddle the boundary between the regions dominantly affected by tropical and mid-latitude weather patterns, respectively, a wide range of weather patterns affect the Bulloo, Paroo and Warrego Rivers, northern Murray-Darling division, and the Gascoyne, Murchison and Greenough Rivers. The northeast Murray-Darling Rivers received flood events from the largest range of weather patterns of any dryland river drainage division in Australia.

During autumn and early winter, easterly dips (continental) produced a significant number of flood events within the Lake Eyre division, Bulloo, Paroo and Warrego Rivers, and northeast and southern Murray-Darling Rivers. The Bulloo, Paroo and Warrego Rivers receive the highest number of floods from easterly dips (continental). A significant number of the largest floods are also the result of easterly dips (continental) in these areas.

Northwest cloudbands, mostly in combination with other synoptic classes, resulted in floods in the Lake Eyre division, Bulloo, Paroo and Warrego Rivers, and in the Murray-Darling division. However, it is only in the Greenough, Murchison and Gascoyne Rivers, particularly the Gascoyne River, where northwest cloudbands caused a significant number of large magnitude floods. These northwest cloudbands are a winter phenomenon occurring predominantly from May to August.

Cut-off low-pressure systems result in a significant number of floods in the Greenough, Murchison and Gascoyne Rivers, the southwest coastal division and the southern Murray- Darling Rivers. Their occurrence in the Greenough, Murchison and Gascoyne Rivers is probably most significant due to these rivers being located a long distance to the north. The Gascoyne River is approximately 1,000km north of Perth and would not be considered in area affected by mid-latitude weather patterns. It has been found that the area off the coast of Western Australia is a preferred location of formation for cut-off lows (Leslie and Zhao 1999).

136

The southern Murray-Darling Rivers had a larger number of floods resulting from cut-off lows than from frontal systems and these tend to occur later (from August to September) than in the southwest coastal division. This possibly related to snowmelt, although this aspect was not included in the analysis. In the latter division, frontal systems tended to cause more floods despite the major formation-zone for cut-off lows being off the southwest coast of Australia. These frontal systems occur during the winter months of June, July and August.

In the southern Murray-Darling Rivers and the southwest coastal division, cut-off lows and frontal systems, particularly in combination with other synoptic classes, resulted in a number of very large floods. However, the influence of tropical weather patterns on flood magnitudes in these southern dryland rivers is clearly evident, with a significant number of the largest floods resulting from tropical weather patterns such as tropical trough/lows and deep tropical lows.

In summary, tropical trough/lows, deep tropical lows and tropical cyclones cause the majority of the flood events in the far north drainage divisions, with tropical cyclones being more dominant in the northwest and deep tropical lows in the northeast. In the Lake Eyre division through to the northeast Murray-Darling sub-division the range of weather patterns causing floods increases dramatically with rivers in the northeast Murray-Darling sub-division affected by the largest range of weather patterns. Floods in the southern Murray-Darling Rivers and southwest coastal division are the result of mid-latitude weather patterns with cut-off lows being more dominant in southern Murray-Darling Rivers relative to the southwest coastal division where frontal systems dominate. However, tropical systems sometimes generate large floods in these southern catchments.

137 Chapter 5

The Spatial Properties of Rainfall Events in Dryland Australia

5.1 Introduction and objectives

It has long been regarded that rainfall in semi-arid to arid areas is the result of more localized (convective) rather than widespread rainfall typical of humid-temperate regions. Sharon (1972) found that for rainfall totals in an arid region of southern Israel, 50-60% were highly localized and coming mostly from small convective cells, this pattern being most pronounced in late spring and fall with uniform rainfall more likely in winter. Sharon (1981) also found that in a hyper-arid region of central Namib, rainfall is of highly localized nature, as well as being temporally irregular. Osburn et al (1979) showed in an arid part of southwestern United States almost all runoff was the result of intense short duration rainfall of limited aerial extent. Wheater et al (1991) demonstrated that rainfall in Saudi Arabia is generally of 1-2 hour duration, highly localized and initiated in the late afternoon as a result of thunderstorms. Bell (1979) concluded that rainfall in the world’s desert areas is highly localized, of short duration and high intensity. It is thought that high surface temperatures lead to instability but, with a general lack of moisture, little widespread rainfall occurs. However, in one of the only Australian studies to date, Cordery et al (1983) and Cordery and Fraser (2000) noted that rainfall totals (annual, monthly and storm) were predominantly of a widespread rather than localized nature in arid western New South Wales.

This chapter describes the spatial rainfall characteristics in a semi-arid and an arid catchment in relation to both large flood events and smaller storm totals. The aim is to determine whether or not the findings by Cordery and Fraser (2000) of widespread rather than localised rainfall events apply in other semi-arid to arid regions of Australia. Two catchments, one semi-arid the other arid, and substantially different size, were chosen for detailed investigation.

138 5.2 Catchment description

The Thomson River catchment with an area of 58,000km2 at Longreach is one of the major tributaries of Cooper Creek within the Lake Eyre basin (Figure 5.1). Mean annual rainfall ranges from 650mm in the headwaters to 300mm where the Thomson and Barcoo Rivers merge to form Cooper Creek which flows on toward Lake Eyre well to the south. Based on mean annual rainfall totals this catchment would be best described as semi-arid. Topography within the Thomson River catchment is flat to gently undulating with the northeast corner being situated on the western edge of the Great Dividing Range.

The Todd River catchment is much smaller, with an area of 400km2 at Anzac Oval in Alice Springs, and is situated within the MacDonnell Ranges in central Australia (Figure 5.1). Mean annual rainfall ranges from 450mm in the higher parts of the MacDonnell Ranges, where the Todd River begins, but declines to 200mm where the Todd River disappears into the sands of the Simpson Desert. The region is best described as arid. Figure 5.2 shows the monthly rainfall distribution for Alice Springs and Longreach.

5.3 Methods

Two approaches were adopted to assess the rainfall properties of both regions. Firstly, the rainfall properties associated with the flood events selected for Chapter 4 were examined. For the Thomson River at Longreach there were 16 flood events with magnitudes greater than or equal to 50% Annual Exceedence Probability (AEP) between 1971 and 2001. Flood magnitudes varied from 811 m3s-1 to 12,958 m3s-1. Within this catchment (above Longreach) there are 22 daily-read rainfall stations with reliable, relatively complete, records spanning this time period (Figure 5.1). Distances between gauges ranges from 10km to 280km. Flooding rainfalls occurred over durations ranging from 24 hours to five days and were totalled accordingly. Rainfall totals were correlated for all possible binary combinations of the 22 rainfall stations for each of the 16 flood events. For the 22 rainfall stations, there are 254 possible combinations.

139 In the Todd River catchment between 1965 and 2001 there were 25 flood events with magnitudes greater than or equal to 50% AEP. The spatial rainfall properties associated with these flood events were analyzed to assess the correlation between rainfall totals at different rainfall stations. Within the Todd River catchment there are 19 rainfall stations (Figure 5.1) with reliable long-term records. The distance between these stations ranges from 7km to 222km. The 19 rainfall stations provide 190 possible binary combinations.

The second approach was designed to focus upon much smaller areas in both regions. Relatively dense networks of five adjacent rainfall stations were selected (Figure 5.1) and the storm rainfall totals were analysed when any of the five rainfall stations registered ≥5mm of rainfall over 24 hours. In the Thomson River catchment, distances between the rainfall stations in this spatially more restricted network ranged from 20 to 76km. To assess whether any seasonal factors could be identified, rainfall events occurring in the wet season (October to April) were analyzed separately from those occurring in the dry season (May to September). For the Thomson River between 1990 and 1995 there was a total of 99 events where ≥5mm of rainfall was recorded at one or more of the five rainfall stations, with 77 occurring during the wet season and 22 during the dry season. In the Todd River catchment, distances between these rainfall stations ranged from 19km to 80km. For the period from 1990 to 1995 there were 78 rainfall events where at least one of the rainfall stations recorded ≥5mm of rainfall in 24 hours, of which 55 occurred in the wet season and 23 occurred in the dry season.

For both the Thomson River and Todd River catchments, the distances between all possible combinations of rainfall stations were computed. This allowed a simple linear correlation between distance and rainfall totals (flooding and storm totals) to assess whether rainfall totals between widely spaced rainfall stations are correlated.

140

Figure 5.1: Rainfall stations in the Todd River and Coopers Creek Catchment areas Source: Map data supplied by AUSLIG

Alice Springs Average Monthly Longreach Average Monthly Rainfall Rainfall 50 90 80 40 70 60 30 50 20 40 30 Rainfall (mm) 10 Rainfall (mm) 20 10 0 0

ul J ar ay Jul Jan Mar May Sep Nov Jan M M Sep Nov

Figure 5.2: Average monthly rainfall totals for Alice Springs (1940-2000) and Longreach (1949-2000), respectively.

141 5.4 Results

5.4.1 Thomson River catchment

The flooding rainfall totals as recorded by the 22 daily rainfall stations across the catchment, which were associated with the flood 16 events in the catchment, were found to be highly correlated. These correlations were statistically significant at the 1% confidence level for rainfall stations that were up to 220km apart, with correlation coefficient values as high 0.96. Figure 5.3 shows the linear regression of the flooding rainfall totals for all possible combinations of the 22 rainfall stations. For the linear regression, the P value is significantly less than 0.01 (1% probability) and the t value is significantly higher than the t critical (0.01), further evidence of strong relationship between correlation coefficient values and distance between rainfall stations. Table 5.1 provides the statistics for the Thomson and Todd River catchments for the analysis of the flood event rainfalls across the catchments, and for the detailed analysis of storm rainfall totals for the five rainfall stations.

With the detailed analysis of the five rainfall stations over the 1990-1995 time period, high correlation coefficient values were also found. All correlations were statistically significant at the 1% level. For linear regression the P value is less than 0.01 (1% probability) and the t value is significantly higher than the t critical (0.01) (Table 5.1). Table 5.2 and 5.3 show the individual correlations for the wet and dry season, respectively. Figure 5.4 shows the correlation values for all possible rainfall stations combinations for the five gauges selected for the detailed analysis. There is a general increase in the correlation coefficient value as the distance between rainfall stations decreases. Also, the correlations are noticeably higher during the dry season compared to the wet season. The wet season correlation values ranged from 0.75 to 0.91 as where the dry season values ranged 0.76 to 0.97.

142 5.4.2 Todd River catchment

In the Todd River catchment, correlation coefficients values were high between the 19 rainfall stations, associated with the 25 flood events between 1965 and 2001. Rainfall stations that were from up to 200km apart returned values that were within the 1% confidence level, with correlation values as high as 0.97 for stations in close proximity. For the linear regression between correlation coefficient values and separating distance, the P value is significantly less than 0.01 (1% probability) and the t value is significantly higher than the t critical (0.01) giving evidence of a strong relationship (Table 5.1). Figure 5.5 shows the correlation coefficients values for all possible combinations of the rainfall stations.

Detailed examination of the five rainfall stations located within the close proximity network for this catchment also returned high correlation coefficient values. In general there was an inverse relationship between the degree of correlation and distance apart, with values varying from 0.90 to 0.65 for distances of between 18.6 and 80.1km for the wet season, and 0.97 and 0.84 for the dry season. Figure 5.6 shows the correlation coefficient values for all possible combinations of the selected rainfall stations. The P value for the linear regression is significantly less than 0.05 (5% probability) and the t value is significantly higher than the t critical (0.01). Table 5.4 and 5.5 show the correlation coefficients for pairs of rainfall stations during the wet season and dry season, respectively.

Thomson Thomson River Todd River Todd River River (Storm rainfalls (Flood event (Storm rainfalls (Flood event – all seasons) rainfalls) – all seasons) rainfalls) Anova F statistic 175 26 60 4

Significance F 3.8 x 10-30 0.0008 6.3 x 10-13 0.06 value (<0.01) (<0.01) (<0.01) (<0.10) P value 2.6 x 10 –86 0.0003 3.1 x 10-35 0.03 (<0.01) (<0.01) (<0.01) (<0.05) t Stat 31.15 6.81 28.05 7.09 (t critical at 0.01) (2.33) (2.55) (2.33) (2.55) Table 5.1: Regression statistics for the Thomson and Todd River catchments.

143 Correlation between Flood Rainfall Totals in The Thomson River Catchment

1.2 y = -0.0023x + 0.842 1 R2 = 0.4356 0.8

0.6

0.4

0.2 Correlation Coefficient 0 0 50 100 150 200 250 300 Distance Between Stations (km)

Figure 5.3: Linear regression fitted to correlations coefficient values for rainfall stations in the Thomson River Catchment for flood-event rainfall totals for all possible distance combinations between rainfall stations based on 16 floods.

Correlation between Storm Rainfall Totals in The Thomson River Catchment

0.8

0.6 y = -0.0025x + 0.9602 2 0.4 R = 0.7691

0.2 Correlation Coefficient

0 0 20406080 Distance Between Stations (km)

Figure 5.4: Correlation coefficient values in the Thomson River catchment for storm rainfall totals between 1990-1995 (All seasons) based on the total 99 events.

144 Stations Separating Correlation Distance (km) coefficient 36066 & 36024 19.8 0.80 36031 & 36024 24.4 0.91 36002 & 36013 24.5 0.87 36013 & 36066 32.8 0.89 36031 & 36066 35.0 0.89 36013 & 36024 49.5 0.80 36013 & 36031 50.5 0.80 36066 & 36002 61.1 0.88 36002 & 36031 71.1 0.75 36024 & 36002 76.0 0.77 Table 5.2: Wet Season (October-April) correlations between individual rainfall stations for the Thomson River Catchment 1990-1995. Note: There are 77 events in each correlation

Stations Separating Correlation Distance (km) coefficient 36066 & 36024 19.8 0.97 36031 & 36024 24.4 0.97 36002 & 36013 24.5 0.91 36013 & 36066 32.8 0.89 36031 & 36066 35.0 0.94 36013 & 36024 49.5 0.92 36013 & 36031 50.5 0.92 36066 & 36002 61.1 0.76 36002 & 36031 71.1 0.83 36024 & 36002 76.0 0.82 Table 5.3: Dry Season (May-September) correlations between individual rainfall stations for the Thomson River Catchment 1990-1995 Note: There are 22 events in each correlation

145 Correlation Coefficients for Flood Rainfall totals in the Alice Springs Area

1.2

y = -0.0022x + 0.9406 1 R2 = 0.2671 0.8

0.6

0.4

Correlation Coefficient 0.2

0 0 50 100 150 200 250 Distance Between Stations (km)

Figure 5.5: Linear regression fitted to correlations coefficient values for flood-event rainfall totals for all possible rainfall station combinations (all seasons) in the Todd River catchment based on 23 flood events.

Correlations Between Storm Rainfall Totals in the Todd River Catchment 1

t 0.8

0.6 y = -0.0014x + 0.8836 2 0.4 R = 0.3648

Correlation Coefficien 0.2

0 0 20406080100 Distance Between Stations (km)

Figure 5.6: Correlation values for storm rainfall totals for the period from 1990-1995 (all seasons)

146 Stations Separating Correlation Distance (km) coefficient 15590 & 15623 18.6 0.90 15631 & 15623 20.0 0.76 15590 & 15631 28.9 0.77 15564 & 15590 43.4 0.84 15553 & 15564 45.2 0.69 15564 & 15631 59.6 0.78 15564 & 15623 60.6 0.83 15631 & 15553 67.1 0.70 15590 & 15553 70.7 0.68 15623 & 15553 80.0 0.65 Table 5.4: Wet season correlations between individual rainfall stations for storm rainfall totals in the Todd River catchment between 1990-1995 (Note: there were 55 events used in these correlations)

Stations Separating Correlation Distance (km) coefficient 15590 & 15623 18.6 0.95 15631 & 15623 20.0 0.95 15590 & 15631 28.9 0.95 15564 & 15590 43.4 0.89 15553 & 15564 45.2 0.84 15564 & 15631 59.6 0.85 15564 & 15623 60.6 0.92 15631 & 15553 67.1 0.93 15590 & 15553 70.7 0.97 15623 & 15553 80.0 0.91 Table 5.5: Dry season correlations between individual rainfall stations for storm rainfall totals in the Todd River catchment between 1990-1995. (Note: there were 23 events used in these correlations)

5.5 Discussion

In both the Thomson and Todd River catchments, for both the flood event rainfall totals and storm rainfall totals (throughout all seasons), the majority of correlation coefficient values were high with most being statistically significant at the 1% confidence level. What is noteworthy is that this relationship exists over very large distances. In the Thomson River catchment, correlation coefficients values for flooding rainfalls were found to be statistically significant at the 1% confidence level over distances of up to 220km, whilst for the Todd River

147 catchment such values were returned for rainfall stations up to 200km apart. This means that a large percentage of the area in each catchment contributes runoff when river flow is initiated rather than discrete storms delivering runoff from just small sections of each catchment. In Australia the uniform nature of rainfall within dryland river catchments is intrinsically linked to the monsoon, for most of Australia’s dryland areas receive their rainfall directly from monsoon related synoptic patterns. Clearly, when rainfall is able to penetrate into inland regions of Australia, often as a result of a monsoon depression or rainfall depression (ex- tropical cyclone), this rainfall occurs over very large areas.

However, significant variations in rainfall totals do exist across each basin, particularly in the arid Todd River catchment. A maximum rainfall center is generally evident for each event, hence rainfall is not completely uniform. Also, regression analyse reveal that the relationship between correlation coefficient values and distance between rainfall stations is stronger in the semi-arid Thomson River catchment than the arid Todd River catchment. For the flood-event rainfall totals, the correlation values had a strong relationship with the distance apart of rainfall stations in the Todd River catchment, demonstrating that flood events are the result of big synoptic-scale rainfall events. However, for storm rainfalls the relationship between correlation coefficients and separating distance was weaker demonstrating that storm rainfall totals over arid Australia, although still predominantly of a widespread nature, are at least partly the result of convective localized activity.

Rainfall in semi-arid to arid Australia is still predominantly the result of widespread rainfall and not generally associated with convective or localized rainfall patterns. When rainfall is of a convective or localized nature, a handful of stations can be expected to return high rainfall totals, whilst the remaining rainfall stations within the catchment will receive little or nothing. For instance, Sharon (1972) found that in arid parts of Israel storm cells had diameters of about 5km. The arid regions of Israel also receive uniform rainfall from specific synoptic patterns, but there the largest observed rainfall totals, those that would lead to flood events, are predominantly of a localized high intensity nature (Sharon 1972).

148 The high correlation coefficients for the detailed analysis of the five network rainfall stations within both the Thomson and Todd River catchments support the findings of Cordery and Fraser (2000) who obtained similarly high values (0.70 to 0.97). Cordery and Fraser (2000) used data from the five rainfall stations collected over a 30-year period from Fowlers Gap research station within an arid region of Western NSW. The distance between these rainfall stations ranged from 6 to 20km with correlation coefficients ranging from 0.72 to 0.85 for storm rainfall, with similarly high values for monthly and annual rainfall totals. In the context of the work of Cordery and Fraser (2000), the particularly important finding of this study is the widespread nature of rainfall over large distances within semi-arid and arid areas. Over distances of 20km, correlation coefficients ranged from 0.76 to 0.95 for the Todd River catchment in arid central Australia, with coefficients from 0.80 to 0.97 in the more semi-arid part of the Thomson River catchment. Even at a distance of 80km, rainfall stations are highly correlated, with significant correlation coefficient values ranging from 0.65 to 0.91. Osborn et al (1979) found that in southwestern United States, coefficients of 0.7 to 0.9 were found for rainfall stations up to 4km apart, but over a distance of 10km these coefficients dropped to between 0.2-0.4.

Sharon (1979) found that in the Jordan valley, rainfall totals were relatively heterogeneous (correlation values of 0.60 over 20km) compared to the hilly areas to the east and west of the valley that received 600-700mm of rainfall per annum with correlation coefficients of 0.8 for distances over 20km. Low values were also found for arid areas in Tanzania, Africa (Sharon 1974).

Detailed analysis of the five network rainfall stations in the Todd and Thomson catchments showed that the correlation coefficients were noticeably higher during the dry relative to the wet season. This reflects the fact that during the wet season thunderstorm occurrence (resulting from higher surface heating) results in more localized convective rainfall in contrast to that occurring during the dry season. Some of the rainfall in the dry season results from northwest cloudbands, which tend to result in large bands of uniform mid-high level clouds bringing uniform rainfall usually of small totals, particularly over central Australia. The general lack of

149 thunderstorms during the dry season means that any rainfall that does occur is of a widespread and relatively uniform nature.

5.6 Conclusion:

This study confirms the results obtained by Cordery and Fraser (2000) on the spatial pattern of rainfall in dryland Australia, although here the study areas and distances between rainfall stations are much larger. For both the semi-arid to arid catchment of the Thomson River in southwest Queensland and the arid catchment of the Todd River in central Australia, rainfall is predominantly of a widespread and uniform nature. This is the case for rainfall associated with moderate to large floods and for lower magnitude storm totals. This is in contrast to many other dryland regions of the world where rainfall has been found to be more of a localized convective nature. Rainfall in dryland Australia, particularly the northern areas including central Australia, is intrinsically linked with the monsoon, or more correctly the wet seasons that typify northern Australia. In the southern areas mid-latitude or temperate weather patterns become more common although monsoon related weather patterns, such as tropical trough and rainfall depressions, frequently track over southern areas giving large rainfall totals. Rainfall in the wet season, due to the presence of thunderstorms, is slightly less widespread compared to the rainfall in the dry season.

150 Chapter 6

Sea Surface Temperatures and Australia’s Dryland Rivers: Links to Flooding and flow patterns

6.1 Introduction

There is a large amount of research that has related Sea Surface Temperatures (SSTs) and the directly related El Nino Southern Oscillation (ENSO) phenomenon to rainfall and streamflow in Australia. Global gridded monthly SST data in relation to ENSO is available for period from 1871 until present. This data, together with the southern Oscillation Index (SOI), has been widely used to establish relationships between rainfall streamflow and ENSO. Ropelewski and Halpert (1987) identified various regions of the world where relationships between low values of the SOI and reduced precipitation exist. They found that areas that showed consistent ENSO related precipitation patterns included the Pacific Ocean, four regions in Australia, two regions in North America, South America, the Indian subcontinent, Africa and one region in Central America. Following on from that study, Ropelewski and Halpert (1989) found that the majority of the regions of the world where rainfall was reduced during low SOI values also had increased precipitation when SOI values were high. The high index phase of the SOI was associated with increased monsoon precipitation in India and northern Australia. In addition, eastern Australia and Tasmania all have a tendency for wetter than average precipitation during high index phases of the SOI.

In an early study, Pittock (1975) used area-averaged annual rainfall for 107 rainfall districts and found that eastern Australia is strongly correlated with the SOI. McBride and Nicholls (1983) studied seasonal relationships between Australian rainfall and the Southern Oscillation Index between 1932 and 1974. The strongest correlations were found in eastern and northern Australia. Seasonally, spring and summer have the strongest and weakest correlations, respectively. For northern Australia this means that the peak monsoon months (December- February) are only weakly related to the SOI relative to the months leading up to the monsoon.

151 Stone and Auliciems (1992) examined SOI phase relationships with rainfall in eastern Australia. They found that the phase which represents a rapid rise in SOI values indicates above median rainfall during spring and summer. This relationship was evident even when the rapid rise occurred during a period when SOI values were not necessarily highly positive. The phase where SOI values are consistently positive is generally associated with above median rainfalls whilst the phase where SOI values are consistently negative corresponds with below median rainfalls.

Nicholls and Kariko (1993) examined the number, average length and average intensity of rainfall events for five rainfall stations in eastern Australia. They found that the ENSO phenomenon affected the number, length and the intensity of the rainfall events at all stations. However, only the number and intensity of the rainfall events showed consistently positive correlations with the SOI.

Drosdowski (1993) found that May-June-July rainfall over parts of southern and eastern Australia could be predicted using SST values from the Indian Ocean off the southwest coast of Australia observed for the period of the previous summer.

Nicholls (1989) used a rotated principal component analysis of Australian winter rainfall to identify two large-scale patterns of variation that together accounted for approximately 50% of rainfall variance. The first pattern is a broad band of strong positive correlations stretching from the northwest to the southeast corners of Australia and is related to the difference in SST values between the central Indian Ocean and the Indonesian region. This pattern of variation is thought to be largely independent of the SOI. During autumn and winter, cold water in the Indian Ocean is associated with warm water off northern Australia. The second rainfall pattern consists of strong positive correlations concentrated in the eastern third of Australia and is related to SST values in the equatorial Pacific Ocean and is clearly related to the SOI.

Smith (1994) also found that winter rainfall over large parts of Australia is related to the dominant SST patterns in Indian Ocean. This relationship was extended revealing evidence of

152 significant correlations between autumn SST values and winter rainfall, allowing more accurate forecasts than autumn SOI values.

Simpson et al (1993) found that, between 1895 and 1985, the annual river discharges (Q) of the Murray-Darling River are inversely related to SSTs in the eastern equatorial Pacific Ocean. In the past, this inverse relationship with SST values has been more commonly correlated with drought as opposed to natural river discharge responses.

Piechota et al (1998) studied seasonal streamflow forecasting for 10 eastern Australian catchments using 66 years of continuous stream flow data. They developed a seasonal streamflow forecast model derived from climatology, persistence, SOI and equatorial Pacific SSTs. Their results further confirm that during El Nino conditions, when the SOI is strongly negative and the SST index is strongly positive in the eastern equatorial Pacific Ocean, streamflow tends to be below normal whereas during La Nina conditions streamflow is above normal. However, this relationship was not found to be true for each ENSO event, thereby indicating significant seasonal and regional variation. Chiew et al (1998) found that the link between streamflow and ENSO was statistically significant for most regions of Australia but is not strong enough for robust, accurate prediction. They concluded that ENSO parameters could be used with some success in forecasting spring runoff in southeastern Australia and summer runoff in northeast and eastern Australia. Puckeridge et al (2000) examined the 48-year hydrograph record for Cooper Creek at Currareva in Southwest Queensland. They found that floods occurred in clusters associated with La Nina events and that this has a profound affect on the ecology of dryland rivers.

Drosdowsky and Chambers (1998) carried out extensive testing of SST anomaly patterns as predictors of Australian seasonal rainfall variation with the view to improving of forecasting. Ten principal components covering the Pacific and Indian Oceans, respectively were used as predictors. The first two principal components (SST1 and SST2) represent the SST anomalies of the eastern equatorial Pacific Ocean and the subtropical Indian Ocean respectively (Figure 6.2). Their results suggest that using SST anomaly patterns solely (SST1 - and SST2), or using them in conjunction with the SOI, can improve forecasts. Drosdowsky (2001) suggested that

153 whilst the Pacific Ocean clearly has the dominant influence on Australia’s rainfall, the Indian Ocean has a modulating affect on the rainfall over southeast Australia.

Drosdowski (2001) compared and contrasted the phase data of Stone and Auliciems (1992) with phase data calculated from the SST1 and SST2. The phases for SST1 show a similar relationship to the SOI as that obtained by Stone and Auliciems (1992). The phases for SST2 were consistent with the dipole pattern previously noted by Nicholls (1989) and Drosdowsky (1993).

This chapter examines the relationships between the Southern Oscillation Index (SOI) and various Sea Surface Temperature (SST) indices from the Pacific and Indian Oceans, and dryland river flow patterns.

6.2: Methods

A representative number of streamflow gauges were selected from within Australia’s semi-arid and arid zone regions (Figure 6.3). These were selected based on the length and quality of the record. The streamflow records selected cover the period from approximately 1965 until 2001 owing to availability of quality streamflow records.

Simple linear correlations are used to identify relationships. The streamflow parameters that are correlated include a partial flood series for the length of record (for 20 years of flow data the highest 20 flood peaks are taken), the total monthly flow (Ml) for each month for the length of record, and the seasonal flow volumes (Ml) for each of the four seasons. All the correlations are lagged up to five months, with the indices that are used detailed below. The peak flood discharge or total monthly flow are correlated against the SOI or SST index for the month of occurrence and the previous 5 months.

The SOI used is the Troup (1965) SOI, which is defined as the difference in Mean Sea Level Pressure (MSLP) between Tahiti and Darwin (Australia). There are numerous SST indices

154 available to provide a direct measure of the temperature anomalies associated with the ENSO phenomenon. The widely used Nino 3 SST index covers an area of the equatorial Pacific Ocean between (5° south - 5° north) and (150° west - 90° west) and simply provides a positive or negative value as to whether the SST for this area is above or below average.

The Multivariate Environmental Index (MEI) (Wolter and Timlin 1993) is derived from multiple climate monitors and is thought to reflect the nature of the ocean-atmosphere coupled climate system better than the either the SOI or SST based indices. The calculated index is based on six main observed variables over the tropical Pacific Ocean, including mean sea level pressure, zonal and meridional components of surface winds, surface air temperature, sea surface temperature and total percent cloudiness.

The Bureau of Meteorology Research Center (BMRC) has calculated 12 principal components for the world’s oceans. These are essentially the 12 main SST anomalies. The positive or negative value of the index determines whether SSTs are above or below average for this anomaly. The first two principal components (SST1 and SST2) represent the SST anomalies of the eastern equatorial Pacific Ocean (ENSO) and the subtropical Indian Ocean, respectively (Figure 6.1). The Indian Ocean SST index covers an area from 30° north to 30° south and 46- 100° east (Figure 6.2).

Streamflow gauges from eastern, southeastern and northeastern Australia were correlated with the SOI, Nino 3, SST1 and the MEI, whereas streamflow gauges in western and northwestern Australia were correlated with the SOI, SST2 and the Indian Ocean SST index. It is thought that SSTs from the equatorial Pacific Ocean predominantly affect the eastern half of Australia, whereas SSTs from the Indian Ocean would have more of an effect on the western half of Australia.

In addition to these correlations, some graphical representations of the data are presented. These include comparison of the number and magnitude of the partial series flood flows with El Nino, Neutral and La Nina years. Years are defined as El Nino if the 12-month average (April-March) of the SOI is less than –5. La Nina years are defined as those when the 12-

155 month average is above +5 whilst neutral years are between +5 and –5. These methods of definition of El Nino, neutral and La Nina years are based on the methods used by Chiew et al (1998).

Figure 6.1: Map of the Bureau of Meteorology SST1 and SST2 sea surface temperature anomalies. NB: Pink and blue correspond with positive and negative SST anomalies, respectively

156

Figure 6.2: Indian Ocean SST index averaging area

Figure 6.3: Selected Australian dryland rivers for ENSO and SST correlations

157 6.3 Results: Correlations with indices

6.3.1 Avoca River

The Avoca River is situated in northern Victoria and forms part of the southern Murray- Darling basin (Figure 6.3). Table 6.1 presents the correlation coefficient values for climate indices and flows for the Avoca River at Quambatook, with a catchment area of 4,740km2.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.23 0.22* 0.05 0.34* 0.30* 0.33* Lag 1 Lag 1 Lag 0 Lag 0 Lag 1 Lag 1 MEI -0.09 -0.19* -0.10 -0.08 -0.22* -0.26* Lag 0 Lag 0 Lag 1 Lag 0 Lag 1 Lag 0 NINO 3 0.07 -0.16* -0.17 -0.04 -0.23* -0.23* Lag 0 Lag 0 Lag 1 Lag 1 Lag 1 Lag 0 SST1 -0.05 -0.17* -0.03 -0.18* -0.18* -0.27* Lag 0 Lag 0 Lag 0 Lag 1 Lag 1 Lag 0 * Statistical significant at the 5% confidence level Table 6.1: Highest correlation coefficient values between climatic indices and flow for the Avoca River at Quambatook. (Summer, autumn, winter and spring refer to the total seasonal flow volume) Note: Flows are correlated against the index for the month of occurrence and the previous 5 months.

Based on 29 years of data the partial series flood flows did not return statistically significant correlation values with any of the indices. Total monthly flows were statistically significant at the 5% confidence level for all of the four indices with the SOI returning the highest correlation coefficient of 0.22 (Table 6.1). The MEI returned the second highest value with – 0.19 followed by values of –0.17 and –0.16 for SST1 and NINO 3, respectively. Seasonal correlations show that the Avoca River is not significantly correlated with any of the indices over the summer months. For the autumn months, the SOI and SST1 have statistically significant values of –0.34 and –0.18, respectively. Winter and spring months return significant values for all indices with the SOI returning the highest values. Correlations for the Avoca River show negligible temporal lags, with the longest lag being around one month. This

158 suggests that there is essentially no lag between SST patterns and flooding rainfall for this catchment.

6.3.2 Macquarie River

The Macquarie River, situated in northern inland New South Wales, is one of the main tributaries to the Darling River. Table 6.2 presents the correlation coefficient values for the Macquarie River at Gongolgon, with a catchment area of 28,000km2.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.15 0.23* 0.27 * 0.23* 0.34* 0.26* Lag 1 Lag 0 Lag 0 Lag 0 Lag 1 Lag 1 MEI -0.19 -0.18* -0.17 -0.17 -0.25* -0.21* Lag 3 Lag 0 Lag 0 Lag 0 Lag 1 Lag 0 NINO 3 -0.16 -0.14* -0.14 -0.13 -0.17 -0.19* Lag 3 Lag 2 Lag 1 Lag 1 Lag 2 Lag 0 SST1 -0.24 -0.21* -0.22* -0.20* -0.27* -0.26* Lag 2 Lag 1 Lag 0 Lag 1 Lag 1 Lag 0 * Statistical significance at the 5% confidence level Table 6.2: Highest correlation coefficient values between climatic indices and low for the Macquarie River at Gongolgon

The correlation coefficient values associated with the partial series floods are not significant. Values associated with total monthly flows are all significant with the highest values being 0.23 and -0.21 for the SOI and SST1, respectively. For seasonal correlations, winter and spring correlate strongly with the SOI and SST1. Outside of the winter and spring seasons, the SOI returned a value of 0.27 for the summer season. Lag values associated with the relationship are most frequently zero and one month meaning that the relationship between SSTs and streamflow is roughly simultaneous.

159 6.3.3 Warrego River

The Warrego River, situated in southern inland Queensland, is the northernmost river of the Murray-Darling Basin (Figure 6.3). Table 6.3 presents the correlation coefficient values for the Warrego River at Auguthella, with a catchment area of 8,070km2.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.25 0.11* 0.17 -0.17 0.15 0.21* Lag 0 Lag 0 Lag 0 Lag 2 Lag 1 Lag 0 MEI -0.08 -0.03 -0.12 0.01 0.03 -0.10 Lag 0 Lag 0 Lag 0 Lag 1 Lag 1 Lag 0 NINO 3 -0.12 -0.04 -0.11 0.04 0.04 -0.16 Lag 0 Lag 1 Lag 1 Lag 1 Lag 0 Lag 1 SST1 -0.14 -0.01 -0.10 0.08 -0.13 -0.10 Lag 0 Lag 1 Lag 0 Lag 0 Lag 1 Lag 0 * Statistical significance at the 5% confidence level Table 6.3: Highest correlation coefficient values between climatic indices and flow for the Warrego River at Auguthella

From table 6.3 it can be seen that the partial series floods did not return significant correlations with any index although the SOI had a value of 0.25. Correlations were also weak for the total monthly flows, with only the SOI returning a significant value. For the seasonal monthly flows the only significant correlation, with a value of 0.21, was in spring with the SOI. It appears that relationships between streamflow and SSTs in the Warrego catchment are relatively weak.

6.3.4 Cooper Creek

Cooper Creek and its tributaries extend from the western side of the Great Dividing Range in Queensland to Lake Eyre in South Australia (Figure 6.3). Table 6.4 presents the correlation coefficient values for Cooper Creek at Currareva with a catchment area of 150,220km2.

Table 6.3 shows that the partial flood series is significantly correlated with the SOI, the MEI and SST1, with values of 0.45, 0.39 and 0.38, respectively. Further significant correlations

160 were recorded for all indices when correlated with total monthly flows with correlation values of between 0.23 for the SOI and –0.17 for NINO 3. Significant values occurred in all seasons although summer and autumn have moderate correlations whilst winter returned lower values in general. Lag values were generally around one to two months. While, this may partially reflect longer flood travel times in this catchment there appears to be a slight lag between favourable SSTs and flooding rainfall.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.45* 0.23* 0.31* 0.30* 0.21* 0.15 Lag 2 Lag 1 Lag 1 Lag 1 Lag 1 Lag 2 MEI -0.39* -0.21* -0.32* -0.22* -0.20* -0.26* Lag 0 Lag 0 Lag 1 Lag 0 Lag 0 Lag 2 NINO 3 -0.29 -0.17* -0.25* -0.20* -0.10 -0.29* Lag 0 Lag 3 Lag 2 Lag 1 Lag 1 Lag 1 SST1 -0.38* -0.22* -0.27* -0.31* -0.21* -0.27* Lag 0 Lag 1 Lag 1 Lag 1 Lag 1 Lag 1 * Statistical significance at the 5% confidence level Table 6.4: Highest correlation coefficient values between climatic indices and flow for Cooper Creek at Currareva

6.3.5 Thomson River

The Thomson River is a major Tributary to Cooper Creek and is situated in western Queensland (Figure 6.3). Table 6.5 presents the correlation coefficient values for the Thomson River at Longreach, with a catchment of 58,000km2.

The partial series floods were only correlated significantly with the NINO3 index, with a value of –0.34. Significant correlation values for the total monthly flows, ranged from 0.16 for the SOI to 0.14 or NINO 3 with SST1 not being correlated significantly. Moderate correlation values were achieved for the summer season. The SOI and MEI have the highest values with 0.36 and 0.30, respectively. The only significant correlations during autumn and winter were a value of 0.19 for the SOI in winter. Spring values were all significant and ranged from 0.27 for the SOI to 0.17 for the MEI. Lag values are between one and two months, although a lag of

161 one month dominates, suggesting this is the likely delay between favourable SSTs and flooding.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.26 0.16* 0.36* 0.08 0.19* 0.27* Lag 2 Lag 0 Lag 2 Lag0 Lag 1 Lag 0 MEI -0.24 -0.15* -0.30* -0.04 -0.05 -0.17* Lag 0 Lag 0 Lag 0 Lag 0 Lag 1 Lag 1 NINO 3 -0.34* -0.14* -0.25* -0.04 -0.08 -0.21* Lag 1 Lag 1 Lag 2 Lag 1 Lag 1 Lag 0 SST1 -0.15 -0.09 -0.24* 0.06 0.08 -0.19* Lag 0 Lag 0 Lag 0 Lag 1 Lag 0 Lag 1 * Statistical significance at the 5% confidence level Table 6.5: Highest correlation coefficient values between climatic indices and flow for the Thomson River at Longreach

6.3.6 Flinders River

The Flinders River flows to the Gulf of Carpentaria and is located in far north Queensland (Figure 6.3). Correlation coefficient values for the Flinders River at Walkers Bend, with a catchment area of 107,150km2, are presented in Table 6.6.

All indices correlate with the partial series floods with the SST1 and the SOI having the highest values of 0.50 and 0.42, respectively. Significant values were also returned for the indices correlated with the total monthly flows. The SOI had the highest value of 0.17. Only summer and autumn have significant correlations, with the SOI and SST1 having moderate values. Correlation values for all indices were poor during winter and spring, with the SOI having the highest values of 0.16 and 0.14 for winter and spring, respectively. The dominant lag period is one month for all indices, suggesting a short lag between higher than average SSTs and flooding rainfall.

162 All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.42* 0.17* 0.28* 0.26* 0.16 0.14 Lag 0 Lag 1 Lag 1 Lag 1 Lag 1 Lag 1 MEI -0.39* -0.11* -0.16 -0.18* -0.02 -0.08 Lag 0 Lag 1 Lag 1 Lag 0 Lag 2 Lag 2 NINO 3 -0.40* -0.11* -0.16 -0.15 0.04 -0.05 Lag 0 Lag 1 Lag 1 Lag 1 Lag 1 Lag 1 SST1 -0.50* -0.13* -0.17 -0.30* 0.07 0.03 Lag 0 Lag 1 Lag 1 Lag 1 Lag 1 Lag 0 Statistical significance at the 5% confidence level Table 6.6: Highest correlation coefficient values between climatic indices and flow for the Flinders River at Walkers Bend

6.3.7 Fitzroy River

The Fitzroy River is located in northwest Australia (Figure 6.3). Table 6.7 provides the correlation coefficient values for the Fitzroy River at Philips Range, with a catchment area of 5,020km2.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.31 0.15* 0.20* 0.24* 0.30* 0.27* Lag 0 Lag 1 Lag 0 Lag 1 Lag 1 Lag 2 IOI -0.28 -0.19* -0.27* -0.23* -0.30* -0.24* Lag 3 Lag 1 Lag 1 Lag 0 Lag 1 Lag 1 SST2 -0.20 -0.14* -0.15 -0.24* -0.50* -0.15 Lag 1 Lag 0 Lag 0 Lag 1 Lag 0 Lag 1 * Statistical significance at the 5% confidence level Table 6.7: Highest correlation coefficient values between climatic indices and flow for the Fitzroy River at Philips Range IOI: This is the Indian Ocean SST index

The partial series floods are not significantly correlated although the SOI has a correlation value of 0.31. Correlation values for total monthly flows were all significant with the Indian Ocean Index (IOI) having the highest value of -0.19. The winter months have the highest correlation values with –0.50 for SST2 and –0.30 for both the SOI and IOI. The SOI and the IOI both recorded significant values in all the other seasons whilst SST2 only has significant

163 values for autumn and winter. Lags values were typically around 1-2 months. This demonstrates a slight lag between favourable SSTs and rainfall leading to significant streamflow.

6.3.8 Ashburton River

The Ashburton River is located in the Pilbara region of Western Australia (Figure 6.3). Table 6.8 presents the correlation coefficient values for the Ashburton River at Capricorn Range, with a catchment area of 41,400km2.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.43* 0.15* 0.20* 0.18* 0.22* 0.25* Lag 3 Lag 3 Lag 3 Lag 0 Lag 2 Lag 1 IOI -0.22 -0.15* -0.25* -0.39* -0.09 -0.16 Lag 1 Lag 1 Lag 1 Lag 1 Lag 2 Lag 0 SST2 -0.15 0.07 0.20* 0.13 -0.10 -0.12 Lag 0 Lag 0 Lag 2 Lag 1 Lag 0 Lag 0 * Statistical significance at the 5% confidence level Table 6.8: Highest correlation coefficient values between climatic indices and flow for the Ashburton River at Capricorn Range

The partial series floods are moderately correlated with the SOI, with a correlation coefficient of 0.43, whereas the IOI and SST only returned values of –0.22 and –0.15, respectively. The total monthly flows were significantly correlated to the SOI and the IOI. SST2 correlated weakly with the seasonal flows, with only the summer flow having a statistically significant correlation value. The SOI had significant correlation values for all seasons with spring having the highest value of 0.25. The IOI performed strongly in autumn and summer with correlations of –0.39 and –0.25, respectively. Monthly lag values ranged from zero to three with a lag of one month being dominant. The lag values suggest a complicated relationship between favourable SSTs and flooding or higher than normal streamflow.

164 6.3.9 Gascoyne River

Figure 6.3 shows the location of the Gascoyne River, which is situated in the central region of western Australia. Table 6.9 shows the correlation coefficient values for the Gascoyne River at Nine Mile Bridge, with a catchment area of 71,000km2.

The partial series floods were only significantly correlated with the IOI, having a correlation value of -0.34. The IOI was also the only index to be significantly correlated with the total monthly flows. For the seasonal correlations, Table 6.9 shows that summer flows were weakly correlated. The IOI returned a correlation coefficient of –0.27 for the autumn months. Only SST2 had a significant value for the winter months, with the SOI and the IOI having significant values for the spring months. The IOI had lag values of 3 months for the partial series flows, total monthly flows and autumn season flows. This suggests that significant rainfall leading to flooding in the Gascoyne River lags behind favourable SSTs by around 2-3 months.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.13 0.02 0.08 0.13 0.06 0.21* Lag 0 Lag 0 Lag 0 Lag 3 Lag 0 Lag 2 IOI -0.34* -0.15* -0.09 -0.27* -0.12 -0.19* Lag 3 Lag 3 Lag 1 Lag 3 Lag 3 Lag 1 SST2 0.22 -0.07 0.10 0.04 -0.21* -0.04 Lag 1 Lag 2 Lag 1 Lag 1 Lag 3 Lag 2 * Statistical significance at the 5% confidence level Table 6.9: Highest correlation coefficient values between climatic indices and flow for the Gascoyne River at Nine Mile Bridge

6.3.10 Murchison River

The Murchison River is located within Western Australia’s mid-west region (Figure 6.3). Table 6.10 provides the correlation coefficients for the Murchison River, Emu Springs, with a catchment area of 82,300km2.

165

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.10 0.12* 0.04 0.22* 0.15 0.26* Lag 0 Lag 0 Lag 0 Lag 0 Lag 0 Lag 2 IOI -0.05 -0.20* -0.07 -0.44* -0.08 -0.20* Lag 0 Lag 0 Lag 1 Lag 0 Lag 0 Lag 0 SST2 0.29 -0.02 -0.06 0.15 -0.18* -0.08 Lag 1 Lag 0 Lag 1 Lag 1 Lag 0 Lag 0 * Statistical significance at the 5% confidence level Table 6.10: Highest correlation coefficient values between climatic indices and flow for the Murchison River at Emu Springs

The partial series flood flows were not significantly correlated with any of the indices. The IOI and SOI were significantly correlated with total monthly flows with values of –0.20 and 0.12, respectively. Autumn produced the highest correlation for the seasonal flows with the IOI and the SOI having correlation coefficients of –0.44 and 0.22, respectively. Other significant correlations occurred in the spring months with the values of 0.26 and –0.20 for the SOI and IOI, respectively. SST2 returned weak correlation across all seasons with only winter having a significant value of –0.18. Lag values were between zero and 2 months, although a lag period of zero months dominated. There appears to be no delay between favourable SSTs and rainfall leading to higher than average stream flows for the Murchison River.

6.3.11 Murray River (Western Australia)

The Murray River is located in the southwest corner of Western Australia (Figure 6.3). Table 6.11 provides the correlation coefficient values for the Murray River at Baden Powell, with a catchment area of 6,840km2.

The partial series floods were significantly correlated with the SOI with a value of 0.32. The IOI and SST2 both failed to produce significant correlations with the partial series flood flows, however, the SOI and SST2 produced significant correlations with the total monthly flows. The winter monthly flows have significant correlations with the SOI and SST2 having values of 0.29 and –0.20, respectively. The SST2 had a significant correlation with summer flows,

166 with a value of –0.23, whilst the SOI had a significant correlation during spring with a value 0.20. There appears to be no apparent lag between favourable SSTs and stream flows in the Murray River, Western Australia.

All Total Summer Autumn Winter Spring Partial Monthly floods Flows SOI 0.32* 0.15* 0.09 0.09 0.29* 0.20* Lag 0 Lag 0 Lag 0 Lag 0 Lag 0 Lag 0 IOI -0.17 -0.01 -0.09 -0.11 -0.12 0.06 Lag 2 Lag 0 Lag 1 Lag 0 Lag 0 Lag 1 SST2 -0.17 -0.12* -0.23* -0.14 -0.20* -0.10 Lag 0 Lag 0 Lag 1 Lag 3 Lag 0 Lag 0 * Statistical significance at the 5% confidence level Table 6.11: Highest correlation coefficient values between climatic indices and flow for the Murray River at Baden Powell

6.4 Flooding and the SOI

6.4.1 Avoca River

Figure 6.4 shows the percentage of partial series floods for each of the SOI classes from a total of 30 floods. These SOI classes are defined in section 6.2. The histograms represent the largest flood in each class for the Avoca River at Quambatook. This shows that the vast majority of floods occur in La Nina or Neutral SOI years. The largest flow for the 28 years of record was 5,090 Ml/day and occurred in a neutral year. During La Nina years the largest flood was 4,303 Ml/day whilst for El Nino years the largest event was just 1,693 Ml/day. From the total number of floods, 51% occurred in neutral year with 43% and 27% occurring in La Nina and El Nino years, respectively.

167 Flood Magnitude and SOI Class

6000 60 Flood Magnitude 5000 50 % SOI Class

4000 40

3000 30

2000 20 SOI Class

(Partial Series Floods) 1000 10 Flood Magnitude ML/Day Precentage of Floods per

0 0 La Nina Neutral ENSO

Figure 6.4: The largest flood magnitude and the percentage of floods per SOI class for the Avoca River at Quambatook

6.4.2 Thomson River

Figure 6.5 shows the percentage of partial series floods per SOI class from a total of 32 floods for the Thomson River at Longreach, with the histograms representing the largest flood in each SOI class The largest flood for the full 32 years of record was 12,958 m3/s in a La Nina year. During neutral years the largest flood magnitude was 4,096 m3/s whilst the largest flood during El Nino years was just 1,195 m3/s. La Nina years accounted for 50% of all floods whilst neutral and El Nino years accounted for 37.5% and 12.5%, respectively.

168 Flood Magnitude and SOI Class

14000 60 Flood Magnitude 12000 50 % SOI Class 10000 40 8000 30 6000

20 SOI Class 4000 (All Partial Floods) 10 Flood Magnitude in M3/S

2000 Percentage of floods by

0 0 La Nina Neutral ENSO

Figure 6.5: The largest flood magnitude and the percentage of floods per SOI class for the Thomson River at Longreach

6.4.3 Flinders River

Figure 6.6 shows the percentage of partial series floods per SOI class from a total of 32 floods for the Flinders River at Walkers Bend, with the histograms representing the largest flood in each SOI class. Neutral and La Nina years recorded the highest number of floods with 46.9% and 37.5%, respectively. El Nino years recorded just 15.6% of the floods. The largest flood for the 32 years was 8,413 m3/s recorded during a La Nina year, with a slightly smaller flood of 7,916 m3/s recorded in a neutral year. The largest flood during El Nino years had a magnitude of just 4,921 m3/s.

169 Flood Magnitude and SOI Class

9000 50 Flood Magnitude 8000 45 % SOI Class 7000 40 6000 35 30 5000 25 4000 20 SOI Class 3000 15 (Partial Series) 2000 10 Flood Magnitude M3/S 1000 5 Percentage of floods by 0 0 La Nina Neutral ENSO

Figure 6.6: The largest flood magnitude and the percentage of floods per SOI class for the Flinders River at Walkers Bend

6.4.4 Fitzroy River

Figure 6.7 shows the percentage of partial series floods per SOI class from a total of 34 floods for the Fitzroy River at Philips Range, with the histogram representing the largest flood in each SOI class. The largest number of floods occurred in neutral years (55.9% of the total) whilst La Nina years and El Nino years recorded 29.4% and 14.7%, respectively. The largest flood for the 34 years of record was of 3,885 m3/s recorded in a neutral year. For La Nina and El Nino years the largest floods were 2,003 m3/s and 742 m3/s, respectively.

170 Flood Magnitude and SOI Class

4500 60 Flood Magnitude 4000 50 % SOI Class 3500 3000 40 2500 30 2000

1500 20 SOI Class (Partial Series) 1000

Flood Magnitude M3/S 10 500 Percentage of flood by 0 0 La Nina Neutral ENSO

Figure 6.7: The largest flood magnitude and the percentage of floods per SOI class for the Fitzroy River at Philips Range

6.4.5 Gascoyne River

Figure 6.8 shows the percentage of partial series floods per SOI class from a total of 35 floods for the Gascoyne River at Fishy Pool, with the histogram representing the largest flood in each SOI class. The largest number of floods and the largest flood magnitude (9,175 m3/s) occurred within neutral SOI years. Forty percent of flood flows have occurred during La Nina years with the largest flood magnitude during a La Nina year being 4,688 m3/s. El Nino years accounted for just 8.6% of floods with the largest flood being 2,373 m3/s.

171 Flood Magnitude and SOI Class

10000 60 Flood Magnitude 9000 8000 50 % SOI Class 7000 40 6000 5000 30 4000 SOI Class 3000 20 2000 (Partial series floods)

Flood Magnitude M3/S 10 1000 Percentage of floods by 0 0 La Nina Neutral ENSO

Figure 6.8: The largest flood magnitude and the percentage of floods per SOI class for the Gascoyne River at Fishy Pool

6.4.6 Murray River (Western Australia)

Figure 6.9 shows the percentage of partial series floods per SOI class from a total of 48 floods for the Murray River at Baden Powel Station, with the histogram representing the largest flood in each SOI class. The largest number of floods occurred in neutral years with 56.2% of the total 48 events. La Nina years recorded 27.1% of the floods whilst El Nino years accounted for just 16.6%. The largest flood flow of 553 m3/s occurred during a La Nina year whilst the largest flood for neutral and El Nino years were 493 m3/s and 210 m3/s, respectively.

172 Flood Magnitude and SOI Class

600 60 Flood Magnitude 500 50 % SOI Class

400 40

300 30

200 20 by SOI Class (Partial Series) Percentage of floods

Flood Magnitude M3/S 100 10

0 0 La Nina Neutral ENSO

Figure 6.9: The largest flood magnitude and the percentage of floods per SOI class for the Murray River at Baden Powel Station

6.5 Discussion

For the correlations between the partial series floods and the SOI and SST indices, only north and northeastern Australia show a clear and consistent relationship. Correlation coefficients were as high as 0.40 and 0.50 for Coopers Creek and the Flinders River, respectively, meaning that flood flows are relatively predictable in this region. The SOI and SST1 (the SST1 index reflects the mature phase of El Nino event) were shown to be amongst the more consistently useful predictors for flooding in north and northeastern Australia.

For seasonal correlations it was found that in general, spring, summer and autumn monthly flow is related to Pacific Ocean SST indices in north and northeastern Australia. This demonstrates that rainfall that is related to the monsoon trough is also in some way related to eastern Pacific Ocean SSTs. The link is likely the onshore movement of the trade winds from the Coral Sea. In southeastern Australia it was found that it was the winter and spring monthly stream flows that are correlated with SSTs, whilst summer and autumn have lower correlation coefficients. On the Avoca River in southeast Australia, eastern Pacific Ocean SSTs predominantly correlated with autumn, winter and spring monthly flows. On the Macquarie

173 River further to the north, not surprisingly they are correlated with summer monthly flows, although the stronger correlations are during the winter and spring months, with correlations of up 0.34. This demonstrates that the strongest correlations, between the eastern Pacific Ocean SSTs, occur during periods when river flows are typically at their highest, i.e. during the winter-spring period in the south and associated with the summer monsoon trough in the north. This means that the dominant synoptic weather patterns that cause floods are related to the SOI and SST indices. In southeastern areas, frontal systems and cut-off lows produced the majority of floods during spring and autumn whereas, in the more central and northern regions of eastern Australia, synoptic weather patterns that are associated with the monsoon trough caused the majority of the floods. The latter further validates the relationship between ENSO and the monsoon, with its related weather patterns, as identified by Evans and Allan (1992). This also suggests that, in both areas, if the SOI and SST indices are not favourable, the number and/or the intensity of flood producing synoptic weather patterns are reduced.

Although numerous correlation coefficients (including those as high as 0.4 and 0.5) are statistically significant, a large percentage of the variance is clearly attributable to other factors. Pittock (1973) found significant correlations between rainfall, particularly in the springtime, for eastern Australia and variations in the mean latitudes of the surface high- pressure belt and other atmospheric parameters such as total ozone, winds and temperatures at various levels in troposphere and stratosphere. Pittock (1984) found that at best only about one-third of the variance in Australian rainfall, even at favoured sites and locations, is accounted for through its correlation with the SOI. For this research the strengths of the relationships are similar to those found in other studies. For example, Chiew et al (1998) found a statistically significant link between streamflow and ENSO. They concluded that although this link was not sufficiently strong for reliable prediction in most areas of Australia, reasonably accurate forecasts could be given for summer rainfall in northern Australia and spring rainfall in eastern Australia.

In Western Australia, correlations show that Pacific Ocean SSTs (SOI) and Indian Ocean SSTs are both factors driving climatic variability. Compared to eastern Australia, especially northeastern Australia, Western Australia appears to be more complex with regards to those

174 factors leading to climatic variability. There is no clear north-south continuum in Western Australia as there is in eastern Australia. Eastern Australia shows a clear spring-summer correlation in the north and an autumn-winter correlation in the south, but with the more centrally located rivers, such as the Macquarie, showing a mixture of these two relationships. In Western Australia, particularly in the north, it is the seasonal correlations that show a clear relationship with the SOI and the Indian Ocean SSTs. It is predominantly the autumn and winter stream flows that are more highly correlated. The correlations are strongest in the northwest with the Fitzroy River being strongly correlated with the SST2 index in autumn and winter (statistically significant correlation coefficients of –0.24 and –0.50, respectively). The Ashburton, Gascoyne and Murchison Rivers all had strong autumn correlations with the Indian Ocean SST index. The Murray River in the southwest of Western Australia had a statistically significant correlation with the SST2 index in winter.

Interestingly, in Western Australia correlations with the Indian Ocean SSTs and SST2 are negative, with large floods and higher than average monthly stream flows related to below average SSTs for the Indian Ocean region. Importantly, Nicholls (1989) and Smith (1994) discovered that winter rainfalls in northwest Australia (together, but less so, with other regions of Australia) are related to differences in SSTs between the central Indian Ocean and the Indonesian region. They found that cold water in the Indian Ocean is associated with warm water in the Indonesian region. The strong correlations (-0.50 for SST2) for monthly stream flows in the Fitzroy River during the winter months are further evidence of the effect of the Indian Ocean SSTs on streamflow. Research into the rainfall of southwest Western Australia by England et al (2006) demonstrates that in dry (wet) rainfall years anomalously warm (cool) water is located off southwest Western Australia, while cooler (warmer) than average water is situated above this in the eastern tropical/sub-tropical Indian Ocean (Figure 6.10).

Except in north and northeastern Australia, where flood events are strongly linked to the ENSO phenomenon, it is difficult to demonstrate the relationship between flood events and SST patterns. However, this does not mean that flooding in other regions of Australia is not affected by the SOI and other SST indices. The effect of the ENSO phenomenon on Australian streamflow is demonstrated qualitatively when flooding and SOI classes are compared for

175 several rivers around Australia. Figures 6.4 to 6.9 show that larger flood magnitudes occur during La Nina and neutral years whilst smaller floods occur during El Nino years. There is no clear trend as to whether the larger floods occur in La Nina as opposed to neutral years, only that smaller floods occur during El Nino years. With the exception of the Thomson River, there are generally more floods occurring during neutral years, then occur in El Nino or La Nina years. El Nino years have far fewer flood events than La Nina or neutral years, despite the number of defined El Nino years being similar to that for La Nina years. The strength of an El Nino event appears to have enough influence over climatic variability to reduce the likelihood of a flood event occurring, and to ensure if a flood event does occur it will generally be a small one.

This study demonstrates that Australian dryland river-flow patterns are linked in a complex manner to both the eastern Pacific Ocean SSTs and the Indian Ocean SSTs. This was first shown when Drosdowsky and Chambers (1998) found that, compared to using the SOI alone, SST anomaly patterns (SST1 and SST2) or SOI with Indian Ocean SSTs improved forecasts. Further to this, Drosdowsky (2001) suggests that whilst it is clear that eastern Pacific SSTs dominate Australia’s climatic variability, Indian Ocean SSTs have a modulating effect on rainfall patterns over eastern Australia. Western Australia is obviously related to Indian Ocean SSTs with a complex interaction between SSTs in the Indonesian region. Remarkably, in terms of climatic variability, the SOI signal is clearly visible in the southwest corner of Western Australia. In the same way that Indian Ocean SSTs affect winter rainfalls in southeast Australia, it appears that eastern Pacific SSTs affect rainfall variability in southwest Western Australia, although the mechanism isn’t clear.

This principle, although not statistically shown, appears to be consistent for all of Australia’s dryland river basins and agrees with what Puckeridge et al (1998) found for Cooper Creek (southwest Queensland), where flood events tend to occur in clusters during La Nina events. Similarly, Piechota et al (1998) found that for eastern Australia streamflow was below normal during El Nino periods and above normal during La Nina conditions. Roshier et al (2001) suggested that the irregular and episodic nature of wetland filling rainfall events in dryland Australia was linked to ENSO.

176

Figure 6.10. Schematic diagram showing the connection between Indian Ocean climate variability and (a) dry, (b) wet years over southwest Western Australia. Sea surface temperature anomalies are shown as actual observed composite fields (color shaded in °C). Wind anomalies are shown as bold arrows, pressure anomalies by H (high) and L (low), and rainfall anomalies by sun/cloud symbols. Source: This diagram is reproduced with permission from England et al (2006)

177 6.6 Conclusion

Only north and northeast Australia shows a clear, consistent and predictable relationship between flood events, the SOI and SST indices. The SST1 anomaly and the SOI are the indicators with the most predictive capacity in this area. Seasonally in north and northeastern Australia, spring, summer and autumn monthly river flows are correlated to eastern Pacific Ocean SST indices whereas in southeastern Australia, winter and spring monthly flows are correlated. This demonstrates that the relationships between monthly flows and SOI and SST indices exist when river flows are, in general, at their highest. Further to this, the synoptic weather patterns that cause high seasonal river flows are related to Pacific Ocean SST indices. These synoptic weather patterns are the monsoon trough and its related weather systems in the northern and northeastern areas, and frontal and cut-off low-pressure systems in the south.

Western Australian monthly river flows are influenced by Pacific Ocean SST indices (SOI) and Indian Ocean SST indices in a complex manner. In the south of Western Australia the autumn, winter and spring seasonal flows are generally more strongly correlated, with summer seasonal flows becoming more strongly correlated in the north, although the relationship in not as clear or predictable as in eastern Australia.

It is difficult to show a predictable pattern between flooding and SST indices, except in north and northeast Australia where the relationship is sufficiently strong. However, it has been shown qualitatively that larger floods occur in La Nina and neutral years, with smaller floods occurring in El Nino years. There is no clear trend for larger floods in La Nina years than in neutral years. However, there are a higher number of floods in La Nina and neutral years compared to El Nino years.

178 Chapter 7

Summary of Flooding Synoptic Weather Patterns, Climatology and Spatial Rainfall Patterns for Australia Dryland Rivers

7.1 Introduction

About 80% of the Australian landmass is classified as semi-arid to arid (dryland) with a large number, type, and diversity of dryland rivers occupying these drainage basins. These rivers are amongst the most variable in the world, with long periods of low or no flow interrupted by some very large floods.

The climatic and rainfall patterns of Australia are extremely variable and indeed erratic in nature. Average annual rainfall totals range from up 8000mm in north Queensland near Innisfail to approximately 125mm near Lake Eyre in South Australia. Generally rainfall totals decrease with distance inland, although on the west coast of Western Australia, aridity extends to the coastline. Rainfall variability increases dramatically with decreasing average annual rainfall totals, meaning that the interior areas of Australia not only receive the lowest rainfall totals but also have the most variable rainfall. Temporal rainfall variability in particular is linked to the ENSO phenomenon that is a response to conditions in the tropical Pacific Ocean and its overlying atmosphere. Driven by ENSO, Australia experiences periods of drought followed by very wet periods and flooding in a somewhat erratic fashion, but broadly cyclical over 2-10 years (Allan et al 1996).

The synoptic weather patterns that result in dryland river flood events vary widely around Australia, from the hot tropical north where floods are the result of monsoon-related synoptic weather patterns from December to April, to southern Australia where floods result from mid- latitude weather patterns during winter and spring. The drainage divisions that are more centrally located, between the hot northern areas and milder southern regions, are affected by flooding caused by a combination of tropical and mid-latitude weather patterns

179 7.2 Synoptic weather patterns leading to flooding

In the far north of Australia, in the Timor Sea and Gulf of Carpentaria divisions, floods are the result of tropical trough/lows, deep tropical lows (also known as monsoon lows or tropical lows) and tropical cyclones, separately or in combination. In the Gulf of Carpentaria division, eighty percent of floods, including the largest magnitude floods, are the result of tropical trough/lows and deep tropical lows. This is the highest percentage of any drainage division for these two synoptic classes. The largest number of floods in the Gulf of Carpentaria division occur during February, in the height of the northern wet season. Tropical cyclones produce more floods in the Timor Sea division (32% of the total) than in the Gulf of Carpentaria division (7% of the total), although tropical trough/lows still produce the highest number of floods in both divisions. The rivers of the Indian Ocean division, further to the south in the Pilbara (De Gray, Fortescue and Ashburton Rivers) and Greenough-Gascoyne region, also have a high proportion of floods resulting from tropical cyclones. Given that there are approximately equal numbers of cyclones generated in the Coral Sea, the Gulf of Carpentaria and off the northwest coast of Australia, it appears obvious that those in northwest Australia are more effective at producing floods than are those in other regions of Australia affected by tropical cyclones. In the Timor Sea and Indian Ocean divisions (including the Pilbara region and Greenough, Murchison and Gascoyne Rivers) of Western Australia, they also tend to produce the largest floods.

Despite tropical weather patterns being the dominant cause of floods in the Lake Eyre division (northeast Murray-Darling Rivers; Bulloo, Paroo and Warrego Rivers) and the Greenough, Murchison and Gascoyne Rivers, there is a greater variety of flood-producing weather patterns compared to those occurring in the Timor Sea and Gulf of Carpentaria divisions, or in the Pilbara Rivers further to the north. Because they straddle tropical and temperate weather systems the northeast Murray-Darling Rivers receive flood events from the largest variety of weather patterns of any dryland river drainage division in Australia.

The Lake Eyre division has a high number of floods as a result of tropical trough/lows (27% of the total) and deep tropical lows (17% of the total), with a smaller number as result of tropical

180 cyclones (14% of the total). In the Bulloo-Paroo-Warrego and northeast Murray-Darling Rivers, tropical trough/lows cause the highest number of floods with tropical cyclones being more prominent in the Bulloo, Paroo and Warrego Rivers where they caused more floods than deep tropical lows. In the Lake Eyre division, and the Bulloo, Paroo and Warreo Rivers, deep tropical lows and tropical trough/lows tend to cause the largest floods whereas in the northeast Murray-Darling Rivers, easterly dips (continental) produced some of the largest floods. Easterly dips (continental) tend to occur in the winter months in association with favourable upper atmosphere conditions and strengthening high-pressure systems. Easterly dips (continental) are also significant flood-producing weather patterns, causing 7% of floods in the Lake Eyre division, 16% of floods in the Bulloo, Paroo and Warrego Rivers, 12% of floods in the northeast Murray-Darling Rivers, and 5% of floods in the southern Murray-Darling Rivers.

In Western Australia northwest cloudbands are significant flood-producing weather patterns. In the Greenough, Murchison and Gascoyne Rivers, particularly the Gascoyne River, northwest cloudbands cause a significant number of large floods. The Gascoyne Region appears to the only area of Australia where there are floods generated by northwest cloudbands, particularly when combined with cut off-low pressure systems. When the alignments and lengths of northwest cloudbands were plotted by Tapp and Barrel (1984), they found that the Gascoyne Region is affected by more cloudbands than other regions of Australia. Having come straight off the Indian Ocean, these cloudbands have abundant moisture. They occur predominantly from May to August in combination with other synoptic weather patterns. In eastern Australia northwest cloudbands cause a small number of flood events in the Lake Eyre and Murray Darling divisions, and in the Bulloo, Paroo and Warrego Rivers.

The southern rivers of the Murray-Darling division and the southwest coastal division have a small number of floods caused by tropical weather patterns, but the dominant flood producing patterns are mid-latitude in origin; frontal and cut-off low-pressure systems are dominant. In the southern Murray-Darling Rivers there are a higher number of floods from cut-off lows relative to frontal systems than in the southwest coastal division. They occur between August and September. Frontal systems cause more floods in the southwest coastal division despite the major zone for the formation of cut-off lows being off the southwest coast of Australia. These

181 frontal systems occur during the winter months of June, July and August. Cut-off lows were also found to result in a significant number of flood events in the Greenough, Murchison and Gascoyne Rivers, the southwest coastal division, and in the southern Murray-Darling Rivers. Their occurrence in the Greenough, Murchison and Gascoyne Rivers is particularly interesting due to these rivers being located well to the north. It has been found that the area off the coast of Western Australia is a preferred location of formation for cut-off lows (Leslie and Zhao 1999). Despite the majority of the largest floods in the southwest coastal division and the southern rivers of the Murray-Darling division being the result of frontal and cut-off low pressure systems, tropical weather patterns, such as tropical troughs/lows and deep tropical lows, still caused a small number of the largest floods, demonstrating the significant influence of such weather patterns in mid-latitude regions. The majority of deep tropical lows remain near or embedded in the monsoon trough, and it is not uncommon for them to travel in a coherent form across most of the continent before decaying in the southern regions of Australia. This was the case in January 1999 when a deep tropical low caused the largest flood on record in the Lort River at Esperance. The 1974 floods that affected most of eastern Australia was due in part to a deep tropical low that drifted slowly across inland Australia. The ability of deep tropical lows to develop, intensify and remain active over land areas makes them highly effective flood-producing weather patterns.

7.3 Synoptic weather patterns; past and future climate change

In the past, during the mid to late Quaternary period (last 300 ka), the arid zone of Australia has shown distinct wet and dry phases, superimposed on a clear drying trend over the last 130 ka (Hesse et al 2004). These wet and dry phases are linked to interglacial and glacial periods, respectively (Nanson et al 1992). During wet phases where runoff is enhanced, a stronger monsoon is inferred with this monsoon activity rapidly increasing around the last glacial termination (14 ka). Mega-lakes have been attributed to this activity (Bowler et al 2001).

It is generally argued that global warming will result in the poleward movement of climate zones, jet streams and major temperature gradients (Sturman and Tapper 1996). Pittock (2005)

182 states that there has been an observed strengthening and poleward movement of atmospheric low-pressure belts around the north and south poles during the late twentieth century. Gibson (1992) discovered a poleward shift in sub-tropical wind maximum in the southern hemisphere, which he states is in agreement with the expected consequences of an enhanced greenhouse effect. The other major zones of the general global circulation, including the sub-tropical high- pressure belt, mid-latitude westerlies and the trade winds, would be also likely to shift towards the poles. The weather patterns would shift with regards to their maximum seasonal extent. The monsoon trough, that currently extends over far northern Australia during the summer months, would extend further south resulting in a greater area of Australia becoming affected by tropical weather patterns. Tropical trough/lows, deep tropical lows and tropical cyclones (predominantly as rainfall depressions) would move over more southerly areas possibly providing larger more frequent floods in central Australian river basins such as the Lake Eyre Basin. Roshier et al (2001) has suggested that increases in rainfall as a result of climate change are possible.

Magee et al (1995; 2004), Magee and Miller (1998) and Nanson et al (1998) have found evidence for high lake-levels sometime between 90-130ka with Croke et al (1999) proposing that enhanced monsoon circulation was responsible for increased rainfalls. They suggest two possible mechanisms for enhancing the influence of the monsoon in the western Lake Eyre basin. These include the southerly displacement of the monsoon trough, and the enhancement of transient monsoon depressions. Allen (1985) originally put forward the idea of a southerly- displaced monsoon trough to explain Late Quaternary climatic conditions within the Lake Eyre basin. The position of the monsoon trough, or more correctly the ‘monsoon sheer line’, determines the area that experiences a true monsoon climate. The area north of approximately 25°S experiences a summer wet and winter dry climatic regime, which is regarded as monsoonal. McBride and Keenan (1982) have found that over 95% of cyclones, and many monsoon depressions, in the Australian region form on the ‘monsoon sheer line’. This sheer line is anchored to two continental heat lows present over northern areas of Australia during the summer months, one in northwest Australia (the Pilbara low) and the other in northeastern Queensland (the Cloncurry low). The monsoon sheer line is skewed southward, more towards the Pilbara low. Allen et al (1992) suggests that that a southerly displacement of the monsoon

183 sheer line is possible through a deepening of the Pilbara heat low. Logically, if the ‘monsoon sheer line’ is displaced further to the south then heavy rainfall associated with the wet season, and associated tropical cyclones and monsoon depressions, will accompany this. It has been demonstrated that present day flooding in the Lake Eyre basin is linked to extreme rainfall events resulting from tropical cyclone remnants or other synoptic patterns occurring together with enhanced monsoon conditions (Allen et al 1986; and Allen 1990).

Early reports by Pittock (1988) suggested that tropical cyclones could track further south due to warmer sea surface temperatures. However, the CSIRO (2001) report suggests that the areas affected by cyclones are likely to remain largely unchanged, although average wind speeds are likely to increase. The Intergovernmental Panel on Climate Change (IPCC) concluded in its report in 2001 that it is likely that cyclones would be, on average, 5-10% more intense by 2050, with peak rainfall intensities increasing by up to 25%. If there are changes in the ENSO phenomenon, the region of formation for tropical cyclones could change substantially (Pittock 2005). However, to date there is no evidence for a substantial increase in poleward movement of tropical cyclones, either in observations or in model projections. There is much uncertainty regarding the future behaviour of ENSO. Global climate models, given warmer temperatures, provide a divided answer, with some suggesting more El Nino like average conditions, whilst others show little change from present conditions. ENSO extremes may be linked to more extreme floods and droughts due to the intensification of variation in the hydrological cycle. Droughts will likely be more intense in El Nino years due to enhanced evaporation and La Nina rains could be more intense due to the greater water holding capacity of warmer air. Drainage basins in central and northern regions of Australia receive the majority of their floods from monsoon related weather patterns, such tropical trough/lows, tropical cyclones and deep tropical lows, and on average receive larger floods during La Nina years. Enhanced monsoon circulation and increased intensity rainfall in La Nina years, as a result of climate change, could mean these synoptic weather patterns produced larger and more frequent floods in these areas.

The CSIRO (2001) report suggests that for many areas in Australia, their models are predicting an increase in extreme daily rainfall leading to more frequent heavy rainfall events and flooding. Areas of Australia that are affected by flash flooding are likely to receive more of

184 such events. Pittock (2005) has suggested that more intense precipitation events are likely over many areas, with more frequent flooding.

During the cooler dryer glacial periods, it is inferred that the monsoon was far less vigorous and probably retreated further to the north. Hesse et al (2004) has suggested that reduced monsoon rain in global cold stages caused lakes and rivers to dry, vegetation to become more sparse, sand dunes to be more active, and dust transport to be increased. It would be expected during cooler glacial periods in Australia that frontal systems and cut-off low pressure systems would affect more central and northern areas as the warmer tropical weather patterns became restricted to the lower latitudes of far northern Australia. However, frontal systems and cut-off low-pressure systems in general produce higher rainfall totals when and where they interact with tropical air masses. During the cooler periods of the Quaternary, warmer tropical air masses may not have been present over the continent, or were less pronounced, possibly preventing high rainfalls even from mid-latitude weather systems. Under conditions warmer than present, the mid-latitude weather patterns, such as frontal and cut-off low pressure systems that currently only seriously affect the southern half of Australia, would move further south, reducing the number of floods they cause. The more centrally located drainage divisions that currently only receive a small percentage of floods from mid-latitude weather patterns, such as the northern Murray-Darling and Lake Eyre Divisions, the Paroo, Warrego and Bulloo Rivers, and the Gascoyne, Murchison and Greenough Rivers, would probably receive all of their flood events from tropical weather patterns. Under such warmer conditions CSIRO (1992 and 1994) suggest that rainfall over south-central areas of Australia will be reduced due to temperate weather systems tracking further to the south. Pittock (2005) suggests that areas that have a typical Mediterranean climate, which are those areas on the low-latitude edge of the mid-latitude westerlies, will receive less rainfall. This would include parts of southern Australia such as Victoria, South Australia and the southern regions of Western Australia. Some models are suggesting that east coast lows might be more intense with higher sea surface temperatures. This is of importance for central and southern areas of eastern Australia given that such low-pressure systems are currently capable of generating wind speeds in excess of 200km/h, and causing severe flooding due to rainfall and/or storm surge (Bryant 1991). A statistically significant increase in east coast lows in the southern hemisphere has been detected

185 between 1979 and 1999 (Pittock 2005), a period also known to have been subject to global warming.

It is very likely that the monsoon played a crucial role, during the warmer interglacial periods, in the climate of the dryland areas of Australia, over the last 300 ka. There have been wet and dry phases, with their warmer and cooler temperatures, respectively. In Australia it seems logical that global warming will initiate a wet phase with higher temperatures and potentially increased runoff as the monsoon trough affects more southerly areas. At the peak of interglacial periods, dryland river basins in central Australia, such as the Lake Eyre basin, probably received increased runoff and floods as a result of more frequent monsoon depressions and tropical cyclones (as rainfall depressions). Dryland river basins, such as the northern sections of the Murray-Darling basin, that are currently affected by a wide range of weather patterns, would become predominantly affected by tropical weather patterns, with mid-latitude weather only affecting the very southern areas of Australia.

7.4 Rainfall properties in arid to semi-arid regions

In this study rainfall events in all seasons in arid and semi-arid areas, have been found to be widespread and uniform rather than localized and convective. A particularly important finding of this study, supporting the more regionally limited work of Cordery and Fraser (2000) in NSW, is the widespread nature of rainfall over distances greater than 200km across dryland regions of Australia. These results are in contrast to dryland regions studied elsewhere in the world where rainfall has been found to be mostly localised and convective in nature (Sharon 1972; Osburn et al 1979; Bell 1979; Sharon 1981; Wheater et al 1991). However, it has been found that in the arid Todd River catchment around Alice Springs rainfall is slightly less widespread than rainfall in the semi-arid Thomson River. This was more apparent for the smaller storm rainfall totals relative to the larger flooding rainfall totals. It appears that the smaller rainfall totals in more arid areas of Australia are the result of slightly more convective, smaller scale weather patterns that lead to more discrete, localized rainfall totals.

186 In dryland Australia, particularly in the northern areas, the generally uniform nature of rainfall (particularly higher rainfalls often leading to flooding) within river catchments is intrinsically linked to monsoon-related synoptic weather patterns. When such weather patterns are able to penetrate inland regions of Australia, often as a result of a deep tropical low or rainfall depression (ex-tropical cyclone), the resulting rainfall occurs over very large areas. Wet season thunderstorms are less widespread compared to the somewhat more widespread rainfall in the dry season. These wet season thunderstorms are associated with the position of the monsoon trough and the associated West Coast and Queensland troughs.

7.5 The role of ENSO and SSTs in the flooding of Australian arid to semi-arid rivers

The El Nino Southern Oscillation (ENSO) phenomenon and the closely related Sea Surface Temperatures (SSTs) are linked with rainfall and flooding in Australia in a complex manner. Only north and northeastern Australia show a clear, consistent and predictable relationship between flood events and the SOI and SST indices. Other studies confirm that, although there is a large amount of variance explained by factors other than the SOI or SSTs, it is possible to provide predictions of rainfall in northern and northeast Australia (McBride and Nichols 1983 and Chiew et al 1998). Although other areas of Australia don’t show a statistically clear relationship, it’s apparent that flooding is qualitatively related to the SOI. The majority of floods, including some of the largest flood magnitudes, occurred in La Nina and neutral years, whilst El Nino years generally have a lower number, and they are relatively small events.

Correlations between monthly stream flows and the SOI and various SST indices show that in northern and northeast Australia, summer and autumn flows are significantly correlated. In southeastern areas the winter and spring flows show stronger, clear relationships. In other words, correlations are most significant when river flows are seasonally highest as a result of specific synoptic weather patterns, such as frontal systems and cut-off lows in the south and tropical trough/lows, deep tropical lows and tropical cyclones in the north. If the SOI and SST indices are not favourable then the number and/or the intensity of flood producing synoptic weather patterns are reduced.

187

Flooding in Western Australia is correlated with both the SOI and Indian Ocean SST indices but in a much more complex way than in eastern Australia. Seasonal river flows are strongly correlated with autumn and winter Indian Ocean SSTs, particularly in central, north and northwestern Australia. These correlations are negative, in that seasonal river flows are higher when Indian Ocean SSTs are below average. Research by England et al (2006) into the rainfall of southwest Western Australia demonstrates that in dry (wet) rainfall years anomalously warm (cool) water is located off southwest Western Australia, while cooler (warmer) than average water is situated above this in the eastern tropical/sub-tropical Indian Ocean. Further to this Nichols (1989) and Smith (1994) have shown that cold water in the Indian Ocean is related to warm waters around Indonesia.

7.6 Conclusion

The majority of Australia is classified as arid or semi-arid with these areas being drained by a wide variety, of highly variable dryland rivers. Flow in many of these rivers varies from no- flow for many years interspersed by occasional large floods. Flooding of Australian dryland rivers is the result of specific synoptic weather patterns. Such flood producing weather patterns vary widely around Australia, from the hot tropical north where floods are the result of monsoon-related synoptic weather patterns from December to April, to southern Australia where floods result from mid-latitude weather patterns during winter and springtime. Man induced climate change has the potential to shift weather patterns to influence areas further south. The monsoon and its related weather patterns could potentially affect dryland river basins further south. Enhanced monsoon circulation together with potential increases in La Nina rains could produce larger, more frequent flooding. Further to this dryland rivers in the south may receive a smaller number of floods resulting from mid-latitude weather patterns, such as frontal systems and cut-off lows, and larger number from tropical weather patterns such as tropical troughs/low, deep lows and cyclones.

In contrast to many other dryland regions of the world where rainfall is of a localized convective nature, Australia’s dryland rainfall patterns are of a more widespread and uniform

188 nature. These widespread rainfall patterns in the north are related to the Australian monsoon. When rain-producing weather patterns penetrate inland areas as a result of a deep tropical low or ex-tropical cyclone they produce rainfall over large areas. Seasonally rainfall is more widespread in the dry season, when thunderstorms are less likely, compared to the wet season. In western NSW Cordery and Fraser (2000) found that rainfall in arid Western NSW was very different to other regions of the world. Rainfall was spatially uniform and even though thunderstorms did occur, they usually resulted in very little rainfall. They concluded that arid NSW receives the majority of its rainfall from more widespread, non-convective weather patterns.

Flooding of many of Australia’s dryland rivers is intrinsically linked with, although not always conclusively correlated with, the ENSO phenomenon and SSTs. In north and northeast Australia, floods show a statistically predictable relationship with the SOI and SST indices allowing some degree of prediction. Many other Australian dryland rivers show a clear statistical relationship between flooding, monthly streamflow, and the SOI and SST indices, although it is not sufficiently strong to allow accurate and reliable prediction. Qualitatively, flooding is linked with the SOI and SST indices. Larger and more numerous floods occur in La Nina and neutral years in contrast to El Nino years.

189 References

ADAMS, M. 1986. A theoretical study of the inland trough of northeastern Australia. Australian Meteorological Magazine 34, 85-92.

ALLEN, R.J. 1985. The Australian summer monsoon, teleconnections and flooding in the Lake Eyre basin. South Australian Geographical Papers 2, Royal Geographic Society. Australasian (South Australia Branch), Adelaide, 47pp.

ALLEN, R.J., BYE, J.A.T., HUTTON, P. 1986. The 1984 flooding of Lake Eyre South, Trans. Roy. Soc. S. Aust. 110, 81-87.

ALLEN, R.J. 1990. Climate. In: Natural history of the north east desert. Eds: TYLER, M.J., TWIDALE, C.R., DAVIES, M., WELLS,C.B. pp 81-84. Royal Society of South Australia, Adelaide.

ALLEN, R.J., LINDESAY, J., PARKER, D. 1996. El Nino Southern Oscillation and Climatic Variability. CSIRO, Melbourne, Australia.

ALLEN, R.J., MITCHELL, C.D., PITTOCK, A.B.1992. The greenhouse effect: regional implications for Western Australia. Annual report 1990-1991. Environment Protection Authority, Perth.

ARMOSTRONG, J.G., COLQUHOUN, J.R. 1976. Intense rainfalls from the thunderstorms over the Sydney metropolitan and Illawarra districts – 10 and 11 March 1975. Department of Science, Bureau of Meteorology Technical Report No 19.

BAKER, V.R. 1977. Stream Channel Response to Floods, with examples from Central Texas, Geological Society of America, Bulletin, 88, 1057-1071.

BOWLER, J.M., WYRWOLL, K.H., LU, Y. 2001. Variations of the northwest Australian monsoon over the last 300,000 years: the palaeohydrological record of the Gregory (Mulan) Lakes system. Quaternary International Volume 83-85, 63-80.

BROADBRIDGE, L.W., HANSTRUM, B.N. 1998. The relationship between tropical cyclones near Western Australia and southern oscillation index. Australian Meteorology Magazine 47, 183-189.

BRUNT, A.T. 1958. Analysis of Two Queensland Storms, Proceedings of the Conference on Extreme Precipitation: Bureau of Meteorology, Melbourne, 1958, pp 191-244.

BRYANT, E.A. 1991. Natural Hazards, Cambridge University Press.

BUREAU OF METEOROLOGY, 1998. Gascoyne-Murchison, Western Australia: Climatic Survey. Published by Bureau of Meteorology, Melbourne, Australia.

190 BUREAU OF METEOROLOGY, 2000. Climatic Atlas of Australia: Rainfall. Published by Bureau of Meteorology, Melbourne, Australia.

BUREAU OF METEOROLOGY, 2003. Climate of Australia, 12th May 2003 (online), http://www.bom.gov.au/lam/climate/levelthree/ausclim/ausclim.htm

CARACENA, F., FRITSCH, J.M. 1983. Focusing mechanisms in the Texas Hill Country flash floods. Monthly Weather Review 111, 2319-2332.

CHIEW, F.H.S., PIECHOTA, T.C., DRACUP, J.A., MCMAHON, T.A. 1998. El Nino/Southern Oscillation and Australian rainfall, streamflow and drought: Links and potential for forecasting. Journal of Hydrology 204, 138-149.

COMMONWEALTH SCIENTIFIC AND INDUSTRIAL RESEARCH ORGANISATION (CSIRO) 1970. The Australian Environment. Melbourne University Press, Melbourne, Australia.

CSIRO 1992. Climate change scenarios for the Australian region, Climate Impact Group, CSIRO Division of Atmospheric Research, 6pp.

CSIRO 1994. Climate change scenarios for the Australian region, Supplementary Statement for Greenhouse 94, Climate Impact Group, CSIRO Division of Atmospheric Research, 10pp.

CSIRO 2001. Climate change projections for Australia. CSIRO Division of Atmospheric Research, 2pp.

CORDERY, I., FRASER, J. 2000. Characteristics of rainfall and transmission losses in the Australian Arid Zone. Third International Hydrology and Water Resources Symposium, November 2000, Barton, Australian Capital Territory, Australia.

CORDERY, I., PILGRIM, D.H., DORAN, D.G. 1983. Some hydrological characteristics of arid western New South Wales. Hydrological and Water Resources Symposium, Institute of Engineers Australia, National Conference Publication 83/13, 287-297.

COSTA, J.E. 1987. A Comparison of the Largest Rainfall-Runoff Floods in the United States with those of the Peoples Republic of China and the World, Journal of Hydrology 96, 101-115.

CROKE, J.C., MAGEE, J.W., WALLENSKY, E.P. 1999. The role of the Australian Monsoon in the western catchment of Lake Eyre, central Australia, during the Last Interglacial. Quaternary International 57/58, 71-80.

DAVIDSON, N.E., HOLLAND, G.J.1987. A diagnostic analysis of two intense monsoon depressions over Australia. Monthly Weather Review 115, 380-392.

191 DHAR, O.N., RAKHECHA, P.R. 1979. Incidence of heavy rainfall in the Indian Desert region. The hydrology of areas of low precipitation. December 1979. IAHS-AISH Publication no 128.

DROSDOWSKY, W. 1993: Potential predictability of winter rainfall over southern and eastern Australia using Indian Ocean sea-surface temperature anomalies. Australian. Meteorological Magazine 42, 1-6.

DROSDOWSKY, W., CHAMBERS, L.E. 1998: Near global sea surface temperature anomalies as predictors of Australian seasonal rainfall. BMRC Research Report No. 65. 39 pp. Bureau of Meteorology, Australia.

DOUPE, R.G., PETTIT, N.E. 2002. Ecological perspectives on regulation and water allocation for the Ord River, Western Australia. River Research Application 18, 307-320.

ENGLAND, M.H., UMMENHOFER, C.C., SANTOSO, A. 2006. Interannual rainfall extremes over southwest Western Australia linked to Indian Ocean climate variability. Journal of Climate 19, 1948-1969.

EVANS, J.L., ALLEN, R.J. El Nino/Southern Oscillation modification to the structure of the monsoon and tropical cyclone activity in the Australian region. International Journal of Climatology, Vol. 12, 611-623.

EVAN, J., BEWICK, B. 2001. Technical Report 73: The Wollongong Flash Flood Event, 15- 19 August 1998. Bureau of Meteorology, Australia.

FINLAYSON, B.L., MCMAHON, T.A. 1988. Australia v the World: a comparative analysis. In Warner, R.F., editor, Fluvial Geomorphology of Australia. Sydney Academic Press, 1-16.

FOSTER, I.J., LYONS, T.J. 1984. Tropical cyclogenesis: a comparative study of two depressions in northwest Australia. Quarterly Journal of the Royal Meteorological Society 110, 105-119.

GENTILLI, J. 1986. Climate. In Jeans, D.N., editor, The Natural Environment. Sydney University Press, 14-47.

GIBSON, T.T. 1992. An observed poleward shift of the southern hemisphere sub-tropical wind maximum – a greenhouse symptom? International journal of climatology, vol 12, 637-649.

GILLET, N.P., KELL, T.D., JONES, P.D. 2006. Regional climate impacts of the Southern Annualar Mode. Geophysical Research Letters 33, doi:10.1029/2006GL027721.

GRAF, W.L. 1988. Fluvial processes in dryland rivers. Springer-Verlag, Berlin.

192 GROOTEMAAT, G.D. 2000. Spatial characteristics of high magnitude rainfall events and channel response in the Illawarra region. BSc (Hons) thesis, School of Geosciences, University of Wollongong: 155pp.

HANSTRUM, B.N., WILSON, K.J., BARRELL, S.L. 1990. Prefrontal troughs over southern Australia. Part I: a climatology. Weather and Forecasting 5, 22-30.

HEATHCOTE, R.L. 1994. Australia. Second Edition. Longman Scientific and Technical, Essex, England.

HENDON, H.H., THOMPSON, D.W.J, WHEELER, M.C. 2007. Australian rainfall and surface temperature variations associated with the southern hemisphere annular mode. Journal of climate 20, 11, p 2452.

HESSE, P.P., MAGEE, J.W., SANDER, V.D.K. 2004. Late Quaternary climates of the Australian arid zone: a review. Quaternary International 118-119, 87-102.

HOLLAND, G.J., LYNCH, A.H., LESLIE, L.M.1987. Australian east coast cyclones. Part I: Synoptic overview and case study. Monthly Weather Review 115, 3024-3036. HOPKINS, L.C., HOLLAND, G.J. 1997. Australian heavy-rain days and associated east coast cyclones: 1958-1992. Journal of Climate 10, 621-635.

HIRSCHBOECK, K.1987. Catastrophic flooding and atmospheric circulation anomalies. In Mayer, L.& Nash, D., Editors, Catastrophic Flooding. Allen and Unwin, Winchester, USA. JURY, M.R., LINDESAY, J.A., WITTMEYER, I.1993. Flood episodes in central South Africa from satellite and ECMWF data. Suid-Afrikaanse Tydskrif Vir Wetenskap Voulume 89, 263- 269.

KANE, R.P. 1997. On the relationship of ENSO with rainfall over different parts of Australia. Australian Meteorological Magazine 46, 39-49.

KEIM, B.D. 1996. Spatial, synoptic and seasonal patterns of heavy rainfall in the southeastern United States. Physical Geography 17, 313-328.

KINGSFORD, R.T., NORMAN, F.I. 2002. Australian water birds - products of the continent's ecology. EMU 102 (1): 47-69 2002.

KNIGHTON, A.D., NANSON, G.C. 1994. Water holes and their significance in the anastomosing channel system of Cooper Creek, Australia. Geomorphology 9, 311-324.

KNIGHTON, A.D., NANSON, G.C. 1997. Distinctiveness, diversity and uniqueness in the arid zone river systems. In Editor: Thomas, D.S.G., Arid Zone Geomorphology: Process, Form and Change in Drylands. John Wiley and Sons.

KNIGHTON, A.D., NANSON, G.C. 2000. Waterhole form and process in the anastomosing channel system of Cooper Creek, Australia. Geomorphology 35, 101-117.

193

KNIGHTON, A.D., NANSON, G.C. 2001. An event based approach to the hydrology of arid zone rivers in the channel country of Australia. Journal of Hydrology 254 (1), 102-123.

KONRAD, C.E. 1997. Synoptic scale features associated with warm season heavy rainfall over the interior southeastern United States. Weather and Forecasting 12, 557-571.

KOTWICKI, V. 1986. Floods of Lake Eyre. Engineering and Water Supply Department, Adelaide, South Australia.

KOTWICKI, V., ISDALE, P. 1991. Hydrology of Lake Eyre: El Nino link. Palaeogeography, Palaeoclimatology, Palaeoecology 84, 87-98.

KOMUSCU, A.E., SEYFULLAH, C. 1998. Analysis of meteorological and terrain features leading to the Izmir flash flood, 3-4 November 1995. Natural Hazards 18, 1-25. LESLIE, L.M., SKINNER, T.C.L. 1994. Real-time forecasting of the Western Australian summertime trough: evaluation of a new regional model. Weather and Forecasting 9, 371-383.

LESLIE, L.M., HOLLAND, G.J., LYNCH, A.H. 1987. Australian east coast cyclones. Part II: Numerical modeling study. Monthly Weather Review 11, 3037-3053. LESLIE, L.M., ZHAO, S.X. 1999. Cut-off low-pressure systems over southern Australia: climatology and case study. International Journal of Climatology 19, 1633-1649.

LINDESAY, J.A., JURY, M.R. 1991. Atmospheric circulation controls and characteristics of a flood event in central South Africa. International Journal of Climatology 11, 609-627.

LYNCH, A.H. 1987. Australian east coast cyclones III: case study of the storm of August 1986. Australian Meteorological Magazine 35, 163-170.

MADDOX, R.A., CHAPPELL, C.F 1978. Meteorological aspects of twenty significant flash flood events. In “ Conference on flash floods events: Hydrometeorological Aspects and Human Aspetcs,” Prepared by: American Meteorological Society, Boston. pp 1-9.

MADDOX, R.A., CHAPPELL, C.F., HOXIT, L.R. 1979. Synoptic and meso-scale aspects of flash flood events. Bulletin American Meteorological Society 60, 115-123.

MADDOX, R.A, CANOVA, F., HOXIT, L.R. 1980. Meteorological characteristics of flash flood events over the western United States. Monthly Weather Review 108, 1866-1877.

MAGEE, J.W., BOWLER, J.M., MILLER, G.H., WILLIAMS, D.L.G. 1995. Stratigraphy, sedimentology, chronology, and palaeohydrology of Quaternary lacustrine deposits at Madigan Gulf, Lake Eyre, South Australia. Palaeogeography, Palaeoclimatology, Palaeoecology 113, 3-42.

194 MAGEE, J.W., MILLER, G.H. 1998. Lake Eyre palaeohydrology from 60ka to present; beach ridges and glacial maximum aridity. Palaeogeography, Palaeoclimatology, Palaeoecology 144, 307-329.

MAGEE, J.W., MILLER, G.H. SPOONER, N.A. QUESTIAUX, D. 2004. Continuous 150 k.y. monsoon record from Lake Eyre, Australia: Insolation-forcing implications and unexpected Holocene failure. Geology, 32 (10): 885-888.

MCBRIDE, J.L., KEENAN, T.D. 1982. Climatology of tropical cyclone genesis in the Australian region. Journal of Climatology 2, 13-33.

MCBRIDE, J.L, NICHOLS, N. 1983. Seasonal relationships between Australian rainfall and the southern Oscillation. Monthly Weather Review 111, 1998-2004.

MCMAHON, T.A., FINLAYSON, B.L., HAINES, A.T. & SRIKANTHAN, R. 1992. Global runoff-continental comparisons of annual flows and peak discharges, Catena Verlag, Germany.

MCMAHON, T.A., FINLAYSON, B.L 2003. Droughts and anti-droughts: the low flow hydrology of Australian rivers. Freshwater Biology 48 (7): 1147-1160 Jul 2003. MCILWAIN, S.A., 1999. Wollongong City storm August 1998, BSc (Hons) thesis, School of Geosciences, University of Wollongong, 110p.

MENEGHINI, B., SIMMONDS, I., SMITH, I.N. 2007. Association between Australian rainfall and the Southern Annular Mode. International Journal of Climatology 27, 109-121.

NANSON, G.C. & HEAN, D. 1985. The West Dapto flood of February 1984: Rainfall characteristics and channel changes, Australian Geographer 16, 4, 249-257.

NANSON, G.C., PRICE, D.M., SHORT, S.A.1992. Wetting and drying of Australia over the past 300ka. Geology 20, 791-794.

NANSON, G.C., CALLEN, R.A., SHORT, S.A. 1998. Hydroclimatic interpretation of Quaternary shorelines on South Australian playas. Palaeogeography, Palaeoclimatology, Palaeoecology 144, 281-305.

NICHOLLS, N. 1989. Sea surface temperatures and Australian winter rainfall. Journal of Climate 2, 965-973.

NICHOLLS, N., KARIKO, A. 1993. East Australian rainfall events: Interannual variations, trends, and relationships with the Southern Oscillation. Journal of Climate 6, 1141-1152.

OSBOURNE, H.B., RENARD, K.G., ROGER SIMANTON, J.1979. Dense networks to measure convective rainfall in southwestern United States. Water Resources Research 15 (6) 1701-1711.

195 PAGE, K.J. 1988. Bankfull discharge frequency of the Murrumbidgee River, New South Wales. In Warner, R.F. (Editor) Fluvial Geomorphology of Australia. Academic Press, Sydney, 267- 281.

PIECHOTA, T.C., CHIEW, F.H.S., DRACUP, J.A., MCMAHON, T.A.1998. Seasonal streamflow forecasting in eastern Australia and the El Nino-southern oscillation. Water Resources Research 34, 3035-3044.

PIERREHUMBERT, C.L., KENNEDY, M.R. 1982. The use of Adjusted United States data to Estimate Probable Maximum Precipitation, Proceedings of the Workshop on Spillway Design: Melbourne 7-9 October 1981.

PITTOCK, A.B. 1973. Global meridional interactions in stratosphere and troposphere. Quart. J.R. Met. Soc. Vol 99, 424-437.

PITTOCK, A.B. 1975. Climatic change and the patterns of variation in Australian Rainfall, Search Vol. 6 No. 11-12.

PITTOCK, A.B. 1984. On the reality, stability, and usefulness of southern hemisphere teleconnections. Aust. Met. Mag. 32, 75-82.

PITTOCK, A.B. 2005. Climate change, turning up the heat. CSIRO Publishing Melbourne, Australia.

PUCKRIDGE, J.T 1999. The role of hydrology in the biology of dryland rivers. In Kingford R.T., Editor, A Free Flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service NSW.

PUCKRIDGE, J.T., WALKER, K.F., COSTELLOE, J.F. 2000. Hydrological persistence and the ecology of dryland rivers. Regulated Rivers Research and Management 16, 385-402.

PUCKRIDGE, J.T., SHELDON, F., WALKER, K.F., BOULTON, A.J. 1998. Flow variability and the ecology of large rivers. Marine and Freshwater Research 49, 55-72.

REINFELDS, I., NANSON, G.C. 2001. Torrents of terror: the August 1998 storm and the magnitude, frequency and impact of major floods in the Illawarra region of New South Wales. Australian Geographical Studies 39 (3) 335-352.

RENWICK, J., THOMPSON, D.W.J. 2006. The Southern Annular Mode and the New Zealand climate. Water and Atmosphere (online). Vol 14, 2. url: http://www.niwa.cri.nz/pubs/wa/ma/14-2/sam

ROPELEWSKI, C.F., HALPERT, M.S. 1987. Global and regional scale precipitation patterns associated with the El Nino\southern oscillation. Monthly Weather Review 115, 1606-1626.

196 ROPELEWSKI, C.F., HALPERT, M.S. 1989. Precipitation patterns associated with the high index phase of the southern oscillation. Journal of Climate 2, 268-284.

ROSHIER, D.A., WHETTON, P.H., ALLEN, R.J., ROBERTSON, A.I. 2001. Dustribution and persistence of temporary wetland habitats in arid Australia in relation to climate. Austral Ecology, 26, 371-384.

SHARON, D. 1972. The spottiness of rainfall in a desert area. Journal of Hydrology 17, 161- 175.

SHARON, D. 1974. The spatial patterns of convective rainfall in Sukumaland, Tanzania-a statistical analysis. Arch. Meteor. Geophys. Bioklim., 22B, 201.

SHARON, D. 1979. Correlation analysis of the Jordan Valley rainfall field. Monthly Weather Review 107, 1042-1047.

SHARON, D. 1981. The distribution in space of local rainfall in the Namib Desert. Journal of Climatology 1, 69-75.

SHARON, D., 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.

SHEPHERD, D.J., COLQUHOUN, J.R. 1985. Meteorological aspects of an Extraordinary Flash Flood Event near Dapto, New South Wales, Australian Meteorological Magazine, 33, 87-102.

SIMPSON, H.J., HERCZEG, A.L., CANE, M.A., ZEBIAK, S.E. 1993. Annual river discharge in southeastern Australia related to El Nino-southern oscillation forecasts of sea surface temperature. Water Resources Research 29, 3671-3680.

SMITH, I. 1994. Indian Ocean sea-surface temperature patterns and Australian winter rainfall. International Journal of Climatology 14, 287-305.

SMITH, R.K., REEDER, M.J., TAPPER, N.J., CHRISTIE, D.R. 1995. Central Australian cold fronts. Monthly Weather Review 123, 16-38.

SPEER, M., GEERTS, B. 1994. A synoptic-mesoalpha-scale climatology for the Southern Hemisphere. Monthly Weather Review 122, 2239-2256.

STONE, R., AULICIEMS, A. 1992. SOI relationships with rainfall in eastern Australia. International Journal of Climate 12, 625-636.

STURMAN, A.P., TAPPER, N.J. 1996. The weather and climate of Australia and New Zealand. Oxford University Press, Melbourne, Australia.

197 TAPP, R.G., BARREL, S.L. 1984. The northwest Australian cloudband: climatology, characteristics and factors associated with development. Journal of Climatology 4, 411-424.

TAPPER, N., HURRY, L. 1993. Australian weather patterns and introductory guide. Dellasta PTY LTD, Victoria, Australia.

TRELOAR, A. B. A., AND STERN, H. 1993: A climatology and synoptic classification of Victorian severe thunderstorms. Preprint volume: 4th International Conf. on Southern Hemisphere Meteorology and Oceanography, Hobart, Australia.

TWIDALE, C.R., CAMPBELL, E.M. 1993. Australian Landforms: Structure, Process and Time. Gleneagles Publishing, Adelaide, Australia.

WARNER, R.F. 1986. Hydrology. In Jeans, D.N., editor, The Natural Environment. Sydney University Press, 14-47.

WARNER, R.F. 1988. Fluvial Geomorphology in Australia: some perspectives. In Warner, R.F., editor, Fluvial Geomorphology of Australia. Sydney Academic Press, 1-16.

WATTS, R.J. 1999. Biodiversity in the Paroo River and its wetlands. In Kingford R.T., Editor, A Free Flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service NSW.

WEEKS, B. 1992. Flooding in the Illawarra: an historical perspective. BSc (Hons) thesis, School of Geosciences, University of Wollongong: 196p.

WHETTON, P.H., PITTOCK, A.B. 2001. Fluctuations of the relationship between ENSO and northeast Australian rainfall. Climate Dynamics 17, 421-432.

WILLIAMS, G.E. 1970. The central Australian stream floods of February-March 1967. Journal of Hydrology 11, 185-200.

WILSON, K.J., STERN, H. 1985. The Australian summertime cool change: synoptic and sub- synoptic scale aspects. Monthly Weather Reviews 113, 177-201.

WOLTER, K., TIMLIN, M.S. 1993. Monitoring ENSO in COADS with a seasonally adjusted principal component index. Proc. of the 17th Climate Diagnostics Workshop, Norman, OK, NOAA/N MC/CAC, NSSL, Oklahoma Clim. Survey, CIMMS and the School of Meteor., Univ. of Oklahoma, 52-57.

WRIGHT, W.J. 1989. A synoptic climatological classification of winter precipitation in Victoria. Australian Meteorological Magazine 37, 217-229.

198 YOUNG, W.J. 1999. Hydrologic descriptions of semi-arid rivers: An ecological perspective. In Kingford R.T., Editor, A Free Flowing River: The Ecology of the Paroo River. National Parks and Wildlife Service NSW.

199