IMPACT OF CLIMATE CHANGE ON

DESIGN WIND SPEEDS IN

CYCLONIC REGIONS

IMPACT OF CLIMATE CHANGE ON DESIGN WIND SPEEDS IN

CYCLONIC REGIONS

(revised edition)

J.D. Holmes

June 2011

1

CONTENTS Page

Executive summary …………………………………………………………………………………………… 2 1. Introduction .………………………………………………………………………………………………… 4 2. History of cyclonic design wind speeds and regional system in AS/NZS1170.2.. 9 3. Simulation methods for prediction of cyclonic wind speeds …………………………. 18 4. Vertical profiles in tropical cyclones and hurricanes …………………………............. 24 5. Inland weakening and penetration of tropical cyclones ………………………………… 31 6. Observed effects of climate change on tropical cyclones worldwide …………….. 39 7. Observed trends in tropical cyclone activity in the Australian region ….…………. 45 8. Predicted future effects of climate change on tropical cyclones ……………………. 50 9. Observations and projections for the Northern Territory …………………………….… 57 10. Comments and observations from Cyclone ‘Yasi’ ………………………………………… 66 11. Conclusions and recommendations…………………………………………………………….. 70 Appendix A - Maximum recorded wind gusts from tropical cyclones at selected locations in Australia …………………………………………………………………………………………. 77 Appendix B – Proposal for a new definition of the peak gust for AS/NZS1170.2 …. 81 Appendix C – Wind field model for Cyclone ‘Yasi’ ………………………………………………. 92 Acknowledgements …………………………………………………………………………………………… 105

2

Executive Summary

The objective of the review described in this report, carried out by JDH Consulting on behalf of the Australian Building Codes Board, is to determine whether there is sufficient information and justification to change design wind speeds in cyclonic regions of Australia in the Australian Standard for Wind Actions (AS/NZS1170.2): a) with reference to currently available wind data, b) with reference to recent extreme events, such as Cyclone ‘Yasi’, and c) with particular reference to climate change. The original (2008) version of this report recommended that the situation be reviewed every 3 to 5 years to consider improved model predictions, and any evidence of changing patterns of climate change on tropical cyclones in the Australian region. This report is the first of those reviews. All chapters of the original report have been revised and an additional chapter (Chapter 10 on Cyclone ‘Yasi’) and two new Appendices have been added for the 2011 version.

The history of the regional wind speed system for the tropical-cyclone affected coastlines of Australia, developed since the 1970s, is described, and the probabilistic simulation approach used for hurricane wind speed prediction in the United States, and used in the past for guidance in Australia, is outlined. Available data on the vertical profile of cyclonic wind gusts with height is reviewed, and the background to the gust profiles in the Standard is given. The inland weakening and penetration of tropical cyclones is reviewed; in comparison with U.S. data, on average the current regional widths in AS/NZS1170.2 appear to be adequate. Recent studies of climate change effects on tropical cyclones are reviewed, including work reported between 2008 and 2011. These indicate that in the Australian Region, the total number of cyclones has diminished. This can be related to the preponderance of El Nino events affected Australia’s climate during the last few decades. However, there is evidence that the number of the more severe events has increased. Simulations of future climate, with projected increases in CO2 concentrations, also predict fewer cyclones, but further increases in the more severe tropical cyclones and a southerly drift in the genesis region on the Queensland coast. This may indicate a greater risk of a severe cyclone affecting Brisbane and South-east Queensland than was assumed in the past. A review of the recent cyclones affecting the Northern Territory indicates that the risk to the northern coastal strip and offshore islands may also be greater than currently implied by the Standard. However, upgrading of Darwin or the Northern Territory Coast to Region D is not justified on present evidence. A review of the wind field overland in Cyclone ‘Yasi’ (February 2011) is given. It was found that the maximum wind gust speeds over the land were up to 94% of the design wind speed (V500) for the majority of buildings in Region C.

3

A number of recommendations have been made both with respect to the Standard, AS/NZS1170.2, and more generally with regard to the database and measurements of the Bureau of Meteorology are made. The main recommendations are as follows:

The peak wind gust used as the basis for the Australian/New Zealand Standard AS/NZS1170.2 needs re-definition, as it clearly is not the ‘3-second’ gust used by the Bureau of Meteorology and the World Meteorological Organization. A proposal for the re-definition of the peak gust in AS/NZS1170.2 is given in this report (Appendix B).

The current FC and FD factors in Section 3.4 of AS/NZS1170.2 should be incorporated into the regional wind speeds for cyclonic Regions C and D.

An extension of the current Region D boundary in the northeast direction along the coast of Western Australia to 15oS is recommended. There is also a case to upgrade the most northerly offshore islands of the Northern Territory and the northerly part of the Cobourg Peninsula. However, a decision on the latter should be deferred until a detailed re-assessment of the maximum wind speeds in Cyclones ‘Thelma’, ‘Ingrid’ and ‘Monica’ has been made.

An upgrading of the current Region B in Queensland between 25oS and 27oS to Region C, as previously proposed, should also be put ‘on hold’ until further evidence, from climate model simulations with higher resolution, is available. Although there is a consensus from current models that cyclones will occur on average about 2 degrees further south, they also agree that fewer cyclones will occur by up to 30%. However, an informative note should be added to AS/NZS1170.2 and AS4055, warning of the possibility of cyclones up to Category 3 penetrating further south along the Queensland coast.

Ongoing monitoring should be carried out of predictions emerging from global and regional climate models as their resolution and reliability improves.

Further rectification of errors and removal of inconsistencies in the assessment of wind speed in the tropical cyclone database maintained by the Bureau of Meteorology is required. Also an indication of the averaging time. accuracy and reliability of wind speed values given in this database should be included, particularly those that are derived indirectly from satellite observations.

A recent project (funded in part by the Department of Climate Change and Energy Efficiency) on the response of the anemometer systems used in the past and present by the Bureau of Meteorology has shown there are significant differences between the pre-1990 readings from Dines anemometers and those from the current Automatic Weather Stations, particularly at stations affected by tropical cyclones. The Bureau of Meteorology should provide details of the change-over dates of the anemometers so that appropriate corrections of the historical database can be made.

More resources should be devoted to the instrumentation of cyclonic wind speeds in Australia, including the adoption of proven technologies from overseas, such as dropwindsondes and mobile tower arrays.

4

1. Introduction 1.1 Project scope Buildings and many other structures in Australia are designed for wind loads using

the Australian/New Zealand Standard ‘Structural design actions, Part 2: wind actions’

(Standards Australia, 2011). This document is called up by the Building Code of

Australia (BCA) administered by the Australian Building Codes Board (ABCB).

The Commentary to AS/NZS1170.2:2002 (Standards Australia 2002, but now

withdrawn) stated: “The Standard does not attempt to predict the effects of possible

future climatic changes, as the evidence for changes in wind speeds is inconclusive.”

With clear evidence from climate scientists that global warming is occurring with

apparent effects on the frequency and intensity of windstorms, it is clear that

community and design professionals expect more guidance in the Standard when

planning and designing structures with anticipated lives of 25 to 100 years.

The Australian Building Codes Board has particular concerns about possible

increased cyclonic wind speeds for the northern coastline of Australia, and has

specified the following scope in 2008 for the study which is the subject of this report.

There is also some concern about the implications of the severity of the winds

produced by Tropical Cyclone ‘Yasi’, which crossed the coast of north Queensland on

February 3rd 2011.

Objective The objective of the study is to determine whether there is sufficient information

and justification to change design wind speeds in cyclonic regions of Australia a) with

reference to currently available wind data, and b) with particular reference to

climate change.

Work Plan (2008 report) To achieve the above objective, the study should include two parts:

o A review of available cyclonic wind speed data and other relevant data to determine whether the current definitions of cyclonic regions and wind speeds are appropriate including the boundaries of the cyclonic regions.

o An assessment on how climate change would impact on design wind speeds. This should include

A review of probable scenarios for climate change

A review of cyclonic wind models

5

An analysis of how climate change would impact on design wind speeds in cyclonic regions including the possibility of extending the cyclone region boundaries

Reference Sources The study is to take into account any known relevant information sources, including

but not limited to, the following:

o AS1170.2-1989 o AS/NZS1170.2:2011 o Bureau of Meteorology data on cyclonic regions o Cyclone Testing Station Report on Cyclone ’George’ o Nicholls report (Community Group for the ‘Review of NT Cyclone Risks’) o CSIRO Research Reports on Climate Change o BRANZ Report ‘An assessment of the need to adapt buildings for

unavoidable consequences of climate change’ o Data from Indian Ocean Climate Change Initiative o Data from Queensland Climate Change Centre for Excellence

For the current (2011) revision of the report, the following additional tasks were proposed:

To consider the comments made by Dr. Jeff Kepert of the Bureau of Meteorology on the 2008 report

To consult with the climate change experts who were consulted in 2008, regarding more recent predictions of the computer models on cyclone occurrences and strengths, and amend the report if necessary.

Re-assess the boundaries of the cyclonic regions following tropical cyclones that have occurred since 2008, particularly Cyclone ‘Yasi’ in February 2011.

To make recommendations on the future values of the Fc and FD, including whether or not these factors should be in fact be retained, or amalgamated with the regional wind speed, VR.

Revise Chapter 9 on ‘Observations and predictions for the Northern Territory’ following publication of a paper in 2009 in the Journal of Applied Meteorology and Climatology of the American Meteorological Society by G. Cook and M. Nicholls: “Estimation of tropical cyclone wind hazard for Darwin: comparison with two other locations and the Australian wind loading code”, and subsequent submission of a discussion paper commenting on it.

Revise Chapter 4 on ‘Vertical profiles in tropical cyclones and hurricanes’ following recent work on dropwindsondes from aircraft flying through hurricanes and on radar measurements.

Consider the implications for the cyclone regions of the re-definition of the 3-second gust by the Bureau of Meteorology in recent years (as revealed by a recent project carried out for the Department of Climate Change and Energy Efficiency).

6

Produce firm documented proposals for submission for discussion and recommendation by subcommittee BD006-02 of Standards Australia.

1.2 Australian Tropical Cyclone database The National Climate Centre of the Bureau of Meteorology maintains a database of

information on tropical cyclones from 1906 onwards recorded in the Australian

Region, which is defined as the region south of the Equator between 105o and 160oE

longitude. The database has extensive column fields for parameters relevant to the

tropical cyclones that have occurred in the Region. The database lists daily

observations, or estimations, of many variables. The following are relevant to the

present report: latitude and longitude of the centre of the storm, central pressure,

outer radius, eye diameter, maximum wind speed and direction, direction and speed

of storm movement.

Since about 1970 satellite information has been available. This has enabled virtually

all cyclones in the Australian Region to be identified and their position to be located

fairly accurately. Using a technique devised by Dvorak (1984) satellite images have

also been used to estimate the strength of cyclones. The cloud patterns in the

vicinity of the eye are used to estimate the maximum wind speed (assumed to be a

sustained wind speed at 10 metres above the surface), and hence through a formula,

the central pressure. Data derived in this way are generally those found in the

database, and only rarely are reliable direct observations available in the Australian

Region. This contrasts with the Atlantic Ocean where, in most cases, aircraft flights

have been made to determine the storm strength with reasonable accuracy. It has

been suggested that changes to satellite images interpreted by the Dvorak technique

have resulted in false trends in intensity estimates over the last 30-40 years in the

Australian database.

It is also known that there are many errors in the Tropical Cyclone database, and the

Bureau of Meteorology currently has a programme of work to correct these.

1.3 Occurrence of tropical cyclones in the Australian Region

Figure 1.1 shows the average number of occurrences of tropical cyclones in 2o 2o

squares in Australian waters based on data from the tropical cyclone database

maintained by the Bureau of Meteorology, (Abbs et al., 2006).

Figure 1.1 shows four regions of significant tropical cyclone occurrence: the north-

west coastline of Western Australia near Broome, the Timor Sea off the Northern

Territory coast, the Gulf of Carpentaria, and the Coral Sea offshore from Queensland.

7

It is noted that cyclonic activity is closest to the coast line offshore from the Western

Australian coastline near Broome.

Figure 1.1. Average number of days per year that tropical cyclones occurred in the Australian region in 1970-2000 (from Abbs et al., 2006).

The regions of cyclonic activity shown in Figure 1.1 are largely reflected in the

present regional zoning system for wind speeds in the Australian Standard

AS/NZS1170.2:2011, although consideration of cyclonic wind speeds are not

explicitly required by the Standard for Perth in Western Australia, or for the coastline

of Northern New South Wales. It is noted that Cyclone ‘Alby’ produced significant

wind speeds (up to 75 knots) in the Perth area in 1978.

1.4 Structure of this report In the following Chapter 2, the history of the development of design wind speeds

from tropical cyclones, and the associated regional boundaries in

AS/NZS1170.2:2011 and its predecessors, is outlined. In Chapter 3, probabilistic

simulation methods that are currently used to determine hurricane design wind

speeds in the United States, and have been used for guidance in Australia in the past,

are discussed. Vertical profiles of wind speeds in tropical cyclones are discussed in

Chapter 4. This is relevant, not only for conversion of design wind speeds at the

standard reference height of 10 metres in open over-land terrain to other heights

and other terrains, but also to convert upper-level wind speeds from simulation

models, and those given in the Bureau of Meteorology database, to surface

conditions. Chapter 5 reviews recent work on the inland penetration and weakening

8

of tropical cyclones and hurricanes and compares it with the weakening implied by

the regional system in Australia.

Climate change issues are discussed in Chapters 6-8. Chapter 6 reviews observations

of global changes in tropical cyclone activity in relation to increased greenhouse gas

concentrations and sea surface temperatures. In Chapter 7, studies of changes in

cyclonic activity in the Australian Region during the last few decades are reviewed.

Recent simulations of future changes in tropical-cyclone activity in the Australian

Region are discussed in Chapter 8. Chapter 9 reviews some recent work from

Darwin concerning cyclone risk specific to the Northern Territory. Chapter 10

reviews recent investigations of the windfield of Cyclone ‘Yasi’ which crossed the

coast of North Queensland in February 2011, and discusses implications with respect

to the Australian Standard for Wind Actions AS/NZS1170.2:2011 (Standards Australia

2011). Chapter 11 contains conclusions and recommendations, including comments

on possible future changes to the cyclonic regional boundaries in AS/NZS1170.2.

There are three Appendices providing respectively, charts of maximum recorded

wind gusts from tropical cyclones at selected locations, a proposal for a new

definition of the peak gust for AS/NZS1170.2, and a wind field model for Cyclone

‘Yasi’.

This version of the report (June 2011) replaces the first edition (June 2008) and

incorporates developments and events of the last three years.

References D.J. Abbs, S. Aryal, E. Campbell, J. McGregor, K. Nguyen, M. Palmer, T. Rafter, I. Watterson and B. Bates (2006), Projections of extreme rainfall and cyclones, Report to the Australian Greenhouse Office, CSIRO. V.F. Dvorak (1984), Tropical cyclone intensity analysis using satellite data, NOAA Technical Report NESDIS-11, National Oceanic and Atmospheric Administration, Silver Spring, Maryland. Standards Australia (2002), Structural design actions-Wind actions-Commentary. (Supplement to AS/NZS 1170.2:2002). Standards Australia (2011), Structural design actions. Part 2: Wind actions. AS/NZS 1170.2:2011.

9

2. History of cyclonic design wind speeds and the regional system in AS/NZS1170.2 2.1 Whittingham’s analyses The first analysis of extreme gust wind speeds in Australia was carried out by

Whittingham (1964). This work included analysis of recorded annual maximum wind

gust data from a number of stations in tropical-cyclone-affected Northern Australia:

Broome (WA), Cairns (Qld.), Carnarvon (WA), Darwin (NT), Groote Eylandte (NT),

Karumba (Qld.), Onslow (WA), Port Hedland (WA), Rockhampton (Qld.), Townsville

(Qld.) and Willis Island (Qld). Whittingham carried out extreme value analyses for

each station and then drew contour maps for extremes at various return periods up

to 100 years; however he analyzed annual maximum gusts at each station

irrespective of their source, and did not specifically extract those produced by

tropical cyclones.

It is interesting to compare Whittingham’s predictions, which were based on quite

short record lengths in most cases, with current values in AS/NZS1170.2:2011 for the

wind gust with an average recurrence interval of 100 years – i.e. V100. This is done

in Table 2.1, in which Whttingham’s values have been converted from knots to

metres per second. The values from AS/NZS1170.2 include the FC and FD

‘uncertainty’ factors.

Table 2.1 Comparison of design gust wind speeds Whittingham’s values versus AS/NZS1170.2:2011

Station Whittingham (1964)

V100 (m/s)

AS/NZS1170.2:2011 V100 (m/s)

Ratio

Broome, WA 50.4 58.8 1.17

Cairns, Qld 43.7 58.8 1.35

Carnarvon, Qld 65.8 72.6 1.10

Darwin, NT 55.0 58.8 1.07

Groote Eylandte, NT 48.4 58.8 1.21

Karumba, Qld 31.9 58.8 1.84

Onslow, WA 70.0 72.6 1.04

Port Hedland, WA 64.3 72.6 1.13

Rockhampton, Qld 48.4 58.8 1.21

Townsville, Qld 44.2 58.8 1.33

Willis Is., Qld * 75.1 58.8 0.78 * Willis Island has been assumed to be in Region C, although this is not explicit in

AS/NZS1170.2

10

The ratios of the values in the 2011 Standard to Whittingham’s values for these

stations are shown in the extreme right column in Table 2.1. All these values except

one are greater than 1.0, in some cases significantly so. The exception is Willis

Island which is several hundred kilometres offshore from Queensland.

Part of the reason for the ratios being greater than 1.0 in Table 2.1 can be attributed

to the FC and FD uncertainty factors used in AS/NZS1170.2:2011. However the main

reason for the large differences from 1.0 is the fact that Whittingham attempted to

make predictions with so few years of maximum gusts; the maximum number

available to Whittingham for any station in Table 2.1 was 21, and in one case as few

as 7 years were available. It is significant that the V100 values in several of the

stations were exceeded significantly within a few years of Whittingham’s analyses –

for example at Townsville during Cyclone ‘Althea’ (1971), at Darwin during Cyclone

‘Tracy’ (1974), and at Karumba during Cyclone ‘Ted’ (1976).

2.2 Development of cyclonic design wind speeds in the Australian Standard

2.2.1 1952 to 1973

The first loading standard in Australia was a simplified structural loading Standard,

known as SAA Interim 350 (1952). Part II of this document gave information on

wind loads. There was no zoning or contour map but a simple table with six

numbers gave advice on design wind speeds. For the coastline (definition not given)

north of Latitude 25o S, a value of 110 mph (49.2 m/s) was specified for all structures

up to 300 feet (91m) in flat country.

The first ‘real’ stand-alone wind loading standard was CA34 Part II -1971 (Standards

Australia 1971). This contained a table of regional basic design wind velocities in

miles per hour for about fifty locations, including many on the cyclone-affected

coastlines and all except one of those in Table 2.1. Values were given for return

periods of 5, 25, 50 and 100 years, and were corrected to the standard height of 33

feet (10m) in flat open terrain (defined in the Standard as ‘Terrain Category 2’). It

also appears that a re-analysis of the annual maximum wind gusts was carried out,

using the extra data available since the work of Whittingham (1964), as the values

differ somewhat from those listed in Table 2.1.

A metric version of the Standard, AS 1170, Part 2-1973 (Standards Australia, 1973)

was published shortly after, with the same design wind speeds as in CA34, converted

to metres per second and rounded. Table 2.2 summarizes the values in both CA34

and AS 1170.2-1973 in comparison with the current values.

11

The ratios of the current (2011) values of V100 to the 1971-3 in Table 2.2 are similar to

those in Table 2.1. Cyclone ‘Althea’ had occurred in Townsville in late 1971, and

produced a maximum recorded gust of 55 m/s at Townsville Airport, but this had not

been included in the 1971 re-analysis, and hence in the values in the 1973 edition.

However the recorded value exceeded both V50 and V100 in both editions of the

Standard.

Table 2.2. Comparison of design gust wind speeds CA34.2-1971 and AS1170.2-1973 versus AS/NZS1170.2:2002

Station AS1170.2-1973 V100 (m/s)

AS/NZS1170.2:2011 V100 (m/s)

Ratio

Broome, WA 48 58.8 1.23

Cairns, Qld 49 58.8 1.20

Carnarvon, Qld 49 72.6 1.48

Darwin, NT 53 58.8 1.11

Karumba, Qld 31 58.8 1.90

Onslow, WA 67 72.6 1.08

Port Hedland, WA 48 72.6 1.51

Rockhampton, Qld 53 58.8 1.11

Townsville, Qld 51 58.8 1.15

Willis Is., Qld 63 58.8 0.93

Both the 1971 and 1973 Standards had a contour map showing values of V50. That in

AS1170.2-1973 is shown in Figure 2.1. In AS1170.2-1973 these ranged up to 60 m/s

in the north of Western Australia, and near Willis Island off Queensland. In the

Northern Territory near Darwin, the value of V50 reached 50 m/s.

Also in both the 1971 and 1973 Standards, a ‘cyclone’ factor of 1.15 to be applied to

the V50 wind speeds in an area up to 50 kilometres inland from the coastline north of

Latitude 27oS, was specified. This was regarded as an additional load factor for

cyclonic wind loads to reduce the risk of exceedence of the final design wind loads

on engineered buildings. Load factors are an important part of the specification of

design wind loads.

2.2.2 1975 to 1983 The occurrence of Cyclone ‘Tracy’ in Darwin in 1974 produced an immediate reaction

in the wind loading Standard. Although the actual peak gust at Darwin Airport was

not recorded, it has been estimated to be about 70 m/s. This greatly exceeded the

value of V100 of 53 m/s - the highest value shown in AS1170.2-1973.

12

Figure 2.1. Regional basic wind speeds V50 (m/s) – AS1170.2-1973

The response in AS1170.2-1975 was the specification of a single cyclonic strip within

50 kilometres of the ‘smoothed’ coastline north of Latitude 27oS, with a value of V50

of 55m/s. This value reduced by 5m/s over each of another two 10 kilometre strips

(Figure 2.2).

The map shown in Figure 2.2 was retained in the 1981 and 1983 editions of the

Standard. The ‘Cyclone Factor’ of 1.15 was also retained in the 1975, 1981 and

1983 editions. Tabulated values at return periods of 5, 25, 50 and 100 years were

also given for all the stations shown in Table 2.2. For all the stations except Onslow

and Willis Island the value of V100 was specified as 60 m/s. These locations retained

the same values given in AS1170.2-1973; thus for Onslow the value of V100 was given

as 67 m/s, and for Willis Island 63 m/s was specified.

Table 2.3 shows the values of V100 specified between 1975 and 1983 compared to

the present values. The ratios are now much closer to 1, with the exceptions being

Carnarvon and Port Hedland which are now rated much higher than previously

(together with Onslow they are now zoned in the higher Region D).

13

Figure 2.2. Regional basic wind speeds V50 (m/s) – AS1170.2-1975 to AS1170.2-1983

Table 2.3. Comparison of design gust wind speeds AS1170.2-1975 to -1983 versus AS/NZS1170.2:2011

Station AS1170.2-1983 V100 (m/s)

AS/NZS1170.2:2011 V100 (m/s)

Ratio

Broome, WA 60 58.8 0.98

Cairns, Qld 60 58.8 0.98

Carnarvon, Qld 60 72.6 1.21

Darwin, NT 60 58.8 0.98

Karumba, Qld 60 58.8 0.98

Onslow, WA 67 72.6 1.08

Port Hedland, WA 60 72.6 1.21

Rockhampton, Qld 60 58.8 0.98

Townsville, Qld 60 58.8 0.98

Willis Is., Qld 63 58.8 0.93

14

2.2.2 1989 to present New cyclonic regional boundaries were introduced in the 1989 edition of AS1170.2,

at the same time that contours were discontinued and a regional system introduced

for the whole of Australia (Figure 2.3). Also, with the introduction of limit states

design into Australian structural standards, separate wind speeds were designated

for ultimate and serviceability limit states. In the former case, an annual risk of

1/1000 was specified for Vu (i.e. Vu was V1000 for all structures in current

terminology). The Cyclone Factor of 1.15 was discontinued, being effectively

incorporated in the values for Vu given for the cyclonic regions C and D.

Regions C and D were essentially the same as those in the 2002 version and in the

current (2011) Standard. There was no extensive re-analysis of cyclonic wind speeds,

but the justifications for the values given were described by Holmes, Melbourne and

Walker (1990). Values of V1000 of 70 m/s and 85 m/s were specified for Regions C

and D respectively and represent approximately the middle of the Category 4 and 5

ranges of the Australian tropical cyclone scale.

The only changes in the 2002 Standard (AS/NZS1170.2:2002) from the 1989 edition

were the additions of two small areas of Region B: firstly a strip between 100 and

150 kilometres from the coastline between 20o and 25oS near the Western

Australian coastline, and secondly east of 142oE and north of 11oS covering the

Torres Strait Islands. The regional map in the 2002 Standard is reproduced in Figure

2.4.

The regional wind speeds and regional boundaries in AS/NZS1170.2:2011 (Figure 2.5)

are unchanged from those in AS/NZS1170.2:2002. The only changes to Table 3.1 in

2011 are a number of additional average recurrence intervals, for which regional

wind speeds are given.

2.3 Summary and discussion The changes in the cyclonic wind speeds and regional boundaries in the Australian

Standards since 1971 have come about from the realization that extreme value

analyses of recorded wind speeds in regions affected by tropical cyclones will give

large uncertainties when tropical cyclones only rarely occur, and only short records

of wind data are available in relation to the interval between direct strikes by

cyclones at a particular location. This reaction was primarily provoked by the

occurrence of Cyclone ‘Tracy’ at Darwin in 1974 which produced maximum gusts

greatly exceeding those given in the Standard at the time for building design.

Specification of design wind speeds in cyclonic onwards from 1975 onwards was

15

mainly guided by computer-based simulation approaches (see Chapter 3), although

these have not been used greatly in Australia since the early 1980s.

Figure 2.3. Regional boundaries and basic wind speeds (m/s) – AS1170.2-1989

Figure 2.4. Regional boundaries – AS/NZS1170.2:2002

16

Figure 2.5. Regional boundaries – AS/NZS1170.2:2011

The introduction of limit states design in the 1980s focused on the need for

estimates of design wind speeds at high return periods of 500 to 1000 years – these

represent risk of exceedence of 0.05 to 0.10 for typical buildings with design lives of

50 years.

It is noted that the Commentary to AS/NZS1170.2:2002 Standards Australia 2002)

stated: “The Standard does not attempt to predict the effects of possible future

climatic changes, as the evidence for changes in wind speeds is inconclusive.” As

noted in the Introduction to this report, this statement is now becoming

unsustainable as significant changes in wind speeds, due to climate change, are

predicted within the lifetime of buildings currently in the design or construction

phases.

17

References J.D. Holmes, W.H. Melbourne and G.R.Walker, A commentary on the Australian Standard for Wind Loads, Australian Wind Engineering Society, 1990. Standards Association of Australia (1952), Minimum design loads on structures, SAA Int. 350. Standards Association of Australia (1971), SAA Loading Code. Part II – Wind forces, AS CA34, Part II - 1971. Standards Association of Australia (1973), SAA Loading Code. Part 2 – Wind forces, AS 1170.2 - 1973. Standards Association of Australia (1975), SAA Loading Code. Part 2 – Wind forces, AS 1170.2 - 1975. Standards Association of Australia (1981), SAA Loading Code. Part 2 – Wind forces, AS 1170.2 - 1981. Standards Association of Australia (1983), SAA Loading Code. Part 2 – Wind forces, AS 1170.2 - 1983. Standards Association of Australia (1989), SAA Loading Code. Part 2 – Wind loads, AS 1170.2 - 1989. Standards Australia (2002a), Structural design actions. Part 2: Wind actions. AS/NZS 1170.2:2002. Standards Australia (2002b), Structural design actions-Wind actions-Commentary. (Supplement to AS/NZS 1170.2:2002). Standards Australia (2011), Structural design actions. Part 2: Wind actions. AS/NZS 1170.2:2011. H.E. Whittingham (1964), Extreme wind gusts in Australia, Bureau of Meteorology, Bulletin No. 46.

18

3. Simulation methods for prediction of cyclonic wind speeds 3.1 Introduction This chapter reviews probabilistic methods for prediction of extreme wind speeds

due to tropical cyclones. These methods have developed due to the shortcomings

of methods based on the analysis of historical recorded data on gust wind speeds

due to tropical cyclones from anemometer stations. The methods aim to make use

of all information available on the tropical cyclones in the vicinity of a site. Thus

track information, such as heading and translation speeds, as well as information on

central pressures and storm size, as indicated by the radius to maximum winds, are

used. This data is used with empirical wind field models to make predictions of

extreme wind speeds; the latter normally give upper level mean speeds (gradient

winds), and assumed factors to convert these to a gust wind speed near the surface

must be applied.

Methods described Vickery et al. (2000a, 2000b), have been used to directly develop

design wind speeds for hurricane regions of the United States in ASCE 7 (the

equivalent of AS/NZS1170 in Australia). These approaches are described in Section

3.3.

Some similar work by Harper (1999) for Australia is described in Section 3.4.

3.2 History of simulation approaches 3.2.1 Work in the United States Vickery and Twisdale (1995a) describe the development of simulation methodologies

for the hurricane-prone coastline of the United States. The first of these was

implemented for the Texas coastline by Russell (1971). Shortly after Russell and

Schueller (1974), Tryggvason et al. (1976), Georgiou et al. (1983) and Twisdale and

Dunn (1983) used similar approaches for portions of the United States coastline.

Batts et al. (1980) were the first to apply the methodology to the entire U.S. eastern

and southern coastlines affected by hurricanes; these predictions were the basis for

the wind speed contours in ASCE 7-1988. Improved estimates were made by

Vickery and Twisdale (1995a, 1995b) and Vickery et al. (2000a, 2000b) and these are

the basis for the hurricane wind speed contours in the present American Loading

Standard ASCE 7-10 (ASCE, 2010).

In all the approaches, probability distributions are developed for the central pressure

difference (p), translation speed, c, and radius of maximum winds, Rmax, for

approaching hurricanes in the historical databases. Most of the approaches

19

mentioned above model hurricanes in a circular sub-region centred on the site of

interest. However, Batts et al. (1980) and Twisdale and Dunn (1983) used a coast

crossing technique to derive the basic probability distributions.

3.2.2 Work in Australia In the late 1970s and early 1980s, Australia was active in developing probabilistic

simulation methods for tropical cyclone wind speeds through the work of Gomes and

Vickery (1976), Martin and Bubb (1976) and Tryggvason (1979). Probably partly as

a result of criticism of the approach by Dorman (1984), and partly because of the

demise of publicly-funded research into wind loading generally, there was little work

of this type from the mid 1980s to the end of the 1990s.

However, more recently there has been an interest in such predictions from the

insurance industry. The work described by Harper (1999) is an example of this; the

latter work is discussed in Section 3.4.

Simulation methods have also been used in Australia for storm surge and wave

prediction – e.g. James and Mason (1999), and unpublished commercial studies by

Global Environmental Modelling Systems Pty. Ltd. (http://www.gems-aus.com/).

3.3 Methods by Vickery et al. (1995-2000) The methodology used by Vickery and Twisdale (1995a) for predictions for the U.S.

coastline is as follows. The statistical distributions of central pressure difference,

p, translation velocity of the hurricane, c, angle of approach, , and the minimum

distance of the centre of the hurricane from the site, dmin, are derived from data

obtained from the National Climatic Data Center of the U.S. These distributions are

site-specific and vary significantly with location along the coastline.

Using site-specific probability distributions of p, c, and dmin, in conjuction with a

wind-field model, thousands of hurricanes were simulated. The distributions used

for p, c, and dmin were Weibull, Lognormal, Bi-Normal and uniform, respectively.

The wind-field model used was that of Shapiro (1983). Each simulated hurricane

was assumed to travel along a straight line path throughout the simulation region

defined using the sampled values of dmin and . The sampled value of p was held

constant up to landfall, after which it was reduced using the filling model described

by Vickery and Twisdale (1995b).

The radius of maximum wind, Rmax, was assumed to have some correlation with the

central pressure difference for storms south of 30oN, and in the case of northern

20

storms, with the latitude. Rmax was modeled with a lognormal distribution, with the

mean value a function of central pressure difference or latitude.

Using the above methodology, Vickery and Twisdale (1995a) made predictions of 50,

1000 and 2000-year return period fastest mile wind speeds at 30 coastal and 16

inland locations. These differed significantly for many stations from the earlier

predictions of Batts et al. (1980) – this was attributed mainly to the use of the

Shapiro windfield model and a new filling model (Vickery and Twisdale 1995b).

As well as a number of sensitivity studies, Vickery and Twisdale (1995a) compared

area and point exceedence probabilities – thus for a single point in the Miami area a

fastest mile wind speed of 65m/s was associated with a return period of about 300

years. However, the return period of this wind speed for anywhere in Dade County,

Florida, is only about 100 years.

The Shapiro wind field model described by Vickery and Twisdale (1995b) consisted of

numerical solutions to the vertically integrated equations in coordinates moving with

the hurricane. An imposed pressure distribution for the vortex was used.

Comparisons with measurements in five hurricanes were favourable and

considerably better agreement was shown than for the wind field model of Batts et

al. (1980). The filling model used by Vickery and Twisdale (1995b) was an earlier

version of that described by Vickery (2005) and discussed in Section 5.4 of this

report.

A new technique for modelling hurricane wind risk in the United States was

described by Vickery et al. (2000a). A storm track modelling approach was adopted

in which the entire track of a hurricane as it crosses the ocean and makes landfall

was modelled. The advantage of this approach over the circular region approach

used previously resulted from not having to assume uniform climatology over the

subregion. Using the empirical storm track modelling approach, storm intensities

change with time and storms change both direction and speed as they pass by a site.

This is more realistic compared to earlier methods. This new technique, together

with the new wind field model described by Vickery et al. (2000b), was used as the

basis for the revised wind speed contours for the hurricane coastlines in ASCE 7-02,

ASCE 7-05 and ASCE 7-10. The wind field model showed excellent agreement with

recorded mean and gust time histories for numerous locations in several hurricanes.

The work described by Vickery et al. has been very well documented and uses

soundly based wind field and filling models, and is calibrated with the

comprehensive data set of measurements available in the United States from both

aircraft flights and surface measurements. However, the predictions for wind

21

speeds in the future are still dependent on the characteristics of past hurricanes

being representative of those in the future. At present future climate change effects

are not included.

3.4 Work by Harper (1999) Harper (1999) describes a similar numerical simulation approach to the prediction of

wind speeds due to tropical cyclones in Australia, and of insurance losses in a city

from a single event. There is little detail given of the methodology used in terms of

the probability distributions used for relevant parameters etc, except that apparently

a circular sub-region approach (see Section 3.2.1) with a radius of 500 kilometres

was used. However, examples are given of predicted wind speeds for Cairns,

Townsville, Mackay and Brisbane in Queensland. This apparently used cyclone track

information from 1959 onwards. Reasonable agreement of the simulated

predictions with those from the recorded gusts due to tropical cyclones for these

locations was found, although the latter may not, in themselves, be representative of

the long-term climate.

3.5 Work by Geoscience Australia Geoscience Australia has recently developed a deterministic-probabilistic Tropical

Cyclone Risk Model for wind speeds (Arthur et al. 2008a). It has been calibrated

against Cyclone ‘Tracy’ (Arthur et al., 2008b).

The model has recently been combined with climate model simulations by CSIRO and

the University of Melbourne, in a collaborative project (Lavender et al., 2011) that

has been reviewed in Section 8.6.

3.6 Summary and conclusions

This Chapter has outlined the development in probabilistic simulation methods for

prediction of wind speeds from hurricanes or tropical cyclones. This method has

been used extensively in the United States where the results are used directly for

design wind speed specification in ASCE 7 - the equivalent of AS/NZS1170.2.

Australia suffers from a lack of aircraft flights into tropical cyclones leading to an

inaccurate cyclone database, and also a lack of surface measurements required for

calibration of wind field and filling models. This means the potential accuracy of

probabilistic simulation models could not be expected to match that in the U.S.

22

Recently these simulation methods have been integrated with larger scale climate

modelling (in Australia and elsewhere), incorporating climate change effects.

Discussion of these methods as used in Australia is given in Chapter 8.

References American Society of Civil Engineers (2010), Minimum design loads for buildings and other structures, ASCE Standard ASCE/SEI 7-10. W.C. Arthur, A. Schofield, R. P. Cechet and L. A. Sanabria (2008a), Return period cyclonic wind hazard in the Australian Region. 28th AMS Conference on Hurricanes and Tropical Meteorology, 28 April - 2 May 2008, Orlando, FL, USA. W. C. Arthur, A. Schofield and R. Cechet (2008b), Severe wind hazard assessment of Cyclone Tracy using a parametric tropical cyclone model, 15th National Australian Meteorological and Oceanographic Society (AMOS) Conference, Geelong, January 29-February 1, 2008. M.E. Batts, M.R. Cordes, L.R. Russell, J.R. Shaver and E. Simiu (1980), Hurricane windspeeds in the United States, Report No. BSS-124, National Bureau of Standards, U.S. Dept. of Commerce, Washington D.C. C.M.L. Dorman (1984), Tropical cyclone wind speeds in Australia, Civil Engineering Transactions, Institution of Engineers, Australia, Vol.26, pp 132-139. L. Gomes and B.J. Vickery (1976), Tropical cyclone gust speeds along the north Australian coast, Civil Engineering Transactions, Institution of Engineers, Australia, Vol.18, pp 40-48. B.A. Harper (1999), Numerical modelling of extreme tropical cyclone winds, Journal of Wind Engineering & Industrial Aerodynamics, Vol.83, pp 35-47. M. James and L.B. Mason (1999), Generation of a synthetic tropical cyclone. In Coasts and Ports ’99. Institution of Engineers Australia. pp 407-412. S.L. Lavender, K.J.E. Walsh, D.J. Abbs, M. Thatcher, W.C. Arthur and R.P. Cechet (2011), Regional climate tropical cyclone hazard for infrastructure adaption to climate change, Final Report, June 2011. G.S. Martin and C.T.J. Bubb (1976), Discussion of ‘Tropical cyclone wind speeds along the North Australian coast’, Civil Engineering Transactions, Institution of Engineers, Australia, Vol.18, pp 48-49. L.R. Russell (1971), Probability distributions for hurricane effects, Journal of Waterways, Harbors and Coastal Engineering, ASCE, Vol.97, pp 139-154.

23

L.R. Russell and G.F. Schueller (1974), Probabilistic models for Texas Gulf Coast hurricane occurrences, Journal of Petroleum Technology, pp 279-288. L.J. Shapiro (1983), The asymmetric boundary layer flow under a translating hurricane, Journal of Atmospheric Sciences, Vol.40, pp 1984-1998. B.V. Tryggvason, D. Surry and A.G. Davenport (1976), Predicting wind-induced response in hurricane zones, Journal of Structural Division, ASCE, Vol.102, pp 2333-2350. B.V. Tryggvason (1979), Computer simulation of tropical cyclone wind effects for Australia, James Cook University, Wind Engineering Report 2/79, April. L.A. Twisdale and W.L. Dunn (1983), Extreme wind risk analysis of the Indian Point Nuclear Generation Station, Final Rep. 44T-2491, Research Triangle Institute, North Carolina, U.S.A. P.J. Vickery (2005), Simple empirical models for estimating the increase in central pressure of tropical cyclones after landfall along the coastline of the United States, Journal of Applied Meteorology, Vol. 44, pp 1807-1826. P.J. Vickery and L.A. Twisdale (1995a), Prediction of hurricane wind speeds in the United States, J. Struct.Engg.,Vol.121, pp 1691-1699. P.J. Vickery and L.A. Twisdale (1995b), Wind-field and filling models for hurricane wind-speed predictions, J. Struct.Engg.,Vol.121, pp 1700-1709. P.J. Vickery, P.F. Skerlj, A.C. Steckley and L.A. Twisdale (2000a), Hurricane wind field model for use in hurricane simulations, J. Struct.Engg.,Vol.126, pp 1203-1221. P.J. Vickery, P.F. Skerlj, and L.A. Twisdale (2000b), Simulation of hurricane risk in the U.S. using empirical track model, J. Struct.Engg.,Vol.126, pp 1222-1227.

24

4. Vertical profiles in tropical cyclones and hurricanes 4.1 Introduction The Australian Standard specifies a Regional wind speed as a 3-second gust at a

height of 10 metres in open country terrain (Terrain Category 2). For buildings of

greater or lesser height than 10 metres, or located in a different terrain type, this

wind speed must be adjusted. This is accomplished by the ‘Terrain-height

Multiplier’, Mz,Cat. Currently in the Standard a different set of Terrain-height

Multipliers is specified for Tropical Cyclone Regions C and D in Table 4.1(B) of the

Standard.

Clearly the values specified for these multipliers in the Standard for buildings that

are not close to 10m in height are very significant in determination of the wind loads

in tropical cyclone regions. The ratio between upper level (gradient) winds and

surface level winds is also important when simulation methods are used to predict

wind speeds due to tropical cyclones (Chapter 3).

In the following reviews are given of the original tower measurements at N.W. Cape

that were the basis of the gust profiles (Mz,Cat) used in AS/NZS1170.2, and of the

recent dropwindsonde measurements that have been carried out in Atlantic

hurricanes.

4.2 Tower measurements at NW Cape by Wilson Some of the few available tower measurements of vertical wind profiles in tropical

cyclone, hurricane or typhoon, were made by Wilson (1979a, 1979b) at the North-

West Cape near Exmouth in Western Australia. Observations were made from

anemometers mounted on guyed tower at heights of 60m, 191m, 279m and 390m

(with anemometers operating at all heights only during one cyclone). Another

anemometer was mounted at 9m height on a pole about 350m away from the main

tower. The fetch was open water with only a short land fetch of less than 5

kilometres for a large range of wind directions from NW through N to S.

In a period of four and half years the anemometers were able to record velocities

from four tropical cyclones during the nineteen-seventies: ‘Beryl’ (1973), ‘Trixie’

(1975), ‘Beverley’ (1975) and ‘Karen’ (1977). The highest 10-minute wind speed at

the top anemometer was about 57 m/s during Cyclone ‘Beverley’.

A significant feature of several of the profiles recorded was the high values of wind

speed recorded on the 60 metre anemometer compared to those at both the 9m

and 191m heights. The surface roughness lengths for the open water fetch and for

25

the land surrounding the tower were estimated to be similar in magnitude (1 to 3

mm – corresponding to Terrain Category 1 in AS/NZS1170.2), and it was concluded

that the inner boundary layer resulting from the water to land transition had little

effect on the measured wind profiles. This appeared to be confirmed by the

measurements which showed little variation with the changing land-water fetch of

different wind azimuths.

However the shear (i.e. the difference in wind speeds) between 9 and 60m was

significantly underestimated by the logarithmic law with a roughness length of 1-3

mm. It would have to be concluded that the boundary layer wind flow was not in

equilibrium with the underlying terrain, or was not neutrally stable.

In the absence of other comparable tower measurements in tropical cyclones,

hurricanes and typhoons, profiles of Mz, Cat, based on the NW Cape measurements,

were incorporated into the 1989 edition of the Australian Standard on Wind Actions

(AS1170.2-1989) for both Regions C and D, and the gust profile was continued

without modification in the 2002 edition.

Figure 4.1 shows the average maximum gust profiles derived from the 10-minute

mean windspeeds and gust factors obtained for the four recorded cyclones at NW

Cape. The profile in the Standard is assumed to increase monotonically up to 100m

and above that height takes a constant value of 1.40 for all Terrain Categories. The

Standard is conservative with respect to the averaged measurements above 100 m.

However, these profiles need reviewing in the light of the more recent

dropwindsonde measurements in U.S. hurricanes described in the following section.

Mz cat comparisons

0

100

200

300

400

500

0.0 0.5 1.0 1.5

Mz,Cat1

Heig

ht

(m)

Avg. measured -

NW Cape

AS/NZS1170.2

Figure 4.1 Comparison of average measured maximum gust profiles for four

cyclones (1973-7) at North-west Cape and Mz,Cat 1 from AS/NZS1170.2:2011

26

4.3 Dropwindsonde measurements The dropwindsonde (Figure 4.2) is a probe containing sensors and a GPS satellite

receiver that enables profiles of various atmospheric variables, including wind speed,

to be monitored as it falls, after being dropped from an aircraft.

Since 1997 the National Oceanic and Atmospheric Administration (NOAA) and the

U.S. Air Force have been deploying GPS-based dropwindsondes into hurricanes in the

Atlantic and eastern North Pacific oceans. These have generated vertical profiles of

wind and thermodynamic parameters from flight level (typically about 3000 metres)

down to sea level. It should be noted that dropwindsondes are not normally

deployed over land.

The development of the dropwindsonde has been described by Hock and Franklin

(1999). Detailed analyses of wind profiles from 1997-1999 have been described by

Franklin et al. (2003); the main purpose was to obtain a ratio between surface (10

metres) and flight level wind speeds for forecasting purposes. Wind profiles were

averaged separately for the eyewalls of hurricanes, and for the outer vortex regions.

Powell et al. (2003) analyzed similar data and fitted logarithmic profiles for various

speed ranges. In contradiction to previous extrapolations for wind over the ocean,

they found that the surface drag coefficient and roughness length did not continue

to increase beyond U10 equal to 30 m/s, in fact, the roughness length decreased from

about 3 mm to 1mm between 33 and 50 m/s.

Figure 4.2 Schematic of a dropwindsonde probe

27

4.3.1 Nature of dropwindsonde profiles The horizontal wind speed is determined from the position of the sonde sampled

every 0.5 seconds, and assumes that the sonde is travelling at the wind speed.

However, corrections are made based on the horizontal and vertical accelerations as

described by Hock and Franklin (1999).

Since the probes translate circumferentially through the hurricane, as well as fall

vertically, the individual dropwindsonde profiles reflect the larger scale turbulent

eddies. Smaller scale turbulence as would be observed by a high-response

anemometer on a fixed mast, is not included due to the sampling interval, the

response of the probe and its movement in ‘following the eddies’. However,

ensemble averaging of many individual normalized profiles, as undertaken by

Franklin et al. and Powell et al., will give expected mean vertical velocity profiles of

horizontal velocity.

There is some debate about the validity of the corrections to the profiles when the

probe horizontal speed is different to the wind speed – for example as the sonde

falls it will tend to have a higher wind speed associated with a greater height.

However, the corrections are already quite small, and improvements to the

correction method will be unlikely to significantly change the profiles, and have

negligible effect on the ensemble-averaged mean profiles. (It is noted that due to

the vertical falling motion, the relative wind speed to the sonde will normally be

vertical and parallel to the cylinder axis. A relative horizontal wind speed will deflect

the total wind vector, but the cylinder will then tend to align itself with the new wind

vector direction).

4.3.2 Results from dropwindsonde profiles in hurricanes From a engineering design point of view, the analysis of Powell et al. and Franklin et

al. has produced the following significant results :

A logarithmic law for the mean wind speed appears to hold in the eyewall

region for heights above the surface up to about 300 metres. Above that

height, the rate of increase is much smaller with a maximum mean velocity at

about 500 metres. Above 500 metres, the velocity falls significantly in the

eyewall region.

For mean velocities at 10 metres height above water surface of about 33 m/s,

the roughness length reaches a maximum value of about 3 mm, falling to 1

mm for velocities of 50 m/s (Powell et al. 2003).

28

Franklin et al. (1999) gives velocity ratios, Uz /U10 for the eyewall region. These

values are reproduced in Table 4.1, and compared with a logarithmic law with z0

equal to 0.0001 m (0.1 mm).

Although there is an apparent discrepancy between the roughness lengths quoted by

Powell et al., and implied by Franklin et al. (i.e. a factor of 10), in both cases, the

values are significantly lower than those currently used for design of buildings for

off-water winds from hurricanes or tropical cyclones.

Since dropwindsondes are only deployed over the ocean, there are no profiles

available for wind over land from hurricanes, as required for the design of most

structures. It also should be noted that the dropwindsonde profiles do not give any

useful information on turbulence intensities, and hence on peak gust envelope

profiles as used in many design codes or standards, (such as ASCE 7 or

AS/NZS1170.2). However, it would not be expected that the latter would differ

significantly from the mean velocity profiles, (and may have an even lower slope.)

Table 4.1. Mean velocity profile in the hurricane eyewall from Franklin et al. (2003) (compared with a logarithmic law with z0 = 0.1 mm)

Height (m) Uz/U10 loge(z/0.0001)/loge(10/0.0001)

10 1.000 1.000

15 1.027 1.035

20 1.048 1.060

30 1.081 1.095

50 1.128 1.140

75 1.169 1.175

100 1.198 1.200

150 1.229 1.235

200 1.261 1.260

250 1.288 1.280

300 1.305 1.295

4.3.3 Recent results from Doppler radar profilers Some recent measurements from both dropwindsondes and Doppler radar profilers

have been described by Giammanco et al. (2011). The radar measurements from

weather radar (WSR-88D) operated by the U.S. National Weather Service were

carried specifically to obtain vertical wind velocity profiles in landfalling hurricanes,

since profiles from dropwindsondes are only available over the ocean. However, the

29

lowest height for which Doppler radar measurements could be obtained was about

35 metres.

The results described by Giammanco et al. for off-land winds indicated that the

higher roughness of the wind blowing over land had significant effect on the mean

velocity profiles – indicating a significantly higher roughness length than for over-

water winds. The average power law exponent up to 400 metres height was found

to be about 0.18. The ratio of wind speed at 100 m height to that at 35m was about

1.20. This ratio is greater than the ratio of 1.14 obtained from Table 4.1(B) in

AS/NZS1170.2:2011 (i.e. 1.40/1.225). However, the radar method measures wind

speed averaged over a volume considerably larger than that for the maximum gust in

the Standard.

4.4 Summary and discussion The Australia/New Zealand Standard AS/NZS1170.2, since 1989, has had a gust

profile for cyclonic regions based on measured tropical cyclone data from the late

1970s and early 1980s from the NW Cape mast near Exmouth Western Australia,

described in Section 4.2. These showed a sharp increase in wind speed from the

lowest anemometer (9 m height) to 60 metres, and then relatively uniform wind

speeds to the highest anemometer on the mast. Although above 100m, the

measured profiles from the NW Cape and the dropwindsonde profiles are near-

uniform, the high shear below 100m is not seen by the dropwindsondes, as

discussed in Section 4.3. It is probable that the lowest (9m) anemometer reading

from the NW Cape mast was heavily influenced by the inner boundary layer as the

surface changed from water to land, although this possibility was discounted at the

time of the measurements (Wilson, 1979b).

The Australian Standard also assumes increasing roughness length with increasing

wind speed over water – this has been shown to be an incorrect assumption over 30

m/s in hurricanes by the dropwindsonde data.

References J. L. Franklin, M.L. Black and K. Valde (2003), GPS dropwindsonde wind profiles in hurricanes and their operational implications, Weather and Forecasting, Vol.18, pp32-34. I.M. Giammanco, J.L. Schroeder and M.D. Powell (2011), Observed characteristics of tropical cyclone vertical wind profiles, accepted for publication in Wind and Structures.

30

T.F. Hock and J.L. Franklin (1999), The NCAR GPS dropwindsonde, Bull. Amer. Met. Soc., Vol. 80, pp 407-420. M.D. Powell, P.J. Vickery and T.A. Reinhold (2003), Reduced drag coefficient for high wind speeds in tropical cyclones, Nature, Vol. 422, pp 279-283, 20 March 2003. K.J. Wilson (1979a), Wind observations from an instrumented tower during Tropical Cyclone Karen, 1977, 12th Technical Conference on Hurricanes and Tropical Meteorology, New Orleans, April 1979. K.J. Wilson (1979b), Characteristics of the subcloud layer wind structure in tropical cyclones, International Conference on Tropical Cyclones, Perth, W.A., November 1979.

31

5. Inland weakening and penetration of tropical cyclones 5.1 Introduction As the high convection regions of a tropical cyclone cross a coastline from a warm

ocean, the storm loses its source of energy and starts to weaken. There is also a

reduction due to surface friction for winds blowing over land. These effects results

in maximum wind speeds falling progressively with distance from the coastline. This

is reflected in AS/NZS1170.2 by 50 kilometre-wide strips that designate the wind

regions in the Standard. At present, the width of these strips is the same on the

Western Australian coastline as on the more topographically complex east

(Queensland) coast. The validity of this has been questioned and will be addressed

in this report.

This chapter reviews recent studies of the inland penetration of tropical cyclones and

hurricanes, with consideration of the relevance to the current regional zoning

system.

The current Standard defines the regional strips with respect to the ‘smoothed’

coastline without defining it. A recommendation in Chapter 11 of this report

addresses this point and attempts to establish a workable definition.

5.2 Observations in Cyclone ‘George’ (2007) Boughton and Falck (2007) have surveyed the available information on wind gusts,

for up to 120 kilometres inland, from Tropical Cyclone ‘George’ which crossed the

Pilbara coast of Western Australia east of Port Hedland in March 2007.

Unfortunately for this event, there was a lack of anemometer data, and few simple

road signs, that have been used in past cyclones as an indicator of upper and lower

limits of wind gust speeds, were available. Hence Boughton and Falck used tree

damage and damage to vehicles, buildings and masts at isolated mining camps to

establish wind speeds at various distances from the coastline. The authors claim,

optimistically, accuracy of +/- 10% for estimates of gust wind speeds made from tree

damage and of +/-5% for predictions made from structures and vehicles.

Maximum gust speeds at Port Hedland on the coast were estimated at 55m/s from

the failure of road signs. However the township was located 60 to 70km west of the

storm track and outside of the eye wall of the cyclone. Tree damage in Port

Hedland was subsequently used to calibrate observed damage at other parts of the

track.

32

Estimates of 200 to 270 km/hour (55 to 75 m/s) at locations along the storm track

about 50 kilometres from the coast were made, but these are quite wide limits, and

it is difficult to judge the weakening of the storm without also having wind speeds at

landfall.

Calculations from the failure of the top of a radio mast at Strelley about 50 km inland

along the track gave a peak gust of 64 to 78 m/s. However, the authors apparently

did not consider possible resonant vibrations of this slender structure which would

have amplified the stresses. Ignoring this possibility may have resulted in

overestimates of the peak gusts.

A road-sign failure near the eye wall about 90 kilometres along the track after

landfall, but about 50 kilometres from the coastline indicated a lower bound of 55

m/s and a probable maximum gust of 60 m/s. This observation is significant, as

although this location is in Region C in AS/NZS1170.2, it would have been in Region B

had the cyclone travelled at right angles to the coastline. For Region B, V500 is 57

m/s, less than the probable maximum gust in this event.

A lower limit of 50 m/s was obtained for the FMG Rail Camp based on overturning of

a piling rig and a toilet block, although the values assumed for force and pressure

coefficients were not stated. This site is about 90 km from the smoothed coastline,

but about 140 km from landfall along the track of the cyclone. This wind speed

exceeds estimates made independently, based on the movement and overturning of

buildings within the camp.

Boughton and Falck stated that without the current FC and FD factors in AS/NZS, their

estimated maximum gust wind speeds would have been exceeded in Cyclone

‘George’ based on distances measured along the cyclone track from landfall, but this

was not the case for the actual regional location of the sites based on the shortest

distance from the smoothed coastline, for which the current Standard appeared to

be quite adequate.

5.3 Studies by Kaplan and DeMaria (1995, 2001) Kaplan and DeMaria (1995, 2001) developed a simple empirical model for predicting

the decay of tropical cyclone winds after landfall. This work was done primarily for

evacuation and emergency management application.

The model is based on a least squares fit to the maximum sustained (1-minute)

surface wind estimates made by the National Hurricane Center to land-falling

33

tropical storm and hurricanes in the United States. The equation for maximum

sustained wind as a function of time takes the form.

t

bb eVRVVtV )()( 0 (5.1)

where Vb is a ‘background’ wind speed V0 is the maximum wind speed just before landfall R is a reduction factor to account for the immediate effect of surface

roughness immediately on landfall

is a decay parameter For storms making landfall closer to the Equator than 37o, Kaplan and DeMaria

established values for Vb, and R of 26.7 knots, 0.095 hour-1 and 0.9, respectively.

For storms land-falling on the Atlantic coast north of 37oN the values obtained were:

Vb = 29.6 knots, = 0.187 hour-1 and R = 0.9.

5.4 Studies by Vickery (2005) For application to hurricane simulation models, Vickery (2005) studied the decay of

the central pressure rather than wind speed, after landfall. An exponential decay

function of the form of Equation (5.2) was used.

ateptp 0)( (5.2)

where p(t) is the difference between the central pressure and the far field

atmospheric pressure at time t after landfall, p0 is the pressure difference at the

time of landfall, and a is a filling constant.

The filling constant, a, was modeled as:

RMW

cpaaa

.0

10 (5.3)

where c is the translation speed of the storm at the time of landfall, and RMW is the

radius to maximum winds.

The characteristics of up to 57 storms which made landfall were studied and grouped

according to the sections of the U.S. coastline where they made landfall. Best fits to

the observed decay of central pressure for up to 36 hours after landfall were made

using Equation (5.3). For example, for the Florida Peninsula, the following

expression fitted the data with a correlation coefficient (r2) of 0.84.

34

RMW

cpa

.00167.00225.0 0 (5.4)

5.5 Application to decay with distance, and widths of regional boundaries The results of Kaplan and Maria, as summarized in Section 5.3, are intended be used

to predict the decay of wind speeds tropical cyclones with respect to time after

landfall. However, they can be converted to the decay rate with distance from the

coastline by multiplying by an average or representative storm translation speed.

Based on these decay rates, appropriate regional boundary widths may also be

determined.

Equation (5.1) can be modified in terms of decay with distance instead of time, as

follows:

x

bb eVRVVxV )()( 0

w

bb eVRVVwV cos/0 )()( (5.5)

where is a decay constant based on the distance inland x

w is the shortest distance from the smoothed coastline equal to x cos

is the angle of the cyclone track from the normal to the smoothed coastline

can be estimated from the decay constant of Kaplan and Maria as follows:

avc

(5.6)

where cav is an average translation speed of the storm. An average value of the angle to the normal can be calculated as follows:

2/

2/

2cos

1cos

dav (5.7)

Hence substituting in Equation (5.5),

avcwbb eVRVVwV

2/0 )()(

(5.8)

Then the values found by Kaplan and DeMaria for Vb and were applied, and

average values used for storm translation speed from Vickery (2005) of 6.2 m/s and

10.6 m/s for storms in the United States less than and greater than 37o latitude,

respectively. R was taken as 0.9 and V0 as 70 m/s. A factor of 1.3 was used to

convert the ‘1-minute maximum sustained wind speeds’ to the maximum gust used

in the Australian Standard.

35

The curves given by Equation (5.8), using parameters found by Kaplan and DeMaria

for U.S. hurricanes, for latitudes less than and greater than 37o, are plotted in Figures

5.1 and 5.2 respectively, and compared with the step changes resulting from the

regional boundaries in AS/NZS1170.2 between 20oS and 25oS in Western Australia.

The differences in the decay rates for the hurricanes north of 37oN in the U.S.

(including the more topographically complex New England region) and those south

of 37oS are small. However, it is noted that decay rates with time are higher for the

northern storms (Kaplan and DeMaria, 2001), but that the average translation

speeds are also higher (Vickery, 2005); these effects tend to compensate and give a

similar rate of decay with distance after landfall as the southern storms.

Figures 5.1 and 5.2 indicate that the current boundaries in AS/NZS1170.2 are

conservative with respect to the average decays found in the U.S. hurricanes, and

appear to be quite satisfactory. However, the lines shown from the U.S. hurricanes

are averages, and individual hurricanes will, of course indicate faster or slower

decays.

5.6 Data from Cyclone ‘Yasi’ A limited amount of data to check the inland weakening of an Australian cyclone is

available from the wind field of Cyclone ‘Yasi’, which made landfall in North

Queensland in early February 2011 (Boughton et al., 2011, and Appendix C). Using

wind gust estimations from failed road signs, the ratios of the estimated upper and

lower limits of gusts from Tully (approximately 17 kilometres inland), Jarra Creek (27

kilometres inland) and Munro Plains (33 kilometres inland) were compared with the

average of the upper and lower limits obtained from South Mission Beach, near the

point of landfall of Cyclone ‘Yasi’.

These values are plotted in Figure 5.3 and compared with the Kaplan and DeMaria

line (Eqn. 5.8), for latitudes less than 37o. A downward arrow on Figure 5.3 indicates

an upper limit, derived from a non-failed road sign, whereas an upward arrow

indicates a lower limit estimated from a failed road sign.

The agreement between the Cyclone ‘Yasi’ values and the U.S. data in Figure 5.3 is

quite good. The upper and lower limits for Tully are above the line, but this can

probably be attributed to local topographic effects for north and south winds at that

location.

5.7 Summary Research from the United States on the decay in wind speeds in hurricanes after

landfall indicates that the current regional boundaries are generally conservative

36

with respect to the average decay of both low latitude (less than 37o) and high

latitude (greater than 37o) storms. However to account for cyclones that decay less

faster than the average, the current boundaries appear to be adequate.

The difference in decay rate with distance for the more northerly storms in the U.S.

with topographically complex terrain after landfall, is little different to those in the

south which cross the coast at the flat coastline of Florida, Louisiana and Texas.

Hence, there is currently little evidence to differentiate the current regional

boundaries between Queensland and Western Australia, as some have suggested.

However, an analysis of the Australian tropical cyclone database (Section 1.2) could

be usefully carried out to determine whether there is a significant difference in the

decay rates between the Queensland and Western Australian coastlines.

Figure 5.1. Decay in gust wind speed for U.S. hurricanes south of 37o N compared to

the Australian Standard

0.0

0.2

0.4

0.6

0.8

1.0

0 50 100 150 200

Decay

Normal distance from coastline (km)

Decay in windspeed

<37deg

AS1170.2:2011 (WA)

37

Figure 5.2. Decay in gust wind speed for U.S. hurricanes north of 37o N compared to the Australian Standard

Figure 5.3. Decay in gust wind speed for U.S. hurricanes south of 37o N compared to measurements from Cyclone ‘Yasi’.

0.0

0.2

0.4

0.6

0.8

1.0

0 50 100 150 200

Decay

Normal distance from coastline (km)

Decay in windspeed

>37deg

AS1170.2:2011 (WA)

38

References G.N. Boughton and D. Falck (2007), Tropical Cyclone George – Wind penetration inland, Cyclone Testing Station, James Cook University, Technical Report No.53, August 2007. G.N. Boughton, D.J. Henderson, J.D. Ginger, J.D. Holmes, G.R. Walker, C.L. Leitch, L.R. Somerville, U. Frye, N.C. Jayasinghe and P.Y. Kim (2011), Tropical Cyclone Yasi – Structural damage to buildings, Cyclone Testing Station, James Cook University, Technical Report No.57, March 2011. J. Kaplan and M. DeMaria (1995), A simple empirical model for predicting the decay of tropical cyclone winds after landfall, Journal of Applied Meteorology, Vol. 34, pp 2499-2512. J. Kaplan and M. DeMaria (2001), On the decay of tropical cyclone winds after landfall in the New England area, Journal of Applied Meteorology, Vol. 40, pp 280-286. P.J. Vickery (2005), Simple empirical models for estimating the increase in central pressure of tropical cyclones after landfall along the coastline of the United States, Journal of Applied Meteorology, Vol. 44, pp 1807-1826.

39

6. Observed effects of climate change on tropical cyclones worldwide 6.1 Introduction Recent international literature concerning observed trends in global tropical cyclone

activity and correlation with increasing sea surface temperature is reviewed in the

following. In Section 6.6 consensus statements by expert groups are reproduced in

an endeavour to give a balanced view of opinions on this topic.

6.2 Interpretations by Webster et al. (2005) Webster et al. (2005) conducted a comprehensive analysis of global tropical cyclone

statistics for the satellite era (1970-2004) in each tropical ocean basin in which they

occur. They found significant increasing trends in sea surface temperature (SST) in

each of the ocean basins except for the South Pacific. Since it is well established

that a SST of 260C is required for tropical cyclone formation in the current climate, it

might be expected that there would be an increase in the number of tropical

cyclones. However there was no significant trend in global cyclones of all strengths.

Some studies have suggested that the 260C threshold may increase in a warmer

climate (J. Kepert – personal communication).

The North Atlantic region did show a statistically significant increase since 1995 in

tropical cyclones of hurricane strength (defined by Webster et al. as having wind

speeds greater than 33 m/s). However an attribution of the increase to increasing

SST is not supported because of the lack of correlation of the number of tropical

cyclones of this strength in other basins.

When Webster et al. examined the number of hurricanes by allocated category

(using the Saffir-Simpson scale), they found a significant increasing trend in the

numbers of Category 4-5 storms, and found that the numbers of these strongest

storms increased from about 50 globally per 5-year period in the 1970s to nearly 90

per five-year period in the decade 1995-2004 (see Figure 6.1). This conclusion has

been controversial and is further discussed in the following. Also as seen in Figure

1 there was no trend from the 1990-1994 period to the 2000-2004 period (as

pointed out by Klotzbach, 2006).

40

Figure 6.1 Apparent increase in number of Category 4 and 5 hurricanes world-wide (from Webster et al., 2005)

6.3 Interpretations by Emanuel (2005) Emanuel’s interpretations of the effects of increasing sea surface temperatures were

expressed in terms of a ‘power dissipation index’ (PDI) defined as follows.

0

3max dtVPDI

where Vmax is the maximum sustained wind speed at 10 metres height, and is the

lifetime of a storm.

On the assumption that the economic loss in windstorms varies as the cube of the

wind speed, the PDI is assumed to represent the ‘destructiveness’ of a tropical

cyclone.

Emanuel summed the PDIs over all storms in each calendar year for the North

Atlantic and North-west Pacific Basins, and found significant increases from about

1990 and 1980 respectively. Significant apparent correlations with changing sea

surface temperatures were found. However, thermodynamic considerations would

indicate only a 6-9% increase in PDI for the 0.5oC observed increase in sea surface

temperature, whereas the observed changes in PDI, with about a 50% increase,

greatly exceeded this range in both basins. Emanuel therefore concluded that only

part of the observed apparent increase in PDI can be attributed to increased sea

surface temperatures.

41

6.4 Interpretations by Klotzbach (2006) Klotzbach (2006) extended the analysis to all basins with tropical cyclone activity,

and excluded data before 1986 on the basis that, before the mid 1980s, only visible

satellite information was available and hence night-time observations were

excluded; also the quality and resolution of satellite imagery had improved greatly by

the later period.

Klotzbach used an ‘Accumulated Cyclone Energy Index’ (ACE) as an indicator of

trends. The ACE is similar to the PDI but incorporates the square instead of the

cube of the maximum surface wind speeds.

Only in the North Atlantic Basin was a significant increasing trend in ACE found. In

fact, decreasing trends were found in the Northeast, Northwest and Southwest

Pacific Basins. Klotzbach’s analysis, using the more recent (and more reliable) data,

found only a small increase in Category 4-5 hurricanes in the North Atlantic and

Northwest Pacific during the 20-year study period. Klotzbach’s findings were stated

to be ‘..contradictory to those of Emanuel (2005) and Webster et al. (2005)’.

6.5 Interpretations by Kossin et al. (2007) To eliminate the variability in global hurricane intensity records due to

improvements in satellite technology, Kossin et al. (2007) constructed a more

homogeneous data by first constructing a consistently analyzed satellite database for

1983 to 2005, and then applying a new objective algorithm to form hurricane

intensity estimates.

Although an increasing trend in PDI, and frequency and percentage of strongest

storms was found for the North Atlantic in agreement with Emanuel (2005) and

Webster et al. (2005), this was not the case for any other basin. In the South Indian

Ocean (including tropical cyclones affecting the WA coastline), no significant trends

were found in the reanalyzed homogeneous data set; in the South Pacific, a

decreasing trend was observed for the period 1983-2005 (Figure 6.2).

The combined global records re-analyzed by Kossin et al. showed no consistent long-

term trend in PDI, number of Category 4-5 storms or in the percentage of strongest

storms. This was in contradiction to conclusions by Webster et al. (2005) and

Emanuel (2005).

42

Figure 6.2 Trend in PDI for Southern Indian Ocean (left) and South Pacific Ocean (right). Red lines from uncorrected database; blue lines from corrected

homogeneous database (from Kossin et al., 2007)

6.6 Consensus statements by IPCC and IWTC The following statements from the International Panel on Climate Change (IPCC), and an expert group at the International Workshop on Tropical Cyclones, summarize consensus expert scientific opinion on the possible global impact of climate change, particularly sea surface temperature increases on tropical cyclones. IPCC (2001) ‘... there is some evidence that regional frequencies of tropical cyclones may change but none that their locations may change. There is also evidence that the peak intensity may change by 5% to 10% .... ’ IWTC (2006) ‘The scientific debate … is not as to whether global warming can cause a trend in tropical cyclone intensities. The more relevant question is how large a trend: a relatively small one decades into the future or large changes occurring today. Currently published theory and numerical modeling results suggest the former, which is inconsistent with the observational studies of … Webster et al. (2005) by a factor of 5… .’ (McBride et al., 2006). IPCC (2007) ‘There is observational evidence for an increase in intense tropical cyclone activity in the North Atlantic since about 1970, correlated with increases in sea surface temperatures. There are also suggestions of increased intense tropical cyclone activity in some other regions where concerns over data quality are greater. … There is no clear trend in the annual numbers of tropical cyclones.’

43

6.7 Recent developments (2008-2011) Studies have continued in the last three years to attempt to identify trends in the

number and intensities of tropical cyclones for all parts of the world where they

occur. Recent work has been summarized by the international Working Group on

‘Tropical-cyclone activity on climate time scales’ (Knutson et al., 2010). However,

before commenting on the studies, the Group noted: ‘Data homogeneity issues

continue to be a concern in studies of past trends and low-frequency variability of

tropical cyclones.’

Elsner and Jagger (2010) studied the increasing intensity of Atlantic hurricanes over

the period 1943 to 2008, and found the upward trend correlated with increasing sea-

surface temperature. They found larger increases over the Gulf of Mexico and

Caribbean Sea, where sea temperatures are highest.

Recent studies for the South Indian and South Pacific Oceans are discussed in

Sections 7.4 and 7.5.

6.8 Summary and Conclusions There has been some speculation that the frequency of severe tropical cyclones

world-wide has already been increasing as a result of global warming. This view was

promoted by two well-known publications in 2005. In 2011, there remains a view

held by some that hurricanes in the North Atlantic and Gulf of Mexico have been

increasing in intensity as a result of rising sea-surface temperatures. This view is not

shared universally, however.

The main problem appears to be observational errors, in both the numbers of

tropical cyclones observed and their intensity, prior to about 1980. Before the

advent of regular satellite observations in about 1970, many storms that did not

cross a coastline were not observed or not recorded. It is only since 1980 that

satellite observations have been regular enough and of sufficient quality to produce

reasonable and consistent estimates of storm strengths. Currently it is only storms

in the North Atlantic and North-West Pacific basins, that the strengths of hurricanes

and typhoons are being regularly observed by means of aircraft penetration flights.

In other parts of the world where frequent tropical cyclones occur, a technique

based on a qualitative interpretation of satellite images has been used. This may

have resulted in some mis-classification of tropical cyclones.

44

There is also the possibility that the observed trends in the north Atlantic since 1970

have been caused by a multi-decadal natural cycle (personal communication – J.

Kepert, Bureau of Meteorology)

References J.B. Elsner and T.H. Jagger (2010), On the increasing intensity of the strongest Atlantic hurricanes, in Hurricanes and Climate Change, Vol. 2, Springer. K. Emanuel (2005), Increasing destructiveness of tropical cyclones over the past thirty years, Nature, Vol. 436, pp686-688. P.J. Webster, G.J. Holland, J.A. Curry and H.R. Chang (2005), Changes in tropical cyclone number, duration and intensity in a warming environment, Science, Vol. 309, pp1844-1846. CSIRO (2007). Climate change in Australia. Technical Report. CSIRO. P.J. Klotzbach (2006), Trends in global tropical cyclone activity in the last twenty years (1986-2005), Geophysical Research Letters, Vol. 33, L10805. T.R. Knutson and 12 others (2010). Report of Working Group on ‘TC activity on climate time scales’, Seventh International Workshop on Tropical Cyclones, Reunion Island, November 15-20, 2010. J.P. Kossin, K.R. Knapp, D.J. Vimont, R.J. Murnane, B.A. Harper (2007), A globally consistent reanalysis of hurricane variability and trends, Geophysical Research Letters, Vol. 34, L04815. J. McBride, J. Kepert, J. Chen, J. Heming, G. Holland, K. Emanuel, T. Knutson, H. Willoughby and C. Landsea (2006), Statement on tropical cyclones and climate change, 6th WMO International Workshop on Tropical Cyclones (IWTC-VI), Costa Rica, November.

45

7. Observed trends in tropical cyclone activity in the Australian Region 7.1 Introduction This chapter reviews trends in tropical cyclonic activity in the Australian Region

during the last thirty years, and possible explanations.

7.2 Observations by Nicholls et al. (1998) A review of cyclonic activity in Australia from 1969/70 to 1995/6 was given by

Nicholls et al. (1998). A downward trend in the total number of tropical cyclones

observed in the Australian Region was observed. However, there was a slight

upward trend in the number of intense cyclones, defined as those with a central

pressure of 970hPa or less, as shown in Figure 7.1.

Nicholls et al. noted that at least part of the downward trend in overall cyclone

numbers can be explained by changes in the way storms are classified as cyclones.

For example Cyclone ‘Wanda’ in 1974 had a central pressure of 1000 hPa, and would

not have been classified as a cyclone in later years. This has an effect on the trend

in the number of cyclones with central pressures higher than 990 hPa, but negligible

effect on stronger ones.

The downward trend in overall numbers was primarily explained by the negative

trend in the Southern Oscillation Index (SOI) – otherwise known as the El Nino

phenomenon – over the time period in question. The SOI is the difference in mean

sea level atmospheric pressure between Tahiti and Darwin, standardized to a mean

of zero and standard deviation of 10, and low values are associated with droughts in

Australia, as well as fewer tropical cyclones. No explanation was given for the

increase in more intense cyclones, although an increase is also predicted by climate

models for the east coast of Australia as a consequence of increasing greenhouse gas

concentrations (see Section 8.4).

7.3 Study by Ramsay et al. (2008) A detailed study Ramsay et al. (2008) provided evidence that the sea surface

temperature in the east and central Pacific Ocean is the main contributing factor to

cyclonic activity in the Australian Region. In fact, the average sea surface

temperature in North Australian waters is only weakly correlated with tropical

cyclone activity near Australia. Cyclone numbers are also affected by the

monsoonal trough and by vertical shear in the atmosphere.

46

Figure 7.1 Trend in number of cyclones with central pressures of 970 hPa or less in the Australian Region during 1969-96 (from Nicholls et al., 1998).

Thus warming of the central Pacific Ocean (coinciding with El Nino conditions) is

unfavourable for tropical cyclone generation in the Australian Region. The

connection between sea surface temperatures in the central Pacific Ocean and

conditions in the Australian Region are illustrated in Figure 7.2. Thus it is possible

that increased ocean temperatures outside of the Australian Region, produced by

greenhouse effects, may result in fewer tropical cyclones affecting Australia.

However, Ramsay et al. note that the numbers of severe tropical cyclones (with

central pressure less than 965 hPa) are much less influenced by El Nino conditions,

and explain the more frequent occurrence of severe cyclones off north-west of

Western Australia compared to the East coast, by this.

Figure 7.2. Schematic showing warmer sea surface temperatures in the central Pacific Ocean, associated with the El Nino phenomenon, and conditions associated with reduced cyclone activity in the Australian Region (from Ramsay et al., 2008)

47

7.4 Study by Kuleshov et al. (2008, 2010) Kuleshov et al. studied tropical cyclone activity in the Southern Hemisphere between

1981/2 and 2005/6, and its dependence on the El Nino Southern Oscillation

phenomenon. As observed previously by others, they found that fewer tropical

cyclones occurred in both the South Pacific and South Indian Oceans during El Nino

than during La Nina years. On average 25 and 29 tropical cyclones occurred in El

Nino and La Nina years respectively in the Southern Hemisphere. Kuleshov et al.

also found that an area of cyclogenesis located between 60oE and 85oE in El Nino

years in the Indian Ocean shifted eastwards (i.e. closer to Australia) in La Nina years.

In contrast, the focus for cyclogenesis in the South Pacific shifted eastwards (away

from Australia) in El Nino years. These two effects explain the predominance of

tropical cyclones in the Australian Region during La Nina years. The decline in total

number of tropical cyclones affecting Australia in recent years can be explained by

the fewer La Nina events compared with El Nino periods.

As shown in Figure 7.3, taken from their paper, Kuleshov et al. (2008) identified

statistically significant increasing trends in severe tropical cyclones, defined as those

with a central pressure less than 945 hPa in the Southern Hemisphere (SH), southern

Indian Ocean (SIO), and in the South Pacific Ocean (SPO). This observation confirms

the earlier one by Nicholls et al. (Figure 7.1), and is also consistent with predictions

of the effect of CO2 induced global warming discussed in the next chapter.

Figure 7.3 Occurrences of tropical cyclones with central pressures of 945 hPa or less

in the Southern Hemisphere between 1981/2 and 2005/6 (from Kuleshov et al., 2008).

The above conclusions have recently been reinforced by Kuleshov et al. (2010);

however they qualified the observations of long-term trends, due to differences and

uncertainties in the data quality over the time of the record.

48

7.5 Study by Goebbert and Leslie (2010) Goebbert and Leslie (2010) examined the tropical cyclone variability of the

northwest Australian sub-basin (0°–35°S, 105°–135°E), using a dataset for the 39-

year period of 1970–2008. The major findings were that for that sub-basin, there

are 5.6 tropical cyclones and 42.4 tropical-cyclone days on average, with standard

deviations of 2.3 storms and 20.0 days. For intense cyclones (Category 3 and higher),

the annual mean frequency is 3.0, with a standard deviation of 1.6, and the annual

average intense tropical-cyclone days is 7.6 days. However, they found no

significant linear trends in either mean annual frequencies of tropical cyclones or

tropical-cyclone days over the 39-year period. The factors influencing the

development of tropical cyclones are less dependent upon standard El Niño–

Southern Oscillation (ENSO) variables than many other basins, including the rest of

the Australian region.

7.6 Summary and conclusions Studies of tropical cyclone activity in the Coral Sea have found that the numbers of

tropical cyclones are negatively affected by the Southern-Oscillation Index – El Nino

phenomenon. Fewer cyclones occur when higher sea surface temperatures occur in

the central and east Pacific Ocean. However the more intense cyclones (i.e. those

of interest for structural design) are much less influenced by the El Nino

phenomenon, and there is evidence of increasing trends in the stronger (Category 3-

5) events during the last 30-40 years. However, these conclusions should be re-

visited in the future to ensure that no fictitious increase in intensity has occurred in

the tropical cyclone database (Section 1.2) due to changes in interpretation of

satellite data over the years (see also Section 6.5).

Development of tropical cyclones in the southern Indian Ocean, however, are less

affected by the El Niño–Southern Oscillation, and no clear trend in the annual

number of tropical cyclones affecting the northern coastline of Western Australia

since 1970 has been identified.

To the extent that global warming will affect the sea surface temperature in the east

and central Pacific and hence the El Nino-La Nina phenomena, it may be expected

that there will be changes to the number and intensities of cyclones in the Australian

Region. Recent predictions are discussed in the following chapter.

References K.H. Goebbert and L.M. Leslie (2010), Interannual variability of northwest Australian tropical cyclones, Journal of Climate, Vol. 23, pp 4538-4555.

49

Y. Kuleshov, L. Qi, R. Fawcett and D. Jones (2008), On tropical cyclone activity in the Southern Hemisphere: trends and the ENSO connection, Geophysical Research Letters, Vol. 35, L14S08. Y. Kuleshov, R. Fawcett, L. Qi, B. Trewin, D. Jones, J. McBride and H. Ramsay (2010), Trends in tropical cyclones in the South Indian Ocean and South Pacific Ocean, J. Geophysical Research, Vol. 115, D01101. N. Nicholls, C. Landsea and J. Gill (1998), Recent trends in Australian Region tropical cyclone activity, Meteorological and Atmospheric Physics, Vol. 65, pp 197-205. H.A. Ramsay, L.M. Leslie, P.J. Lamb, M.B. Richman and M. Leplastrier (2008), Interannual variability of tropical cyclones in the Australian Region: Role of large-scale environment, Journal of Climate, Vol. 21, pp 1083-1103.

50

8. Predicted future effects of climate change on tropical cyclones

8.1 Introduction A number of simulation studies have been carried out in recent years to try to

determine the effects of enhanced greenhouse gas concentrations on the

occurrence and intensities of tropical cyclones in the Australian region. These are

based on Global Climate (computer) Models (GCM), derived from Numerical

Weather Prediction Models developed for weather forecasting purposes. The

smallest grid size used in the studies reviewed here is 14 kilometres. This, of course,

is insufficient to resolve tropical cyclones (with eye diameters of the order of 50

kilometres) in detail, and the conclusions of these studies must be taken as

preliminary.

These studies have identified ‘tropical-cyclone like vortices’ in the output of GCM

models, and drawn conclusions about actual cyclones in the real world. The

advantage of such studies is that enhanced greenhouse gas concentrations in the

upper atmosphere can be simulated.

A recent approach to improve the resolution of the models is ‘downscaling’ – i.e. the

outputs of the coarser resolution GCMs are used as inputs to Regional Climate

Models (RCMs) of finer resolution.

Regarding the general state-of-the art for simulation of future changes in tropical

storms by GCMs and regional climate models (RCMs), the international Working

Group on ‘Tropical-cyclone activity on climate scales’ state that “….there is a large

overall uncertainty in future changes in tropical cyclone frequency as projected by

climate models forced with future greenhouse gases….it is likely that the global

frequency of tropical cyclones will either decrease or remain essentially unchanged

due to greenhouse warming. In addition, recent climate model studies consistently

project a decrease in formation rates averaged over the Southern Hemisphere.”

(Knutson et al., 2010)

In the highest resolution GCM study to date, with a 14 kilometre grid, Yamada et al.

(2010) found a 40% reduction in tropical storms globally for a late 21st Century

climate change scenario. However this model also predicted an increase in the

number and intensity of the most intense storms. For example, the minimum

central pressure for the most intense storm decreased from 902 hPa under the

current climate, to 871 hPa under the warmer climate.

51

8.2 Simulations by Walsh et al. (2000-11) Walsh and Ryan (2000), using a RCM (125km resolution nesting to 30 km) of the

Australian region, inserted idealized tropical cyclones and examined their intensity

evolution under current and enhanced CO2 concentrations. Small increases in

intensity were observed – for example after 2 days, the average central pressure

reduced from about 970 hPa to 965 hPa. However the error bars representing the

standard deviations of the observed changes exceeded the predicted changes.

Walsh et al. (2004) used 30-kilometre horizontal resolution climate models to

simulate the 1967-96 climate, and with a climate with enhanced (3) carbon dioxide

concentrations in the upper atmosphere. Under the enhanced greenhouse

conditions the numbers of tropical cyclones formed and their region of formation do

not change much. However there was a 26% increase in the number of storms with

central pressures less than 970 hPa, and an increase in the number of intense storms

occurring south of 30o S latitude. There was no discussion of any change in the

number coast crossings of the eastern Australian coast, however.

Lavender and Walsh (2011) summarize recent ‘downscaled’ fine-resolution

simulations by CSIRO and the University of Melbourne. Using three different

simulation models, and projected increases in sea-surface temperatures, they found

an average 30% reduction in the total number of cyclones occurring in the Australian

Region – i.e. from an annual rate of 12.5 to about 9. However, as in previous studies,

they also found a poleward shift in both the genesis and dissipation of tropical

cyclones – potentially threatening more populated areas of Queensland, for

example. In addition they project a slight increase in the numbers of most severe

storms. The latter seems to be a common prediction by Global Climate Models for

all basins worldwide, as discussed in Section 8.1.

8.3 Simulations by Abbs et al. (2006) Simulations by Abbs et al. (2006) covered the Indian Ocean off north-west of

Western Australia, and the Timor Sea region, as well as the Coral Sea off the

Queensland coast. These simulations indicated a significant reduction in the

number of cyclones off Western Australia, little change off the Northern Territory

coast, but some increase off the North Queensland coast by 2070. Increases in

storm strength for the most severe storms were found by this model.

8.4 Simulations by Leslie et al. (2007) The study described by Leslie et al. (2007) was funded by the Insurance Australia

Group (IAG), but was limited by a relatively coarse grid spacing of 50 kilometres.

52

The model was used to simulate the climate in the south-west Pacific Ocean in a

‘control’ period of 1970-2000. The authors found a favourable agreement between

the ‘tropical-cyclone-like vortices’ in the simulated outputs from the GCM model and

the observed cyclones, with respect to the number of tropical cyclones in various

intensity categories, lifetime of storm, monthly distribution, and distribution with

latitude.

The model was then applied to the 2000-2050 period with both current and

enhanced greenhouse gas concentrations. No significant change was found in the

total tropical cyclone numbers in the south-west Pacific during 2000-2050.

However, there was a marked increase (about 22%) in the number of Category 3-5

storms in response to increasing greenhouse gases. A southerly shift of over 2

degrees of latitude in the tropical-cyclone genesis region was found.

The paper concluded that there is a potential for tropical cyclones to develop during

the next fifty years that are more intense that any so far recorded in the south-west

Pacific, including ‘super cyclones’ with central pressures below 900hPa. The latter

would produce extreme winds considerably in excess of those specified for building

design for any return period up to 2000 years along the Queensland coast (Regions B

and C in AS/NZS 1170.2).

8.5 Research at Geoscience Australia Geoscience Australia undertook a study for the Garnaut Review of the future impacts

of tropical cyclones in the Australian region under a range of climate change

scenarios. The Garnaut Climate Change Review was an independent study by

Professor Ross Garnaut, which was commissioned by the Commonwealth, State and

Territory Governments. The Review examined the impacts of climate change on the

Australian economy, and recommended medium to long-term policies and policy

frameworks to improve the prospects for sustainable prosperity.

The impact of climate change on tropical cyclone risk for the states of Queensland,

and Western Australia, and also the Northern Territory, was examined utilizing

global climate models available from the Intergovernmental Panel on Climate

Change (IPCC) Fourth Assessment Report. The model output was used to estimate

the influence of climate change on tropical cyclone activity following the technique

described in Vecchi and Soden (2007). The current generation of models lacks the

horizontal resolution necessary to resolve the intense inner core of tropical cyclones

so a thermodynamic approach aimed at identifying changes in large-scale

environmental factors (that are known to affect cyclone development/strength) is

53

employed. The maximum potential intensity (MPI) of a tropical cyclone was utilised

to determine changes in intensity. The MPI sets a theoretical upper limit for the

distribution of tropical cyclone intensity at a given point, given a vertical

temperature and humidity profile (Holland, 1997; Emanuel, 1999). A range of

climate change scenarios have been employed to assess the wind and storm surge

hazard and risk compared to current climate levels. The overall trend in MPI for the

climate change simulations is one of increased levels across northern Australia,

resulting in higher maximum wind speeds and elevated storm surge levels. This work

was reported in September 2008, and updated in March 2011, and is published on

the Garnaut Review website http://www.garnautreview.org.au/.

Geoscience Australia was also contracted by the Federal Department of Climate

Change to conduct a similar study to that Knutson et al. (2008) in the Australian

region. The regional modelling was undertaken by K. Emanuel (MIT) utilising the

IPCC Fourth Assessment Report simulations. A tropical-cyclone climatology (cyclone

tracks with associated intensity information) was produced for each IPCC model

simulation considered. The climatology is utilized by the Geoscience Australia

Tropical Cyclone Risk Model (TRCM) (Arthur et al., 2008; Arthur, 2010) through the

synthetic track generation module which can be seeded with either historical data or

alternatively the tracks of tropical cyclone-like vortices (TCLV’s) extracted from

climate simulations. In this way, and utilizing a statistical sampling process (Monte-

Carlo simulation), the model can rapidly increase the catalogue of TC events under

future climate regimes.

Finally, Geoscience Australia has recently collaborated with CSIRO Marine and

Atmospheric Research and University of Melbourne on a study of the effects of

climate change on wind speeds with high return periods in Australia. This study is

summarized in the following section.

8.6 Collaborative study of tropical cyclone wind risk in Australia A collaborative study was carried out in 2008-2011 to make practical

recommendations on the changes to extreme wind speeds in Australia as a result of

various climate change scenarios (Lavender et al., 2011). The future climate was

simulated using a range of global climate models to drive a regional climate model

of high resolution. This part of the work was summarized by Lavender and Walsh

(2011), and has already been discussed in Section 8.2. The results from all the

models predicted decreases of up to 30% in total numbers of tropical cyclones in

the Australian Region by the end of the 21st Century. In addition, the percentage of

very intense storms is predicted to increase. An average poleward (southerly) shift

54

in tropical cyclone occurrences of between 1 and 3 degrees of latitude is also

predicted.

The climate model projections were then used to produce spatial and intensity

distributions which in turn were used to generate thousands of simulated cyclones

with the same characteristics and hence values of V500 as used for most structures in

AS/NZS1170.2. Synthetic tracks of tropical cyclones were generated used an

autoregressive model similar to that proposed by Hall and Jewison (2007). The

wind-field model of Powell et al. (2005) and the boundary-layer model of Kepert

(2001) were adopted to generate gust wind speeds, which were then fitted with a

generalized extreme-value distribution. An important component of the modeling

carried out in this work was that all the simulations were ‘calibrated’ with observed

cyclone occurrences and wind speeds in the current climate prior to making

projections about future changes.

The report states that ‘3-second gust speeds’ were generated, but it is not clear if

these were 3-second moving-average gusts, or the somewhat higher (and shorter

duration) gusts used in AS/NZS1170.2 (see Appendix B).

An interesting aspect of the results in this report is that existing differences

between the current Regions C and D in AS/NZS1170.2 would be further enhanced

in future climate scenarios.

8.7 Summary By comparing and calibrating the simulations with the historical record of tropical

cyclone occurrences in number and intensity over the last 30 years or so, reasonable

predictions of events in a warmer climate may be made for the Australian region.

Only one of the studies to date has included the Indian Ocean basin off the north-

west of Western Australia, and the Timor Sea region, near Darwin, but there have

been several studies of the south-west Pacific and Coral Sea regions off Queensland.

In relation to the SW Pacific simulations, some consensus has emerged, with the

following general conclusions:

The overall number of tropical cyclones in the SW Pacific is expected to decrease by up to 30% in a warmer climate.

There is expected to be an increase in the number and frequency of the strongest storms in the next fifty years – i.e. in the Category 3 to 5 storms that are interest in structural design of buildings and other structures.

It is predicted that a southward shift of 1-30 in the genesis region and track locations will occur in the next 90 years.

55

The quality of these predictions is subject to the current resolution of the prediction

models, although these appear to have improved in recent years. However, the

increased resolution has tended to reinforce the above trends that were originally

identified in studies with coarser resolution, and the predicted trends for the

Australian region are similar to those predicted for other parts of the world. It is

expected that the resolution and hence the quality of the predictions will improve

further over the next few years.

The collaborative modelling of CSIRO, Geoscience Australia and the University of

Melbourne has considerable potential as a tool to refine estimates of design wind

speeds in Australia, as global climate model simulations improve in the future with

greater resolution.

References D.J. Abbs, S. Aryal, E. Campbell, J. McGregor, K. Nguyen, M. Palmer, T. Rafter, I. Watterson and B. Bates (2006), Projections of extreme rainfall and cyclones, Report to the Australian Greenhouse Office, CSIRO. W.C. Arthur (2010), Projected changes in cyclonic wind hazard in the Australian Region. 29th AMS Conference on Hurricanes and Tropical Meteorology, 10 - 14 May 2010, Tucson, Arizona, USA. W.C. Arthur, A. Schofield, R. P. Cechet and L. A. Sanabria (2008), Return period cyclonic wind hazard in the Australian Region. 28th AMS Conference on Hurricanes and Tropical Meteorology, 28 April - 2 May 2008, Orlando, Florida, USA. K.A. Emanuel (1999), Thermodynamic control of hurricane intensity, Nature, Vol.401, pp665-669.

T.M. Hall and S. Jewson (2007), Statistical modelling of North Atlantic tropical cyclone tracks, Tellus A, Vol. 59, pp 486-498. G.J. Holland (1997), The maximum potential intensity of tropical cyclones, Journal of Atmospheric Sciences, Vol.54, pp 2519-2541. J. D. Kepert (2001), The dynamics of boundary layer jets within the tropical cyclone core. Part 1: Linear theory, J. Atmos. Sciences, Vol.58, pp 2469-2484. T.R. Knutson, J. J. Sirutis, S. T. Garner, G. A. Vecchi, and I. M. Held (2008), Simulated

reduction in Atlantic hurricane frequency under twenty-first-century warming

conditions, Nature Geoscience, Vol.1, pp 359-364.

56

T.R. Knutson and 12 others (2010). Report of Working Group on ‘TC activity on climate time scales’, Seventh International Workshop on Tropical Cyclones, Reunion Island, November 15-20, 2010. S.L. Lavender and K.J.E. Walsh (2011), Dynamically downscaled simulations of Australian region tropical cyclones in current and future climates, Geophysical Research Letters, Vol.38, L10705. S.L. Lavender, K.J.E. Walsh, D.J. Abbs, M. Thatcher, W.C. Arthur and R.P. Cechet (2011), Regional climate tropical cyclone hazard for infrastructure adaption to climate change, Final Report, June 2011. L.M. Leslie, D.J. Karoly, M. Leplastrier, and B.W. Buckley (2007), Variability of tropical cyclones over the Southwest Pacific Ocean using a high-resolution climate model, Meteorology and Atmospheric Physics, Vol. 97, pp 171-180.

K.C. Nguyen and K.J.E. Walsh (2001), Interannual and decadal and transient greenhouse simulation of tropical cyclone-like vortices in a regional climate model of the South Pacific, Journal of Climate, Vol.14, pp 3043-3054. M. Powell, G. Soukup, S. Cocke, S. Gulati, N. Morisseau-Leroy, S. Hamid, N. Dorst and L. Axe (2005), State of Florida hurricane loss projection model: atmospheric science component, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 93, pp 651-674. G.A. Vecchi, and B. J. Soden, (2007), Increased tropical Atlantic wind shear in model projections of global warming, Geophysical Research Letters, Vol.34, L08702. K.J.E. Walsh and J.J. Katzfey (2000), The impact of climate change on the poleward movement of tropical-cyclone like vortices in a regional climate model, Journal of Climate, Vol. 13, pp 1116-1132. K.J.E. Walsh and B.F. Ryan (2000), Tropical cyclone intensity increase near Australia as a result of climate change, Journal of Climate, Vol. 13, pp 3029-3036. K.J.E. Walsh, K.C. Nguyen and J.L. McGregor (2004), Finer-resolution regional climate model simulations of the impact of climate change on tropical cyclones near Australia, Climate Dynamics, Vol. 22, pp 47-56. Y. Yamada, K. Oouchi, M. Satoh, H. Tomita and W. Yanase, (2010), Projections of changes in tropical cyclone activity and cloud height due to global warming, Geophysical Research Letters, Vol.37, L07709.

57

9. Observations and projections for the Northern Territory 9.1 Introduction This chapter reviews some studies by individuals and groups for cyclonic wind speeds

in Darwin and the Northern Territory. Comments are made about the appropriate

zoning of Darwin and other parts of the Northern Territory.

9.2 Reports by M. Nicholls A community Group for the ‘Review of NT Cyclone Risks’ was formed in 2005 and

received a grant from Emergency Management Australia in 2006 to study cyclone

risks in the Northern Territory. Mike Nicholls was the Secretary of the Group and

also the subcontractor for most of the work. The work was completed in early 2007

and a web site (www.cyclone.org.au) contains the report of the work consisting of a

main report and fifteen lengthy Appendices (Nicholls et al., 2007).

A two-page summary of the report was supplied to Sub-Committee BD006-02 of

Standards Australia in August 2007. The following comments relate primarily to that

summary.

The findings of the report were summarized as follows (with wording only slightly

altered):

a) Three of the most intense cyclones that have been observed in Australian waters since satellite observations began in 1960 (‘Thelma’ 1998, ‘Ingrid’ 2005 and ‘Monica’ 2006) all came within 350 km of Darwin when they were at maximum intensity and within a nine-year period. b) These ‘TIM’ cyclones may have been the result of global warming, or they may signal the return of a more active period similar to one that appears to have existed in the first 90 years of (European) NT settlement. c) NT buildings should probably be designed for wind loads that are at least 60% higher than the minimum loads permitted under the current AS/NZS1170.2.

a) is generally not disputed, although the exact strength of the three storms at

landfall is subject to speculation. All three were rated Category 5 at some point in

their lives by the Bureau of Meteorology, but they did not retain maximum strength

throughout their lifecycle.

58

Figure 9.1 (compiled by Dr. J. Kepert of the Bureau of Meteorology) shows estimated

(using the Dvorak imaging technique) values of the 10-minute mean wind speeds for

Cyclones ‘Thelma’, Ingrid’ and ‘Monica’ at various points along the tracks, together

with a circle of radius 350 km centred at Darwin. The large changes in intensity over

relatively short distances during their life-cycle can be noted from this Figure.

Figure 9.1. Tracks of Cyclones ‘Thelma’, ‘Ingrid’ and ‘Monica’, showing estimated 10-minute mean wind speeds in knots at 10 metres height, and a circle of 350km

radius, centred at Darwin, (source: Dr. J. Kepert, Bureau of Meteorology)

59

b) The first 90 years of European settlement produced a number of severe cyclones but without the benefit of modern satellites and ground-based instrumentation it is difficult to be categorical about their intensities. It is possible that the intensity of TIM cyclones was related to global warming. c) This conclusion is justified by some ‘ball-park’ probabilistic estimates. These are discussed in the following.

Estimates of the exceedence probability for given wind speeds in Darwin are stated

to be the product of the probabilities of three independent events:

a) the ‘time probability’ of a cyclone having at least that gust speed occurring in any one year (within 350 kilometres in Darwin), b) the ‘spatial’ probability of the region of maximum winds of the cyclone enveloping a building site, c) an ‘intensity’ probability that the cyclone will maintain its intensity ‘all the way to Darwin’ (this probability was assumed have a value of 0.5). Based on this approach, and the occurrence of the TIM cyclones in a 9-year period,

the Nicholls’ group derives a relationship between wind gust speed and average

recurrence interval (return period) which approximates Region D in AS/NZS1170.2.

Darwin is currently located in Region C in the Standard.

The methodology is, in fact, a simplified version of the probabilistic simulation

approaches described in Chapter 3. A similar approximate approach is used for the

Queensland coast in Section 10.4 of the present report. It is a valid approach, in

principle, but has many shortcomings in its implementation in the case of the NT

group. Some of these are as follows.

Estimated (from satellite images using the Dvorak method) gust wind speeds at 10 metres given in the Bureau of Meteorology database have been used, instead of using reported surface values of wind speed (see further discussion in the next section). The need to use validated surface data, instead of estimated values, was very clear in the case of Cyclone ‘Yasi’, in Chapter 10.

No wind field model (such as the well-established Holland (1980) model) has been adopted.

In the case of the Cyclone ‘Monica’, most of the 350 kilometres between Darwin and the centre of the cyclones was over land, not water (see Figure 9.1).

The spatial of ‘geometric’ probability (b) should be based on area, not radius, as used by Cook and Nicholls. That is the probability of intersection

60

of a point (building site) with the footprint area of maximum winds is required.

The intensity factor of (c) is very over simplified. As discussed in Chapter 5, the weakening in intensity depends on the distance travelled by a storm over land (see Chapter 5). Thus this factor would be very different for a storm approaching Darwin overland compared to those approaching from the sea.

The method does not take account of the preferred tracks of cyclones affecting the

Northern Territory coastline. In fact ‘Thelma’ and ‘Ingrid’ followed a generally east

to west track along the NT coast and were all well north of Darwin. As shown in

Section 9.5 the observed gust wind speeds at Darwin from ‘Thelma’ and ‘Ingrid’ were

all quite low. ‘Monica’ (2006) also produced quite a low gust at Darwin.

9.3 Report by Cook G.D. Cook (2007), in an unpublished manuscript, made predictions of wind gust

speed as a function of return period for three coastal locations in the Northern

Territory: Darwin, Maningrida and Nhulunbuy. These appear to have been made

partly on historical experience based on the cyclone database of the Bureau of

Meteorology, and partly using a simulated database generated by a U.S. group,

WindRiskTech. Cook concluded that the cyclone risk to Northern Territory stations

are better described by the Region D line AS/NZS1170.2 rather than that for Region

C. Surprisingly given the cyclones of recent years (i.e. the TIM cyclones), Cook

found Darwin to have a slightly higher risk than Maningrida and Nhulunbuy.

Cook makes the valid points that the Arafura Sea has shallower depths and warmer

temperatures than the North-West Shelf (WA) and Coral Sea (Queensland). Also

quoting references based on data from Atlantic hurricanes, wind fields are more

peaked with smaller radii of maximum winds in lower latitudes, so that maximum

gust speeds will be higher for a given central pressure differential. These factors

need to be included in simulation models to predict tropical cyclone wind speeds for

the Northern Territory.

In the introduction to his paper, Cook gives ‘estimated’ maximum gust speeds of 87,

91 and 99 m/s for Cyclones ‘Thelma’, ‘Ingrid’ and ‘Monica’. These values are derived

from those in the Bureau of Meteorology database, which gives estimates of

sustained wind speeds at 10 metres height, indirectly made using satellite images,

and some reported surface readings. The reported surface values (i.e. measured at

surface level) in the database are significantly lower than those used by Cook.

61

The probabilistic method used by Cook to predict wind speed is similar to that used

by Nicholls as described in the previous section; the associated criticisms of this

approach given above are valid.

9.4 Papers in Journal of Applied Meteorology and Climatology Since the first (2008) version of this report was published, a paper by Cook and

Nicholls (2009) has been published in the Journal of Applied Meteorology and

Climatology of the American Meteorological Society (AMS). Following the earlier

work described in this chapter, this paper applied three approaches to assess the

cyclone risk for Darwin, and concluded that the risk at Darwin for high return periods

was greater than that at both Townsville and Port Hedland.

The methods Cook and Nicholls used for this paper to assess the cyclone risk were as

follows:

a) An analysis of simulated synthetic cyclone tracks by WindRisk Tech of Boston

U.S.A. b) An analysis of the qualitative historical record of intense storms passing

within 50 kilometres of Darwin, Townsville and Port Hedland since European settlement, and,

c) An analysis of the record of intense storms in the ‘Cyclones’ database of the Bureau of Meteorology passing within 350 kilometres of each of the three locations since 1985.

Subsequent to the publication of the paper by Cook and Nicholls, a discussion paper

(Harper et al., 2010) has been submitted to the AMS commenting on the Cook and

Nicholls (2009) paper in detail. The latter paper found serious flaws in the

arguments of Cook and Nicholls. Specifically Harper et al. found that:

i) Cook and Nicholls made an invalid assumption for Method (c) above, by

assuming that cyclones are at their maximum strength along their entire path crossing the sampling circle, even after they have crossed extensive land areas.

ii) In Method (a), the annual rate of simulated cyclones at each station greatly exceeded the average annual rate for those recorded for the entire Australian region.

iii) Key cyclones were omitted from the analysis for Townsville and Port Hedland when comparing the risk at those location with that at Darwin.

In addition to the above flaws, Cook and Nicholls ignored any ground observations in

their analyses of historical events, and relied completely on satellite-based estimates

of wind speeds. As has been shown recently in Cyclone ‘Yasi’ (see Chapter 10), such

62

estimates are, in many cases, over-conservative and primarily used for warning

purposes, and should not be relied on for quantitative risk assessment.

As discussed in Section 9.5 of this report, Harper et al. subsequently showed that the

number of tropical cyclones, for any specified threshold, affecting Port Hedland

(Region D) greatly exceeded the number affecting Darwin over the same period.

Analysis of recorded gusts from anemometers at Port Hedland and Darwin

supported this, and it was concluded that based on the best available evidence from

the current climate, Darwin is adequately treated by its current location in Region C.

The data provided by WindRiskTech in Approach (a) by Cook and Nicholls, appears to

be very similar to that provided to Geoscience Australia (GA) for a cyclone risk study

for Australia (see Section 8.5). Comparisons of predictions by GA from this data set

generated by a Global Climate Model, and those based on the Bureau of

Meteorology’s ‘best-track’ database indicated over-estimates of values of design

wind speeds (V500) by more than 25% (Arthur, 2010) for the northern part of

Australia (north of 15o S). These differences are very close to the difference between

V500 for Regions D and C (88/69 = 1.27), and may explain the conclusions of Cook and

Nicholls based on the simulated data provided by WindRiskTech.

9.5 Observed wind speeds and overview of cyclone risk for Darwin and the NT The predictions of significantly higher wind speeds by Nicholls and Cook for Darwin,

than those that the Standard AS/NZS1170.2:2002 predicts, are not supported on

strong probabilistic arguments, and are partly based on overestimates of gust wind

speeds from recent cyclones. Also, as noted by Nicholls, his group’s predictions

contradict significantly earlier predictions for Darwin by Georgiou (2000) and Harper

(2005), (although these earlier studies did not include the full effect of the TIM

cyclones). Studies by Geoscience Australia (Arthur et al. 2008, Lavender et al., 2011)

showed a similar risk for Darwin as for the North Queensland coast, but significantly

lower risk than for the Port Hedland-Onslow coastal strip in Western Australia.

It is of interest to consider the observed gust wind speeds at Darwin Airport due to

tropical cyclones. This is shown in Figure 9.2 for the period 1960 to 2005. Maximum

wind gusts from the daily database during all cyclones in the Bureau of Meteorology

cyclone database are shown. The ‘official’ value of 61 m/s for Cyclone ‘Tracy’ (1974)

shown in the daily database for Darwin Airport, is shown although it is widely

believed that the true maximum is greater (the anemometer failed during this

event).

63

Figure 9.2. Maximum wind gusts speeds at Darwin Airport due to cyclones 1960-2005

The gust speed for ‘Tracy’ is clearly much larger than all others recorded. Those for

‘Thelma’ (1998) and ‘Ingrid’ (2005) at Darwin are both no more than 20 m/s. Figure

9.2 can be compared with similar figures for Onslow and Port Hedland (Region D)

shown in Appendix A. With seven cyclones producing gusts greater than 40 m/s,

and three events producing 50 m/s over a similar period, Port Hedland is clearly

historically much more ‘active’ than Darwin. Onslow had two events producing 60

m/s and four above 50 m/s during a similar period. However, Learmonth WA, also

currently in Region D, has a similar chart to Darwin with only one large event

(Cyclone ‘Vance’ in 1999).

It should be noted that the values of gust wind speeds in Figure 9.2 and in Appendix

A, since about 1990 (the exact date is uncertain), were recorded with automatic

weather stations (AWS), and a 3-second moving average was applied to the digital

outputs from the cup anemometers. This change results in recorded gusts, on

average being 15-20% lower than those from earlier years recorded by the Dines

anemometer (Ginger et al. 2011). Gusts from the latter instrument are the basis for

AS/NZS1170.2. However, application of increases of 15-20% to the later gusts in

Figure 9.2 by 15-20% would still result in them being all less than 30 m/s.

Similar data to that in Figure 9.2 are not available for stations on the northern

coastline and offshore islands of the Northern Territory, but it is possible that such

charts, if they were available, would be more similar to those for Port Hedland or

Onslow, based on recent cyclonic activity.

Cyclonic wind speeds - Darwin 1960-2005

0

10

20

30

40

50

60

70

1960

1964

1965

1968

1971

1974

1980

1981

1985

1987

1990

1995

1998

2002

Year

Gu

st

sp

eed

(m

/s)

Thelma 1998

Ingrid 2005 Tracy 1974

64

9.6 Discussion and conclusions Recent studies by Nicholls and Cook in Darwin have cast doubt on the current

(Region C) zoning for Darwin in AS/NZS1170.2. Apparently Nicholls and Cook

significantly over-estimated the gust speeds near ground level, by using estimated

wind speeds from satellite technology instead of the reported surface wind gust

speeds, in the Bureau of Meteorology database for Cyclones ‘Thelma’, ‘Ingrid’ and

‘Monica’ (the ‘TIM’ cyclones). The latter events had little effect on Darwin, but a

more significant effect on the northern coastline and islands of the Northern

Territory, particularly Cyclone ‘Ingrid’ – see Figure 9.1. Furthermore, some

simulated data used by Nicholls and Cook to support their case has been shown by

Geoscience Australia to greatly overestimate the cyclone risk in the northern part of

Australia (Arthur, 2010).

The need to use wind speeds from ‘ground truth’ studies, instead of estimated wind

speeds derived from satellite observations, and primarily used for warning purposes,

is reinforced by the analysis after Cyclone ‘Yasi’ (Chapter 10 in this report), which

found significantly lower wind speeds overland than those estimated from satellite

observations for that event.

Recent detailed comments by Harper et al. (2010) have reinforced and extended the

arguments given in this Chapter. This paper found that, based on the best

evidence available for the current climate, Darwin is correctly located in Region C in

AS/NZS1170.2.

Finally it should be noted that the recent simulations by Lavender et al. (2011) for

the whole Australian region, gave very similar predictions of design wind speeds for

Darwin as for the Queensland coast (i.e. Region C) and lower values than those

found for the north-west coast of Western Australia (i.e. the present Region D), for

both the current climate and for projected future climates under global warming

scenarios, and largely supported the present regional zoning in AS/NZS1170.2.

Consideration of the strength of the recent TIM cyclones and of the propensity of the

tracks to cross the islands along the northern extremity (i.e. Croker, Bathurst and

Melville Islands and the Cobourg Peninsula) of the Northern Territory, from east to

west, with little weakening suggests there may be a case for the allocation higher

design wind speeds to these locations. However, projections by Lavender et al.

(2011) (Fig. 19) suggest that extreme wind speeds in these areas will in fact

significantly reduce under future climate change scenarios for these islands, as they

will for Darwin itself.

65

References W.C. Arthur (2010), Projected changes in cyclonic wind hazard in the Australian Region. 29th AMS Conference on Hurricanes and Tropical Meteorology, 10 - 14 May 2010, Tucson, Arizona, USA. W.C. Arthur, A. Schofield, R. P. Cechet and L. A. Sanabria (2008), Return period cyclonic wind hazard in the Australian Region, 28th AMS Conference on Hurricanes and Tropical Meteorology, 28 April - 2 May 2008, Orlando, FL, USA. G.N. Boughton, D.J. Henderson, J.D. Ginger, J.D. Holmes, G.R. Walker, C.J. Leitch, L.R.Somerville, U. Frye, N.C. Jayasinghe, P.Y.Kim (2011), Tropical Cyclone Yasi – Structural damage to buildings, James Cook University, Cyclone Testing Station, Technical Report No. 57, March 2011. G.D. Cook (2007), Has the hazard from tropical cyclone gusts been underestimated for northern Australia? personal communication (M.J. Syme), CSIRO Sustainable Ecosystems. G.D. Cook and M.J. Nicholls (2009), Estimation of tropical cyclone wind hazard for Darwin: Comparison with two other locations and the Australian Wind Loading Code, Journal of Applied Meteorology and Climatology, Vol. 48, pp 2231-2340. P.N. Georgiou, (2000), On the probability of Darwin being struck by a category 5 cyclone. A report to the Northern Territory Department of Transport and Works and Colless & O'Neill Pty Ltd. Environment and Climate Risk Assessment, Pymble, NSW. B.A. Harper, (2005), Darwin TCWC Northern Region Storm Tide Prediction System. System Development Technical Report, pp. 112 + appendices. Systems Engineering Australia Pty Ltd, Bridgeman Downs, Qld. B.A. Harper, J.D. Holmes, J.D. Kepert, L.M. Mason, P.J. Vickery, (2010), Comments on “Estimation of tropical cyclone wind hazard for Darwin: Comparison with two other locations and the Australian Wind Loading Code”, by G.D. Cook and M.J. Nicholls, submitted to Journal of Applied Meteorology and Climatology. G.J. Holland, “A analytic model of the wind and pressure profiles in hurricanes ”, Monthly Weather Review, Vol. 108, pp 1212-1218, 1980. S.L. Lavender, K.J.E. Walsh, D.J. Abbs, M. Thatcher, W.C. Arthur and R.P. Cechet (2011), Regional climate tropical cyclone hazard for infrastructure adaption to climate change, Final Report, June 2011. M. Nicholls et al (2007), Review of NT cyclone risks, Report by the Community Group for the review of NT cyclone risks, April 2007. (available on CD-ROM and at www.cyclone.org.au).

66

10. Comments and observations from Cyclone ‘Yasi’ 10.1 Introduction

Tropical Cyclone ‘Yasi’ was a severe tropical cyclone, with a relatively large diameter,

that crossed the Queensland coast near Mission Beach just after midnight on

Thursday 3 February 2011. It moved into the Coral Sea from the east, having been

named in the Fiji region. ‘Yasi’ received extensive coverage by the media due to its

size, and it being called a Category 5 event by the Bureau of Meteorology. Damage

from wind forces and storm surge occurred between Townsville to Cairns, with the

most severe damage occurring between Ingham and Innisfail. A satellite image of

Cyclone ‘Yasi’ just before it crossed the coast is given on the front cover of this

report.

A full report on the wind and storm surge damage resulting from Cyclone ‘Yasi’ is

available (Boughton et al., 2011)

10.2 Windfield and maximum wind gusts Three approaches were used to estimate the maximum values of gusts reached at

the main centres affected by Cyclone ‘Yasi’:

a) Use of anemometer data from the Bureau of Meteorology, or other agencies,

where available. However, these were sparsely distributed and had some

siting problems.

b) A field investigation of failed and non-failed road signs (‘windicators’)

c) Use of the Holland wind field model (Holland, 1980) to predict wind speeds,

and interpolate between the ‘windicator’ estimates.

The ‘windicator’ technique, which had been used in Australia since Cyclone ‘Althea’

in 1971, was further developed in this event. Over 100 failed road signs were

inspected during the course of the field investigation. Many of these were found to

have failed as a result of a footing failure and were ignored. Detailed dimensions

were obtained from those that had shown a permanent deformation resulting from

generation of a plastic moment at, or near, ground level. In those cases, a suitable

non-failed sign was sought in the general vicinity, although this was not always

possible. When the information was available, lower and upper limits of gust wind

speed could be derived. The methodology for the ‘windicator’ approach is described

in TR57 (Boughton et al., 2011, Appendix A2)

The wind field for the study area was also mapped by the author, using the Holland

model to interpolate between the recordings from the ‘windicators’. The time

67

histories of wind speeds and directions predicted by the model were compared with

the available recorded data from four automatic weather stations. This approach

and the results for ‘Yasi’, are discussed in Appendix C (this is a revised version of that

given in TR57, Boughton et al., 2011).

The wind field modelling suggests that the maximum gusts experienced by structures

in the study area were about 65 m/s (standardized to 10 m height over flat, open

terrain, and approximately the gust to which a road sign of about 1m2 responds to).

This represents about 94% of the design wind gust speed (V500) for most structures in

Region C.

10.3 Gust wind speeds in ‘Yasi’ compared to VR in AS/NZS1170.2 Although the maximum gust (acting on a 1 square metre area) predicted to have

occurred anywhere during ‘Yasi’ of 65 m/s, was below the design wind speed (V500)

of 69.3 m/s for Level 2 buildings in Region C, it exceeded slightly the value (64 m/s)

for Level 1 buildings. Of course, the predicted maximum of 65 m/s was for a

relatively small area near Cardwell, and other locations received lower gusts.

A number of coastal locations on the south side of the point of eye crossing, such as

South Mission Beach, Tully Heads and Dunk Island, are predicted to have

experienced gusts equivalent to V100 (59 m/s or 212 kilometres per hour) or more –

see Table C3 in Appendix C. Only these plus a few other coastal towns such as

Bingil Bay and Kurrimine experienced V50 (55 m/s or 197 km/h) or greater gusts.

Nearly all locations listed in Table C3 in Appendix C3 experienced gust wind speeds

greater than V25 (49 m/s or 178 km/h), and Townsville, about 170 kilometres away

from the centre of ‘Yasi’ at its closest point, experienced wind gusts of 44 m/s

(approximately V15) as recorded by a Dines anemometer.

10.4 Risk of exceedence of ‘Yasi’ level winds in the future ‘Yasi’ was a strong event with winds approaching design values according to

AS/NZS1170.2, and the knowledge of its wind field characteristics can be used to give

rough estimates of the annual risk of exceedence of cyclonic winds of say 60 m/s.

Hence, an estimate can be made of the annual risk of exceedence of winds of 60 m/s

due to tropical cyclones on the Queensland coast, by assuming that a tropical

cyclone of the magnitude of ‘Yasi’ – i.e. capable of generating wind gusts of 60 m/s,

or greater, impacts on that coastline on average every 25 years. This is probably

reasonable based on the experience of the last 50 years, since the only other storm

to impact in that time capable of generating 60 m/s gusts was ‘Larry’ in 2006.

68

Assume the strip of coastline affected by 60 m/s gusts or greater, during such events

is 100 kilometres – i.e. roughly twice the diameter of the eye of ‘Yasi’. Then,

assuming a total length of coastline of 2000 kilometres – roughly the distance from

Bundaberg to Cape York, and the length of Region C in Queensland, the approximate

annual probability or exceedence of 60 m/s gusts is:

(Annual rate of occurrence of Yasi-level storms) (chance of intersection of any

given location with the 50 kilometre strip)

= (1/25) (100/2000) = (4/2000) = 1/500

This rough calculation suggests wind gusts with an average recurrence interval (or

‘return period’) of 500 years is about 60 m/s, somewhat less than the value of 69 m/s

given for V500 in AS/NZS1170.2:2011.

Of course, there may well be a greater chance of a storm of ‘Yasi’’s magnitude

striking the middle section on the Queensland coast, say between Cooktown and

Mackay, than there is to the north and south of those locations, respectively. Thus,

the length of the target strip might be reduced to 1000 kilometres. Also for climatic

reasons, such as an increase in La Nina events, or due to global warming, the

frequency of events such as ‘Yasi’ may increase – let us say the annual rate of

occurrence doubles to (2/25). In those conservative scenarios, the revised annual

probability of exceedence of 60 m/s gusts is:

= (2/25) (100/1000) = (1/125)

The value of V125 calculated from the equation for VR given for Region C below Table

3.1 in AS/NZS1170.2:2011 is:

Fc (122-104R-0.1) = 1.05 [122 – 104(125)-0.1 ] = 61 m/s

The above calculations, although very approximate, suggest that the values given in

AS/NZS1170.2:2011 are adequate, and, in fact, are conservative based on the

current climate for at least the Queensland east coast part of Region C. Of course,

this statement does not necessarily apply to other coastlines in Australia.

10.5 Conclusion

The extensive post-event analysis of Cyclone ‘Yasi’, the strongest event to strike the

east coast of Queensland in the last fifty years, has established to a good accuracy

the maximum gust wind field of Cyclone ‘Yasi’, with verified gust speeds based on

‘ground truth’. This leads to approximate estimates of average recurrence intervals

(i.e. ‘return periods’) for wind gusts of 60 m/s or greater. Such estimates indicate

69

that values currently given in AS/NZS1170.2:2011 for Region C are conservative, with

some allowance for future climate changes.

The analysis in this chapter indicates the need to use validated gust wind speeds,

based on measurements at ground level, and supported by wind field modelling,

such as that provided by the well-established Holland model. Such analysis and

modeling should be carried out whenever a severe event such as ‘Yasi’ strikes

anywhere along the tropical coastline of Australia.

It should be noted that estimates of wind gust speeds stated universally in the

media, both before and after ‘Yasi’, far exceeded those given in Appendix C and TR57

(Boughton et al., 2011); this led to unwarranted concerns regarding the adequacy of

the various Australian Standards and the Building Code of Australia to provide

adequate safety in events such as ‘Yasi’. Such concerns proved to be unfounded, at

least for buildings constructed since 1980, as the damage surveys have shown.

References

G.N. Boughton and nine others (2011), “Tropical Cyclone ‘Yasi’ – structural damage to buildings”, James Cook University, Cyclone Testing Station, CTS Technical Report 57, April 2011.

G.J. Holland (1980), “A analytic model of the wind and pressure profiles in hurricanes”, Monthly Weather Review, Vol. 108, pp 1212-1218.

70

11. Conclusions and recommendations 11.1 General conclusions The Australian Standard for Wind Actions (Standards Australia, 2011) sets wind

speeds for structural design for return periods specified by the Building Code of

Australia. The Standard has designated special regions for design wind speeds for

tropical cyclones since 1975. However, possible effects of climate change due to

global warming have, up to now, been deliberately excluded (Standards Australia

2002).

JDH Consulting has reviewed scientific literature and other sources, from both

Australia and overseas, relevant to the problem of predicting cyclonic wind speeds in

a warming climate. The trends in the last thirty years in which satellite images are

available and consistent are still somewhat inconclusive. The following statement

by the International Panel on Climate Change (IPCC) summarizes the global situation:

‘There is observational evidence for an increase in intense tropical cyclone activity in

the North Atlantic since about 1970, correlated with increases in sea surface

temperatures. There are also suggestions of increased intense tropical cyclone

activity in some other regions where concerns over data quality are greater. … There

is no clear trend in the annual numbers of tropical cyclones.’

Regarding projections into the future from models, the expert Working Group on

‘Tropical-cyclone activity on climate time scales’ reported to the Seventh

International Workshop on Tropical Cyclones (Knutson et al., 2010):

‘It is likely that the global frequency of tropical cyclones will either decrease or

remain essentially unchanged due to greenhouse warming. We have very low

confidence in projected changes in individual basins. Current models project changes

ranging from -6 to -34% globally, and up to +/-50% or more in individual basins by

the late 21st Century. Some increase in mean tropical cyclone maximum wind speed is

likely (+2 to +11% globally) with projected 21stCentury warming although increases

may not occur in all tropical regions. The frequency of the most intense storms will

more likely than not increase by a substantially larger percentage in some basins.’

In the Australian Region there has been a fall in the frequency of tropical cyclones

during most of the last thirty years. This is correlated with the Southern Oscillation

Index and the well-known El Nino phenomenon. The latter also manifests itself in

the increase in sea surface temperature in the central Pacific Ocean, and may be a

consequence of global warming. However, a strong La Nina event re-appeared in

2011 with Cyclone ‘Yasi’ (see Chapter 10) and the floods in southeast Queensland

being results of that.

71

Modelling by Global Climate Models (GCMs) for the Australian Region predict an

average fall in the number of cyclones in the Australian Region of 30% by the end of

the 21st Century (Lavender and Walsh, 2011).

There is evidence that the number of more intense cyclones (Category 3 and above)

in the Australian Region has increased slightly in the last thirty years. These more

intense storms are those of relevance for ultimate limits state design of buildings and

other structures. The apparent trend may be affected, at least partly, by changes in

the observational practices of the Bureau of Meteorology, and reflected in the

tropical cyclones database maintained by the National Climate Centre of the BoM.

However, an increase in more intense cyclones due to global warming, for the Coral

Sea region off the Queensland coast, is predicted by GCMs by several independent

studies. Three studies also predict a 2-3o southward shift in average tropical cyclone

occurrences off the Queensland coast.

Studies of the average inland decay of hurricanes in the United States and for

Cyclone ‘George’ (2007) in Australia, indicate the current widths of the regional

boundaries are slightly conservative but adequately specified. However, a clearer

definition of ‘smoothed coastline’ needs to be given in the Standard.

Recommendations for the design wind speeds for Regions B, C and D in

AS/NZS1170.2 are given in the Section 11.3 Other related recommendations are

listed in Section 11.4.

11.2 Estimates of the effects of climate change on design wind speeds An approximate estimate of the change in return periods of high wind speeds can be

made based on predicted changes in numbers and intensities of tropical cyclones as

a result of global warming.

The return periods of high wind speeds due to tropical cyclones can be written

approximately as:

Rv [.pL. pS Gv,c(V) ]-1 (11.1)

where is the average annual rate of occurrence of tropical cyclones in the region pL is the probability of a cyclone making landfall somewhere along the coastline of interest pS is the probability of the region of maximum winds affecting a particular site on the coastline,

72

Gv,c(V) is the probability of exceedence of the wind speed, V by the maximum gust in the maximum wind region of the cyclone. It should be noted that Eq. (11.1) is only valid for high return periods and the

corresponding high wind speeds, which can be assumed are caused by a single very

severe cyclone. The annual probability of exceedence of lower wind speeds is

obtained from an accumulation of contributions from a number of lesser events, and

in those cases Eq. (11.1) is not valid. However, Eq. (11.1) is sufficient to estimate the

changes in return period for the high wind speeds of ultimate limit states design

resulting from predicted changes in tropical cyclones.

For example, consider Coral Sea cyclones with the potential to affect 2000

kilometres of Queensland coastline. The average rate of occurrence of tropical

cyclones in the Coral Sea in recent years is about 3. Assume the probability of any

one of these making landfall somewhere along the eastern coastline is pL = 0.5.

Given a landfall, the geometric probability of a ‘diameter of maximum winds’ of 100

kilometres to envelope a given building site along the coastline is about 100/2000 or

0.05. Then substituting the known values for the current climate in Eq. (11.1) for a

return period R of 500 years and a gust wind speed at 10 metres of 69 m/s,

500 [3(0.5) (0.05) Gv,c(V) ]-1 from which we find Gv,c (V) = 0.0267, for V500= 69 m/s in the current climate. Assume that the annual number of cyclones of any strength in the region reduces by

30% (as predicted by the modeling of Lavender and Walsh (2011) for the Australian

Region), then reduces to an average of 2.1 per annum. Also, assume that the

intensity of the maximum wind speed, V, increases by 7%, so that V for the same

probability of exceedence, Gv,c(V) = 0.027, increases from 69 m/s to 74 m/s (this

increase is typical of what is being predicted by models as a result of global

warming).

Then the return period of 74 m/s in the new climate is

[2.1(0.5) (0.05) 0.0267]-1 = 714 years However the return period of 74 m/s for Region C in AS/NZS1170.2 in the current

climate is about 1150 years. Thus stronger winds are expected to occur more

frequently in this scenario.

Although this calculation is approximate, it is insensitive to the assumptions; for

example if the assumed probability of landfall, pL is changed to 0.4 in both the

current and future climates, there is no change in the calculated return period of 74

73

m/s of 714 years. However, of course, if the probability of landfall increases in the

future climate, the return period of any given wind speed will fall further.

This approach can be developed further and generalized, enabling interesting

scenarios to be investigated, e.g. it can be shown that for the above case the

frequency of cyclone occurrences in the Coral Sea would need to fall by about 60% to

compensate for an average increase in wind speed of the strongest storms by 7%.

On the other hand, a 30% fall in the average number of cyclones is sufficient to

compensate for a 3% increase in maximum wind speed of the strongest events.

11.3 Recommendations for the Australian Standard and BCA Based on the reviews carried out for this report, the following recommendations are

made.

1. The peak wind gust used as the basis for the Australian/New Zealand Standard

AS/NZS1170.2 needs re-definition, as it clearly is not the ‘3-second’ gust used by the

Bureau of Meteorology and the World Meteorological Organization. A proposal for

the re-definition of the peak gust in AS/NZS1170.2 is given in this report (Appendix

B).

2. The current regional boundary for Region D should be extended north east along

the Western Australian coast to 15oS. This will incorporate Broome and Derby. This

change reflects the occurrence zone of tropical cyclones (see Figure 1.1), and is also

recommended in the report by the Cyclone Testing Station on Cyclone ‘George’

(Boughton and Falck, 2007). Note this recommendation is based on the current

climate – not on any future projections resulting from global warming.

3. There is also a case to upgrade the most northerly offshore islands of the

Northern Territory and the northerly part of the Cobourg Peninsula. However, a

decision on the latter should be deferred until a detailed re-assessment of the

maximum wind speeds in Cyclones ‘Thelma’, ‘Ingrid’ and ‘Monica’ has been made.

The experience from Cyclones ‘Larry’ and ‘Yasi’ in Queensland for which detailed

analyses of the wind gusts over land were made, indicated much lower maximum

wind gusts than those predicted from satellite imaging approaches.

On the basis of existing information, it is not recommended that Darwin be included

in Region D.

4. An upgrading of the current Region B in Queensland between 25oS and 27oS, as

previously proposed, should also be put ‘on hold’ until further evidence is available,

and improved simulations based on climate models are available. However, it is

74

recommended an informative note should be added to AS/NZS1170.2 and AS4055,

warning of the possibility of cyclones up to Category 3 penetrating further south

along the Queensland coast.

5. The existing factors for Regions C and D, FC and FD, should be incorporated into

the regional wind speeds. Severe cyclones since 1999 such as ‘Vance’, ‘Larry’, and

Yasi’ have justified incorporating the increases into the design wind speeds.

However, there is evidence to increase the design wind speeds beyond the present

values from climate change predictions. However, it appears that projected

increases in the intensities of the strongest cyclones are at least partly mitigated by

the projected reduction in the average number of cyclones of about 30%.

6. The ‘smoothed’ coastline may be defined by applying a ‘moving average’ filter

with a length of 50 kilometres to the actual coastline. This value is characteristic of

the eye diameter of a typical cyclone, and ensures that minor features that will not

affect the storm characteristics, are not included. A project should be initiated to

provide a map of Australia with this ‘smoothed’ coastline on an internet web site, for

ease of use by designers and local authorities.

7. The vertical profiles of wind gust speed for Regions C and D in AS/NZS1170.2

have been derived from measured data for a particular location, 3-5 kilometres from

the coastline, and incorporate a sea-land transition that may not be appropriate

further inland. The wind profiles (Mz,cat) for tropical cyclones need to be re-

considered taking account of recent dropwindsonde measurements in U.S.

hurricanes.

8. The Standard currently recommends that Terrain Category 2 (roughness length =

0.02 m) should be used for off-ocean winds for both cyclonic and non-cyclonic

regions for ultimate limit states design. However the recent dropwindsonde

measurements in hurricanes off the United States (discussed in Chapter 4) indicate

that the water surface is much smoother at high wind speeds near the radius of

maximum winds. It is therefore recommended that Terrain Category 1 be specified

for off-water winds for coastal locations in Regions C and D.

11.4 Other recommendations 1. Global climate models (such as those described by Lavender and Walsh, 2011)

should be extended with surface wind field models to give predictions of extreme

gust wind speeds at 10 metres height for structural design. Such work has already

commenced with the recent collaborative work reported by Lavender et al. (2011).

It is important that such models be ‘calibrated’ against the current climate.

75

2. As climate change effects on tropical cyclone activity is the subject of ongoing

studies by several groups in Australia and overseas, a similar review to the one

carried out for this report should be undertaken every 3 to 5 years. The present

report is a starting point for these revisions, following the original report produced

by JDH Consulting in 2008.

3. The historic tropical cyclone database maintained by the Bureau of Meteorology

is most important for the prediction of cyclonic wind speeds and other effects. It is

known to have significant errors and is currently being revised by the Bureau. There

also appear to be systematic errors in the wind speeds listed in the database derived

from satellite images by the Dvorak technique. These values are unrealistically high

and inconsistent with surface observations, including some of those listed elsewhere

in the database. These errors have led to inaccurate conclusions to be drawn by

some users of the database, and it is recommended that the methodology used to

derive the wind speeds be reviewed.

4. A recent project (funded in part by the Department of Climate Change and

Energy Efficiency) on the response of the anemometer systems used in the past and

present by the Bureau of Meteorology has shown there are significant differences

between the pre-1990 readings from Dines anemometers and those from the

current Automatic Weather Stations, particularly at stations affected by tropical

cyclones. The Bureau of Meteorology should provide details of the change-over

dates of the anemometers so that appropriate corrections of the historical database

can be made.

5. The quality and quantity of observations of the strength and effects of tropical

cyclones in the Australian Region is vastly inferior to that available in the United

States. Accepting that resources will not allow the same level of instrumentation to

be deployed in Australia, some increased effort should be made in this country in the

following areas:

Deployment of mobile anemometers before land-falling cyclones to obtain accurate spatial wind speed information from future cyclones. (This is currently practised by several groups in the U.S.; however in Australia damage and post-event surveys have had to rely on wind damage assessment from structures such as road signs, and even from tree damage).

Deployment of dropwindsondes (see Chapter 4) in Australia, including the possibility of dropping them over remote land sites (deployment over land is currently not permitted in the United States)

76

References G.N. Boughton and D. Falck (2007), Tropical Cyclone George – Wind penetration inland, Cyclone Testing Station, James Cook University, Technical Report No.53, August 2007. CSIRO (2007). Climate change in Australia. Technical Report. CSIRO. T.R. Knutson and 12 others (2010). Report of Working Group on ‘TC activity on climate time scales’, Seventh International Workshop on Tropical Cyclones, Reunion Island, November 15-20, 2010. S.L. Lavender and K..J.E. Walsh (2011), Dynamically downscaled simulations of Australian region tropical cyclones in current and future climates, Geophysical Research Letters, Vol.38, L10705. S.L. Lavender, K.J.E. Walsh, D.J. Abbs, M. Thatcher, W.C. Arthur and R.P. Cechet (2011), Regional climate tropical cyclone hazard for infrastructure adaption to climate change, Final Report, June 2011. Standards Australia (2002), Structural design actions-Wind actions-Commentary. (Supplement to AS/NZS 1170.2:2002). Standards Australia (2011), Structural design actions. Part 2: Wind actions. AS/NZS 1170.2:2011.

77

APPENDIX A

MAXIMUM RECORDED WIND GUSTS FROM TROPICAL CYCLONES AT SELECTED LOCATIONS IN AUSTRALIA

The following charts show gusts primarily recorded by Dines anemometers. However later values since about 1990 (the actual date varies from station to station) have been recorded by cup anemometers associated with Automatic Weather Stations (AWS). In a few cases at Onslow and Port Hedland, modeled values have been substituted when the anemometer has failed. It should be noted that the gusts recorded by the AWSs have also been digitally averaged with a 3-second moving-average ‘filter’, following a World Meteorological Organization (WMO) recommendation. A recent study (Ginger et al., 2011) has found that this processing, together with the change of anemometer, produces maximum gusts that are, on average, 15-20% less than earlier values generated by Dines anemometer which are the basis of AS/NZS1170.2. For example, the Dines anemometer at Learmonth during Cyclone ‘Vance’ read 15% greater than the gust recorded by the AWS; (the Dines reading is shown in the chart for Learmonth following). The later gusts shown in the following charts have not been corrected for the change of measurement systems; however such corrections will be required in the future, once change-over dates have been established.

Cyclonic wind speeds - Broome 1957-2007

0

10

20

30

40

50

60

70

1957

1961

1963

1970

1973

1975

1977

1980

1983

1986

1988

1993

1997

1999

2001

2004

2007

Win

d g

ust

sp

eed

(m

/s)

Broome (Region C) has been fortunate in not experiencing the full brunt of a tropical cyclone although it is located close to the most active part of the Australian region.

78

Cyclonic wind speeds - Cairns 1956-2001

0

10

20

30

40

50

60

70

1956

1959

1961

1969

1971

1976

1977

1979

1981

1983

1986

1997

2000

Win

d g

ust

sp

eed

(m

/s)

Cairns (Region C) has not received a direct strike by a tropical cyclone since 1956.

Cyclone ‘Winifred’ in 1986 did significant damage to Innisfail to the south of Cairns but only produced a 30 m/s gust at Cairns Airport. More recently there have been

some effects from Cyclone ‘Larry’ (2006) and ‘Yasi’ (2011).

Cyclonic wind speeds - Darwin 1960-2005

0

10

20

30

40

50

60

70

1960

1964

1965

1968

1971

1974

1980

1981

1985

1987

1990

1995

1998

2002

Year

Gu

st

sp

eed

(m

/s)

Darwin (Region C) was directly hit by Cyclone ‘Tracy’ in 1974, but no other direct

strikes between 1959 and 2005.

79

Cyclonic wind speeds - Learmonth 1979-2006

0

10

20

30

40

50

60

70

80

1979

1981

1982

1983

1984

1985

1986

1986

1987

1989

1991

1995

1995

1996

1997

2000

2006

Year

Win

d g

ust

sp

eed

(m

/s)

Learmonth/Exmouth (Region D) experienced the largest gust in Australia recorded by

a Dines anemometer during Cyclone ‘Vance’ in 1999 (74 m/s)

Cyclonic wind speeds - Onslow 1958-2004

0

10

20

30

40

50

60

70

80

1958

1960

1962

1963

1964

1966

1968

1970

1971

1973

1975

1976

1980

1981

1984

1986

1988

1990

1995

1996

1999

2000

2004

Win

d g

ust

sp

eed

(m

/s)

Onslow (Region D) has experienced several strikes between 1959 and 2004, with a highest gust of 68.5 m/s from Cyclone ‘Trixie’ in 1975.

80

Cyclonic wind speeds - Port Hedland 1958-2004

0

10

20

30

40

50

60

70

1958

1960

1962

1963

1968

1970

1973

1974

1975

1979

1981

1983

1984

1987

1991

1996

1998

2000

2002

2004

Win

d g

ust

sp

eed

(m

/s)

Port Hedland (Region D) experienced several high cyclonic gusts between 1960 and 2004, including those from Cyclone ‘Joan’ in 1975, and ‘Leo’ in 1977.

81

APPENDIX B

PROPOSAL FOR A NEW DEFINITION OF THE PEAK GUST FOR AS/NZS1170.2

B1. Introduction The Australian/New Zealand Standard for Wind Actions AS/NZS1170.2:2011 is based on a basic wind speed which is currently described in Section 3.2 of the Standard as ‘a 3 second gust’ (Standards Australia, 2011). The original version of AS1170.2-1973 (Standards Australia, 1973), and its predecessor CA 34 Part 2 (Standards Australia, 1971), both refer to ‘a gust of 2 to 3 seconds duration’ as the basic wind speed. This definition appears to have originated from the report by Whittingham (1964), who stated that ’the Dines anemometer gives a good indication of the speed of strong gusts of 2 to 3 seconds duration’. The ‘2-3 second definition’ continued up the 1989 edition of AS1170.2 (Standards Australia, 1989), which stated that ‘the basic design gust wind speed is defined in this standard as being the maximum 2-second to 3-second gust occurring within 1 hour...’. With the benefit of modern equipment and techniques the recent DCEE-supported project (Ginger et al. 2011) has shown that the first natural mode of the low-speed Dines anemometer has a period of about 2 seconds, and the high-speed type, used mainly in cyclonic regions, has a natural period of about 3 seconds. However, these values are not equivalent moving, or block, averaging times. It can be shown, from the results in the DCEE project, that the equivalent moving averaging time for the damped resonant response of the Dines anemometer, in relation to the maximum gusts produced, is actually about 0.1 to 0.3 seconds (depending on the type, and on the mean wind speed) – i.e. much shorter than the 2-3 seconds stated in the Standard. Since the Australian Standard is largely based on uncorrected gusts from the Dines anemometers, it can therefore be asserted that the basis of the Standard is a gust with a much shorter averaging time. Shortly after the introduction of Automatic Weather Stations with cup anemometers (about 1990), the Bureau of Meteorology incorporated a 3-second moving average based on a World Meteorological Organization specification (Beljaars, 1987), in addition to the inherent filtering of the cup anemometer itself. This ‘double filtering’ results in severely attenuated wind spectra at the high-frequency end, and in maximum gusts 10-20% less than those previously recorded by the Dines anemometers, and forming the basis for the Standard. In this Appendix, an alternative definition of maximum gust is proposed, which is generally consistent with the implied definition used in Australian and New Zealand Standards for the last 40 years. However, new gust data generated since 1990, incorporating the 3-second moving average filter, will require correction before being used for basic regional wind speeds in the Standard.

82

The alternative option of re-defining the gust wind speed in AS/NZS1170.2 to that used by the Bureau of Meteorology and the WMO, would require a reduction in wind speeds that would need to vary from region to region, and a corresponding increase in shape factors.

B2. Proposed definition The proposed definition of the basic maximum gust in AS/NZS1170.2 is as follows: ‘The largest expected gust averaged over an area of 1 square metre normal to the wind flow, centred at a height of ten metres over flat, open country, terrain’. The above definition incorporates an ‘aerodynamic admittance’ function for the area of 1 square metre, which effectively filters the turbulence in the approaching wind. This is a small enough area for the assumed quasi-steady relationship between peak gusts in the approaching wind, and peak wind pressures and forces to apply, and it also results in little reduction in the maximum gust as recorded by an ‘ideal’ anemometer. An appropriate function for the aerodynamic admittance is discussed in the following section, and subsequently used to develop correction values for recorded gust data. B3. Methodology for deriving correction factors Random process theory can be used to predict the wind gust factors recorded by 3-cup and Dines anemometers in a turbulent wind of known intensity and spectral density. It can also be used to derive the gust factor for the wind spectral density ‘filtered’ by the aerodynamic admittance for the area in the proposed definition. Using this approach, the spectral density of the wind turbulence was modelled using the well-known von Karman form, which, in non-dimensional form, can be written as follows:

6/52

2

8.701

4)(.

U

n

U

n

nSn

u

u

u

u

(B1)

where u is an integral length scale, and u is the standard deviation of turbulence which can be obtained from the mean wind speed and the

intensity of turbulence (i.e. u = Iu U).

For a height of 10 metres, AS/NZS1170.2 gives a value of u of 85 metres. The aerodynamic admittance was incorporated by multiplying the spectral density by a transfer function |H (n)|2. The following expression was used, based on numerous measurements on small plates in grid turbulence (Vickery, 1968):

83

23/4

2

21

1|)(|

U

An

nH

(B2)

where A is the exposed area. For calculation of the gust factors associated with the Dines and cup anemometers, including the effect of 3-second moving averaging, transfer functions have been given in Appendix II-4 of the DCCEE report (Ginger et al., 2011).

The cycling rate or ‘average frequency’, , of the filtered process can be calculated as follows.

(B3)

Note: It should be noted that the numerator in Eq. (B3) with use of Su(n) from Eq. (B1),unfiltered, i.e. with |H(n)|2 equal to 1.0 for all n, will not converge. Thus, some form of filtering is required for a cycling ratio to be calculated. At very high frequencies in the atmosphere, the spectrum decays faster than given by Eq. (B1) due to the formation of a viscous sub-range. The expected peak factor can be calculated using the well-known formula for Gaussian random processes of Davenport (1964):

(B4)

where is Euler’s Constant (0.5772), and T is the sample time for which the expected peak is to be determined.

Finally the expected gust factor was obtained from: (B5)

where u,f is the standard deviation of the filtered process given by:

(B6)

B4. Spectral densities The spectral density of the turbulence as modified by either the 1 square metre area, or by an anemometer, is shown, for two different mean wind speeds, in Figures B1 to B4. It can be seen that the modified spectrum in Figure B1, under the aerodynamic admittance for 1 square metre, i.e. that associated with the proposed definition, moves to the right – i.e. increasing frequencies with increasing wind speed, along with the underlying wind spectrum.

84

The spectra modified by the low speed (100 knot) and high-speed (200 knot) types of Dines anemometer are shown in Figures B2 and B3, respectively. These show peaks due to the resonant frequencies of the float system, and hence ‘distorted’ spectra at the high-frequency end. The resonant frequencies have been assumed to remain constant with changing wind speed. Work by Kepert for the DCCEE project (2011) has indicated that this is the case for the dominant low-frequency peak. The spectra from the cup anemometers, with a 3-second moving average filter, are shown in Figure B4. The high-frequency end of the spectral density is truncated, with more removed as the mean wind speed increases. This is not a good characteristic for a definition of a wind gust associated with wind loads on structures. B5. Calculated correction factors

Correction factors for maximum gusts for various averaging times (i.e. 1 minute and 10 minutes) mean wind speeds and turbulence intensities, for the low-speed Dines anemometer, high-speed Dines anemometer and 3-cup anemometer with 3-second moving average filtering are shown in Tables B1, B2 and B3 respectively. The averaging time of 60 seconds with lower turbulence intensity is more appropriate for thunderstorm winds (Holmes et al. 2007). The 10-minute (600 seconds) averaging time is more applicable to synoptic scale winds, including tropical cyclones. The correction factors for both types of Dines anemometer (Tables B1 and B2) are small, and insensitive to averaging time, mean wind speed and turbulence intensity. The correction increments are generally less than 3% for the low-speed Dines, and less than 1% for the high-speed version of the Dines. The correction factors required for the gusts from cup anemometers, with 3-second moving average filtering, as used by the Bureau of Meteorology since the mid-1990s, are shown in Table B3. These range from 1.09 to 1.19, and are dependent on the mean wind speed and turbulence intensity. The dependence on the mean wind speed is not surprising, given that the moving average time is unchanged at 3 seconds for all wind speeds, and hence truncates more and more of the high frequency end of the wind spectrum, as the wind speed increases, as shown in Figures 4(a) and 4(b).

85

Table B1. Correction factors for maximum gusts from Low-speed Dines

anemometers

Mean wind speed

Turbulence intensity

Sample time

(secs)

Avg. Gust factor (1m2

area)

Max. gust (m/s)

Avg. Gust factor

(l.s. Dines)

Correction factor

20 0.10 60 1.276 25.5 1.248 1.022

25 0.10 60 1.279 32.0 1.249 1.024

30 0.10 60 1.281 38.4 1.251 1.025

35 0.10 60 1.283 44.9 1.251 1.026

40 0.10 60 1.285 51.4 1.252 1.026

20 0.10 600 1.347 26.9 1.323 1.018

25 0.10 600 1.350 33.7 1.324 1.019

30 0.10 600 1.351 40.5 1.324 1.021

35 0.10 600 1.353 47.3 1.324 1.022

40 0.10 600 1.354 54.2 1.324 1.022

20 0.15 600 1.521 30.4 1.485 1.024

25 0.15 600 1.524 38.1 1.486 1.026

30 0.15 600 1.527 45.8 1.486 1.028

35 0.15 600 1.529 53.5 1.486 1.029

40 0.15 600 1.531 61.2 1.486 1.030

20 0.20 600 1.694 33.9 1.647 1.029

25 0.20 600 1.699 42.5 1.648 1.031

30 0.20 600 1.703 51.1 1.648 1.033

35 0.20 600 1.705 59.7 1.648 1.035

Table B2. Correction factors for maximum gusts from High-speed Dines

anemometers

Mean wind

speed Turbulence

intensity Sample

time (secs)

Avg. Gust factor

(1m2 area)

Max. gust (m/s)

Avg. Gust factor

(hs. Dines)

Correction factor

20 0.10 60 1.276 25.5 1.263 1.010

25 0.10 60 1.279 32.0 1.266 1.011

30 0.10 60 1.281 38.4 1.268 1.011

35 0.10 60 1.283 44.9 1.270 1.011

40 0.10 60 1.285 51.4 1.271 1.011

20 0.15 600 1.521 30.4 1.514 1.004

25 0.15 600 1.524 38.1 1.517 1.005

30 0.15 600 1.527 45.8 1.520 1.005

35 0.15 600 1.529 53.5 1.522 1.005

40 0.15 600 1.531 61.2 1.523 1.005

45 0.15 600 1.532 68.9 1.524 1.005

50 0.15 600 1.533 76.7 1.525 1.005

20 0.20 600 1.694 33.9 1.685 1.005

25 0.20 600 1.699 42.5 1.690 1.006

30 0.20 600 1.703 51.1 1.693 1.006

35 0.20 600 1.705 59.7 1.696 1.006

40 0.20 600 1.708 68.3 1.698 1.006

45 0.20 600 1.709 76.9 1.699 1.006

50 0.20 600 1.711 85.5 1.700 1.006

86

B6. Discussion and conclusions The changeover in the early to mid 1990s from Dines anemometers to Automatic Weather Systems has resulted in a significant step change in the peak gusts routinely recorded by the Bureau of Meteorology, and supplied as ‘maximum daily gusts’ by the National Climate Centre. This change has required a re-assessment of the peak gust used as a basis for the Australian/New Zealand Standard on Wind actions, AS/NZS1170.2. One option would be to switch the basis of the Standard to the ‘3-second gust’ based on a moving average of that duration, currently being used by the Bureau. However, this option has three disadvantages:

a) Existing regional wind speeds in the Standard would need to be reduced by amounts of about 8 to 20%, depending on the region, together with corresponding increases in shape factors of double those amounts to compensate.

b) The Standard is based on a ‘quasi-steady’ model of wind loading in which pressure coefficients from a variety of wind-tunnel test results can be adapted for use in the Standard. This model requires little, or no, correction to the calculated wind loads for bodies with small frontal areas – say from 1 to 5 square metres, if a gust of a short effective duration is used. However the ‘3-second gust’ as defined by the WMO, requires a much larger frontal area of more than 1000 square metres, for the quasi-steady assumption to apply. Factors to increase the calculated loads would be required if they were applied to smaller structures.

c) The use of a 3-second moving average, irrespective of the wind speed, has undesirable effects on the apparent spectrum of turbulence as shown by Figure 4, truncating larger and larger parts of the high-frequency end of the spectrum, with this effect increasing with increasing wind speeds. This would require special factors to be introduced for regions with higher design wind speeds to compensate.

The alternative proposal presented here does not have any of the above disadvantages and would preserve the existing format of the Standard used since 1971. Very small corrections, possibly considered as negligible, to the ‘old’ data recorded by the two types of Dines anemometers, would be required, (see Tables B1 and B2). The proposed new ‘1 square metre’ definition would require significant corrections to the ‘new’ maximum gust data emerging from the Automatic Weather Systems, but this would only be required to be applied by those persons analyzing extreme wind gust data for use with the Standard. However, a possible disadvantage, perceived by some, would be a departure from ‘the international standard

87

definition’ for a peak gust. Confusion might also arise for users who interpret ‘on-line’ gust data from the AWSs, without making the appropriate corrections. B7. References A.G. Davenport (1964), Note on the distribution of the largest value of a random function with application to gust loading, Proc., Institution of Civil Engineers (U.K.), Vol. 28, pp187-196. A.C.M. Beljaars (1987), The measurement of gustiness at routine wind stations – a review, Instruments and Observing Methods Report, World Meteorological Organization, No.31, W.M.O. Geneva. J.D. Kepert (2011), Modelling the transient response of the Dines Anemometer, Appendix II-5 in , ‘Extreme windspeed baseline climate investigation project’, Report to the Department of Climate Change and Energy Efficiency, April 2011. Standards Association of Australia (1971), SAA Loading Code. Part II: Wind forces, AS CA34, Part II-1971, Standards Association of Australia, North Sydney. Standards Association of Australia (1973), SAA Loading Code. Part 2: Wind forces, AS1170.2-1973, Standards Association of Australia, North Sydney. Standards Association of Australia (1989), SAA Loading Code. Part 2: Wind forces, AS1170.2-1989, Standards Association of Australia, North Sydney. Standards Australia (2011), Structural design actions. Part 2: Wind actions, Australian-New Zealand Standard, AS/NZS1170.2:2011. B.J. Vickery (1968), Load fluctuations in turbulent flow, ASCE, J. Eng. Mech. Div, Vol. 94, pp31-46. H.E. Whittingham (1964), Extreme wind gusts in Australia, Bureau of Meteorology, Bulletin No. 46.

88

0

0.1

0.2

0.3

0.4

0.001 0.01 0.1 1 10

Unfiltered spectrum

1 square metre

n.Su(n)/2

Frequency (Hertz)

0

0.1

0.2

0.3

0.4

0.001 0.01 0.1 1 10

Unfiltered spectrum

1 square metre

n.Su(n)/2

Frequency (Hertz)

(a) (b)

Figure B1. Spectral density of turbulent velocities, as ‘filtered’ by the 1 m2 sample

area. (a) Mean wind speed = 20 m/s. (b) Mean wind speed = 40 m/s

89

0

0.05

0.1

0.15

0.2

0.25

0.3

0.001 0.01 0.1 1 10

Unfiltered spectrum

Dines anemometer -low speed

n.Su(n)/2

0

0.05

0.1

0.15

0.2

0.25

0.3

0.001 0.01 0.1 1 10

Unfiltered spectrum

Dines anemometer -low speed

n.Su(n)/2

(a) (b)

Figure B2. Spectral density of turbulent velocities, as recorded by a low-speed Dines anemometer.

(a) Mean wind speed = 20 m/s. (b) Mean wind speed = 40 m/s

90

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.001 0.01 0.1 1 10

Unfiltered spectrum

Dines anemometer -high speed

n.Su(n)/2

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.001 0.01 0.1 1 10

Unfiltered spectrum

Dines anemometer -high speed

n.Su(n)/2

(a) (b)

Figure B3. Spectral density of turbulent velocities, as recorded by a high-speed Dines anemometer.

(a) Mean wind speed = 20 m/s. (b) Mean wind speed = 40 m/s

91

0

0.1

0.2

0.3

0.4

0.001 0.01 0.1 1 10

Unfiltered spectrum

Filtered cup anemometer

n.Su(n)/2

Frequency (Hertz)

0

0.1

0.2

0.3

0.4

0.001 0.01 0.1 1 10

Unfiltered spectrum

Filtered cup anemometer

n.Su(n)/2

Frequency (Hertz)

(a) (b)

Figure B4. Spectral density of turbulent velocities, as recorded by a cup anemometer with 13m distance constant, plus 3-second moving averaging

(a) Mean wind speed = 20 m/s. (b) Mean wind speed = 40 m/s

92

APPENDIX C

WIND FIELD MODEL FOR CYCLONE ‘YASI’

C.1 Introduction

A report on structural damage from ‘Yasi’ was published a few weeks after the event

by the Cyclone Testing Station of James Cook University (Boughton, et al., 2011).

This report (TR57) includes an analysis of the windfield generated by ‘Yasi’ in relation

to distribution of the maximum wind gusts over land, primarily in the zone affected

by the strongest winds.

A revised model and predictions of maximum gusts at key locations is presented

herein. The main differences result from the separation of the 3-second moving

average gusts obtained from the automatic weather stations and the ‘1 square

metre’ gusts obtained from the analysis of road signs (‘windicators’). The former is

the international standard for defining peak gusts proposed by the World

Meteorological Organization (Beljaars, 1987), and the latter is very close to that

recorded by Dines anemometers and currently the basis for the Australian Standard

for Wind Actions (Standards Australia, 2011).

The need to separate the gusts as above is a result of a project recently completed

for the Department of Climate Change and Energy Efficiency (DCCEE) (Ginger et al.,

2011) which identified differences between the two types of maximum gusts, of up

to 20% in cyclonic situations.

C2. The Holland model Outside of the eye of a tropical cyclone, the wind speed at upper levels decays with

the radial distance from the storm centre. The gradient wind equation (Equation

(C1)) can be used to determine this wind speed:

r

pr

4

rf

2

rfU

a

22

(C1)

where f is the Coriolis parameter (=2 sin )

is the angular velocity for the earth’s rotation (rad/sec)

is the angle of latitude r is the radius from the storm centre

a is the density of air p is the atmospheric pressure

93

To apply Equation (1), it is necessary to establish a suitable function for the pressure

gradient. A commonly assumed expression is the following (Holland, 1980):

(C2)

where pc is the central pressure of the tropical cyclone pn is the atmospheric pressure at the edge of the storm A and B are scaling parameters

An earlier version of Equation (C2) with the B parameter fixed at 1.0 was proposed

by Myers (1957), and was, in fact, used by the author to estimate wind speeds at

Burketown during Cyclone ‘Ted’ (Holmes, 1977).

The pressure difference (pn – pc) can be written as p, and is an indication of the

strength of the storm.

Differentiating Equation (C2) and substituting in (C1), we have:

(C3)

This is an equation for the mean wind field at upper levels (gradient height) in a

tropical cyclone as a function of radius from the storm centre, r, the characteristic

parameters, A and B, the pressure drop across the cyclone, p and the Coriolis

parameter, f.

Near the centre of a tropical cyclone, the Coriolis forces, i.e. the first two terms in

Equations (1) and (2), are small, and it can be shown by differentiating the remaining

term that the maximum value of U occurs when r equals A1/B. Thus A1/B is to a good

approximation, the radius of maximum winds in the cyclone. The exponent B is

generally found to be in the range 1.0 to 2.5 (Holland, 1980).

C3. The Holland model as applied to ‘Yasi’ Equation (C3) requires a number of parameters to be provided:

The central pressure of the cyclone, pc. In this case, it was taken as 930 hPa, based on the measurement at Clump Point, close to the point of landfall of Cyclone Yasi.

The ambient pressure, far from the centre of the cyclone, pn. In this case, an average value of barometric pressure well before the cyclone made landfall was used. Thus at 3 a.m. on February 2, the barometric pressures at Cairns, Townsville, Lucinda Point and South Johnstone were respectively: 1006 hPa,

B

cn

c

r

Aexp

pp

pp

)r

Aexp(

r

AB

ρ

Δp

4

rf

2

rfU

BBa

22

94

1008 hPa, 1007 hPa and 1007 hPa. An average value of 1007 hPa was used for pn.

The radius of maximum winds, rmax. This is somewhat greater than half the diameter of the eye as visible from radar (estimated as 30 nautical miles, or 50 kilometres) and a value of 32.5 kilometres was estimated for rmax.

Holland ‘B’ parameter. A value of B of 1.7 was used in the original modelling of Cyclone ‘Yasi’ winds (Boughton et al., 2011) – however in the revised model a reduced value of 1.2 was found to give the best fit to the recorded maximum gusts, as discussed in the following section.

Table C1 gives the lowest recorded values of barometric pressure at eight locations

during the event. Figure C1 compares these values with Equation (C2) with values of

pc, pn, A and B given above. Good agreement is seen; this together with the

calibration against the recorded maximum wind gusts as described following gives

general validity to the modelling approach adopted.

Table C1. Recorded minimum barometric pressures and approximate distances from the

cyclone centre

Location

Pressure (hPa)

Radial distance, r (km)

Clump Point 930 ~ 0

South Johnstone 955 30

East Innisfail 967 40

Holmes Reef 969 60

Flinders Reef 976 75

Lucinda Point 980 77.5

Cairns 983 100

Townsville 993 170

Equation (C3) is an equation for the gradient wind, and requires factors to convert

this to a 10-minute mean wind speed at 10 metres height, and a gust factor to

convert the latter to a 3-second gust. There is also a change in direction from the

gradient to the surface wind, towards the low-pressure centre of the cyclone.

In the present case, the ratio of 10-minute mean winds to gradient wind was taken

as 0.7, and the gust factor was taken as 1.4 for overland winds and 1.3 for overwater

winds for the 3-second gusts, and 1.65 and 1.55 respectively for the ‘1 square metre’

gusts. The wind direction at the surface was adjusted by an ‘inflow’ angle of 30

degrees clockwise.

Once a cyclone makes landfall, there is an immediate weakening in strength as the

eye collapses. This continues progressively as the storm moves further inland. In the

95

case of the modelling of Cyclone ’Yasi’, the weakening factors in Table C2 were

applied to the wind speeds, as a function of the distance of the centre of the storm

from landfall.

Table C2. Weakening factors after landfall

Distance from landfall (km)

Weakening factor

0 1.0

10 0.90

20 0.875

30 0.85

40 0.83

The factors in Table C2 are based on data from land falling U.S. hurricanes analyzed

by Kaplan and de Maria (1995 and 2001).

Outside the radius of maximum winds, the vortex gradient winds produced by the

Holland model were summed vectorially with a forward motion component, taken as

10 m/s (36 km/h) in a direction 24 degrees south of west, based on observations of

radar and satellite images of ‘Yasi’.

C4. Calibration of the wind field model

Figure C2 shows a cross plot of the peak gusts from the model wind field against the

‘measured’ values, with the latter consisting of a combination of anemometer

readings, and averages of upper and lower limits from non-failed and failed road

signs, or ‘windicators’ (this technique is discussed in detail in Boughton et al., 2011).

All the values used in the calibration are shown in Table C3, including two readings

from reef anemometers close to the approach track. All gusts have been corrected

to 10 metres height in open terrain.

Good correlation is seen in Figure C2, with a correlation coefficient of 0.88, and a

slope very close to 1.0, indicating no systematic bias in the model. This was achieved

by adjustment of the ‘B parameter’ in Equation (3) to obtain the best fit.

The model shows an underestimation of the maximum gusts at Tully, and an

overestimation at Cardwell, compared with the ‘measured’ values derived from the

‘windicators’. Both of these differences may be a result of topographic effects with

channelling between Mount Tyson and Mount Mackay producing an increase at Tully

for north and south winds, and shielding from Hinchinbrook Island reducing gusts

from easterly winds at Cardwell.

96

* average of upper and lower limits from road signs + corrected for terrain, height and topography

C5. Reproduction of wind speed and direction histories

Histories of wind gusts and directions with time, and the position of the cyclone,

were obtained by locating the centre of the cyclone at 10 kilometre intervals along

its track, and evaluating the vectorial sum of the vortex speed and the forward speed

of the storm. Nine positions of the storm were used spanning 40 kilometres before

and after the landfall, and a total time of more than two hours.

Figures C3, C4 and C5 compare the calculated histories with measured values from

four automatic weather stations on land for which recorded values were available at

intervals of 15-30 minutes. For the graphs a coast crossing time of 12.45 am on

February 3rd was assumed, and times for the other positions of the centre of the

cyclone were estimated assuming a forward speed of 36 km/h (10 m/s).

The comparisons of maximum gusts (km/h) in Figure C3 are generally good; this is

perhaps not surprising as the maximum gusts from these stations, together with the

‘windicator’ estimates, were use to optimize the model parameters. The

Table C3

Location Recorded gust (m/s)*

1 sq. m.

Recorded gust (m/s)

3 sec.

Predicted gust (m/s)

1 sq. m.

Predicted gust (m/s)

3 sec.

Mourilyan 47.5 51.6

Tully-Birkala 62.5 51.6

Cardwell 58 65.1

Kurrimine 50.5 56.5

S. Mission Beach 63 60

El Arish 50 51.5

Silkwood 50 51.1

Halifax-Macknade 45.5 50.3

Bingil Bay 55 56.9

Kennedy 59 58.1

Dallachy-Bilyana 50 57.2

South Johnstone 43.5+ 43.8

Cairns 25.8 22.7

Townsville 37.5 30.7

Lucinda Point 45.9+ 46.8

East Innisfail 43.3+ 35.3

Holmes Reef 40.3 33.2

Flinders Reef 43.9 47.9

97

measurements for Townsville indicate greater wind speeds after the eye of the

cyclone made landfall; this feature was not reproduced by the model which

incorporates weakening after landfall.

Figure C4 compares predicted 10-minute mean wind speeds with measured values

from the same four stations. Agreement is quite good for South Johnstone,

Townsville and Cairns, although the increase in wind speeds post landfall at

Townsville is again not reproduced. Agreement is less good for Lucinda Point. In

that case, the anemometer is poorly sited near the top of a 30 metre high shed at

the end of the jetty. No information is available to reliably correct the readings for

the position of the anemometer, although a correction was made for its height

above the sea surface; however, it is likely that the 10-minute mean wind speeds

were more affected by the siting than were the gusts.

Wind directions are compared in Figure C5 (except for Lucinda Point, for which the

direction vane appeared to malfunction). The agreement is good at Townsville.

Although the general trend with time is similar, the predicted values at South

Johnstone and Cairns are lower than the recorded values. The measured directions

at South Johnstone and Cairns appear to indicate significantly greater inflow angles

than the 30 degrees assumed in the model. Local topography may also have

influenced the wind directions at the measurement stations.

C6. Predicted maximum gusts at selected locations

Table C4 shows the predicted maximum gusts, in kilometres per hour, produced by

the revised model, and those shown in TR57 (Boughton et al., 2011). In the latter

case, no distinction was made between the 3-second averaged and ‘1-square metre’

gusts.

If the new 1 square metre gusts are compared with the previous values, the

predicted maxima are largely unchanged at Tully, Mission Beach, Kurrimine and

Bingil Bay. There is a small increase at Cardwell, and larger increases at Lucinda and

Abergowrie. The predicted maximum gust at Innisfail is reduced slightly from the

previous value.

The maximum 1-square metre gusts anywhere in the event at 10-metres height,

without any topographic effects is predicted to be 234 km/h (at Cardwell). However

the maximum 3-second gust is predicted to have been just under 200 km/h (also at

Cardwell).

98

C7. Conclusions

A re-analysis of the overland wind field model for ‘Yasi’ has produced maximum

gusts applicable to AS/NZS1170.2 (i.e. ‘1 square metre’ gusts) that are generally

similar to those given previously in TR57 (Boughton, et al., 2011), although increases

have occurred at locations to the south of the coast crossing point.

The approach used has been validated by comparisons with recorded barometric

pressures, and with time histories of wind speeds and directions at four automatic

weather stations. Comparisons of the predicted variation of wind speeds, both 3-

second gusts and ten-minute means, during the event, with the recorded values

from South Johnstone, Lucinda Point, Townsville and Cairns are favourable. Good

agreement was achieved between modelled and measured wind directions at

Townsville. Less good agreement between modelled and measured wind directions

at South Johnstone and Cairns was obtained, indicating greater inflow angles than

are assumed in the model.

The maximum gust anywhere in the event is now estimated to be 234 km/h, at 10

metres height over flat, open terrain – about 94% of the design value for most

buildings in Region C in AS/NZS1170.2.

However, predicted 3-second moving-average gusts are 18-19% lower than the 1

square metre values, and do not exceed 200 kilometres per hour.

Table C4. Predicted maximum gust at selected centres

Location Predicted gust (km/h) TR57 model

New predicted

gust (km/h) 1 sq. m.

New predicted

gust (km/h) 3 sec.

Abergowrie 161 183 155

Bingil Bay 199 205 172

Cardwell 226 234 197

Dunk Island 204 214 172

El Arish 187 185 157

Ingham 143 172 146

Innisfail 157 150 127

Kurrimine 201 203 171

Lucinda 170 201 168

Mission Beach 204 205 172

Mourilyan 188 186 158

S. Mission Beach 214 216 181

Tully 187 186 158

Tully Heads 218 222 186

99

C8. References A.C.M. Beljaars (1987), The measurement of gustiness at routine wind stations – a review, Instruments and Observing Methods Report, World Meteorological Organization, No.31, W.M.O. Geneva, 1987. G.N. Boughton and nine others (2011), Tropical Cyclone ‘Yasi’ – structural damage to buildings, James Cook University, Cyclone Testing Station, CTS Technical Report 57, April. J.D. Ginger and six others (2011), Extreme windspeed baseline climate investigation report, A report for Department of Climate Change and Energy Efficiency, James Cook University - Cyclone Testing Station, Bureau of Meteorology, Geoscience Australia, JDH Consulting, April. G.J. Holland (1980), A analytic model of the wind and pressure profiles in hurricanes, Monthly Weather Review, Vol.108, pp 1212-1218. J.D. Holmes (1977), Report on structural damage at Burketown after Cyclone ‘Ted’ 19/12/76, James Cook University, Wind Engineering Report, 1/77, April. J. Kaplan and M. de Maria (1995), A simple empirical model for predicting the decay of tropical cyclone winds after landfall, Journal of Applied Meteorology, Vol. 34, pp2499-2512. J. Kaplan and M. de Maria (2001), On the decay of tropical cyclone winds after landfall in the New England area, Journal of Applied Meteorology, Vol. 40, pp 280-286. V.A. Myers (1957), Maximum hurricane winds, Bulletin of the American Meteorological Society, Vol. 38, pp 227-228. Standards Australia, Structural design actions. Part 2: Wind actions, Australian-New Zealand Standard, AS/NZS1170.2:2011.

100

920

940

960

980

1000

1020

0 20 40 60 80 100 120 140 160 180 200

Pre

ssu

re a

t se

a le

ve

l (h

Pa)

Radial distance from centre (km)

Figure C1. Comparison of recorded barometric pressures (Table C1) with Eq.(C2) with A = 65.2 km and B = 1.2

101

y = 1.0015xR² = 0.7693

0

10

20

30

40

50

60

70

0 10 20 30 40 50 60 70

mo

de

l gu

st s

pe

ed

(m

/s)

Figure C2. Cross-plot of measured and predicted maximum gust speeds

102

0

50

100

150

200

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

3-se

c g

ust

win

d s

pe

ed

(k

m/h

)

South Johnstone

0

50

100

150

200

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

3-se

c g

ust

win

d s

pe

ed

(k

m/h

)

Lucinda Point

0

50

100

150

200

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

3-se

c g

ust

win

d s

pe

ed

(k

m/h

)

Townsville

0

50

100

150

200

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

3-se

c g

ust

win

d s

pe

ed

(k

m/h

)

Cairns

Figure C3. Comparison of measured and modelled maximum 3-second (moving average) gusts

103

0

50

100

150

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

10-m

in m

ean

win

d s

pe

ed

(k

m/h

)

South Johnstone

0

50

100

150

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

10-m

in m

ean

win

d s

pe

ed

(k

m/h

)

Lucinda Point

0

50

100

150

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

10-m

in m

ean

win

d s

pe

ed

(k

m/h

)

Townsville

0

50

100

150

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

10-m

in m

ean

win

d s

pe

ed

(k

m/h

)

Cairns

Figure C4. Comparison of measured and modelled 10-minute mean wind speeds

104

0

90

180

270

360

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

win

d d

ire

ctio

n (d

eg)

South Johnstone

0

90

180

270

360

-40 -20 0 20 40

Lucinda Point

Predictions

Distance from landfall (km)

win

d d

ire

ctio

n (d

eg)

0

90

180

270

360

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

win

d d

ire

ctio

n (d

eg)

Townsville

0

90

180

270

360

-40 -20 0 20 40

Measured

Predictions

Distance from landfall (km)

win

d d

ire

ctio

n (d

eg)

Cairns

Figure C5. Comparison of measured and modelled wind directions

105

ACKNOWLEDGEMENTS

The support of Dr. Lam Pham and Mr. Brian Ashe of the Australian Building Codes Board during this project is gratefully acknowledged. The interest and advice from the members of the 2008 Steering Committee are also acknowledged: Michael Syme (CSIRO) Fabio Finocchiaro (Northern Territory Administration) Nabil Yazdani (Western Australia Administration) Lance Glare (Queensland Administration) Cam Leitch (Cyclone Testing Station, James Cook University) Jeff Kepert (Bureau of Meteorology) Bob Cechet (Geoscience Australia) Sergio Detoffi (Standards Australia) The author also gratefully acknowledges the assistance and free sharing of information given by the following persons in face-to-face meetings, telephone calls or e-mails during the course of the project: Debbie Abbs (CSIRO)

Craig Arthur (Geoscience Australia) Bob Cechet (Geoscience Australia)

Jeff Kepert (BoM) Yuri Kuleshov (BoM)

Mark Leplastrier (Insurance Australia Group) Neville Nicholls (Monash University)

Kevin Walsh (University of Melbourne) The author also acknowledges ongoing discussions over several years with Dr. Bruce Harper (GHD) on various aspects of tropical cyclones, including the Dvorak method and the Holland model.