Methodologies for Economic Impact and Adaptation Assessment of Cyclone Damage Risks Due to Climate Change *

121

Methodologies for economic impact and adaptation assessment of cyclone damage risks due to climate change *

MG Stewart † Centre for Infrastructure Performance and Reliability, School of Engineering, The University of Newcastle, NSW, Australia

Y Li Department of Civil and Environmental Engineering, Michigan Technological University, Houghton, Michigan, USA

SUMMARY: Increases in wind damage are expected if the intensity and/or frequency of tropical cyclones increase due to enhanced greenhouse conditions (climate change). The paper proposes a methodology to estimate cyclone damage risks due to enhanced greenhouse conditions using residential construction in the North Queensland cities of Cairns, Townsville and Mackay as a case study, and then assesses the economic viability of several climate adaptation strategies. The analysis includes probabilistic modelling of cyclone intensity and frequency, time-dependent increase in wind speed from enhanced greenhouse conditions (global warming), and vulnerability functions of building damage. Increases in mean annual maximum wind speed from 0% to 25% over 50 years are considered to represent the uncertainty in changing wind hazard patterns as a result of climate change. The effect of regional changes to building inventory over time and space, rate of retrofi tting, cost of retrofi t, reduction in vulnerability, and discount rate will be considered. The risk-cost-benefi t analysis considering temporal changes in wind hazard and building vulnerability can be used to help optimise the timing and extent of climate adaptation strategies.

1 INTRODUCTION scenario is a 5-10% increase in wind speeds by 2050, with an increase of 25% being the worst-case scenario Increases in wind damage are expected if the intensity with a very low probability that this scenario will and/or frequency of tropical cyclones increase due to eventuate. Hence, this paper will consider a range of enhanced greenhouse conditions (climate change). increases in wind speeds from 5% to 25% in North However, the 2007 Intergovernmental Panel on Queensland over the next 50 years, but most results Climate Change (Christensen et al, 2007) provided will be presented for a 10% increase in wind speed as very little guidance as to the expected increase in this appears to be the most likely scenario arising from intensity and/or frequency of tropical cyclones in enhanced greenhouse conditions. Australia and elsewhere. However, Walsh et al (2001) Tropical cyclones and hurricanes can cause signifi cant predicted up to a 25% increase in wind gusts for the sources of economic loss, for example, Cyclone same return period wind speeds for Townsville, Cairns Tracy in 1974 caused over $500 million in damages and other coastal locations in North Queensland by (Holmes, 2001), and Cyclone Larry in 2006 caused 2050. Yet, a more recent report by the same author stated that maximum tropical cyclone winds are over $1 billion in damages. Approximately 60-80% of likely to increase by only 5-10% by some time after damage caused by Cyclone Larry arose from damage 2050 (Walsh et al, 2002). A review of the most current to residential construction in houses built before research (AGO, 2007) suggests that a plausible enhanced building standards were implemented in north Queensland from the early to mid 1980s * Paper S09-023 submitted 29/05/09; accepted for (Ginger et al, 2007). The potential for larger losses publication after review and revision 27/07/09. exists given the increasing development of coastal Published in AJSE Online 2009, pp. 85-100. communities in North Queensland. † Corresponding author Prof Mark Stewart can be Following the devastating damage caused by contacted at [email protected]. Cyclone Althea (Townsville) in 1971 and Cyclone

SS09-02309-023 Stewart.inddStewart.indd 121121 111/01/101/01/10 10:3110:31 AMAM 122 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

Tracy (Darwin) in 1974 changes were made to the vulnerability and discount rate will be considered. Queensland Home Building Code (1981) requiring The case study considers the cyclone damage risks new housing to be strengthened by a set of deemed- for the North Queensland cities of Cairns, Townsville to-comply provisions (eg. Walker, 1980). These and Mackay, where three site exposures are enhanced building standards for houses came into considered: Foreshore, Town and Inland. Results will effect on 1 July 1982, although many new houses be given in terms of annual and cumulative economic built in the years prior to 1982 complied to the risks, and damage loss. Given the uncertainty of the Australian Wind Loading Code AS1170.2 (Standards impacts of global warming, a range of increases in Australia, 2002). This means that houses built wind speed are considered: (i) no climate change since 1980-1985 in North Queensland represent (stationary system) and (ii) mean annual maximum “properly engineered forms of cyclone resistant wind speed increases by 5% to 25% over the next 50 construction” (Reardon & Henderson, 1998) – years (non-stationary system). A particular climate this enhanced type of residential construction is adaptation strategy will be economically viable referred to herein as “post-1980 construction”. Other when the cumulative costs of retrofi t and reduced related standards, such as wind loads for housing damages fall below the cumulative damage costs of AS4055 (Standards Australia, 2006) and residential existing vulnerability (ie. ‘‘do nothing’’ scenario) – in timber framed construction for cyclonic regions other words, the net benefi t of the climate adaptation AS1684.3 (Standards Australia, 2005) are used strategy exceeds zero. The risk-based cost-benefi t in more recent housing design and construction. analysis considering temporal changes in wind Hence, the vulnerability of pre-1980 construction hazard and building vulnerability can be used to is signifi cantly higher than post-1980 construction. help optimise the timing and extent of retrofi tting If damage from cyclones is expected to increase with existing houses to adapt to the potential impact of time due to climate change, then climate adaptation enhanced greenhouse conditions. strategies may be needed. This may be achieved by There is clearly great uncertainty and debate about retrofi tting/strengthening pre-1980 construction to predicted changes in wind hazards due to climate the enhanced post-1980 standard. Another climate change and so it is not the intention of this paper to adaptation strategy may be to further reduce the support any specifi c assumption of climate change. vulnerability of new construction, or to implement Instead, the purpose of this paper is to investigate planning controls to limit development in highly the potential impact of assumed climate change vulnerable coastal locations. Most cyclone (hurricane) scenarios on damage loss estimation and examine risk research has focused on changes to building the cost-effectiveness of various climate adaptation vulnerability and inventory, and its time-dependent strategies. This paper will also provide a tool for risk- effect on damage risk (eg. Harper, 1999; Granger et informed decision making under uncertainty, which al, 2000; Huang et al, 2001; Jain et al, 2005). However, will be of utility to building code and government relatively little attention has been paid to quantifying planning agencies. the costs and benefi ts of climate adaptation strategies (retrofi tting, strengthening), and assessing at what point in time a climate adaptation strategy becomes 2 PROBABILISTIC WIND MODEL economically viable. Cost-benefit analysis for strengthening a residence to withstand cyclones Probabilistic wind field modelling of the North has been used to weigh different retrofi t options on Queensland cities of Cairns, Townsville and Mackay hazard mitigation (Li & Ellingwood, 2009). Stewart et has been conducted by Harper (1999), where the al (2003) and Stewart (2003) developed a cost-benefi t predicted wind speeds compared very well to analysis decision-making framework to assess the measured tropical cyclone data, and that Cairns, economic viability of strengthened construction Townsville and Mackay have similar extreme wind and other damage mitigation strategies for US and climates. Since most site specifi c simulation-based Australian hurricanes and tropical cyclones. In this hazard models are proprietary and not available work, retrofi tting was assumed to occur when cyclone for this study, Stewart (2003) fi tted an EV-Type I damage occurred and so the additional cost of the distribution to the Harper (1999) predictions. The retrofi t was minimised (since the structure had to be EV-Type I (Gumbel) cumulative distribution function repaired anyway), and since damage would occur to for annual maximum gust speeds is thus: the most vulnerable construction then over a long time –α(v – u) period it would be expected that the most vulnerable Fv(V) = exp[–e ] (1) construction would be retrofi tted, thus reducing the where v is the gust wind speed (m/s) for a standard region-wide vulnerability to tropical cyclones. category 2 terrain (AS1170.2, Standards Australia, A cyclone damage risk-cost-benefit analysis is 2002) and a 10 m anemometer height. The parameters developed to assess the economic viability of several u and α are site-specifi c. The statistical parameters climate adaptation (hazard mitigation) strategies. are α = 0.154 and u = 13.60 for North Queensland The effect of regional changes to building inventory, (Stewart, 2003). The parameters correspond to rate of retrofitting, cost of retrofit, reduction in annual mean maximum wind speed of 17.4 m/s and

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 122122 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 123

coeffi cient of variation (CoV) of 0.48. A comparison of Table 1: Terrain and shielding multipliers wind speed predictions based on the Harper (1999) (Harper, 1999). model and the best fi t EV-Type I distribution for North Queensland is shown in fi gure 1. Site Terrain Shielding exposure multiplier, K multiplier, K If we assume a 10% increase in wind speed after 50 t s years, then the annual mean maximum wind speed Foreshore 0.946 1.00 increases to 19.1 m/s. As there is no information Town 0.864 0.85 how the increase will occur, a linear time-dependent increase in mean wind speed is assumed. The COV Inland 0.864 1.00 for wind speed for all years is assumed constant at 0.48. In this case, the Gumbel parameters u(t) and incorporation of the poleward shift of tropical cyclones α (t) are time-dependent (back-calculated from the in the probabilistic framework developed herein is updated mean and COV of wind speed), and so beyond the scope of the present report, but is clearly the time-dependent probability density function of an area in need of further research. the Gumbel distribution for annual maximum gust speeds is: 3 BUILDING VULNERABILITY α –α(t)(v – u(t)) –α(t)(v – u(t)) fv(v, t) = (t)e exp[–e ] (2) FUNCTIONS Terrain and shielding effects are defi ned for the A building vulnerability function relates wind speed following three exposure categories (Stewart, to building damage, which in this paper is expressed 2003): Foreshore (1 km from coast), Town (1-2 km in terms of percentage damage, which can then be from coast), Inland (>2 km from coast). The terrain related to economic loss. Several vulnerability models and shielding multipliers for the three exposure for wind hazard have been developed (Unanwa et categories are listed in table 1 (Harper, 1999). Since al, 2000; Khanduri & Morrow, 2003; Pinelli et al, the risk assessment is to be conducted on a regional 2004; Jain et al, 2005). In Australia, a widely-used scale, local topographic features were not considered. building vulnerability model for North Queensland The probabilistic wind fi eld model described herein is that proposed by Walker (1994), based on insurance is relatively simple, but it will allow the relative industry experience. The vulnerability function for changes in damage risks and losses due to temporal insured damage to residential construction in North changes in wind hazard and building vulnerability Queensland is summarised as (Stewart, 2003): to be estimated. 26 While the present study has focused on a known () 20§·§·vv 1 20 1 FvD ¨¸¨¸ AA cyclonic region subject to assumed increases in wind ©¹©¹ (3) speeds, another consequence of enhanced greenhouse v 1, ( ) 100% !dFvD conditions is the poleward shift of tropical cyclones A (CSIRO, 2007). A southern shift of 2° of latitude is approximately 200 km, so regions historically less where v is the standard gust speed, and A = 30 subject to cyclones (eg. Southeast Queensland) may for pre-1980 construction and 37.5 for post-1980 in the future be more vulnerable to damage. The construction. The model was developed from

70 AS1170.2 (Region C) 60

20 Cairns Airport (Harper 1999) Townsville Airport (Harper 1999)

Gust Wind Speed (m/s) Mackay Met Office (Harper 1999) 10 Gumbel Fit: D=0.154, u=13.6 0 1 10 100 1000 10000 Average Recurrence Interval ARI (years)

Figure 1: Comparison of probabilistic wind fi eld models (adapted from Stewart, 2003).

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 123123 111/01/101/01/10 10:3110:31 AMAM 124 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

insurance loss data and expert judgment, and this building vulnerability model is subject to includes building and contents damage. Houses considerable uncertainty. However, it is a very useful built in North Queensland after 1980 represent starting point for quantifying the effectiveness of enhanced wind resistant standards as a result of the strengthened building standards (or enforcement). devastating damage caused by Cyclones Althea and The general belief, from experimental testing, Tracy in 1971 and 1974, respectively. The damage is damage surveys and anecdotal evidence, is that expressed as a percentage of insured value. Note many strengthening procedures, if properly designed that the estimation is expressed as expected (mean) and installed, will signifi cantly reduce vulnerability. damage, as the variability of damage is not included. The building vulnerability model shown in fi gure 2 Figure 2 shows the vulnerability model for residential clearly supports this belief. construction. See Stewart (2003) for a full description of the building vulnerability model. 4 ANNUAL AND CUMULATIVE DAMAGES Henderson & Ginger (2007) developed a probabilistic model of component and connection strengths for The annual insured damage risk in terms of high-set houses typically built in the 1960s and 1970s percentage damage D(t) in year t caused by a wind in Townsville, Darwin and other locations in northern hazard can be calculated by: Australia. Their building vulnerability model for Dt() F ( v ) f ( vtdv ,) (4) this type of pre-1980 construction is also shown in ³ Dv fi gure 2, where it is seen to be in good agreement

with equation (3) for pre-1980 construction. The where FD(v) is the vulnerability function defi ned

Henderson & Ginger (2007) building vulnerability in equation (3) and fv(v, t) is the time-dependent model also compared very well with damage data probability density function for cyclone wind speed from Cyclones Althea and Tracy. Figure 2 also shows given by equation (2). Equation (4) assumes that that hurricane damage in the US from Hurricanes damage is caused by the largest wind event in any Andrew and Hugo in 1989 and 1992, respectively, is calendar year, which will slightly underestimate bounded by the vulnerability of pre-1980 and post- damage risks in the event of an (very unlikely) lesser 1980 constructions. These comparisons provide some damaging event occurring in the same season. Of evidence that the vulnerability models proposed more interest to decision-makers may be annual by Walker (1994) are in the “right ballpark”. It also or cumulative monetary damages or losses. The suggests that housing that existed in the southeastern expected annual damage loss expressed in dollars is: US during the period 1989-1992, particularly N () () () () its vulnerability to minor damage, is generally DEjªºDtNtD pre pre post tNtC post I Lt() ¬¼ representative of Australian pre-1980 construction c ¦ t (5) j 1 1 r quality. This is consistent with general observations made by Reardon & Meecham (1993). So the expected cumulative damage costs starting at

It is clearly acknowledged by Walker (1994) that time to and extending over a time period T is:

U.S. Damages (Huang et al. 2001) Henderson and Ginger (2007) 100 Pre-1980 Construction Post 1980 Construction 90 80 70

(%) 60 D 50 40 30 Damage F 20 10 0 20 30 40 50 60 70 80 Peak Gust Wind Speed (m/s)

Figure 2: Damage vulnerability functions for residential construction.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 124124 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 125

T are pre-1980 houses and the rest are assumed to (,) () (6) LtTco¦ Lt c be built according to post-1980 building standards tt o (Stewart, 2003). In the period 2008 to 2025, it is estimated that 50,000 new dwellings will be required where Dpre(t) and Dpost(t) are the damage risk associated with pre-1980 and post-1980 construction; N (t) and to accommodate projected population growth (QG, pre 2008). Over the 50-year period 2001 to 2050, this is Npost(t) are the numbers of houses constructed to pre- 1980 and post-1980 standards in a region in year t; equivalent to a growth of roughly 100%. By 2050, the N is the number of exposure locations; j indicates total house numbers will thus increase by 100% to α 250,000, ie. 2500 new dwellings built each year for exposure; Ej represents the proportion of houses in each exposure site, assumed as 0.2, 0.6 and 0.2 for the next 50 years. For regional damage estimates, the Foreshore, Town and Inland, respectively; r is the proportion of Foreshore-Town-Inland construction is assumed constant at 20-60-20% over the 50-year time discount rate; and CI is the insured value of a house. For all scenarios considered herein it is assumed frame. This scenario is based on several assumptions, that the wind fi eld characteristics are constant across but more accurate demographic and housing studies a region and so the damage risks expected in one can be used to refi ne the scenario assumed herein. location (say Cairns) will be identical to another The median replacement value of a house in North location (say Townsville) in the region. Queensland is approximately $215,000 (Li & Stewart, Note also that the damage risks and losses calculated 2008). The insured value of the house is higher than herein are based on a region-wide analysis of the replacement value due to many homeowners also wind speeds and housing demographics. A more holding contents insurance, which led Huang et al detailed (GIS-based) probabilistic wind fi eld model (2001) to assume that the insured value of a house that considered local topographical factors would is 150% of the (replacement) value of the structure. produce a wider range of damage risks; namely, some It follows that the median insured value of a house

localities within a region would have higher damage in North Queensland is approximately CI = $320,000 risks and others lower, even though they may both in 2008 terms. be located in the same broad exposure category Figure 3 shows the expected annual damage used herein. Hence, although the economic risks to (percentage of insured value) risks D(t) obtained be calculated herein will be subject to considerable from equation (4) for houses built with pre-1980 uncertainty, they are well suited for comparative and post-1980 standards in a Foreshore location, analyses such as that conducted herein. assuming 0%, 10% or 25% increase in wind speeds. According to Australian 2001 census data there are As expected, the annual damage risks for pre-1980 approximately 125,000 dwellings (houses, units, and post-1980 construction increase with time if wind apartments) in coastal regions of North Queensland, speed increases with time. The annual damage risks most of these are located in the large coastal cities for pre-1980 construction are approximately 4-7 times of Cairns and Townsville (ABS, 2001), 50% of which higher than post-1980 construction risks.

pre-1980 (0% increase in wind speed) pre-1980 (10% increase in wind speed) pre-1980 (25% increase in wind speed) 0.9 post-1980 (0% increase in wind speed) 0.8 post-1980 (10% increase in wind speed) post-1980 (25% increase in wind speed) 0.7

0.6

0.5

0.4

0.3

0.2

Annual Damage D(t) (%) 0.1

0 0 1020304050 Time t (years)

Figure 3: Annual damage risk D(t) for pre-1980 and post-1980 construction in Foreshore exposure, with 0%, 10% and 25% increase in wind speed.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 125125 111/01/101/01/10 10:3110:31 AMAM 126 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

If it is assumed that in 50 years time the number of period. A 25% increase in wind speed will increase pre-1980 houses will gradually decrease by 50% to regional damage considerably with annual damage 31,250 houses, either due to the replacement of old costs of up to $41 million, compared to $7 million for houses, or renovations or retrofi t of pre-1980 houses a 0% increase in wind speed. to meet enhanced building standards. Hence, in 50 years time, N (50) = 31,250 houses and N (50) Figure 5 shows the cumulative regional damage costs pre post L (1, 50) for pre-1980 and post-1980 construction in = 218,750 houses. The annual damage costs for c the Foreshore region are calculated from equation the three exposure categories, with the scenarios α of increases in wind speed from 0-25% and no (5) for N = 1 and Ej = 0.2, and are presented in fi gure 4, assuming no discounting (r = 0%). The discounting is assumed. It can be seen that the annual damage costs with no change in wind speed most severe losses in 50 years will occur to pre-1980 decreases with time because of the growth in new construction in Foreshore locations. If a 25% wind housing numbers (with reduced vulnerability). speed increase in expected, then damage to pre-1980 However, for a 10% increase in wind speed the construction in Foreshore locations reaches $600 annual regional damage risks double over the 50-year million over 50 years. Clearly, the majority of wind

50 45 0% increase in wind speed 10% increase in wind speed 40 25% increase in wind speed

(t) ($ million) 35 c 30 25 20 15 10 5 0 0 1020304050 Annual Damage Costs L Time t (years)

Figure 4: Annual damage costs for Foreshore region.

Figure 5: Cumulative regional damage costs in 50 years for pre-1980 and post-1980 construction in Foreshore, Inland and Town exposures.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 126126 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 127

damage occurs to the pre-1980 construction for all only 9% and 25% (for 10% and 25% increases in wind exposure locations in North Queensland. speed) if a more immediate time period is considered (next 10 years). This demonstrates that damage Figure 6 shows the regional cumulative damage costs will accelerate over time, and the longer the costs L (1, T) for North Queensland, over intervals c time period considered, the greater the proportional of T = 10, 25 and 50 years, N = 3 and assuming increase in damage costs when compared to the no no discounting (r = 0%). It is observed that the change in wind speed scenario. cumulative damage costs in 50 years is $690 million if there is no change in wind speed, and the losses As there is signifi cant uncertainty about climate increase to $1.073 billion and $2.017 billion when the change scenarios and their timeframe, a “fragility” cyclone intensity is assumed to increase by 10% and type curve may be useful that shows the increase in 25% in 50 years time, respectively. These are a 56% regional damages for various wind speed increases and 192% increase in total losses for the region over taken over several time periods, in this case 2030, 50 years assuming 10% and 25% wind speed increase 2050, 2070 or 2100, where time is measured from when compared to the no change in wind speed 2010 – see fi gure 7. For example, the increase in scenario. However, increases in damage costs are cumulative damages up to 2100 is based on a wind

Figure 6: Regional cumulative damage costs in North Queensland.

3500

3000

2500

2000 2100 1500 2070 2050 1000 2030 500

Increase in Damages ($ million) 0 0 5 10 15 20 25 30 % Increase in Wind Speed Figure 7: Regional increases in damage costs for various climate change scenarios and time periods.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 127127 111/01/101/01/10 10:3110:31 AMAM 128 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

speed increase occurring over T = 90 years. Figure replacement value and n is the percentage rate of 7 shows that the increase in regional damage costs retrofi tting. Note that if n = 10%, then all pre-1980 may exceed several billion dollars for some extreme construction will be retrofi tted in 10 years, but if n = (worst case) climate change scenarios, but may be as 1%, then only 50% of pre-1980 construction will be low as a few hundred million dollars if the climate retrofi tted in 50 years. change scenario predicts a small increase in wind The regional loss for climate adaptation strategy speed. For more details about cyclone damage risks, 3, where each year all of the 2500 new houses are see Li & Stewart (2008). strengthened to reduce vulnerability by R%, is equation (9), below. 5 COST-EFFECTIVENESS OF CLIMATE The cost of retrofi tting (C ) is very much dependent ADAPTATION STRATEGIES st on the required reduction in vulnerability, structural confi guration, and current design and construction The situation assuming no climate adaptation practices. So it is diffi cult to estimate costs accurately strategies is referred to herein as the “do nothing” (Li & Ellingwood, 2009). Nonetheless, AGO (2007) scenario. While recognising that future changes to estimated that the increase in construction cost of housing demographics is imprecise, a reasonable assumption may be that over the next 50 years there new houses due to an increase in design wind class will be no retrofit to pre-1980 construction, that (say C1 to C2) is approximately $2000 to $6000 per the housing mix is 50-50 (pre-1980 to post-1980) at house. If the median replacement value of a house in year 1 and the rate of new (post-1980) construction North Queensland is $215,000, then these increases in C are 1-3% of the value of the house. A number is 2500 houses per year. Thus, the “do nothing: st regional loss estimation is equation (7), below, where of other studies have found that the additional cost N (t) = N (t) = 62,500 houses. to new housing for increased cyclone resistance is pre post in the range of 1-10% (eg. Stewart et al, 2003) and The cost-effectiveness of various retrofi t strategies to approximately 5% for Australian cyclone-resistant adapt to climate change is investigated by comparing systems (Reardon & Oliver, 1983). There is very little regional loss for the following climate adaptation data on the costs of retrofi tting an existing house for strategies: increased cyclone resistance. However, Leicester 1. retrofit/strengthen pre-1980 construction at (1981) observed that “estimated” additional costs selected high wind exposure sites (Foreshore for houses in Australia range from 15% to 50% for exposure) to enhanced (post-1980) standards retrofi t of existing houses. There can be expected to be 2. retrofi t/strengthen pre-1980 construction in the a relatively wide range of retrofi t costs (Cst) due to the whole region to enhanced (post-1980) standards large choice of strengthening procedures available 3. reduce vulnerability of new construction at selected for housing construction. high wind exposure sites (Foreshore exposure). The analysis assumes that the cost of retrofi t will be The regional loss for climate adaptation strategies 1 an additional cost, borne by the residential home

and 2 are equations (8a) and (8b), below, where Cst is owner, government or other agency. For example, the cost of retrofi t expressed as percentage of house if a climate adaptation measure is likely to reduce

TND 62, 500Dt ( ) 62, 500 D ( t ) 2500 tDtC ( ) (,) Ej^` pre post post I LtTco ¦¦ t (7) tt j 1 1 o r

ªº§·100 nt nt n §·C ½ D ®62, 500Dt ( ) D ( t )st 62, 500 D ( t ) 2500 tDtC ( )¾ TN Ej«»¨¸100 pre 100 post 100¨¸ 1.5 post post I (,) ¯ ¬¼©¹ ©¹ ¿ LtTc adapt o ¦¦ t 1 1 tt o j r nt d 100 (8a)

TND 125,000Dt ( ) 2500 tDtC ( ) (,)Ej^` post post I LtTc adapt o ¦¦ t 1 1 tt o j r nt ! 100 (8b)

½§·100 R §·C D 62, 500Dt ( ) 62, 500 D ( t ) 2500 t D ( t ) 2500 st C TN Ej®¾ pre post¨¸100 post¨¸ 1.5 I (,) ¯¿©¹ ©¹ LtTcadapto ¦¦ t (9) tt j 1 1 o r

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 128128 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 129

damages in a cost-effective manner, then government For example, if the retrofi t of all pre-1980 construction may invest resources into funding the costs of a climate is completed in 10 years (n = 10%), the net benefi t

change adaption program. Alternatively, insurers may is $86.5 million if the retrofi t cost is Cst = 2.5%. The

provide a reduction in premiums for homeowners that adaptation strategy is also cost-effective if Cst = 5%, retrofi t their houses. Either way, these are pro-active but the net benefi t reduces to as little as $4.7 million.

measures that, for appropriate climate adaptation If Cst is 10% or higher, then adaptation strategy programs, will benefi t home owners, insurers, society 1 is not cost-effective. As the retrofi tting process (less social disruption) and government. accelerates (n increases), the cost-effectiveness is more prominent. Figures 9 and 10 show when an The effects of three climate adaptation strategies are adaptation strategy becomes economical viable now discussed, assuming a 10% increase in wind speeds (ie. the net benefi t is positive). When retrofi t cost over the next 50 years and a discount rate of r = 4%. Note is C =1%, it takes only 8 years for the adaptation that the net benefi t of an adaptation strategy is L (t , T) st c o strategy to be cost-effective, regardless of the annual – L (t , T) and the percentage change in net benefi t c-adapt o upgrading rate (n). However, as C increases, it takes is 100(L (t , T) – L (t , T))/L (t , T). The percentage st c o c-adapt o c o a longer time for the adaptation strategy to become change in net benefi t is not affected by the number of economically viable. houses in the region as this will infl uence Lc(to, T) and Table 2 shows that adaptation strategy 1 is cost- Lc-adapt(to, T) equally. For example, if the number of houses effective as long as C is less than 5.81-6.49% of in the region is reduced by 50% then Npre(1) = Npost(1) = st 31,250 houses and new houses increase by only 1250 house value (approximately $12,000 to $14,000). This appears to be a relatively small cost, however, as houses/year, then Lc(to, T) and Lc-adapt(to, T) reduce by 50%, but percentage change in net benefi t is unchanged. discussed above, the cost of retrofi tting an existing

house is likely to be much higher (say Cst = 15-50%), 5.1 Adaptation strategy 1: Effect of which would suggest that ensuring that adaptation retrofi tting Foreshore construction strategy 1 is cost-effective may be diffi cult to achieve in practice. However, retrofit costs are highly Figure 5 shows that the annual damage for Foreshore variable, and so it is not possible for the present paper construction is many times the damage for Inland to assess if retrofi tting suffi cient to enhance a pre-1980 and Town exposures. Thus, an effective adaptation house to post-1980 standards can be undertaken for C less than 5-6%. strategy is likely to be one that focuses on reducing st vulnerability in Foreshore locations, rather than all The “do nothing” scenario assumed herein is believed houses in the North Queensland coastal region. The the most realistic, but there are other possibilities for regional loss for retrofi tting pre-1980 construction time-dependent changes in construction over the in Foreshore locations is estimated from equation next 50 years. One scenario might be that some pre- α (8) where N = 1 and Ej = 0.2. The net benefi t over 50 1980 construction will be retrofi tted over the next years for a wind speed increase by 10% by year 50 50 years due to home renovations, demolition and

with n = 1-10% and Cst = 2.5-25% is shown in fi gure 8. other owner-initiated improvements. In this case, it

Figure 8: Net benefi t for adaptation strategy 1 (retrofi t Foreshore construction).

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 129129 111/01/101/01/10 10:3110:31 AMAM 130 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

n=1% C =2.5% st 100 n=1% C =5% st 80 n=1% C =10% st 60 n=4% C =2.5% st 40 n=4% C =5% st n=4% C =10% 20 st 0 -20 -40 -60 -80 -100 Percentage Change in Net Benefit 0 1020304050 Time t (years)

Figure 9: Percentage increase in net benefi t for adaptation strategy 1.

Figure 10: Time needed for adaptation strategy 1 to be cost-effective.

Table 2: Maximum retrofi t cost Cst for adaptation strategy 1 to be cost-effective.

n (%) 1 2 3 4 5 6.67 10

Maximum retrofi t cost Cst (%) 5.81 5.83 6.31 6.41 6.45 6.47 6.49

may be that (say) 50% of pre-1980 construction will 5.2 Adaptation strategy 2: Effect of retrofi tting all be upgraded over 50 years (ie. n = 1%) at no cost to pre-1980 construction to post-1980 standards government. The net benefi t for this case is reduced from that discussed above, so that retrofi t costs of The regional loss for adaptation strategy 2 are

Cst = 5% are no longer cost-effective. A retrofi t cost of calculated from equation (8), where N = 3 using

only Cst = 2.5% is cost-effective for this alternate “do data from table 1. It can be observed from fi gure 11 nothing” scenario. Finally, if the discount rate is taken that this retrofi t strategy is marginally cost-effective

as less than 4%, then net benefi t will increase as this only if Cst is 2.5% or less and the annual upgrading will increase the present value of future losses, which rate (n) is 4% or higher. Clearly, when compared will make adaptation strategies more cost-effective. to adaptation strategy 1 (see fi gure 8), retrofi tting

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 130130 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 131

all pre-1980 construction is signifi cantly less cost- less than 0.55% (approximately $1200). Given that the effective than retrofi tting houses only in vulnerable additional cost to new housing for increased cyclone exposures, such as Foreshore locations. resistance is in the range of 1-10%, it may be diffi cult to achieve a 50% reduction in existing vulnerabilities 5.3 Adaptation strategy 3: Effect of improving for an additional cost of no more than 0.55% of the new Foreshore construction value of the house.

Figure 12 shows the percentage change in net benefi t 5.4 Other climate change scenarios calculated from equation (9) when vulnerability of new construction is reduced by R = 50%. Note There is significant uncertainty about future that the reduction in vulnerability applies only to predictions in wind speeds due to enhanced new construction in Foreshore locations (N = 1). greenhouse conditions. Hence, percentage increases A reduction in vulnerability of 50% is signifi cant, in net benefi t are calculated for 50-year wind speed but fi gure 12 shows that this adaptation strategy increases of 5% and 25%, which are shown in fi gures

is not cost-effectiveness, even if Cst is as low as 1%. 13 and 14 for adaptation strategy 1. A retrofi t cost

Adaptation strategy 3 is only cost-effective if Cst is of Cst = 10% now becomes cost-effective for a 25%

Figure 11: Cost-effectiveness of adaptation strategy 2.

-25

-50

-75

C =1% -100 st C =2.5% st C =5% -125 st C =10% st -150 Percentage Change in Net Benefit 0 1020304050 Time t (years)

Figure 12: Percentage increase in net benefi t for adaptation strategy 3 (new construction).

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 131131 111/01/101/01/10 10:3110:31 AMAM 132 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

Figure 13: Cost-effectiveness of adaptation strategy 1 for 5% increase in wind speed.

Figure 14: Cost-effectiveness of adaptation strategy 1 for 25% increase in wind speed.

increase in wind speed, but is not cost-effective for there is only a 5% increase in wind speed over the wind speed increases of 5% or 10%. As expected, next 50 years. Adaptation strategy 3 (ie. strengthen the benefi t of adaptation strategies increases as the new construction) will only be cost-effective when

cyclone intensity increases with time. Cst = 1.75% and if the wind speed is assumed to increase by 25% over 50 years. If the wind speed increase is 25%, then the percentage increase in net benefit for adaptation strategy 2 To be sure, the results presented herein are sensitive (retrofi t all pre-1980 construction) varies from 6.2% to the selected or assumed parameter values.

to 27% for Cst = 2.5%, depending on the annual Nevertheless, the results provide a reasonable upgrading rate (n = 1-10%). On the other hand, it is indication of the relative measures of cost- not cost-effective to adopt adaptation strategy 2 if effectiveness for some typical climate adaptation

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 132132 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 133

strategies. If more detailed information becomes by 24% or 56%, respectively, when compared to available, then the risk-based decision support damage costs assuming no change in wind speed. It framework developed herein can be applied to such was found that it is cost-effective for older residential cases to provide improved decision support. construction in Foreshore (high vulnerability) locations in North Queensland to be retrofi tted to higher wind resistant standards if such retrofi tting 6 FURTHER WORK costs less than approximately 6% of the house replacement value, when wind speed is expected to There is clearly much scope for further work. This may increase by 10% in 50 years. If wind speed is expected include developing building vulnerability models for to increase by 25%, it is cost-effective to (i) retrofi t all different housing types or construction techniques pre-1980 construction at all sites with retrofi t cost less (and materials), age profi les, code specifi cations, than 2.5% of house value or to (ii) retrofi t houses in compliance and enforcement, changes in exposure Foreshore locations if such retrofi tting costs less than categories (eg. effect of increased urbanisation), and approximately 10% of house value. so on. The development of such models will require a substantial research effort that may include: fi eld or test data of building performance; component ACKNOWLEDGEMENTS and structural system strength prediction modelling; assessing the effect of component and structural The authors thank Yuejun Yin, doctoral candidate system strength on the integrity of the building at Michigan Technological University, USA, for envelope; and probabilistic structural response his assistance in performing some of the analyses. modelling to develop vulnerability (fragility) curves. Some of this work was undertaken while the The work by Henderson & Ginger (2007) provided a second author was supported by a Centre for framework for such modelling. There is also a need Infrastructure Performance and Reliability Visiting to relate failure of a component, structural system Fellowship. The second author appreciates the or building envelope to economic losses needed financial support provided by the Centre for for vulnerability models. Such economic data may Infrastructure Performance and Reliability at The be obtained from the collection and analysis of University of Newcastle, Australia. insurance loss data or from expert judgements. A risk analysis for a specifi c region will require REFERENCES an accurate and detailed probabilistic wind fi eld model capable of considering topographic, terrain Australian Bureau of Statistics (ABS), 2001, Census roughness and shielding effects. The demographics Data, Canberra, ACT. of housing into age, style, etc., will also be required, and will be influenced by the resolution of the Australian Greenhouse Office (AGO), 2007, An probabilistic wind fi eld model. Finally, more rigorous Assessment of the Need to Adapt Buildings for the economic decision analyses may be developed that Unavoidable Consequences of Climate Change, Final consider the effect of insurance premiums, excess, Report, Commonwealth of Australia, August. insurer incentives, discount rates, exposure periods, life safety, cyclone mitigation and response costs, Christensen, J. H., Hewitson, B., Busuioc, A., Chen, and other costs and benefi ts of cyclone adaptation A., Gao, X., Held, I., Jones, R., Kolli, R. K., Kwon, strategies related to the building owner, insurer, W.-T., Laprise, R., Magaña Rueda, V., Mearns, L., reinsurance company, government agency or society Menéndez, C. G., Räisänen, J., Rinke, A., Sarr, A. & in general. This will require a more detailed multi- Whetton, P. 2007, “Regional Climate Projections”, attribute decision support analysis. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, 7 CONCLUSIONS Chapter 11, Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M. & Miller, H. Cyclone damage risks to residential construction as L. (editors), Cambridge University Press, Cambridge, a result of potential climate change and examining United Kingdom, and New York, NY, USA. the cost-effectiveness of different adaptation strategies are estimated from a risk-cost-benefit CSIRO, 2007, Climate Change in Australia: Technical analysis. Adaptation strategies considered include: Report 2007, Marine and Atmospheric Research (i) retrofi tting older (pre-1980) construction in the Division. North Queensland region to enhanced standards; (ii) retrofi tting only the older construction at selected Ginger, J. D., Henderson D. J., Leitch, C. J. & high wind exposure sites; or (iii) reducing the Boughton, G. N. 2007, “Tropical Cyclone Larry: vulnerability of new construction. An increase in Estimation of wind fi eld and assessment of building wind speed of 5% or 10% over 50 years will increase damage”, Australian Journal of Structural Engineering, expected total insured damage for North Queensland Vol. 7, No. 3, pp. 209-224.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 133133 111/01/101/01/10 10:3110:31 AMAM 134 “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li

Granger, K., Harper, B., Jones, T., Stehle, J., Hayne, Australasian Structural Engineering Conference, M. & Callaghan, J. 2000, “Cyclone Risks”, Community Butterworth, J. (editor), Vol. 2, Structural Engineering Risk in Mackay: A Multi-Hazard Risk Assessment, Society of New Zealand, Auckland, pp. 1007-1014. AGSO, Canberra, pp. 88-130. Reardon, G. F. & Meecham, D. 1993, “US Hurricanes Harper, B. A. 1999, “Numerical modeling of extreme of 1992 – An Australian Perspective”, Proceedings of tropical cyclone winds”, Journal of Wind Engineering ASCE Conference on Hurricanes of 1992: Lessons Learned and Industrial Aerodynamics, Vol. 83, pp. 35-47. and Implications for the Future, New York, pp. 642-651.

Henderson, D. J. & Ginger J. D. 2007, “Vulnerability Reardon, G. F. & Oliver, J. 1983, “The impact of cyclone Model of an Australian High-Set House Subjected to Isaac on buildings of Tonga”, Journal of Wind Engineering Cyclonic Wind Loading”, Wind and Structures, Vol. and Industrial Aerodynamics, Vol. 14, pp. 67-78. 10, No. 3, pp. 269-285. Standards Australia, 2002, AS1170.2 SAA Loading Holmes, J. D. 2001, Wind Loading of Structures, Spon Code—Part 2: Wind Loads, Sydney. Press, London. Standards Australia, 2005, AS1684.3 Residential Huang, Z., Rosowsky, D. V. & Sparks, P. R. 2001, Timber Framed Construction – Part 3 Cyclonic Areas, “Long-term hurricane risk assessment and expected Sydney. damage to residential structures”, Reliability Engineering and System Safety, Vol. 74, pp. 239-249. Standards Australia, 2006, AS4055 Wind Loads for Housing, Sydney. Jain, V. K., Davidson, R. & Rosowsky, D. 2005, “Modeling changes in hurricane risk over time”, Stewart, M. G. 2003, “Cyclone Damage and Temporal Natural Hazards Review, Vol. 6, No. 2, pp. 88-96. Changes to Building Vulnerability and Economic Risks for Residential Construction”, Journal of Wind Khanduri, A. C. & Morrow, G. C. 2003, “Vulnerability Engineering and Industrial Aerodynamics, Vol. 91, pp. of buildings to windstorms and insurance loss 671-691. estimation”, Journal of Wind Engineering and Industrial Aerodynamics, Vol. 91, No. 4, pp. 455-467. Stewart, M. G., Rosowsky, D. V. & Huang, Z. 2003, “Hurricane Risks and Economic Viability of Li, Y. & Ellingwood, B. R. 2009, “Risk-based decision Strengthened Construction”, Natural Hazards Review, making for multi-hazard mitigation for wood- ASCE, Vol. 4, No. 1, pp. 12-19. frame residential construction”, Australian Journal of Structural Engineering, Vol. 9, No. 1, pp. 17-26. Unanwa, C. O., Mcdonald, J. R., Mehta, K. C. & Smith, D. A. 2000, “The development of wind damage bands Li, Y. & Stewart, M. G. 2008, Damage risk assessment of for buildings”, Journal Wind Engineering and Industrial buildings to tropical cyclones under enhanced greenhouse Aerodynamics, Vol. 84, pp. 119-149. conditions, Research Report No. 271.11.2008, University of Newcastle, Newcastle, NSW, Australia. Walker, G. R. 1980, “A review of the impact of Leicester, R. H. 1981, “A risk model for cyclone cyclone Tracy on Australian building regulations and damage to dwellings”, 3rd International Conference on practice”, Civil Engineering Transactions, Institution Structural Safety and Reliability, Trondheim, Norway, of Engineers Australia, Vol. 22, No. 2, pp. 100-107. pp. 761-771. Walker, G. R. 1994, CSIRO Division of Building, Pinelli, J. P., Simiu, E., Gurley, K., Subramanian, Construction and Engineering, Dr. B. A. Harper, 1994, C., Zhang, J., Cope A., Fillibe, J. & Hamid, S. 2004, personal communication. “Hurricane damage prediction model for residential structures”, Journal of Structural Engineering, ASCE, Walsh, K., Hennessy, K., Jones, R., McInnes, K., Vol. 130, No. 11, pp. 1685-1691. Page, C., Pittock, A. B., Suppiah, R. & Whetton, 2001, Climate Change in Queensland under Enhanced QG, 2008, Far North Queensland Draft Regional Plan Greenhouse Conditions, Australian Commonwealth 2025, Department of Infrastructure and Planning, Scientific and Research Organisation (CSIRO) Queensland Government, Brisbane, Queensland, Atmospheric Research, Australia. Australia, April. Walsh, K., Cai, W. J., Hennessy, K., Jones, R., McInnes, Queensland Home Building Code, 1981, Queensland K., Nguyen, K., Page, C. & Whetton, P. 2002, Climate Home Building Code, Appendix 4. change in Queensland under enhanced greenhouse conditions: Final Report, 1997-2002, Australian Reardon, G. F. & Henderson, D. 1988, “Cyclone Commonwealth Scientifi c and Research Organisation risk assessment of houses in North Queensland”, (CSIRO), Atmospheric Research, Australia.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 134134 111/01/101/01/10 10:3110:31 AMAM “Methodologies for economic impact and adaptation assessment ...” – Stewart & Li 135

MARK STEWART

Mark G Stewart is a Professor of Civil Engineering and Director of the Centre for Infrastructure Performance and Reliability, School of Engineering, the University of Newcastle, Australia. He received his PhD in 1988 from the University of Newcastle. His research interests include stochastic deterioration modelling, spatial time-dependent structural and serviceability reliability, probabilistic risk assessment, security risk assessment, and life-cycle cost and decision analysis. Recent work is focusing on the effects of climate change on built infrastructure, particularly the potential for increased corrosion risks and wind damage.

YUE LI

Dr Yue Li joined the Department of Civil and Environmental Engineering at Michigan Technological University as Donald and Rose Ann Tomasini Assistant Professor of Structural Engineering in August 2005. He earned his PhD degree in Civil Engineering, with an emphasis in Structural Engineering, from Georgia Institute of Technology in Atlanta, Georgia, in August 2005. Yue’s research interests include bridge engineering, structural reliability analysis, probabilistic design, natural and manmade hazard mitigation, structural load modelling and combinations of loads, structural health monitoring and condition assessment, performance-based engineering, earthquake engineering, wind engineering, and wood engineering. He received the Michigan Tech Research Excellent Fund Award in 2008. He has published in such journals as Journal of Structural Engineering, Australian Journal of Structural Engineering, Engineering Structures, Structural Safety, Earthquake Engineering & Structural Dynamics, and IEEE Sensors Journal: Special Issue on Structural Health Monitoring. His teaching interests include basic structural engineering, probability, statistical and engineering decision analysis, structural reliability, and performance-based structural design. He has worked as a structural engineer for fi ve years, and was involved in the design of new international terminal at Hartsfi eld-Jackson Atlanta International Airport.

Australian Journal of Structural Engineering Vol 10 No 2

SS09-02309-023 Stewart.inddStewart.indd 135135 111/01/101/01/10 10:3110:31 AMAM