Urban Growth and Long-Term Changes in Natural Hazard Risk: Implications for Insurance

Stephanie E. Chang

University of British Columbia

Question

• How is urban seismic risk changing? ...and why? – total risk – distribution of risk – rate of change

Population growth More high-vulnerability populations More structures at risk Greater interdependency, . . .

Engineering advances Better codes, construction practice Greater awareness Higher incomes, insurance, . . .

Risk = Hazard ⋅Vulnerability

Resilience

probability consequences

Outline

• What evidence do we have on how risk is changing? – Focus on earthquake risk and casualties

• What can models tell us? – Case study of Vancouver, Canada

• How might findings be applicable in Asia/Pacific?

• What about economic disruption risk?

• What are implications for the insurance sector?

0100000200000300000400000500000600000

Empirical Evidence

“As opposed to widely publicised claims of rapidly increasing loss trends, we find decreasing trends for both casualties and [economic] losses, when population growth and urbanisation are accounted for.” (Scawthorn, 2011)

Decadal earthquake fatalities as % of global population, 1950s~2000s (Bilham 2009)

Losses from great natural catastrophes, 1950-2008 (2008 values) (Munich Re 2008)

Overall

Insured

Total earthquake deaths by decade, 1950-2009 (after Spence et al. 2011)

Risk Factor Trends

• Increasing Risk – Suburban sprawl encroaching on hazard-prone areas (NRC

2006) – Federal policies encouraging risk reduction and sharing rather

than risk avoidance (Burby et al. 1999) • Development encouraged by false sense of security

– Planned land use – Los Angeles (Olshanky and Wu 2002) – Population change - coastal migration, aging, race/ethnic

composition, income & housing profiles (Cutter et al. 2007)

• Decreasing Risk – Improved building codes – ?

0200400600800

10001200

Year 0 Year 20

Global Loss Trend

Global v. Local Trends

Global trends do not necessarily translate to local trends

20k

100k

Year 20

20k

100k

Year 0

PrA(EQ)=X City A

City B PrB(EQ)=X

1% loss

1% loss

“Repeat” Events

(earthquake.usgs.gov)

1971 San Fernando

1994 Northridge

1971 San Fernando

1994 Northridge

Magnitude, depth

Mw 6.6 8.4 km

Mw 6.7 18.4 km

Population, L.A. County

7.0 million (in 1970)

8.9 million (in 1990)

Casualties 58 deaths, 2000 injuries

57 deaths, 9000+ injuries

Direct losses (1994$)

1.8 billion 24~44 billion

Sources: SCEC; US Census; CA OES; Eguchi et al. 1998

Codes, retrofits, professional awareness since San Fernando did contribute significantly to reducing losses in Northridge (Olshansky 2001)

Los Angeles earthquakes

Modeling Changes in Urban Disaster Risk

Hazard Inventory

Direct Damage

Induced Damage

Direct Loss Indirect Loss

• Higher frequency of extreme rainfall events Hazard

• Improved building codes

Inventory

Direct Damage

• Increased dependence on overseas suppliers

Direct Loss Indirect Loss

• Suburbanization

Hazard Inventory

Greater Vancouver (Canada)

36%

301%

jamestung.blogspot.com

Population Growth 1971~2006

Tourism BC Tom Ryan City of Surrey

Codes

Building Stock Changes

1971 2006 Population (millions) 1.08 2.12 - in masonry buildings 2.8% (31,000) 0.9% (18,900) - in concrete buildings 6.7% 11.4% Dwellings 256,000 803,000 - single-detached houses 43.8% 35.6%

• NBCC adopted in 1973 (seismic provisions by Vancouver in 1965); revisions in 1985, 1999, 2005 (Finn 2004)

• “...most buildings constructed in British Columbia prior to the 1970s have limited resistance to seismic effects.” (Ventura et al. 2005)

• Currently 1/3 of housing units in metro area built before 1971

Construction

(Census; Ventura et al. 2005)

Earthquake Loss Model

Buildings Damage Casualties

• HAZUS-MH • Deaths and

serious injuries

• Ventura et al. (2005)

• BC buildings • local engineers • MMI

• Census (pop., dwellings)

• Ventura et al. (2005) (structural type)

• Census data – pros and cons • Modeling challenges and solutions • Single scenario • Consistent assumptions for 1971 and 2006 models • Uncertainty and errors

Scenario Event

M7.3

Strait of Georgia

Subcrustal earthquake

4am

BC PEP

VIII

VII

Ground Motions

• Similar to 1946 Vancouver Island earthquake • Strong but realistic event • Same scenario for 1971 and 2006 • Residential casualties only

Damage and Casualties

1971

2006

Deaths 35 22 Fatality rate (deaths per 1,000) 0.032 0.010 Serious injuries 51 38 Serious injury rate (inj. per 1,000) 0.047 0.018 Pop. in significantly damaged dwellings

31,200 2.9%

50,700 2.4%

Spatial Differentials

Population in Significantly Damaged Buildings (Ratio 2006: 1971)

Risk decreased

Risk increased

Discussion

Trend more reliable than static loss estimate

• Significance – Total casualties: net effect neutral – Casualty risk per person: reduced (=safer?) – In some areas, expected losses increasing – Improvements in earthquake engineering have

barely kept up with growth of population at risk – Future growth may outpace marginal improvements

in engineering

• Limitations – Single scenario – Residential building damage only – Computational and data assumptions – Omissions (e.g., code changes)

North Carolina (USA) study

• Hurricane loss model • 7 counties in North Carolina • Cost of damage to wood-

frame houses • 2000 ~ 2020, changes in:

– inventory – vulnerability – economic value

• Net neutral effect on risk

Jain, V.K. and R.A. Davidson. 2007. “Application of a Regional Hurricane Wind Risk Forecasting Model for Wood-Frame Houses,” Risk Analysis 27(1): 45-58.

Comparison with Cities in Asia/Pacific

0102030405060708090

100

1970 1980 1990 2000 2010

Popu

latio

n (%

of 2

009)

Year

Tokyo

Kobe

Memphis

Christchurch

Vancouver

Brisbane

Beijing

Some Differences with Japanese Cities

• Japanese urbanization: – much longer urban history – war damage and post-war reconstruction – higher densities – higher reliance on transit (rail, bus) – lesser reliance on auto – aging, stable/declining populations

Vintage and Damage in 1995 Kobe Earthquake

Housing Damage by Vintage

Source: Hirayama (2000)

Era Built Loss Rate Pre 1945 58.3% 1945-1955 48.7% 1956-1965 48.7% 1966-1975 22.9% 1976-1985 6.1% Post 1986 5.7% All vintages 23.7%

Population by City Ward, 1965~2005

Source: Chang (2010)

Building Stock Replacement

Tokyo • 1923 Great Kanto

earthquake • WWII firebombs • Seismic codes • Lifetime of

buildings, rate of demolition and replacement

Projected Change in Number of Wood-frame Houses in Japan,

2000~2050

Uni

ts (1

0,00

0s)

Year

Existing buildings

(by vintage)

New buildings

(Ohara et al. 2007)

(H. Shindo)

Land Use Changes

Tokyo • Landfill /

reclaimed land in Tokyo Bay since 1600s

Ground failure in areas reclaimed after WWII

>200 miles from epicentral area

Cost to city: $900 m.

photo: Japan Times

Damage to sewer, water, gas pipelines

77,000 hh lost water 1,100+ buildings

damaged /destroyed by liquefaction

2011 Great East Japan Earthquake

geosage.com

USGS

Important Variables

• Era of rapid growth • Building stock replacement rate • Land use change • Infrastructure deterioration

• Geographic setting (coastal, soils) • Population size • Construction practices change • Building codes change • Socio-demographic change • Economic change • Hazard and risk awareness • Mitigation policies, ...etc.

Economic Disruption

• Some key factors 1. damage to infrastructure

2. ripple effects (supply chains)

3. linkages between cities

...also change over time...

Infrastructure Deterioration

0

2

4

6

8

10

12

14

16

18

<10 10~20 20~30 30~40 40~50 50~60 60~70 70~80

Rela

tive

Brea

k Ra

te

(20~

30 y

ears

old

= 1

.0)

Pipe Age (years)

Break Rate by Pipe Age (cast iron)

Vintagepre-19291930 - 19491950-19691970 - 19891990 - 2000

ntagepre-19291930 - 19491950-19691970 - 19891990 - 2000

pe Material (inferred)Cast IronDuctile Iron

Portland, Oregon water system

INVENTORY 2000 2050 % cast iron 72% 49%

% over 80 years old

31% 48%

(Source: Chang, 2003)

Year Weeks to Restore T=0 4 T=20 5 T=50 9

M6.1 Molalla-Canby Fault EQ

Port of Kobe

Rank:

1976 2nd

1994 6th

1997 17th

2010 49th

Chang, S.E. 2010. “Urban disaster recovery: a measurement framework and its application to the 1995 Kobe earthquake,” Disasters 34(2): 303-327. (Also: Chang 2000)

Implications for Insurance

• Importance of anticipating risk dynamics – esp. influence of demographic, social, economic, and built

environment changes – identify emerging insurance needs, opportunities, potential

problem areas – help set actuarially sound, risk-based premiums – understanding correlations across cities in portfolio

• Need for new (dynamic) models – Incorporate projections of population, etc. – Validate actual events with building exposures at that time period – Model vulnerabilities relevant to building vintages (esp. codes,

construction practices)