A Preview of the AIR Typhoon Models for Southeast Asia

Home , Typhoon Haiyan, Typhoon Ketsana, Typhoon Morakot

Apoorv Dabral, Ph.D. Kevin Hill, Ph.D., Ruilong Li, Ph.D.

2014 Non-Life Premium in USD billion (Swiss Re, Sigma) 120 80 40 0

Change in Non-Life Premium from 2005 to 2014 (Swiss Re, Sigma) 300%

200%

100%

Typhoon Morakot, 2009 Typhoon Haiyan, 2013 - Haiyan • One of the strongest typhoons to make landfall • Significant storm surge - Morakot

• Record-breaking Typhoon Ketsana, 2009 precipitation - Ketsana • Significant flooding • Major losses in Philippines and Vietnam

Southeast Asia Typhoon Models – Update and Expansion

Taiwan Hong Kong Currently Supported Territories Territories: Added in 2016 Models • Taiwan • Guam • Hong Kong • Macau • Philippines • Saipan • Vietnam

Philippines

‒ Explicitly capturing large industrial facilities and parks ‒ Tall buildings classification ‒ High geographic resolution ‒ Better risk differentiation

- Part of basinwide catalogue that covers the region - 1 km resolution - Updated wind field and flood modules • Wind: region-specific wind–central pressure relationship, Willoughby decay • Flood: CASC2D, time- dependent - Storm surge module for the Philippines, Hong Kong, and Taiwan

- Conventional buildings, contents, and business interruption - Infrastructure - Marine cargo and hull - Large industrial facilities - CAR/EAR - General auto - 2-wheeled vehicles - Warehouses

April – Philadelphia

June and August • Singapore • Vietnam • Philippines • Beijing • Indonesia • Taiwan

Kevin Hill, Ph.D.

Guam, Model Saipan, Macau, and resolution has been Storm surge Vietnam have increased to component been added a uniform features a 1-km numerical model (SuWAT)

AIR Typhoon Modifications Models for to the stochastic Storm surge Southeast Asia – catalogue for hazard added Hazard Module the Philippines for the and Guam Philippines, Hong Kong, and Taiwan

Flood hazard features a Wind numerical hazard model calculation (CASC2D) updated

Previous Release: 81 unique points Updated model: ~400,000 points (Manila area shown)

Landfall Frequencies – the Philippines

0.40

0.35 Events - Modified landfall 0.30 intensities are included 0.25 0.20

in the models for Guam Landfalling and Philippines 0.15 0.10 0.05 0.00

Proportion of 0 1 2 3 4 5 Saffir Simpson Category

Historical Prev. Stochastic Updated Stochastic

Many New Marquee Events Have Been Added

Taiwan Herb 96 Existing Nari 01 Hong Kong- Toraji 01 Added Macau Haitang 05 Rose 71 Sepat 07 Hope 79 Morakot 09 Ellen 83 Sam 99 Philippines Nuri 08 Joan 70 Guam-Saipan Hagupit 08 Mike 90 Karen 62 Vicente 2012 Angela 95 Pamela 76 Vietnam Vicki 98 Paka 97 Xangsane 06 Durian 06 Pongsona 02 Ketsana 09 Fengshen 08 Nesat 2011 Ketsana 09 Vicente 12 Megi 10 Washi 11 Nesat 11 Bopha 12 Haiyan 13 CONFIDENTIAL ©2016 AIR WORLDWIDE 14 Wind Hazard Is Modelled in a Robust Fashion

- Maximum wind speed is modelled using a wind–central pressure relationship customised for typhoons affecting Southeast Asia • Based on Japan Meteorological Agency (JMA) data 180

160 140 120 100 80 60

40 JMA Data

MaximumWind Speed (mph) 20 SEAsiaSoutheast Best Fit Asia Best Fit 0 880 900 920 940 960 980 1000 1020 Central Pressure (mb)

- Radial decay of wind speed is based upon Willoughby et al. (2006) • Dual exponential wind profile is the most accurate match to observed wind speeds

- Friction and gust factors Rough -- recalculated using same smooth

methodology and data as for other Asia regions • These are dependent on wind direction • Provides detailed wind speed footprint for coastal and inland area

Nuri 2008

Hope 1979 Ellen 1983 Sam 1999

120

100 Hagupit 2008 80

WindSpeed(mph) 20

0 Hope Ellen Sam Nuri Hagupit Vicente Model HK Observatory Vicente 2012

Model 100-year return period wind map Taipei

Hualien

Engineering Study is Jang and Lee,1997: Analysis of Design Wind Speed Distribution of Taiwan Area, J. Marine Science and Techn.

Model 100 year return period wind speeds Regional Severe Wind (1-min sust.) Hazard Map 1% AEP (1/100)

Wind Speed (mph) < 38 38-54 54-87 87-136 136-163 >163

- Modelled precipitation depends on typhoon characteristics and accounts for geographic and local effects - Higher resolution than previous model is more likely to resolve areas of maximum precipitation

Taibei Hualien Kaohsiung GSOD Hong Kong Observed vs Modeled Precip 5 yr RP Ha Noi Model 350

300

Da Nang 250 Manila 200 150 Guam Nha Trang Guiuan 100 Mactan Precipitation(mm) 50 0 Guam Ha Noi Da Nang NhaTrang Hong Guiuan Mactan Manila Taibei Hualien Kaohsiung Kong Station

- CASC2D is time dependent, so flood depth is influenced by rainfall rate and total precipitation - Amount of water that infiltrates depends on soil type and slope - Routing depends on slope and surface friction (vegetation, Manning Coefficient)

Track across The Philippines by Ketsana (Ondoy) as a weak Tropical Storm in Sept 2009.

Ketsana flooded more than 80% of the city of Manila Modelled flood footprint

Chi-Lung River Chi-Lung River Hun-Shui River

Hun-Shui River

Hsin-Tian River Source: National Taiwan University Hsin-Tian River and University of Exeter

Flood Map of Taipei – 200 Year Taipei Region from AIR Model – 250 Year

- Storm surge peril modelled using SuWAT - coupled numerical model of Surge, WAve and Tide (same as Japan typhoon model) • Model developed in Japan and has been used in numerous peer- reviewed publications • Fully dynamical model which takes into account the impact of waves on storm surge • Model utilises high-resolution terrain information for accurate model storm surge inundation

3.5 Observed Modeled 3

2.5

1.5

Water Level(m) 1

0.5

“In Tacloban, the terminal building of Tacloban Airport was destroyed by a 5.2 m (17 ft) storm surge up to the height of the second story. “

5+ 4-5

3-4

2-3

- Model produces a maximum storm surge of 5-6 meters - JSCE-PICE joint survey team estimates the storm surge level was 5-6 meters inside of the Leyte Gulf

Ruilong Li, Ph.D.

Enhanced Wind and Flood Damage Estimation & Accounting for regional and temporal Updated Time vulnerability Element Vulnerability Addition of Storm Module and Surge Damage Probability Module Distributions for AIR Damage Typhoon Models for Southeast Asia – Vulnerability Updates

New Risk Types: High Resolution Damage Functions IFM, Marine, for “Unknown” Infrastructure, Building Features Warehouses, …

Hong Kong - Hong Kong and Macau • Engineered structures • Quality and control referred to

British Standards, with additional AX3 Battery practice notes for engineers - Philippines The Philippines • Good building code • Building standards may not have been followed

- Taiwan CTV

• Robust wind design code Taiwan • Engineered structures • Better enforcement

- Guam and Saipan • Good building code adoption and enforcement

Vietnam Guam JICA NWS - Vietnam - Guam and Saipan • Little quality control • Good building code adoption and • Wind is generally not extreme, but enforcement flooding is more significant

Code of Practice on Wind Effects in Hong Kong 1983

Code of Practice on Wind Vietnam Building Effects in Hong Kong 2004 Guam adopted Code TCVN IBC 2009 2737-1995

NSCP-1992

NSCP-2001

NSCP-2010

Building - Wind vulnerability Wind Wind Hazard Design and Territory Vulnerability • Wind hazard level Level Enforcement Level - High Low Level vulnerability Guam Mid-High High Mid - Low High vulnerability Hong Kong Mid-Low High Low-Mid • Wind design and Macau Mid-Low High Low-Mid enforcement level Philippines High Low Mid-High - High Low vulnerability Saipan Mid-High High Mid

- Low High Taiwan High High Low vulnerability Vietnam Low Low-Mid High

- Temporal vulnerability: Taiwan • Building code evolution • Aging and Building Code 2007

deterioration

Vulnerability Temporal Temporal

Year Built

NSCP 1992 Taiwan Building Code 2007 TCVN-2737-1995

Support regional vulnerability for Philippines, Taiwan, Vietnam

Low-rise

Damage Function Number of Stories Mid-Rise Classes 1-3 Low rise High-Rise

4-7 Mid rise Tall

8-29 High rise Relative Vulnerability

30+ Tall

Height Categories

Hong Kong - Hong Kong: • Building codes and city planning requires flood mitigation Source: http://www.hktdc.com/mis/ires/en/s/hk- HKIE measures gushes-with-urban-solutions.html

• Better flood storage, levee Philippines system - Taiwan: Good sewer system and levee system zambonga

- Philippines and Vietnam: fairly Vietnam poor sewer and drainage system

Flood Flood Flood - Flood vulnerability Territory Hazard Mitigation Vulnerability • Flood hazard level level Level Level - High Low Guam Low Low Low vulnerability Hong Kong Mid High Mid - Low High vulnerability Macau Mid High Mid • Flood Mitigation Philippines Mid-High Low Mid-High level Saipan Low Low Low - High Low vulnerability Taiwan Mid-High Mid-High Low-Mid - Low High Vietnam High Low High vulnerability

1-Story Damage Function Number of Stories Classes 2-Story 3-Story 1 1 Mid-Rise 2 2 High-Rise 3 3 Tall

4-7 Mid rise Relative Vulnerability 8-29 High rise 30+ Tall

Height Categories

Haiyan, 2013, Surge damage in PHP - As opposed to flood, surge is saltwater and has velocity component - Accounting for the hydrodynamic effect of water using an equivalent static Jeff Masters depth Nari, 2001, Flood in Taipei City, TW - Supporting all lines of business, occupancy, and construction

Wikipedia

Cresta 1 – Taipei City

Cresta 2 Changhua

Risk Types Wind Flood Surge*

Conventional Buildings Existing Existing New

Infrastructure New New New Marine Cargo, Hull New New New Large Industrial Facilities New New New Builder’s Risk Existing Existing New Auto Existing Existing New 2-wheeled vehicle New New New Warehouse New New New * Surge for the Philippines, Hong Kong, and Taiwan

Company A

Company B

Company C

60%

50% 40% 30% 20%

Market Share Market 1 10% 0%

1 4-5 6 7 8 9 2-5 CRESTA 6

Replacement Value Distribution in Lines of Business

Res 7 13% Ind 40% 8

9 Com 47%

Philippines Company Books Validation Observed Modelled

CONFIDENTIAL ©2016 AIR WORLDWIDE 45 Major Historical Events Are Benchmarked on Loss Exceedance Probability Curves Flood Only An Example from Taiwan Nari (2001) Nari, 2001 Morakot, 2009 Surge Wind 3% 3%

AAL 2 5 10 20 25 50 100 200 250 500 1000 Return Period Flood 94% All Perils Morakot (2009) Nari, 2001 Wind Morakot, 2009 Surge 14% 20%

April – Philadelphia

June & August • Singapore • Vietnam • Philippines • Beijing • Indonesia • Taiwan