The New Inland Flood Model and Updated Model for

Marc Marcella, Ph.D. Yang “Gabby” Gao, Ph.D. Ruilong Li, Ph.D. .

CONFIDENTIAL ©2017 AIR WORLDWIDE 1 Agenda

- Modeling Flood Hazard for Japan

- Modeling Flood Vulnerability for Japan

- Updates to the AIR Typhoon Model for Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 2 Modeling Precipitation in Touchstone JapanTyphoon Typhoon Related Model Perils

Non Tropical Typhoon Typhoon Cyclone Wind Surge Precip. Flood Precip. Flood

JapanWater Inland Related Flood Perils Model - SOLUTIONS in Touchstone®: • Two models in Touchstone - AIR Typhoon Model for Japan: wind, surge, tropical cyclone precipitation flood - AIR Inland Flood Model for Japan: non tropical cyclone precipitation flood

CONFIDENTIAL ©2017 AIR WORLDWIDE 3 Modeling Flood Hazard for Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 4 Flood Model Components

GEOPROCESSING PRECIPITATION

HYDROLOGY HYDRAULICS VULNERABILITY

CONFIDENTIAL ©2017 AIR WORLDWIDE 5 Flooding in Japan Flood peril is a significant risk for Japan • 50% of population and 75% of assets located in a floodplain • Major evacuations, loss of life, and billions in damage Aso, Kyushu - 2012 Koshigaya, - 2015

Non Tropical Cyclone Tropical Cyclone Rainfall 40% 60%

CONFIDENTIAL ©2017 AIR WORLDWIDE 6 Geoprocessing

GEOPROCESSING

HYDROLOGY HYDRAULICS

CONFIDENTIAL ©2017 AIR WORLDWIDE 7 17,425 ~1,100 river dams & Reservoirs segments

~1,000 103,000 km streamflow total stream gauge stations length

2 20,263 338,000 km catchments modeled CONFIDENTIAL ©2017 AIR WORLDWIDE 8 Geoprocessing Data Layers

Digital Terrain Model Model Boundary: (DTM): Country Border Provided by MLIT

Unit Catchments: River Network: From DTM & River Derived from DTM Networks

Catchment Properties: Land Use: JAXA, MODIS, Cross Sections: AIR , Impervious Surface: MLIT & DTM River Network NOAA & Soil: MLIT

Dams & Reservoirs: Flood Defenses: MLIT, iCold, GRAND & DTM & River Network SRTM CONFIDENTIAL ©2017 AIR WORLDWIDE 9 Japan Digital Terrain Model Availability

➢ 5m DTM covers 60% of Japan

➢ Most major cities and rivers included

➢ 10m DTM used to supplement remaining area 10m5m DTMDTM CONFIDENTIAL ©2017 AIR WORLDWIDE 10 Precipitation Generation

PRECIPITATION

HYDROLOGY HYDRAULICS

CONFIDENTIAL ©2017 AIR WORLDWIDE 11 Overview of Steps for Simulating Precipitation 1. Numerical modeling of historical TC events and non TC rainfall

2. Stochastic simulations of precipitation learning from numerically modeled precipitation 3. Blending of non TC and TC precipitation.

1. 2. 3.

CONFIDENTIAL ©2017 AIR WORLDWIDE 12 Step 1: Numerical Modeling Coarse-Resolution Fine-Resolution Global Model Regional Model

downscaling

General Circulation Model, Numerical Weather Prediction Reanalysis Data Model

Regional Domain Selection for Numerical Weather

Prediction (fine resolution) CONFIDENTIAL ©2017 AIR WORLDWIDE 13 Step 2: Stochastic Simulation of Non TC Precipitation

Numerical Weather Prediction Model Stochastic Simulation Output

Statistically robust perturbation of precipitation patterns

CONFIDENTIAL ©2017 AIR WORLDWIDE 14 Step 2: Stochastic Simulation of TC Precipitation

Group models in bundles representing different stages during a typical TC life cycle: 1. Central pressure 2. Storm evolution time (genesis, dissipation, etc.)

At simulation stage, draw samples from distinct models according to stochastic tracks: 1. Central pressure value 2. Time within storm cycle of stochastically simulated Japan catalog track

CONFIDENTIAL ©2017 AIR WORLDWIDE 15 Step 3: Blending Non TC and TC Precipitation

Tropical cyclone simulations are blended into the non tropical cyclone rainfall simulation

BLENDED

CONFIDENTIAL ©2017 AIR WORLDWIDE 16 Hydrology Model

PRECIPITATION

HYDROLOGY

HYDRAULICS

CONFIDENTIAL ©2017 AIR WORLDWIDE 17 Hydrologic Model: Transforming Precipitation to Flow

Precipitation Runoff and Flow Generation

Probability Distributed Model (PDM)

Snow

Model Flow

Time

Muskingum-Cunge Channel Routing Flow

Time CONFIDENTIAL ©2017 AIR WORLDWIDE 18 Hydrologic Modeling Stages Develop All Model Components River (Snow model, runoff, River Flow Network & reservoir, diversions, etc.) Topography New Bayesian Aproach Compile Input Hydrological Model Data Calibration

Others.. Met Data Calibrated Parameters Simulate Simulate 35 Year 10K Historical Stochastic Dynamically Stochastic Downscaled Events Events

Validate Validate Historical Stochastic Events/ Catalog Catalog CONFIDENTIAL ©2017 AIR WORLDWIDE 19 Reservoirs, Dams, and Diversions

Metropolitan Area Outer Underground Discharge Kurobe Dam, Honshu Channel, Greater Tokyo Tokuyama Dam, Ibigawa • Reservoirs/dams significantly attenuate the flows downstream • The operation rules (reservoir rule curves) determine the desired reservoir stage at any given time

CONFIDENTIAL ©2017 AIR WORLDWIDE 20 Hydraulics Model

PRECIPITATION

Riverine flooding Flash flooding On-floodplain HYDROLOGY HYDRAULICS Off-floodplain FLUVIAL PLUVIAL Widespread Localized

CONFIDENTIAL ©2017 AIR WORLDWIDE 21 Pluvial: New Off-Plain Flood Model

Precipitation Intensity

Urban Area

Drainage

2D Pluvial Model Schematic CONFIDENTIAL ©2017 AIR WORLDWIDE 22 Pluvial Flood Model: Storm Drainage Capacity in Urbanized Areas Drainage sufficiency rate and design precipitation are used to estimate storm drainage capacity

Drainage Sufficiency Rate (%) Precipitationhigh85 7 yr 2 hr (mm) Drainage Capacity (mm) low15 High High High Low Low Low

CONFIDENTIAL ©2017 AIR WORLDWIDE 23 Fluvial: Explicit Two-Dimensional Modeling

withstand after SOP

failure before SOP

Flood defense failures are dynamically simulated with one-sided failure possible

failure point

Wide floodplains modeled more effectively, with a more-robust approach and more-complex conditions CONFIDENTIAL ©2017 AIR WORLDWIDE 24 Summary of Advancements - Unified view of hazard, including tropical cyclone and non tropical cyclone rainfall - New, improved approach to hydrologic model calibration, as well as detailed reservoir representation - New physically based off-plain flood model explicitly simulates pluvial flooding - Explicit modeling of certain failure points, including one-sided flood defense failure

CONFIDENTIAL ©2017 AIR WORLDWIDE 25 Modeling Flood Vulnerability for Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 26 AIR’s Inland Flood Modeling Experience

Inland Flood Model for Inland Flood Model for Germany Central Europe

2008 2011 2014 2015 2017

Inland Flood Model for Inland Flood Model for the Inland Flood & Typhoon Great Britain United States Models for Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 27 Vulnerability Modeling Framework: Overview of Risk Features and Modeling Approach

First Floor Ceiling Height Height Water Depth

Basement Ground Elevation

CONFIDENTIAL ©2017 AIR WORLDWIDE 28 Vulnerability Modeling Framework: Overview of Risk Features and Modeling Approach

CONFIDENTIAL ©2017 AIR WORLDWIDE 29 Primary Risk Features Supported in Japan

Conventional Occupancy, Construction, Height, and Coverage

➢ Composite Construction Class (Fire Codes)

➢ Low Rise: 1-, 2-, and 3-Story

➢ Unknown at Prefecture level

CONFIDENTIAL ©2017 AIR WORLDWIDE 30 Special Properties Supported in Japan Inland Flood Modeling

Marine, Inland Transit, Builder’s Risk, and Railway

~147,000 unique damage functions CONFIDENTIAL ©2017 AIR WORLDWIDE 31 Vulnerability: Component-Level Approach to Developing Damage Functions

• Buildings are divided into 6 components, which total 100% of replacement value • Component-level damage functions (DFs) are then combined in proportion to their contribution to the overall replacement value

CONFIDENTIAL ©2017 AIR WORLDWIDE 32 Vulnerability: Component-Level Approach to Developing Damage Functions

Electrical Interior Foundation Mechanical Plumbing Structures

U.S. Single-Family Home: Japan Single-Family Home: Wood Building Wood Building

CONFIDENTIAL ©2017 AIR WORLDWIDE 33 Data Sources for Vulnerability Model Development

✓ Construction Research Institute Monthly Price Index ✓ Japan Survey and Inter-Risk Research Institute ✓ Japanese Architecture Disaster Prevention Association Ministry of Land, Infrastructure and Japanese & Global Publications Transport Water Management (MLIT) CONFIDENTIAL ©2017 AIR WORLDWIDE 34 Component-Based Damage Functions Evaluation

Single-Family 3-Story Home: Steel Single-Family 3-Story Home: Wood

Damage Function Damage Function

Mean Damage Ratio Damage Mean Mean Damage Ratio Damage Mean

Inundation Depth Inundation Depth

CONFIDENTIAL ©2017 AIR WORLDWIDE 35 Benchmarking Loss Sources for Loss Validation • MLIT National Flood Database • Industry reports and research publications about major historical events losses

CONFIDENTIAL ©2017 AIR WORLDWIDE 36 Regional Distribution of Total Flood Risk

5% 4%

21%

10% 31%

8% 4% 17%

CONFIDENTIAL ©2017 AIR WORLDWIDE 37 Regional Distribution of Flood Risk for TC and Non TC Flood

5% Total 4% Total 2% TC 21% Total 3% Non TC 2% TC 12% TC 2% Non TC 9% Non TC 10% Total 31% Total 7% TC 25 % TC 3% Non TC 6% Non TC

8% Total 4% TC 17% Total 4% Non TC 4% Total 12 % TC 5% Non TC 3% TC 1% Non TC

CONFIDENTIAL ©2017 AIR WORLDWIDE 38 Updates to the AIR Typhoon Model for Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 39 Updates to the Typhoon Model for Japan Updated the IED and Introduced new IED enhanced building and and enhanced Introduce new IED, time element damage building damage new flood module, function (supported old function (supports and updated wind fire codes) new fire codes) and surge modules

2002 2007 2010 2012 2015 2017

Released the first Became part of the Introduced physically AIR Typhoon Northwest Pacific based storm surge Model for Japan Basinwide Typhoon module and minor Model, introduced flood wind vulnerability module, and supported update more LOBs (marine, inland transit, railway, etc.) CONFIDENTIAL ©2017 AIR WORLDWIDE 40 New Data Used to Evaluate the Model AIR has collected detailed claims and loss information from 2001 to 2015

Typhoon Etau

CONFIDENTIAL ©2017 AIR WORLDWIDE 41 Typhoon Wind: Re-Evaluating the Damage Function

Commercial – Fire Class 1 Commercial – Fire Class 1 Observed

Modeled

Ratio Damage Ratio Damage Observed Mean Building

Modeled Content Mean Damage Content 30 50 70 90 110 Building Mean Damage Ratio Wind Speed (MPH)

CONFIDENTIAL ©2017 AIR WORLDWIDE 42 Typhoon Wind: Positive Validation of the Updated Wind Component

Commercial – Fire Class 1 Commercial Observed Observed Modeled

Modeled

Ratio

Insured Loss Insured Building MeanDamage 30 50 70 90 110 Chaba Songda Tokage Nabi Shanshan Wind Speed (MPH) Historical Events

CONFIDENTIAL ©2017 AIR WORLDWIDE 43 Enhanced Flood Hazard and Vulnerability Module

GEOPROCESSING Building

Service PRECIPITATION Foundation Structure Interiors Equipment

Mechanical HYDROLOGY Electrical

Plumbing HYDRAULICS

Flood Hazard Flood Vulnerability

CONFIDENTIAL ©2017 AIR WORLDWIDE 44 Enhanced Storm Surge Hazard and Vulnerability Module

Storm Surge Hazard

• Updated the coastal levee system in Hiroshima and Saga • Increased the storm surge resolution from 250 meters to 25 meters

Storm Surge Vulnerability

• Used the component level methodology • Supported the same secondary features as flood

CONFIDENTIAL ©2017 AIR WORLDWIDE 45 Updated Validation for Historical Events

Observed

Modeled Insured Loss Insured

Historical Events

CONFIDENTIAL ©2017 AIR WORLDWIDE 46 AAL Contribution by Region and Sub-Perils

AAL Distribution by Region AAL Distribution by Sub-Perils Tohuku Surge

Okinawa Kanto Flood

Chubu Kyushu Kinki Wind

Shikoku Chugoku

Typhoon risk in Japan is mainly in the southern region, including Kyushu, Okinawa, Chugoku, Kanto, etc. CONFIDENTIAL ©2017 AIR WORLDWIDE 47 AAL Contribution by Region and Sub-Perils AAL Contribution by Sub-Peril for Key Regions Chubu Hokkaido Surge Surge AAL Distribution by Sub-Perils

Flood Flood Surge

Wind Wind Flood Chugoku Kanto Surge

Surge Flood

Wind Wind Flood Wind

Kyushu Kinki Flood Surge Surge

Flood

Wind Wind

CONFIDENTIAL ©2017 AIR WORLDWIDE 48 In the New AIR Model, Typhoon Risk Varies by Region and Sub-Perils in Japan

Region Total Wind Flood Surge Category Absolute Change % Japan Small Small Large Medium Small <10%

Hokkaido Small Small Medium Small Medium 10%-50% Large >50% Tohoku Medium Small Large Large

Kanto Medium Small Large Medium

Chubu Small Medium Medium Large

Kansai Small Medium Small Large

Chugoku Small Small Large Medium

Shikoku Medium Small Medium Large

Kyushu Small Small Large Medium

Okinawa Medium Medium Large Small

CONFIDENTIAL ©2017 AIR WORLDWIDE 49 Closing Remarks

• Introducing a new model for floods from non tropical cyclones and enhancing flood module in typhoon model • Flood model validates well against observed data from MLIT and other sources • Updated typhoon model • AIR provides a comprehensive view of risk in Japan

CONFIDENTIAL ©2017 AIR WORLDWIDE 50 Thank you!

CONFIDENTIAL ©2017 AIR WORLDWIDE 51