Updates to Modelling Floods in Europe

Alastair Clarke, PhD Richard M Yablonsky, PhD

Introduction

AIR’s Inland Flood Model for Central Europe

• Updates to the Hazard Component • Updates to the Vulnerability Component • Modelled Losses

The Addition of Great Britain Storm Surge to AIR’s Extratropical Cyclone Model for Europe • Updates to the Hazard Component • Hazard Model Validation • Modelled Losses • Upcoming Software Changes

Central Europe Inland Flood

Poland, Switzerland, updates to Italy, Germany Austria, Czech Hungary, other Republic Slovakia countries

2000 2011 2013 2015 2019 Future

Addition of Great Britain Great Britain Great Britain Vulnerability Storm Surge Update to EU ETC Great Britain Storm Surge

Recent New Data Events Poland and Information updates to other countries

Additional Domain 2019 Asset Demand Expansion Types

Addition of Great Britain Storm Surge Building Model Into EU ETC Other Unification Experience Models

• Utilised best available building footprint data sets to accurately locate risks • Updated construction splits, replacement costs, coverage splits, and policy conditions • Improved automobile

valuation methodology 2 • Identified high-value industrial

facilities TIV per km per TIV

• Addition of marine cargo, marine hull, industrial facilities, wind turbine, and builder’s risk Marine • Incorporation of unknown damage functions at CRESTA level • Updated secondary damage distributions Industrial Facilities

Custom Flood Presence of Floor of Protection Basement Interest

First-Floor Custom Height Elevation

Addition of: - Two major river basins: Oder and Vistula - 14,470 catchments - More than 86,873 km of river network - More than 110,649 river cross sections at intervals of roughly 500 metres

1997 Poland

Event Year Existing Events

2002 Austria, Czech Republic, Four Germany 2013 New 2001 Central Poland 2005 Austria, SwitzerlandEurope Events 2007 Switzerland

2010 Poland

Precipitation

Define Events Surface Vulnerability Module Runoff Pluvial

Hydrological Model River Flow Fluvial

Precipitation • Precipitation downscaling upgraded for better consistency across country boundaries leads to more realistic Define Events footprints for extreme events

• Addition of Poland resulted in Hydrological redefining of events Model

• Germany hazard was re-evaluated based on new information Pluvial • Improved pluvial hazard consistencies across political boundaries

• Updated levee information Fluvial

Data Sources • National and State-Level Flood Authorities

• Municipal Flood Protection Authorities

• CORINE Land Cover

Source: Wikipedia

©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 18 18 The Updated Pluvial Module Uses a Statistical Model • Design relative runoff and urban land use are used to estimate pluvial drainage in urban areas

Rostock !

Hamburg !

Berlin ! Hannover ! Münster !

Düsseldorf ! Dresden ! Legend

Frankfurt am Main ! Drainage5 year Protection

Nürnberg ! CapacityRR

Stuttgart 0 ! Low

München 1.1 - 1.4 Freiburg im Breisgau ! ! 1.5 - 1.5 1.6 - 1.6 High1.7 - 1.9

Building Contents

4 4 3.5 3.5 3 3 2.5 2.5 2 2 1.5 1.5

1 1 Relative to to ResidentialRelative 0.5 to ResidentialRelative 0.5 0 0 Commercial Building Industrial Building Commercial Contents Industrial Contents

Germany Switzerland, Austria, and Czech. Rep. Previous Updated

0.9

0.8

0.7 Residential Claims

0.6 Commercial Claims

0.5

0.4

0.3 MeanDamage Ratio 0.2

0.1

0 0 1 2 3 4 5 6 7 8 Flood Depth (Metres)

0.07

0.06 0.8

0.05

0.04 0.6

0.03

Probability 0.4

0.02 Mean Damage Ratio (MDR) Damage Ratio Damage

0.01 0.2 0 0.0 0.2 0.4 0.6 0.8 1.0 0 Damage Ratio Hazard Intensity (Flood Depth)

©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 23 23 Ground-Up AAL Changes on AIR’s Industry Change Impact Exposure Database (IED) <25% 25–50% >50% Germany Switzerland Austria Czech Republic

Impact of Hazard and SoP Updates

Impact of Residential Damage Function Updates

Impact of Commercial Damage Function Updates

Overall Impact on AIR IED

14 Observed Lower Limit

12 Modelled Observed Upper Limit 10

6 Loss (EUR Billions)(EUR Loss

0 CH05 CH07 AT02 AT05 AT13 CZ02 CZ13 DE02 DE13 PL97 PL01 PL10 PL13

20 1997

Loss 2010 2001 Aggregate 10 Occurrence

- AAL 2 5 10 20 40 50 100 250 500 1,0001000 25002,500 Return Period (Years)

Loss Updated

AAL 5 10 20 40 50 100 250 500 1,0001000 2,5002500

Aggregate Occurrence Return Period (Years)

AIR Industry Exposure Database (IED): Germany IED: Gross Insurable AAL Change by CRESTA

200% Gross Insurable Loss Change Increase 150% No Change

Decrease 100%

50%

0% Gross Insurable AAL Change AAL Insurable Gross -50%

Aggregated Annual Ground-Up Loss by Country

Austria Switzerland Czech. R Germany Poland

Previous Updated

2019 Update: Previously: Storm Surge sub-peril in the AIR AIR Coastal Flood Model for Extratropical Cyclone Model Great Britain for Europe

Sea Defences Topography

Shallow Water Deep Sea Wind Pressure

Bathymetry

in Great Britain Historical Storm Tracks Latitude

Storm Anne (2014): West Coast Example* Storm Xaver (2013): East Coast Example** *another example is Storm Undine (1991) **another famous example is North Sea Flood (1953)

Southerly wind pushes water into Northerly wind pushes water into Irish Sea the Channel

Hydrodynamic Domain Coarse • Expanded domain covers all Mesh of England and Wales Fine Mesh: • Uses Delft3D Flexible Mesh Current Coverage New Coverage • Dynamic simulation of tides and surge • Driven by wind and pressure from EU ETC model catalogue (see Keshtpoor, Carnacina, and Yablonsky, 2019: New Statistical Approach to Select Coastal Flood-Producing Extratropical Cyclones from a 10,000-Year Stochastic Catalog. J. Waterway, Port, Coastal, and Ocean Eng.) • Surge depth at 10-metre resolution via downscaling and topography subtraction

• Data can be mapped to a Manning friction coefficient

• Helps to ensure proper flood extent over land

• Particularly challenging in urban areas

UK Environment Agency Spatial Flood Defences Dataset • 2016 vintage defence data set for the entire United Kingdom: defined crest heights, material, quality condition, and design type • Undefined heights were extracted from 2- to10- metre topography (DTM) sources

• The Thames Barrier was completed in 1984

• During construction, flood defences were raised by ~2 metres for ~30 km downstream to match the Barrier

• Defences for ~5 km upstream were also raised to increase protection and match protection of Central London

©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 44 44 Storm Surge Model Validation at Great Britain Tide Gauges for Four Storms Four Storms Combined: 1953, Undine, Xaver, and Anne Total water level Wind setup (no tide)

) Perfect line

Metres Modelled Water Elevation ( Elevation Water Modelled

RMSETWL= 0.32 m RMSE = 0.28 m Tide Gauge Locations WS Observed Water Elevation (Metres)

©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 45 45 Modelled Pressure, Wind, and Surge During North Sea Flood of 1953 a) SLP (mb) b) U (m/s) 20 1030 a) Sea level pressure 60 1020 10 1010 b) West-to-east wind 55 1000 0

speed (U) ~180 metres 990 ) above ground ° 50 980 -10

c) V (m/s) d) Surge (m) c) South-to-north wind 20 2 speed (V) ~180 metres Latitude( 60 10 1 above ground 0 55 0 -10 d) Water level due to -1 -20 wind setup (without 50 -30 -2 tide) -10 0 10 -10 0 10 Longitude(°)

©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 46 46 Modelled Water Level Near Thames Barrier from a Storm Like the 1953 North Sea Flood 8 ~7.3 m (height of Thames Barrier) Tide Only Surge Only 6 Linear Tide & Surge ~5.4 m (like North Sea Total Water Level 1953 storm) 4

-2

Thames Barrier

Hull, UK

Humber Estuary

Boston, UK Boston, UK

The Haven River The Haven River

Insufficient resolution (~220 m) Sufficient resolution (up to 35 m) for water to reach Boston, UK for water to reach Boston, UK

ModelledModel Ground-Up Ground-Up Estimated 9 8 7 6 5 4 3

Losses (GBP Losses (GBP Billions) 2 1 0 North Sea Flood (1953 Vintage)

ModelledModel GroundInsured-Up Observed 450 400 350 300 250 200 150

100 Losses (GBP Millions) (GBP Losses 50 0 Undine (1991) Xaver (2013 Vintage) Anne (2014) Flood Events Observed Data Sources: UK Met Office, Munich Re, ABI ©2019 AIR Worldwide Confidential - For European Seminar Attendees Only 55 55 Exceedance Probability of Insurable Gross Loss 40

35 1953 X 30 X = No Levees 25 1953 V V = Vintage Levees

20 M = Modern Levees Loss

15 1953 M 2013 V 10 2013 M 1991

5 2014

0 AOL 2 4 5 10 15 20 50 100 200 500 1,000 2,000 5,000 10,000 Return Period (Years)

0.8 1953 M V = Vintage Levees

0.6 M = Modern Levees Loss

2013 V 0.4 2013 M

1991 0.2 2014

0 AOL 2 4 5 10 15 20 50 100 Return Period (Years)

Residential, Commercial,Insured Aggregate Industrial, Loss Agriculture, Comparison and Auto Lines of Business 50

Wind Only Loss 20 Wind & Updated Surge

0 AAL 2 5 10 20 25 50 100 250 500 1,000 Return Period (Years)

Loss Wind & Previous 20 Surge

Wind & Updated 10 Surge

0 AAL 2 5 10 20 25 50 100 250 500 1,000 Return Period (Years)

Touchstone 2018

* Europe Extratropical Cyclone is still categorised as a “Tropical Cyclone” peril in Touchstone®

Touchstone 2019

The Updated AIR Inland Flood Model for Central Europe:

• Expanded model coverage includes Poland and new historical events • Updated existing model domain, including standard of protection • Updated vulnerability for coverage and secondary risk characteristics The Addition of Storm Surge to the AIR Extratropical Cyclone Model for Europe: • Expanded model coverage includes all of England and Wales • Updated with modern-day levee information to improve surge footprints • Utilised the Delft3D Flexible Mesh hydrodynamic model • Used wind and pressure from EU ETC model catalogue to drive surge