Earthquake Risk Modelling

Dr. Dirk HollnackGeoRisks Research GroupMunich Reinsurance Company

2

1. Principles of Risk Assessment

2. Hazard Maps

3. Earthquake Scenarios

4. Probabilistic Modelling

5. Insurance Aspects

Contents

3

Principles of Risk Assessment

4

Risk = fRisk = f

Hazard= occurrence probability for event of a certain size

Vulnerability ofbuildings, contents, BI

Values, Liabilities

Earthquake Risk and its Components

5

Natural Catastrophe Modelling

Representation of natural phenomena

(severity, location, probability)

Calculate the consequences of these phenomena

Risk management (preparedness, mitigation)

Estimate loss potentials

Why do we use risk models?

6

Player in EQ Risk Modelling

EQ Risk Modelling is done by:

Consultants

(Re)Insurances

Brokers

Geol. surveys and public agencies

Scientific groups/universities

‘Science’ and

public

‘Insurance Business’

7

NatCat Risk Modelling for Insurance Business

Insurance business uses NatCat risk models since the 80th

Some examples:

- AIR since 1987

- Munich Re since 1987

- RMS since 1988

- EQECAT since 1994

- Benfield since 1999

8

EQ Risk Modelling

Why are university risk models only used for a very limited extend in

insurance business?

The methodology, resolution and parameters to be used vary with the

purpose of risk modelling (i.e. mortality, disaster management, risk

reduction, financial risk)

EQ models for insurances have a kind of standard which meets the

requirements of the business. Research projects are often designed for a

small area (i.e. one city), working on a high resolution and/or are focused

on a detailed problem:

High computational requirements (run-time, memory)

Results are often difficult to adapt for insurance purposes

9

Exposure: What is a “Risk Element”?

Building

Contents

Machinery & equipment

Construction sites

Consequential loss (Business interruption, Advanced loss of profit)

Vehicles, Life, Arts, Social events (Olympic games, rock concerts), etc

=> Much broader sense than in normally used in EQ Engineering

10

Average annual loss (AAL) => rating – site specific

Probable maximum loss (PML) => catastrophe potential - regional scale

An adequate Price and PML must reflect

Risk Location - Hazard

Type of Risk - Vulnerability

Insurance Conditions

(Claims Experience)

Insurance Aspects

11

PML

AAL

12

In Principle two different types of contracts which require different

modelling methods:

Portfolio = large number of risks which are spatially distributed

Facultative = single risk (mainly large industry complexes or

buildings)

EQ Risk Models for Insurances

13

return period

loss

in [

%]

TS

I

single risk

portfolio

Introduction to concepts of loss estimationSingle Risk vs. Portfolio

14

Insurance Conditions

Self-participation

Deductibles

Limits

15

Hazard Maps

16

Nathan- World Map of Natural Hazards

(Maximum Intensity of a 475 years return period)

No information about other return periods

Hazard Maps

17

Hazard Map/ Usage

Basis of building codes/regulations

Basis of tariff zones

Warning signal

Loss potential estimation

Comparison of two locations

18

Hazard map/ problems

Affected region not clear

Verification difficult

No regional differences inside hazard zones

Secondary effects not included

Only for one return period

19

Intensity plots

5

6

7

8

9

10

11

1.00 10.00 100.00 1000.00 10000.00 100000.00

Return period

Inte

nsity

Hazard at other return periods

20

21

22

Earthquake Scenarios

23

Scenario

What loss potentials can hit me in the case of a natural catastrophe?

24

Scenario

Selection of scenarios

Historical

Modified historical

Theoretical possible (virtual)

25

Isoseismal Map / Intensity Scales

Düren (1756) - M = 6.1

xz

xz

xz

xz

xz

xz

xz

xz

xz xz

xz

xz

xz

xz

xz

xz

xz

xz

xz

xzxz

xz

xz

Bonn

Köln

Essen

Neuss

Aachen

Krefeld

Dortmund

Düsseldorf

Leverkusen

Düren_17560 - 5.55.5 - 6.56.5 - 7.57.5 - 8.58.5 - 9.59.5 - 12

26

Scenarios – Historical modified

27

Scenarios/ use

As/if calculations

Comparison to market loss estimates

Verification of probabilistic models

Loss potential estimate/ budgets

28

Scenario / limitations

Is my Scenario ...

realistic ?

adequate ?

out-dated ?

a support in determining the premium level ?

29

Modelling Earthquake Risk

30

Probabilistic modelling

What are the loss potentials I have to expect for my

portfolio?

How frequent do these losses occur?

31

Introduction to concepts of loss estimationProbabilistic modelling

Principle:

Generation of large synthetic event sets

(thousands to hundreds of thousands)

Assignment of occurence probabilities

Calculation of losses

Calculation of exceedence probabilities

Calculation of PML curve and technical rate

32

Event simulation is based on:

Measured events

Historic/ pre-historic events

Regional characteristics

Physical framework

Probabilistic modelling

33

Vulnerability functionMunich Re's

loss experience

Set ofscenarios

Statistics

Expected loss/ lossoccurrenceprobability

Risk curve

Value distribution

Individual exposure

%

%

%%

%

%

%

%

%

%

%

%

%

Sapporo

Aomori

Nagasaki

Fukuoka

Kita Kyushu

HiroshimaKobe

Osaka

KyotoNagoya Yokohama

Kawasaki

Tokyo

Industrial Sum Insured (Earthquake) < 1,0001,000 - 3,0003,000 - 6,0006,000 - 10,000 > 10,000 Mio. ¥

% Major Cities

0 200 400 Kilometers

Scientific

input

Hazardinformation

The holistic Solution for Risk Assessment:

Risk Models

34

Detailed Risk Information

Risk models require high resolution data:

(GPS) coordinates

Geotechnical information

Building characteristics

Age

Height

Occupancy

Construction type

35

CRESTA – An Insurance Standard

CRESTA was set up by the insurance industry

in 1977 as an independent organisation for

the technical management of natural hazard

coverage.

CRESTA's main tasks are:  

Determining country-specific zones for the uniform and detailed reporting of

accumulation risk data relating to natural hazards and creating corresponding

zonal maps for each country   

Drawing up standardised accumulation risk-recording forms for each country

     

Working out a uniform format for the processing and electronic transfer of

accumulation risk data between insurance and reinsurance companies

36

Greece – 16 ZonesGermany – 8270 Zones

The CRESTA Format

37

Quality of Input Data

38

Uncertainties in Risk Modelling

Event (location, size)

Intensity (attenuation, directivity)

Local influence (amplification, frequency)

Risk information (building quality, location)

Vulnerability (average damage, distribution)

Loss (estimation of values, demand surge)

39

Uncertainties in Risk Modelling

There is a general tendency in modelling to increase the

resolution and the number of parameters:

Does this really increase the quality of the models?

40

Secondary Effects

41

Vulnerability: Single Location

42

Izmit/Turkey, Aug 17, 1999

43

Loss Assessment (Exercises)

44

Location of the risk Intensity levels for various return periods Type and quality of the risk to estimate the

vulnerability Value of the risk Insurance conditions applied

Information required

Exercise 1:Estimation of Insurance Rate – Single Risk

45

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

1 10 100 1000 10000

return period

los

s

sum of premiums = sum of loss

(over a certain time) (over a certain time)

Estimation of Insurance Rate

46

sum of premiums = sum of loss

(over a certain time) (over a certain time)

Estimation of Insurance Rate

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

1 10 100 1000 10000

return period

los

s

47

Hazard Maps

As a rule of Thumb (only for earthquakes):If the return period for one Intensity is known, a factor of 3-4 can be used to assess the return period for other Intensities

48

Rate Calculation

Rate (%) = 1/Return Period(1) * Loss%(1) + 1/Return Period(2) * Loss

%(2) ... + 1/Return Period(n) * Loss%(n)

49

• Geographical distribution of the liabilities(Accumulation assessment zones)

• Risk classes (residential, commercial, industrial)

• Insured interests (building, contents, lop)

• Intensity field of the EQ-scenario

• Vulnerabilities

• Values

• Deductibles applied

Information required

Exercise 2Estimation of Scenario Losses

50

Accumulation assessment zones Example

Definition of zones by either geographical regions or provinces or districts or

postal codes

Capital

4

2

1

10

7

9

8

6

5

3

51

Dr. Dirk HollnackGeophysical/Geological RisksGeo Risks Research Dept.

Thank you for your attention!