Building Catastrophe Models using Open Data and Open Source

0
Building Catastrophe Models using
Open Data and Open Source
Chris Ewing
Aon Benfield

1
Chris Ewing
Aon Benfield
Building Catastrophe Models using
Open Data and Open Source

2
Overview
1. What is reinsurance and
catastrophe modelling?
2. Building catastrophe models
 Open Source, Open Data
 Model development
 Visualisation
3. Making models ‘Open for Business’
4. Summary

What is Reinsurance?
 Reinsurance is simply insurance
for insurers - it allows the insurer
to remove risk
 Natural catastrophes can produce
large losses for insurers and are
one type of risk where reinsurance
is needed
 Aon Benfield are reinsurance
intermediaries (brokers)

4
 Aon Benfield is an industry leader in treaty,
facultative and capital markets, and act as
reinsurance intermediaries and
capital advisors
 Within Aon Benfield, Analytics offers clients
catastrophe management, actuarial, rating
agency advisory and risk and capital strategy
expertise
 Within Analytics, Impact Forecasting
develop catastrophe models that
help analyse the financial implications
of catastrophic events
What do we do?

What is Impact Forecasting & ELEMENTS?
 Impact Forecasting Team
– Catastrophe model developers
– Independent, transparent, open, modular and
bespoke models
– Natural and man-made perils
– Filling gaps and main perils
– Global team (60+)
 ELEMENTS platform
– Runs our models and 3rd party models
– Visualisation: uncertainty & mapping
– 20+ programmers over 4 years
– Distributed to and run by re/insurers and
Aon Benfield colleagues

6
Vulnerability Exposure
Hazard
RISK
Peril Frequency and Severity
EQ shaking intensity
Wind strength
Flood depth inundation
Blast radius
Risk Portfolio Data
Structure values
Contents values
Time Element
Number of people
Deductibles / Limits
Reinsurance
Hazard Susceptibility
Structural classifications
Occupancy descriptions
Secondary Characteristics
Loss
Calculation
What is Catastrophe Modelling?

7
 To help understand the risk faced by
corporations to natural catastrophes
 To help with pricing catastrophe cover
 Development of cat models following:
– European windstorms 1987 / 1990
– Hurricane Andrew 1992
– World Trade Centre terrorist attacks
– Hurricane Katrina 2005
 Catastrophe Models also cover
Terrorism, Pandemic Influenza, Workers
Compensation and Crop
Why use Catastrophe Modelling?

8
 Most insurers and reinsurers and
regulators have some form of (basic)
catastrophe model
 Catastrophe Models are required
for regulatory requirements
e.g. Solvency II in EEC
 3 commercial modelling companies
(AIR, EQECAT, RMS)
 Aon Benfield Impact Forecasting
develops ELEMENTS - our in-house
catastrophe model
Who uses Catastrophe Models?

The 3 opens
 For Model Development
 For Visualisation

Model development 1
 Creating model components (Hazard,
Vulnerability, Exposure)
 Developing model components often
faster and cheaper using Open-Source
software
 Hazard examples….
– Creating EQ footprint maps – Python
(matplotlib and basemap)
– Creating 100,000s of EQ
footprints – Python
– Creating a pre-calculated
consolidated EQ footprint table
(Master Table) – Python / R
X
VIII
VII
VI
VI VII VIII IX X
Intensity
VI
VI
VII
VII
X
X
X
IX
IXIX
IX
Comparison of the calculated seismic intensities
(rupture source modelling) vs macroseismic map
(source: Ozmen, 2000)

I = 5I = 3I = 5.5
I = 0
I = 12I = 8
I = 0
Calculating hazard footprints
90%
Commercial
building
9%
1%
I = 10
I = 0
 Which areas are affected in the admin zone?
 Is there any insured exposure?
 Calculating footprints – the intensity of shaking felt
Mw 7.6
Mw 7.2

Model development 2
Cumulative Distribution Function
(CDF)
EventID, X, Y , Intensity X,Y, Admin Levels, Weight
(Urban, Land Scan, no weight)
Event information Event Footprint
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.25 0.35 0.45 0.55
Hazard
CDF
EventID GeoID 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1 40006020 7.5 7.2 7.0 7 0 0 0 0 0 0 0
1 40005020 8.2 8.0 7.5 7.1 6.5 0 0 0 0 0 0

Model development 3
Flood extent (in this case PERILS)
Postal codes
Master table – postal code

Model development 4
 Open Data
 Free but different levels of freeness
– Personal, Educational, Commercial, etc.
 Use depends on model requirements
– detailed scenario model
vs. country wide probabilistic model
 Many sources of global free/low-cost
geospatial data
– check out http://freegisdata.rtwilson.com/ !
Source: Gray, 2011

Model development 5
 Exposure / hazard data
 Proprietary (not free)
– LandScan
– GfK GeoMarketing
– Tom Tom
– Ordnance Survey
 Open (free)
– SRTM – terrain
– Vs30 – earthquakes
– CORINE – floods
– Ordnance Survey Open Data
– OSM
Source: NASA (2013)
Source:
USGS (2013)

Visualisation 1
 ELEMENTS Explorer
– Visualise the background of the
catastrophe model (ELEMENTS)
– Map hazard, exposure, loss
– Built using GeoServer and Open Layers
– OSM inside
– Using WMS, SLD etc.
 For automated visualisation tasks
(e.g. viewing losses for 1,000s of portfolios)
– Open-source allows you to get into the nuts and bolts
– You can extend components (e.g. OpenLayers/GeoServer map printing)

Building Catastrophe Models using Open Data and Open Source

Visualisation 3
 (Some proprietary GIS tools we use)….
– ESRI ArcGIS
– Pitney Bowes Spectrum (Geocoding)
– Global Mapper
– Google Earth

Making models ‘open for business’ 1
 Catastrophe model software
– Generally proprietary and black-box
– Cannot see ‘underneath’ the model
 ELEMENTS is different:
– The components are open,
viewable and customisable
– hazard, vulnerability and exposure
can be modified
Source: opensourceway

 In ELEMENTS
– All model components are pre-calculated
– Change a model by changing a table / file
ELEMENTS Model components
Haz. Exp. Vuln. Portfolio

 ELEMENTS used within Aon Benfield and within insurance and reinsurance
companies……
 Potential other users could include…..
– Government agencies
– Emergency planners
– Emergency responders
– Humanitarian agencies and NGOs
– Commercial businesses
– Academic institutions
– ….the list is endless!
 Combining open and proprietary data and software could be the most
effective solution

Summary
 Catastrophe models allow you to better understand your risk
 The components can be built using Open Data and Open Source
– Open Data sources are especially useful for country-wide probabilistic
models where proprietary data may be too expensive
– Open Source can help speed up big data tasks
 Visualisation is key to understanding model inputs and outputs
– Open Source tools have helped with this
– Open Standards have helped with map styling etc.
 Catastrophe models are used in insurance but could be applied to many
different sectors and industries (emergency planning etc.)
– We’re open for business!

Thank you
Chris Ewing
Impact Forecasting
Aon Benfield

Building Catastrophe Models using Open Data and Open Source

More Related Content

What's hot (19)

Similar to Building Catastrophe Models using Open Data and Open Source (20)

Recently uploaded (20)

Building Catastrophe Models using Open Data and Open Source

Editor's Notes