Slide 1: Oasis Conference 2023
Oasis Conference 2023
Slide 2: Latest on Oasis LMF
Ben Hayes Chief Technology Officer, Oasis Loss Modelling Framework Latest on Oasis LMF Dickie Whitaker Chief Executive, Oasis Loss Modelling Framework
Slide 3: Latest On Oasis LMF
Latest On Oasis LMF Dickie Whitaker, CEO Oasis LMF Ben Hayes, CTO Oasis LMF 6th September 2023 3
Slide 4: Topics
Topics • Update on Oasis Software • Features • Performance • Reliability • Deployment Options • Oasis Family of Software • Risk Explorer • Forcasting • ++ • Open Data Standards • Oasis Projects GXM • Models in Oasis • Oasis, the future and innovation 4
Slide 5: Update on Oasis Software – Features
Update on Oasis Software – Features (BH) • Full OED Validation • Financial Module Completeness • Memory usage optimisation • Stochastic Disaggregation • Post Loss Amplification • Geocoding Plugin • Full Documentation rewrite All driven by Oasis technical steering group 5
Slide 6: Update on Oasis Software: Performance - Standard Models
Update on Oasis Software Performance - Standard Models • Benchmark exercise • 258,518 location real portfolio • 4 production models • 2 standard Oasis Models • 2 Complex Model implementations • GUL, Gross and Reinsurance ELTs • Sampled losses • Comparisons • V1.15 (May 2021) vs. V1.27 (Apr 2023) • S
Slide 7: Update on Oasis Software: Performance - Complex Models
Update on Oasis Software Performance - Complex Models • Complex models don’t use Oasis standard data formats. • Less pronounced improvements seen. • More model-specific work required to optimise performance. • CL USEQ • model runs limited to ~110 chunks. • JBA USFL • Significant improvement in execu
Slide 8: Update on Oasis Software – Reliability
Update on Oasis Software – Reliability (BH) Key metrics from Guy Carpenter Oasis Usage 2022 2,613 • Number of analyses completed 60 • Number of users • Average number of locations 14 million modelled per week 96% • 1st time success rate 8
Slide 9: Update on Oasis Software – Reliability: Further Improvements
Update on Oasis Software – Reliability (BH Further Improvements… Majority of the 4% New tool built to Oasis service to errors due to quickly build Oasis host commercial incorrectly coded environments models and OED files automatically test against new Integration OED Validation Camel Continuous Conf
Slide 10: Deployment Options
Deployment Options (BH) Laptop Cloud On-Prem Hosted 10
Slide 11: Oasis Family of Software
Oasis Family of Software Oasis Forecasting 11
Slide 12: ODS – Steering Committee
ODS – Steering Committee • The ODS SC meet 3-4 times a year and is chaired by Oasis LMF. The following companies are represented at the SC. • Governed using Github semantics for major/minor updates and patches. ‘Day-to-Day’ implementations and issues raised on Github Repo. • Starting to put together
Slide 13: Oasis Model Coverage
Oasis Model Coverage Over 100 models from various providers * * * JBA – ‘Global flood model’ available on the Oasis platform (except Antarctica) CYBER CoreLogic – Committed in making all models available on the Oasis platform 13
Slide 14: Innovation
Innovation * Global Exposure model / IED * Validation * Models airmic BIG TICKET LAUNCHES FIRST CLIENT-FACING ‘OPEN BANKING’ INFRASTRUCTURE FOR INSURANCE Big Ticket enables Powered by ‘API first’ connecting the entire commercial insurance value chain. YO
Slide 15: Using Oasis in Practice
Simos Koumoutsaris Global Head of Cat Research & Development, SCOR Using Oasis in Practice Paul McEwan Risk Analytics Director, Sompo International
Slide 16: Using Oasis in Practice
Using Oasis in Practice Simos Koumoutsaris Global Head of Cat R&D 06/09/2023 Oasis Insight Conference 2023, Zurich
Slide 17: Outline
Outline • OasisLMF is a mature product now that is being used in an operational setting. • Examples of how Oasis is used in insurance: Guy Carpenter, SCOR • The «Nazare» project • Some key points to consider for adopting Oasis • ODS
Slide 18: Oasis at SCOR
Oasis at SCOR • Operationally effective multi-vendor modelling approach developed over the last decade. • Gives us access to more models and provides a more comprehensive View-of-Risk including a measure of uncertainty and helps to build confidence internally and with clients. • OasisLMF is key for
Slide 19: Oasis at SCOR
Oasis at SCOR • Currently, we host a small internal OasisLMF environment. • Operationally embedded in our pricing and accumulation workflows, including a high level of systems integration. • Looking at implementing the new Oasis distributed platform in-house. Vendor Loss Accumulation platforms Tool
Slide 20: Oasis at GC
Oasis at GC • Need for an efficient platform where a broad range of in-house produced and 3rd party models could be hosted and run. • Externally hosted environment, +50 users, +100 analyses per week AWS GC models Risk Analytics NRMC (GC VoR) 3rd party models Vendor Loss platforms
Slide 21: The «Nazare» project
The «Nazare» project • Financial and resource support in Q2 2021 from Oasis members in order to accelerate the development of the Oasis platform. • Technical Steering Group (TSG) was responsible to steer Oasis developments in two directions: 1. Oasis Enterprise platform Worker 1 • Distributed enviro
Slide 22: The «Nazare» project
The «Nazare» project • Financial and resource support in Q2 2021 from Oasis members in order to accelerate the development of the Oasis platform. • Technical Steering Group (TSG) was responsible to steer Oasis developments in two directions: 2. Core Oasis Development • Prioritization of developments
Slide 23: Which deployment option to adopt?
Which deployment option to adopt? • Many options available for adopting Oasis: internal vs. external, stand-alone vs. platform • Different needs depending on number of users, number of models, license and/or build models, internal expertise, cost, etc. • Additional key considerations: • Alignment re
Slide 24: How many samples should I run?
How many samples should I run? • Guidance is not often provided by model 60 developers. Portfolio with 100K risks • Depends on the portfolio, the peril-region, the 50 model, the output levels, etc. 40 • Increasing number of samples increases % Error 30 compute and memory requirements, i.e., closely
Slide 25: Efficiency is key
Efficiency is key • Open Data Standards brings efficiency: • Easy sharing of data, • Using multiple models, • Switching between vendors • Provides opportunities to build tools around it (e.g., visualization tools, tools for model evaluation and model calibration, automated documentation, and others)
Slide 26: Slide 26
Slide 27: Model Vendor Updates
William Forde Senior Director of Industry Solutions, CoreLogic Dr. Lauren Mudd Dr. Joshua Macabuag Model Vendor Updates Senior Engineer, Co-Founder and ARA Chief Product Officer, Renew Risk Chris Ewing Harry Vardigans Head of Client Business Development Management, Impact Manager, Fathom Forecasting
Slide 28: Dr. Lauren Mudd
Dr. Lauren Mudd Senior Engineer, ARA
Slide 29: ARA HurLoss Hurricane Catastrophe Modeling
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS Applied Research Associates, Inc. (ARA) HurLoss Hurricane Catastrophe Modeling OASIS Insight Conference Zurich September 6, 2023 © 2023 Applied Research Associates, Inc. • ARA Proprietary NATIONAL SECURITY INFRASTRUCTURE ENERGY & ENVIRONMENT HEALTH SOLUTIONS
Slide 30: Who We Are
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS Who We Are ▪ ARA is a research and engineering company dedicated to solving critical problems to improve our safety and security ▪ We provide an independent view of hurricane risk with many applications, both within and beyond insurance ▪ First Principles – H
Slide 31: ARA's Explicit Modeling
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS ARA’s Explicit Modeling • Hazard model provides a basis for structural design (buildings, bridges, offshore wind, solar), not just cat modeling • Damage based on 3-D computer modeling – including modeling of individual building envelope components and the eff
Slide 32: ARA – JBA Joint Wind and Flood Model
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS ARA – JBA Joint Wind and Flood Model Tropical cyclone rainfall Simulation of catchment response Data Granularity © 2023 Applied Research Associates, Inc. • ARA Proprietary www.ara.com
Slide 33: ARA Introduces Climate Change Effects
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS ARA Introduces Climate Change Effects • Climate change: Direct link between global circulation model outputs and ARA hurricane hazard model inputs © 2023 Applied Research Associates, Inc. • ARA Proprietary www.ara.com
Slide 34: ARA Pioneering Climate Methodology
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS Storm Parameters ARA Pioneering Climate Methodology at Date & Time i Sea Surface Temp. Couple global circulation model environmental Tropopause Temp. data outputs with hurricane track and wind field Wind Shear models Relative Intensity • Model is able to quan
Slide 35: Transparent and Easy to Use
INNOVATIVE SOLUTIONS TO COMPLEX PROBLEMS Transparent and Easy to Use • Most extensively peer-reviewed hurricane cat model in both open literature and by government agencies and private institutions • Available on Oasis • Open platform with standardized workflow and financial module • Open Data Stand
Slide 36: Harry Vardigans
Harry Vardigans Business Development Manager – Insurance, Fathom
Slide 37: Water & climate risk intelligence
fathom.global Water & climate © Fathom risk intelligence
Slide 38: Our values
fathom.global Our values We are open and We make a Science is in We are transparent difference our DNA dynamic © Fathom
Slide 39: Our research output
fathom.global Our research output 10 9 79620 ? pieces of published papers on track combined ongoing industry accelerator research in 2022 for publication in 2023 research citations of projects: our team's papers SWOT, FCDO, CReDo, Met Office etc. *accurate as of August 2023 © Fathom
Slide 40: How is this possible?
fathom.global How is this possible? — Our team primarily consists of scientists, academics and developers — We're beginning to build out our commercial function — We are getting better at finding new ways of connecting research to industry © Fathom
Slide 41: Our products
fathom.global Our products Flood Maps Risk Scores Catastrophe Models Global Terrain Data *Available via the Fathom API, Fathom Portal, On premise or through a third party platform © Fathom
Slide 42: Communicating the complex
fathom.global Communicating the complex Fathom's Product Methodology Metadata Customer support Stack docs © Fathom
Slide 43: What next?
fathom.global What next? — Focus on closing any gaps in our insurance offering — Working collaboratively with the industry to build relevant and useful products © Fathom
Slide 44: Chris Ewing
Chris Ewing Head of Client Management, Impact Forecasting
Slide 45: Impact Forecasting
Impact Forecasting Oasis Insight Conference Zurich September 2023 Impact Forecasting Proprietary & Confidential
Slide 46: What do Impact Forecasting offer?
What do Impact Forecasting offer? Peak-Peril Custom Data for Models Models Underwriting Climate Event Change Response Solutions
Slide 47: Impact Forecasting
Impact Forecasting Global catastrophe model development team
Slide 48: Supporting the Open Cat Modelling community
Supporting the Open Cat Modelling community Aon and Impact Forecasting support through…. Hazard and Open Exposure Data Using Oasis loss Community Vulnerability model (OED) and Open calculation engine Involvement formats Results Data (ORD) • ELEMENTS • Interoperability • Exposure schemes • Working gr
Slide 49: Are you concerned about peak peril catastrophe risk?
Are you concerned about peak peril catastrophe risk? Explore Impact Forecasting’s peak peril models • US hurricane model with Florida recently certified • US earthquake based on the latest USGS seismic hazard updates • Europe Windstorm model covering 22 countries • Japan typhoon model covering wind,
Slide 50: Does low resolution data restrict your ability to price risk?
Does low resolution data restrict your ability to price risk? Access Impact Forecasting’s high resolution underwriting data Hazard and Integration Better risk insights Decisions Hazard and risk Access through For risk selection maps for flood, your systems or and pricing wind, convective Impact stra
Slide 51: Can you quantify climate change impacts on physical risks?
Can you quantify climate change impacts on physical risks? Model climate scenarios to price risk today and tomorrow Focused Solutions Full suite of platforms and services • Focus on peak territories, physically based studies • Development findings directly incorporated to our platforms • Robust appr
Slide 52: Are you lacking relevant, up-to-date info on events?
Are you lacking relevant, up-to-date info on events? Impact Forecasting’s Catastrophe Insight can help Global Coverage Available for US Hurricane, Euro Windstorm and Japan Typhoon Available for modelled perils Catastrophe Reporting AER Loss Forecasting AER Post-Event Analysis Event Response Package
Slide 53: Thank You
Thank You chris.ewing@aon.com www.impactforecasting.com
Slide 54: Dr. Joshua Macabuag
Dr. Joshua Macabuag Co-Founder and Chief Product Officer, Renew Risk
Slide 55: Risk Analytics and Modelling for Renewable Energy
Risk Analytics and Modelling for Renewable Energy Ashima Gupta Chief Executive Officer Dr Joshua Macabuag Chief Product Officer Prof Subhamoy Bhattacharya Chief Scientist Gaurav Chawla Exec Chair
Slide 56: Renew Risk provides risk analytics for renewable energy
“ “Renew Risk provides world-class risk analytics and modelling solutions for renewable energy asset finance and insurance starting from offshore wind farms” 56
Slide 57: Why Renewables?
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING WHY RENEWABLES? Exponential growth of renewable energy needs financial risk solutions Problem Solution World's first deep-science SaaS for “Capacity Crunch” financial life-cycle of Offshore projects Catastrophe Risk Modelling 1 ✓ For (Re)Insurers & B
Slide 58: The Gap
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING THE GAP Models needed for regions of peak exposure growth • New assets in NatCat regions for first time • No commercially-available Offshore-Wind-Specific Cat Models models currently ➢ US Hurricane available now ➢ In development: ➢ Taiwan Earthquake
Slide 59: Industry Exposure Database
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING INDUSTRY EXPOSURE DATABASE Model unique offshore assets • Turbines, Cables, Substations • All assets, now and scheduled ➢ Enables modelling of risks with minimal information ➢ Significantly saves modelling time 59
Slide 60: Offshore Hazard
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING OFFSHORE HAZARD Unique Offshore Condition • Hurricane – Wind, Wave & Current • Earthquake – Shaking, Landslide, Liquefaction • Independent Validation • Wind and sea modelled with global leader: 60
Slide 61: Vulnerability Modelling
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING VULNERABILITY MODELLING Limited claims exist, so Engineering Methods Turbines, Blades, M&E, Cables, Substations ➢ The best in Engineering ➢ Additional Offerings: ➢ Expert support for OWF Specialty risks ➢ Design/construction reviews and other inputs
Slide 62: Model Deployment
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING MODEL DEPLOYMENT Accessible via Nasdaq Risk Modelling for Catastrophes • Standardized Loss Metrics • Available on Nasdaq and AWS 62
Slide 63: The Team
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING THE TEAM Skillsets: Risk Modelling, Engineering, Insurance, Fintech Software, Sales and Entrepreneurship. ASHIMA GUPTA PROF SUBY BHATTACHARYA DR JOSH MACABUAG GAURAV CHAWLA CEO Chief Scientist Chief Product Officer Exec Chair Product and client Profe
Slide 64: Clients, Partners and Investors
RENEW RISK | IMPROVING RENEWABLE ENERGY FINANCING CLIENTS, PARTNERS AND INVESTORS Going to market with robust ecosystem of insurance companies, academia, partners and investors Clients Investors Angel Investment Partners 64
Slide 65: Contact Us
Improving Renewable Energy Financing CONTACT US Mrs Ashima Gupta Dr Joshua Macabuag Mr Gaurav Chawla Prof Subhamoy Bhattacharya Chief Executive Officer Chief Product Officer Exec Chair Chief Scientist Ashima.Gupta@renew-risk.com Joshua.Macabuag@renew-risk.com Gaurav.Chawla@renew-risk.com Suby@renew-
Slide 66: Open Source Assets in Swiss Re: A Glass Half Full or Half Empty?
Open Source Assets in Martin Bertogg Swiss Re: A Glass Half Head Cat Perils, Swiss Re Full or Half Empty?
Slide 67: Covering the Globe with nat cat modelling capabilities
Covering the Globe with nat cat modelling capabilities: A need to find new ways of procurement of risk insights Swiss Re’s global nat cat book1 Powerful resources Illustrative ~200 models, covering >90% of global insured exposures ~30 secondary perils’ risk views added or enhanced since 2019 Europea
Slide 68: 2023 - a very full glass for Swiss Re in Nat Cat Modelling
2023 - a very full glass for Swiss Re in Nat Cat Modelling Technical Earthquake Flood Steering Group: China: modelling Industry GEM & OASIS capabilities: learning on as game enabling co- neutral changers development grounds Martin Bertogg | OASIS conference Sept. 2023 | Catastrophe Perils, Swiss Re
Slide 69: Accelerated learning and better outcomes at a lower price point
Accelerated learning and better outcomes at a lower price point... -> Give it a try! Martin Bertogg | OASIS conference Sept. 2023 | Catastrophe Perils, Swiss Re
Slide 70: New Ways to look at Flood Risk at the Location Level
New Ways to look at David Schenkel Flood Risk at the Head of Product, Location Level REOR20
Slide 71: AI is the new flood defence
REOR in Latin means think and compute %20 is the URL encoding symbol for the white space AI is the new flood defence David Schenkel Head of Product david@reor20.com Zurich, 11.09.2023 18May 2022 REOR20 All rights Reserved
Slide 72: 3 major issues with flood risk models
3 major issues with flood risk models to understand risk at the location level Zurich, Zurich,411.09.2023 May 2022 REOR20 All rights Reserved
Slide 73: Current Models
1 Zurich, 11.09.2023 Current Models REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 74: How it should be: Calculation with buildings
1 Zurich, 11.09.2023 How it should be: Calculation with buildings REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 75: Maximum depth
1 Maximum depth 2.00m x with buildings 1.00m 0.25m x without buildings Zurich, 11.09.2023 REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 76: Average depth over time around the building
2 Average depth over time around the building 2.00m x with buildings 1.00m 0.25m x without buildings Zurich, 11.09.2023 Time REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 77: same depth same risk?
3 1 1 same depth same risk? 2 2 Zurich, 11.09.2023 REOR20 All rights Reserved
Slide 78: Zurich, 11.09.2023 REOR20 All rights Reserved
3 Zurich, 11.09.2023 REOR20 All rights Reserved
Slide 79: High Fidelity Modelling: Expensive, tedious - and nearly impossible to scale
$ x High Fidelity Modelling: Expensive, tedious - and nearly impossible to scale Zurich, Zurich,11.09.2023 11.09.2023 t REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 80: We taught an AI the physics of flood
$ x We taught an AI the physics of flood making it +100’000s times faster & simpler Zurich, Zurich,11.09.2023 11.09.2023 t REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 81: Where we stand… … where we go
Visual risk exploration for Deepdives Provide rich set of data on the risk in Where we stand… the vicinity of the property to better assess and de-risk … where we go Location & Property API returns property-specific Rule-based Acceptance REOR20 risk scores Flood Policy Characteristics and Pricing Zu
Slide 82: Implications
Implications - From point-based to property-based lookups - Need for up-to date building & defences datasets - Aggregation and disaggregation per building - Re-calculation of risk for new urban developments Zurich, 11.09.2023 REOR20 All rights Reserved
Slide 83: Thank you!
Thank you! David Head of Product david@reor20.com Zurich, 11.09.2023 REOR20 REOR20 AllAllrights rightsReserved Reserved
Slide 84: Handling Complexity in Flood Models: A Close Look at Hazard and Exposure Data
Oliver Wing Petr Puncochar Chief Research Global Head of Flood Officer, Fathom Model Development, Impact Forecasting Handling Complexity in Flood Models: A Close Look at Hazard and Exposure Data Federica Remondi Jane Toothill Managing Director Flood Lead, JBA Risk Swiss Re Management
Slide 85: Handling Complexity in Flood Models: are we tackling it all?
Handling Complexity in Flood Models: are we tackling it all? Federica Remondi, Oasis Conference, 06.09.2023
Slide 86: Current and future flood models face challenges
Current and future flood models face challenges that require a robust while flexible approach Robust, granular and continental- Urbanization scale hydrodynamic model Simulate urban flash flooding in light of High-performance hydrodynamic model that increasing population densities and built-up areas
Slide 87: Increase in built-up areas is a global phenomenon
The increase in built-up areas is a global phenomena also seen in regions exposed to flood risk Federica Remondi | Oasis Conference 06.09.2023
Slide 88: Growth rate of built-up areas in flood-exposed regions
The growth rate of built-up areas in regions highly exposed to flooding is rarely lower than the one in the least exposed regions Federica Remondi | Oasis Conference 06.09.2023
Slide 89: Population growth reflects built-up trends
Population growth is similarly reflecting built-up trends, showing no major correlation to flooding risk Federica Remondi | Oasis Conference 06.09.2023
Slide 90: Sharing flood exposure data at high resolution is not yet the norm
Sharing flood exposure data at high resolution is not a yet the norm also in the re/insurance key markets, increasing uncertainty in the model outcomes % TIV received in treaty submissions at address/exact level* Which assumptions do 80% 20-40% we take? How can we reduce uncertainty? USA, Germany &
Slide 91: Can flood models be used equally from portfolio accumulation to tariffs analysis?
Can flood models be used equally from portfolio accumulation down to tariffs analysis ? Federica Remondi | Oasis Conference 06.09.2023
Slide 92: Model Validation: Are There Opportunities to Improve?
Dickie Whitaker Model Validation: Are Chief Executive, There Opportunities to Oasis LMF Improve How This is Done and the Data Nigel Winspear Malcolm Haylock That’s Used? Head of Catastrophe Head of Natural Catastrophe Modelling, Fathom Analytics Research, Sompo International Oriol Gaspa Rebull Guido
Slide 93: Dr. Nigel Winspear
Dr. Nigel Winspear Head of Natural Catastrophe Analytics Research, Sompo International
Slide 94: Catastrophe Model Validation
Catastrophe Model Validation What and How? Sept 6, 2023 Oasis Zurich Dr. N. Winspear V1.1
Slide 95: Agenda
Agenda 1. Core questions to be addressed 2. Emphasis on both questions 3. Tailor effort to match [the materiality of the peril] 4. Designing the content in advance 5. Remember what it is not 95
Slide 96: Model validation – resolves to 2 core questions
Model validation – resolves to 2 core questions Validation is … the process by which you determine whether the external catastrophe model provides a valid representation of the catastrophe risk for your portfolio [LMA (2012) - Validating external catastrophe models under Solvency II] This resolves t
Slide 97: Tailor effort to match
Tailor effort to match Tailor the work to the materiality of the peril Materiality can be assessed at a high level and then mapped to different categories of model validation study OEP 1:100 VaR % of Gross Materiality Validation Questions (USD Millions) - Written Premium Category Category Addressed
Slide 98: Plan the content in advance
Plan the content in advance Validation Validation Questions Example of Content Category Module Addressed Comprehensive/ Loss Q1, Q2 - Industry back-testing, scenario testing, historic loss comparison (modelled vs. actual). intermediate validation - Own portfolio back-testing, scenario testing, histo
Slide 99: Remember what it is not
Remember what it is not 1. Development of an ‘own view of risk’ 1. Often assumed that model validation will result in a view of risk 2. Model validation lays the foundation for subsequent development of a view of risk 3. View of risk development needs input from a much wider panel of experts – actua
Slide 100: Oriol Gaspa Rebull
Oriol Gaspa Rebull Head of UK Property Analytics, Aon Reinsurance
Slide 101: Oasis Event
Oasis Event Oriol Gaspa Introduction September 2023 Proprietary & Confidential
Slide 102: Aon's perspective on Model Evaluation Needs
Aon’s perspective on Model Evaluation Needs Aon continually evaluates a large range of models on behalf or for clients The risk landscape has become increasingly complex and has led to a proliferation of models; by peril and country globally. Climate change is creating further uncertainty and is a k
Slide 103: Aon's ecosystem of Model Validation
Aon’s ecosystem of Model Validation Large global Super regionals & Regionals Global Re Specialty Casualty, Cyber & Motor Validation of model changes Evaluation and opinion of underlying methodologies Analysis on portfolio model accuracy Validation of actuarial profiling Benchmarking of underlying ex
Slide 104: Guido Felder
Guido Felder Lead Catastrophe Researcher, Zurich Insurance
Slide 105: Simplifying model validation through transparency and standardization
Simplifying model validation through transparency and standardization Oasis Conference 2023, 6 September 2023 Guido Felder Lead Cat Researcher guido.felder@zurich.com © Zurich
Slide 106: Two perspectives on model validation
Two perspectives on model validation Scientific perspective Business perspective Cross-peril correlation Does the model reflect scientific Is the model appropriate for the state of the art? Global consistency business needs? - Publications Secondary perils Consistency with observed data - Claims his
Slide 107: Example for the need for transparency
Example for the need for transparency Question Natural variability Long term average Does the hazard module reflect current conditions? Points to consider Vintage of data set(s) used for event set generation - Varying input time series - Irregular update cycles - On-levelling to current conditions U
Slide 108: Simplifying model validation through transparency and standardization
Simplifying model validation through transparency and standardization Insurer view Model provider view • Internal validation will remain • Exposed to market conditions • Need for transparency on models (incl. limitations) • Low motivation for transparency around limitations and deficiencies • Limite
Slide 109: Malcolm Haylock
Malcolm Haylock InsuranceHead of Catastrophe Modelling, Fathom
Slide 110: Model Validation - Meeting in the Middle
fathom.global Model Validation - Meeting in the Middle © Fathom Oasis Conference, Zurich, 6th September, 2023
Slide 111: Is there a common understanding of what matters?
fathom.global Is there a common understanding of what matters? North Atlantic Basin TC Statistics TC Box cover for Florida (50km radius) Bloemendaal, N. et al (2020). Generation of a global synthetic tropical cyclone hazard dataset using STORM. Scientific Data, 7(1), 1–12. https://doi.org/10.1038/s4
Slide 112: Why the difference?
fathom.global Why the difference? Historical Stochastic Kernel density exceedance for cat-2 and higher (obs/yr/deg2/kn)
Slide 113: Developers and users need to validate models, but can we do this more efficiently?
Developers and users need to validate models, but can we fathom.global do this more efficiently? Event Local Hazard Exposure Vulnerability Risk Developer User
Slide 114: How Do We Get There?
fathom.global How Do We Get There? • What users need from developers • Open channels of communication. • Model documentation that is • Scientific – clear, concise, objective and reproduceable. • Covers key uncertainties and sensitivities. • Key data exhibits in order to interrogate the risk drivers
Slide 115: sigma 1/2022
sigma 1/2022 | 18 May 2022
Slide 116: Legal notice
Legal notice ©2023 Swiss Re. All rights reserved. You may use this presentation for private or internal purposes but note that any copyright or other proprietary notices must not be removed. You are not permitted to create any modifications or derivative works of this presentation, or to use it for