Oasis LMF - Loading
menu

Oasis Conference - The good, the bad and the ugly

05/09/2017 - 06/09/2017

Answering the questions you have always wanted to know


This not-for-profit conference will give you deeper insights into cat models than anywhere else, and is driven by our members.

Spaces are limited, so please sign up as soon as possible to avoid being disappointed.

 

Agenda



Day 1


8.30 
Registration / Welcome coffee

9.00
Opening Address: Overview of the Oasis community today and future developments
Dickie Whitaker, CEO, OasisLMF 

9.30
Getting the most out of Oasis membership
How different members are using Oasis and what benefits they are getting

  • Integrating Oasis into internal systems
  • Experiences and benefits of shared services
  • Building and commissioning your own models

Session Moderator: Paul Nunn, Head of Catastrophe Risk Modelling, SCOR
Jakir Patel, Head of Catastrophe Systems, SCOR
Simon Blaquière, Head of ‎Reinsurance Actuarial and Cat Modelling, AXA Global P&C
Beat Aeberhardt, Head of NatCat tools, Swiss Re
Marc Wüest, Expert Atmospheric Perils, Swiss Re

10.30
Networking / coffee break 

11.00
Morning Breakouts 

Workshop 1: The Four dimensions of correlation in catastrophe models 
Dag Lohmann, Co-Founder & CEO, KatRisk LLC

This session is about the process of building catastrophe models and focuses on the role that temporal, spatial and uncertainty correlations play. I will specifically talk about:
1) Correlation between continents and the relationship to climate models and teleconnections
2) Correlation within a region (e.g. continent) across multiple perils (e.g. wind, storm surge and inland flood) and their representation in cat models
3) Correlation within a single peril model (e.g. temporal correlation when modeling floods)
4) Correlation of uncertainty within a single peril model, e.g. through modeling severity uncertainty correlation
I will make the case that catastrophe models can be abstracted into frequency, severity, and correlation models. This can be linked to the underwriting and porfolio management processes, where we can separate these components. Examples will largely come from the probabilistic KatRisk North America hurricane, storm surge and inland flood models. The session will be followed by Q&A and discussion.


Workshop 2: Flood models: approaches, uncertainties and challenges
Andrew Smith, COO, SSBN 
Chris Sampson, CTO, SSBN
Paul Bates, Chairman and co-founder, SSBN
Emma Raven, Principal Analyst, JBA Risk Management

Over the last 5 years, high resolution flood models have changed from being the preserve of local-scale engineering projects to being available for the types of national and global scale analyses that better serve the insurance industry. This represents a paradigm shift for the field, and has necessitated rapid methodological advances that replace field measurements and manual model construction with big data and computerised automation.
A number of commercial and academic teams around the world are developing large scale flood models, and most currently fall into one of two methodological categories: those based on data from river flow gauges, and those based on data from rainfall-sensing satellites. The differences between the two approaches are important for end users to understand, and this breakout workshop aims to provide a high-level introduction to both before summarising the latest relevant academic research and discussing some of the key challenges faced by developers looking to create probabilistic model frameworks using these models. We encourage attendees to bring questions, feedback and ideas that may help model developers continue to improve their models.


Workshop 3: Vulnerability 
Brad Eccles, Director of Operations, ABS Group
Natural Hazard Vulnerability Curve Generation and Calibration - Opening the Blackbox
This session will summarise the engineering considerations for generating “Raw” natural hazard vulnerability curves for typical facility types. It will set out how engineers, using their knowledge of structures, contents and equipment, materials of construction etc, link physical vulnerability to damage estimates and repair times. However, this approach can significantly over predict the losses from claims data, and so through an interactive workshop, the factors to consider during the model “calibration” process will be investigated. This session is designed to give a tantalizing glimpse inside the vulnerability Blackbox.

Simon Blaquière, ‎Head of Reinsurance Actuarial and Cat Modelling, AXA Global P&C
Leveraging internal claims data to enhance vulnerability curves
AXA pays more than a 1bn€ of natural catastrophes losses every year hence generating massive volume of data and experience. Given these information, specific analysis combining data science, engineering and physics are performed to develop in-house fragility curves.

This session will summarize technics used at AXA to build internal natural hazard models with a dedicated focus on the vulnerability module. Specific examples will be discussed regarding flood, earthquake and windstorm perils and it will be highlighted how internal expertise could improve risk analysis.

12.30 
Lunch 

13.30
Model validation using loss data
Rick Thomas, Managing Director, Beach
Many commercial models are not consistent with historical loss information sourced from both public and confidential sources. We will use a realistic case study to examine methods of demonstrating that the model is wrong at certain . Historically model adjustment was the obvious solution, however with OASIS it may be possible to examine the sources of over or underestimation transparently.

Panel discussion
Moderator: Claire Souch, Director AWHA Consulting
Will Gardner, Managing Director, COMBUS
Rick Thomas, Managing Director, Beach
Paolo Bazzurro, Professor of Hazard and Risk Assessment, University Institute for Superior Studies (IUSS) Pavia 

14.30
Networking / coffee break

14.45
Afternoon Breakouts

Workshop 4: Using climate models to answer the industry's most pressing questions
Richard Dixon, Director, CatInsight
Recently several industry meteorologists prioritised their most important areas of required understanding in European Windstorm. This talk discusses these priorities and how state-of-the-art climate modelling techniques can potentially answer some of the most pressing questions industry would like to have answered.

Len Shaffrey, Professor of Climate Science, University of Reading
An overview of using climate models to estimate hydro-meteorological hazards. Outline of current issues and future opportunities.

Vikki Thompson, Scientist, Met Office
I present a new methodology which has enabled us to assess the risk of unprecedented rainfall in the UK – the UNSEEN method (UNprecedented Simulated Extremes in ENsembles). An analysis of a large ensemble of high-resolution initialised climate simulations show that in the current climate there remains a high chance of exceeding the observed record monthly rainfall totals in many regions of the UK. In south east England there is a 7% chance of exceeding the current rainfall record in at least one month in any given winter. There is scope to apply this method to other regions and climatic variables.

Tom Philp,‎ Science Analyst, XL Catlin
North Atlantic Hurricane (NAHU) modelling is a research field that is advancing rapidly. Coupled with an academic research community that is increasingly embedded within the backbone of the global re/insurance industry, our views of NAHU risk are evolving in many ways. This talk will summarise the latest updates in scientific NAHU research, and will highlight areas that views of NAHU risk may, as a direct consequence, change in the near future.

Workshop 5: Model Evaluation
How can we evaluate Flood Models?
Jessica Turner, Senior Manager, Weather Modelling, Lloyds Banking Group 
The limited historic record along with rapid changes in the built environment over time create challenges for evaluating flood models. Increasing building on flood plains, changes in water management and flood defence schemes can render what historic data exists obsolete. Numerical weather and climate models can be used to extend the record but they typically stop at precipitation modelling, which does not always translate directly to the flood risk. Impacts of climate change in the short-term are uncertain. Footprints of past historic events rarely exist to associate flood depths with losses. Given these challenges, this talk will outline some approaches to the problem of flood model evaluation that can be taken from the perspective of a primary insurer.

Windstorm Risk - Probabilistic, Deterministic and Real-Time Exploration
Ioana Dima-West, Head of Model Research and Evaluation, Willis Re
The presentation will focus on ways of evaluating probabilistic cat models through the application of scientific research. In addition, I will discuss the use of historical storm data and scenario-based analyses, as well as offer a brief description of our Event Response Service.

Earthquake Risk 
John Alarcon, Executive Director, Willis Re

Workshop 6: Loss Impact from Financial Models
Imelda Powers, Global Chief CAT Modeller and Managing Director, Guy Carpenter
Study the impact of losses to individual risks and portfolios arising from different financial models. Considerations will be given to secondary uncertainty, risk correlation, and treatment of conflicting policy terms.

16.15
Wrap-up Day 1
Claire Souch, Director, AWHA Consulting
Dickie Whitaker, CEO, Oasis LMF


16.30
Networking drinks and food


 

Day 2


8.30 
Oasis AGM / Members ONLY 
 

Conference Start
9.30

The IDF Risk Modelling and Mapping Group: Goals, strategy and progress update
Ian Branagan, Group Chief Risk Officer and Senior Vice President, RenaissanceRe Holdings Ltd. and Co-Chair of IDF RMMG

9.45 
Augusto Hidalgo, President & CEO, National Reinsurance Corporation of the Philippines
OASIS Platform - Impact in Philippine Government Strategy and Insurance Sector Participation in Disaster Risk Management
The Philippines has been chosen as one of the pilot countries for OASIS' Platform for Climate and Catastrophe Risk Assessment in Asia. Under this project OASIS will work with various partners in developing a catastrophe model for flood in the country. What needs of this natural disaster-stricken country will this game-changing initiative address? How will this first-ever catastrophe model for flood affect Philippine government's disaster risk management strategies and the insurance industry's role in boosting Filipinos' resilience to disasters?

10.15
Model uncertainty and sensitivity testing - approaches and tools available today and future needs and capabilities
Guillermo Franco, Managing Director & Global Head of CAT Risk Research, Guy Carpenter

  • How to test model sensitivity robustly
  • Real examples of model differences for the same portfolio
  • What is needed to do this better
  • How to use in decision making
  • How will Oasis help in the future

Panel discussion
Moderator: Matthew Jones, Director, Cat Risk Intelligence
Adam Podlaha, ‎CEO of Impact Forecasting, AonBenfield
Matt Foote, ‎Group Head of Exposure Management, Argo Global
Guillermo Franco, Managing Director & Global Head of CAT Risk Research, Guy Carpenter

11.15
Networking / coffee break 

11.45
Morning Breakouts 

Workshop 7: Earthquake Modelling

Mohammad Zolfaghari, ‎Director, CatRisk Solutions 
Challenges ahead to provide more realistic earthquake cat risk models
Use of catastrophe loss modelling to assess losses from future potential natural hazards such as earthquakes have been well established in many decision making and risk management sections. Such demands in conjunction with our understanding of natural hazard and their effect on built environment as well as fast computational power have paved the way for significant improvement in cat risk modelling in recent decades. However, uncertainties associated with results from such models are still of big concern to many end users. It is therefore, important to define the challenges ahead and to address areas which further investment should be made in order to achieve better and more reliable cat risk models.

Paolo Bazzurro, Professor of Hazard and Risk Assessment, University Institute for Superior Studies (IUSS) Pavia Learning from Site-Specific seismic Risk Assessment to Improve Portfolio Risk Estimation
The most critical and volatile step in assessing portfolio losses due to earthquakes is the definition of vulnerability of classes of like-buildings. In most models the vulnerability functions that relate ground motion intensity to repair losses cannot be based solely on loss observations due to the scarcity of data. Hence, it is customary to supplement the development of these relationships with dynamic response analyses of computer models of buildings subject to ground motion records of different severity. The Performance-based Earthquake Engineering (PBEE) framework is a well-established methodology to do so but it was conceived mostly for assessing collapse probabilities of for single structures. The objectives of portfolio loss estimations are different and some adjustments are needed to the PBEE framework to avoid the introduction of biases and inaccuracies. This presentation starts addressing these issues and provides a proposal for decreasing the potential for misrepresentation of portfolio risk.


Workshop 8: Oasis in practice - detailed case studies 

Mark Pinkerton, CTO, Oasis LMF
Oasis technology in context
The IT landscape has changed dramatically in the last five years. There are many innovations from data science and cloud-scale computing that are highly applicable to exposure management, and many are in the Open Source domain. This session will discuss some of the key technology concepts, put Oasis technology in context, and present some thoughts on taking control of your exposure management technology strategy.

Jonathan Meagher, Senior Analyst, Allianz Re
Oasis experience at Allianz
Work will be presented on integrating ktools, the Oasis numerical core, to run as part of AMP, the Allianz in-house Nat Cat modeling platform, which runs on a Kubernetes cluster. As part of this work an HPC version of ktools was developed based on MPI C, so that the computation could be parallelized across a set of pods running on a distributed cluster. The goal of this work is to produce a setup suitable for running complex models with large exposure sets.

Adam Podlaha, CEO of Impact Forecasting, AonBenfield
Impact Forecasting and Oasis: opening up the world of catastrophe modelling
Impact Forecasting have advocated open cat models and more in-depth quantification of uncertainty over the last decade. The arrival of the Oasis platform has further supported the market-wide change in thinking about catastrophe modelling has been a welcome addition. Because of these synergies Impact Forecasting is committed supporting the Oasis ecosystem by adopting the Oasis model format in ELEMENTS for IF as well for 3rd party models and by releasing IF models on the Oasis platform either directly at re/insurers premises or via 3rd party solution providers such as Boat Oasis. At Impact Forecasting we’re excited about quenching our thirst in the Oasis!


REPEAT Workshop 2: Flood models: approaches, uncertainties and challenges
Andrew Smith, COO, SSBN 
Chris Sampson, CTO, SSBN
Paul Bates, Chairman and co-founder, SSBN
Emma Raven, Principal Analyst, JBA Risk Management

Over the last 5 years, high resolution flood models have changed from being the preserve of local-scale engineering projects to being available for the types of national and global scale analyses that better serve the insurance industry. This represents a paradigm shift for the field, and has necessitated rapid methodological advances that replace field measurements and manual model construction with big data and computerised automation.
A number of commercial and academic teams around the world are developing large scale flood models, and most currently fall into one of two methodological categories: those based on data from river flow gauges, and those based on data from rainfall-sensing satellites. The differences between the two approaches are important for end users to understand, and this breakout workshop aims to provide a high-level introduction to both before summarising the latest relevant academic research and discussing some of the key challenges faced by developers looking to create probabilistic model frameworks using these models. We encourage attendees to bring questions, feedback and ideas that may help model developers continue to improve their models.


13.15 
Lunch

14.15 
Commissioning a Model on Oasis - benefits, experiences and lessons learnt 
SCOR + Mohammad Zolfaghari, ‎Director, CatRisk Solutions 

15.15 
Networking / coffee break 

15.30 
Afternoon Breakouts 

REPEAT Workshop 3: Vulnerability 
Brad Eccles, Director of Operations, ABS Group
Natural Hazard Vulnerability Curve Generation and Calibration – Opening the Blackbox
This session will summarise the engineering considerations for generating “Raw” natural hazard vulnerability curves for typical facility types. It will set out how engineers, using their knowledge of structures, contents and equipment, materials of construction etc, link physical vulnerability to damage estimates and repair times. However, this approach can significantly over predict the losses from claims data, and so through an interactive workshop, the factors to consider during the model “calibration” process will be investigated. This session is designed to give a tantalizing glimpse inside the vulnerability Blackbox.

Simon Blaquière, ‎Head of Reinsurance Actuarial and Cat Modelling, AXA Global P&C
Leveraging internal claims data to enhance vulnerability curves

REPEAT Workshop 5: Model Evaluation
How can we evaluate Flood Models?
Jessica Turner, Senior Manager, Weather Modelling, Lloyds Banking Group 
The limited historic record along with rapid changes in the built environment over time create challenges for evaluating flood models. Increasing building on flood plains, changes in water management and flood defence schemes can render what historic data exists obsolete. Numerical weather and climate models can be used to extend the record but they typically stop at precipitation modelling, which does not always translate directly to the flood risk. Impacts of climate change in the short-term are uncertain. Footprints of past historic events rarely exist to associate flood depths with losses. Given these challenges, this talk will outline some approaches to the problem of flood model evaluation that can be taken from the perspective of a primary insurer.

Windstorm Risk - Probabilistic, Deterministic and Real-Time Exploration
Ioana Dima-West, Head of Model Research and Evaluation, Willis Re
The presentation will focus on ways of evaluating probabilistic cat models through the application of scientific research. In addition, I will discuss the use of historical storm data and scenario-based analyses, as well as offer a brief description of our Event Response Service.

Earthquake Risk
TBC, Willis Re


REPEAT Workshop 6: Loss Impact from Financial Models
Imelda Powers, Global Chief CAT Modeller and Managing Director, Guy Carpenter
Study the impact of losses to individual risks and portfolios arising from different financial models. Considerations will be given to secondary uncertainty, risk correlation, and treatment of conflicting policy terms.


17.00 
Closing comments
Dickie Whitaker, CEO, Oasis LMF