Held in the heart of Paris, the Oasis Insights Conference brought together leading catastrophe modelling experts andinsurance professionals to explore emerging natural catastrophe (NatCat) risks across Europe. Hosted by SCOR, the eventfeatured presentations on European windstorms, hailstorm modelling, exposure data quality, extreme weather resilience,historical flood analysis, wildfires, model validation, and rare extreme events.
Across the sessions, a consistent message emerged: transparency around model uncertainties and data limitations isessential, and establishing robust data standards is critical for the industry’s future. Evidence from successful mitigationinitiatives demonstrates how resilience measures, especially those developed through public-private partnerships, cansignificantly reduce losses. As climate change, urbanisation, and interconnected perils such as drought-driven wildfirestransform the risk landscape faster than historical data can track, the industry must prioritise data quality over modelcomplexity and develop forward-looking tools that envision plausible futures rather than rely on an increasinglyinsufficient past.
Navigating Hailstorm Risks in Europe
Speaker: Gilles Andre (RiskWeatherTech)Panel: Aidan Brocklehurst (Impact Forecasting), Pierre Mouilhade (IPE), Daniel Bernet (Moody's RMS)Moderator: Simos Koumoutsaris (SCOR)
The first session, presented scientific perspectives on hailstorm risk modelling, followed by panel discussion addressingcritical industry challenges. Key themes included improving the accuracy of hailstorm frequency and intensity predictionsamid climate change, with current catastrophe models facing limitations in capturing the full range of impacts on modernconstructions such as solar panels and large windows. The panel explored how advanced technologies like remotesensing and machine learning could enhance hail risk assessments, while discussing the role that contract definitions playin the variability of modelled losses and opportunities for standardisation. A major focus was quantifying andincorporating the evolving vulnerability of different building materials and designs to hail damage, recognising thatmodern construction methods and materials respond differently to hail impacts than traditional buildings. The discussionhighlighted the need for ongoing model refinement as both climate patterns and built environment characteristics evolve.
Exposure Data Requirements for Effective Risk Management
Presenter: Henry Bovy (SCOR)
Henry Bovy (SCOR) highlighted exposure data quality challenges facing SCOR's library of 1.5 billion addresses andthousands of billions of euros in sum insured representing broadly one year of Reinsurance submission, identifyingproblems across three dimensions: WHERE (30,000 locations sharing identical coordinates), WHAT (34.7% ofconstruction types unknown, impossible attributes like 166+ story buildings), and HOW (difficulty linking exposure totreaty terms). Henry Bovy argued that 25-year-old data formats designed for risk transfer are inadequate for assessingevolving risks like climate change, political risks, and ESG factors, proposing minimal quality standards including large riskcoding for values exceeding €500 million, perils completeness, data schema allowances for emerging perils and a betterdistinction between factual data and assume exposure qualifier require to execute a catastrophe models. Thepresentation emphasized that exposure data should serve risk assessment purposes along the risk transfer chain beyondsimply running catastrophe models, questioning whether the industry is receiving data structured to anticipate currentand future risk landscapes effectively.
Building Resilience to extreme weather: Lessons from US Initiatives
Presenter: Fanny Pouget, (Milliman) and Simon Blaquière (Generali)Moderator: Marie-Laure (SCOR)
The third session showcased how US resilience programs, particularly the FORTIFIED home standard, have successfullyreduced catastrophe losses by 50% in frequency and cost compared to standard construction. Alabama's public-privatepartnership model (providing up to $10,000 grants for roof retrofitting while requiring insurers to offer 20-50% premiumdiscounts) has achieved over 53,000 resilient homes, demonstrating effective incentive structures. The panel emphasisedthat catastrophe models should evolve into dynamic tools integrating detailed resilience and mitigation data to enablefairer pricing, while noting that wildfire mitigation requires community-level coordination (including "zone 0" vegetationmanagement within 5 feet of structures) rather than individual action alone, offering valuable lessons for Europeanmarkets facing increasing climate-related pressures.
Model Validation and Evaluation
Panellists: Sandra Hansen, (Guy Carpenter), Elisabeth Viktor, (Swiss Re), Dickie Whitaker, (Oasis LMF)
The panel discussed improving the evaluation of catastrophe models across all hazards (severe convective storms,tropical cyclones, earthquakes, floods) by developing standardised questions for vendors to give end users a clearer viewof model strengths, weaknesses, and usability while making evaluations more efficient. Sandra Hansen emphasisedidentifying what is worthwhile to evaluate, suggesting hazard intensity metrics and stressing transparency aroundimpactful assumptions. She proceeded to highlight challenges in comparing models across vendors and even betweenversions from the same vendor, particularly regarding vulnerability data differences. Elisabeth Viktor underlined theimportance of transparency, so underwriters understand both model capabilities and limitations, noting that convectivestorms are particularly difficult to evaluate and that convergence and validation are pressing industry issues. DickieWhitaker pointed out that improved evaluation processes will make the use of multiple models more efficient andinsightful, especially for time-constrained users, and that platforms like Oasis could streamline this process. The panelconcluded that consistent industry requests to vendors would improve comparability while still allowing flexibility andinnovation in model development.
Navigating the risk of rare events
Presenter: Simos Koumoutsaris (SCOR)
Simos Koumoutsaris presented SCOR’s innovative approaches for deriving valuable insights about rare extreme eventswith little historical precedent. Utilizing web scraping and AI technologies, the team was able to assemble acomprehensive dataset on natural disasters dating back to 1888, sourced from online historical archives. He alsointroduced how counterfactual scenario analysis can be integrated in model evaluation and significantly enhancebenchmarking and performance assessment. A key example involved the usage of spatial counterfactuals of major floodsin Germany, such as the 2021 Bernd flood, to benchmark empirical losses, uncover uncertainties, and comparecatastrophe models. To promote consistency across the industry, the Weighted Mean Absolute Percentage Error(wMAPE) is proposed as a standardized evaluation metric. Additionally, climate-conditioned modeling revealed the urgentneed to continuously refine risk understanding, as historical data alone becomes increasingly insufficient for anticipatingfuture extremes.
European Windstorm Risk Assessment
Presenter: Alex Doyle (Verisk Extreme Event Solutions)
Alex Doyle’s presentation examined whether the recent period of relatively low European windstorm activity represents afundamental shift in risk or simply a fortunate lull. The presentation explored how extra-tropical cyclone trends arecomplex and depend on timescale, region, and whether examining frequency, intensity, or insured losses. Throughrandom resampling of seed storms, the analysis showed that annual maximum losses could have been significantly higherthan observed. Key insights included caution about using pre-1979 reanalysis data due to quality concerns and the findingthat there is no clear signal suggesting Europe cannot return to the high-loss environment of the 1990s.
Storm Kyrill Sensitivity Analysis
Presenter: Alexandros Georgiadis (AON Impact Forecasting)
Georgiadis presented a practical methodology for assessing "what if" scenarios by analysing how Storm Kyrill losseswould change with variations in track position and intensity. The analysis revealed dramatic non-linear relationshipsbetween small intensity changes and loss outcomes: a 10% increase in wind speeds can lead to over 200% higher EU-wide losses (from €7.15 billion to €21.94 billion), while a 10% decrease reduced losses by 71%. This sensitivity testingapproach provides valuable insights about modelling uncertainty and portfolio exposure, though it differs from truecounterfactual event simulations which would require full physics-based modelling with forecast ensembles.
Historical Flood Analysis: The 1910 Seine Flood
Presenter: Jane Toothill, (JBA Risk Management)
Jane Toothill examined the 1910 Paris flood (8.62 meters above normal, 200,000 displaced) to assess how would it impactmodern Paris’s infrastructure and financial district if it were to happen today. JBA's analysis showed that modern flood defensedefence infrastructure would reduce flooded area by approximately 45% (from 19 km² to 13 km²), while theproposed Seine Bassée flood management project could provide a 9.6% loss reduction, approximately balancing the 9.3%loss increase expected from 4°C climate warming. Key modelling lessons emphasized that small flood depth changeslead to disproportionately large loss changes, the importance of modelling cellars and basements for business interruptionimpacts, and that extreme events may require "storyline" approaches rather than traditional probabilistic modelling. Thepresentation cautioned against attributing all flood risk changes to climate change, noting that urban development,defence improvements, and value distribution shifts are equally important factors.
European Wildfire Risk Modelling
Presenter: Nicolas Brunea (Reask) and Mathis Joffrain (AXA)
Moderator: Florent Lobligeois (AXA)
This session examined whether European wildfires cancould become catastrophic events, with 2025 emerging aspotentially the largest wildfire season on record. Bruneau's climate-connected stochastic modelling revealed drought-wildfire interconnections in Southern France and Iberia, with climate projections showing increasing burned area trends,longer fire seasons expanding to northern latitudes, and more frequent extreme fire behaviour that can overwhelmsuppression resources and create concomitant events across multiple regions. AXA's modelling suggests 20% of Frenchexposure is at wildfire risk, with ember transport extending 600-1,200 meters beyond the Wildland-Urban Interface(noting that 90% of European ignitions are man-made with 30% intentional in France). Both speakers concluded thatEuropean wildfires should primarily be considered as an Aggregate Exceedance Probability (AEP) risk with correlatedexposure characteristics rather than traditional single-event catastrophes, though the potential for truly catastrophicscenarios cannot be dismissed as climate change continues altering the risk landscape.
The Oasis Insights conference showcased current European catastrophe risk modelling procedures and research whilemaintaining appropriate transparency about remaining uncertainties. The path forward demands more than just bettermodels. It also requires cross-sector collaboration, policy innovation, community-level resilience building, and politicalcommitment to invest in adaptation strategies before catastrophes force the issue.
This event wouldn't have been possible without the dedication of our speakers, moderators, and event coordinators, thankyou all. We're especially grateful to SCOR for their hospitality and support.
Stay Connected
If you're already part of our community, thank you — your contributions make innovations like those showcased in Parispossible.
If you’re new to Oasis, there’s never been a better time to get involved. With over 100 models now available and a growingnetwork of experts, the Oasis ecosystem offers unmatched opportunities for collaboration and innovation.
Missed the conference? You can explore our resources, documentation, and details on upcoming events at Oasis LossModelling Framework | Open source catastrophe modelling platform.
The next chapter of catastrophe modelling is being written by our community — and we’d love you to be part of that story.
Interested in contributing to future newsletters, articles, or events? Get in touch at harriet.smith@oasislmf.org — we loveshowcasing the incredible work our members are doing!