Why your risk models may be underestimating extreme weather losses

If your catastrophe models are built on historical data alone, they could be underestimating risk, leaving your capital reserves and business exposed. Reask’s Unified Tropical Cyclone (UTC) modelling framework takes a dynamic approach that reflects today’s changing climate.

Reask’s latest research, led by Dr. Thomas Loridan and Dr. Nicolas Bruneau, highlights a critical issue: traditional catastrophe models may not be keeping up with today’s climate risks. Their work applies machine learning and climate physics to provide a more accurate way to assess risk, giving insurers and reinsurers a clearer picture of their exposure.

Rather than assuming past events dictate future risk, the UTC framework generates climate-connected tropical cyclone event sets—simulated scenarios that reflect how storms could behave under current and future climate conditions. By incorporating environmental factors like sea surface temperatures, atmospheric circulation and wind shear, it offers a more realistic view of tropical cyclone risk.

US major landfall risk distributions based on different underlying climate conditions using Reask’s UTC.

Validated against historical storm data and independent climate data, the UTC framework presents a scientifically robust alternative to traditional catastrophe models. By integrating climate-connected event sets into catastrophe modelling workflows, insurers and reinsurers can gain a far more comprehensive understanding of their exposure to extreme weather risk.

So, are your catastrophe models accurately pricing risk?

Recent data shows that North Atlantic sea surface temperatures have been significantly warmer than long-term averages. This prolonged warming has increased the intensity and unpredictability of tropical cyclone tail events—the extreme, low-probability storms at the far end (or “tail”) of a probability distribution. These events are often underestimated by catastrophe models that rely solely on historical data and fail to reflect the severity of today’s evolving climate.

Storm frequency, intensity and paths are shifting, making traditional catastrophe models that depend on past trends increasingly unreliable. This gap in forecasting accuracy carries serious financial and regulatory consequences for insurers, reinsurers and capital markets looking to price, manage and mitigate extreme event exposure. In short, many risk professionals may be underestimating their capital at risk, leaving businesses exposed to financial disaster.

A two-part problem for risk professionals

Traditional catastrophe models rely on historically calibrated distributions—statistical patterns derived from past storm data—on the assumption that historical storm patterns provide a reliable foundation for future risk assessment. 

Reask’s UTC framework challenges this assumption, demonstrating that risk is tightly connected to climate variability. To remain accurate, models must adjust dynamically to reflect physics-supported environmental evidence rather than historical trends.

“A model fit to historical data will be best fit to those regions where we have abundant historical records (e.g. the North Atlantic) but will generalise poorly to other basins where data are scarce and TC behaviours may differ (e.g. the South Indian Ocean).”

— Thomas Loridan & Nicolas Bruneau, Reask UTC Journal Article (2024)

Alongside the flawed assumptions of traditional catastrophe models, insurers and reinsurers now face increasing regulatory scrutiny to prove their models are aligned with contemporary climate risks. 

Under Solvency II and NAIC climate risk guidelines, insurers that cannot accurately quantify their exposure may be required to hold additional capital reserves, directly impacting profitability and reinsurance strategies. As regulators push for more climate-aligned financial disclosures and stress tests, insurers relying on static models will struggle to justify their risk exposure to both regulators and investors.

If catastrophe models fail to reflect real climate risk, capital reserves may be at greater risk than many insurers realise, with financial and regulatory consequences that are only growing more urgent.

Why traditional models fall short

Most traditional catastrophe models—driven by historical data rather than physics—were designed for a world in which we believed the past was a reliable indicator of the future. That assumption is now breaking down as climate forcing, the natural and human-driven factors altering the Earth’s climate system, reshapes weather patterns and storm behaviour. 

“By design, [probabilistic] models are best fit to portray risk under conditions consistent with our historical experience. This poses a problem when trying to infer risk under a rapidly changing climate or in regions where we do not have a good record of historical experience.”

— Thomas Loridan & Nicolas Bruneau, Reask UTC Journal Article (2024)

Static models are no match for a dynamic climate

Many catastrophe models still rely on historical storm data to predict future hurricane risks, assuming past patterns will persist. But climate forcing is shifting global weather patterns. This makes the historical baseline conditions scientists have long relied on unreliable.

One of the clearest examples of this disconnect is rising sea surface temperatures. Warmer oceans provide more energy for tropical cyclones, driving higher wind speeds and rainfall. Yet, many models continue to base storm intensity assumptions on pre-2000 climate conditions, failing to account for the well-documented increase in ocean heat content and storm intensification.

Reask takes a different approach. Instead of relying on outdated historical trends, it explicitly links climate data to tropical cyclone activity, producing risk assessments that reflect contemporary and future climate conditions.

“Reask’s UTC modelling framework explicitly connects global climate data to TC activity and event behaviours. When driven by climate data representative of historical conditions, the UTC is able to simulate a global view of risk consistent with historical experience.”

— Thomas Loridan & Nicolas Bruneau, Reask UTC Journal Article (2024)

The financial and regulatory liability of underestimating risk

The accuracy of catastrophe models directly impacts capital planning, reinsurance strategies and overall financial stability. When models fail to reflect the realities of a changing climate, insurers risk mispricing reinsurance, misallocating capital and exposing themselves to catastrophic losses.

Underestimating storm intensity can leave insurers holding too little in reserves, making them unable to cover policy claims when extreme weather hits. At the same time, regulatory scrutiny is tightening. Insurers who fail to justify their risk assessments may face higher capital requirements, reducing profitability and flexibility.

Reask UTC shows how tail risk (measured as TVaR99, the average loss of the worst 1% of outcomes) in a representative hurricane-exposed insurance portfolio changes relative to a static baseline over time, strongly impacted by the observed ocean temperatures.

Reinsurance pricing is another major challenge. Invest too little and firms are left exposed. Invest too much and they could see profits dwindle.

Without a climate-connected framework, insurers may over-concentrate exposure in regions with ample historical data while underestimating risk in areas with limited records. This creates blind spots in portfolio diversification and misallocated capital when storm behaviour deviates from past trends, especially in response to climate cycles like El Niño and La Niña.

And this is key: relying too heavily on historical data without accounting for climate variability can distort capital allocation. Firms that fail to adjust their models may find regulators determining their risk exposure is too high, forcing them to increase capital reserves, limiting investment flexibility and cutting into profitability.

“We’re often told not to use the stock market’s past performance as an accurate predictor of our financial future, so why should we continue to rely on the past performance of something extremely volatile as an indicator for global-level physical catastrophes?”

— Jamie Rodney, CEO, Reask

The solution: AI-driven risk forecasting

Traditional catastrophe models fall short because they rely on outdated assumptions. Risk professionals–underwriters, analysts, actuaries and CROs–know that climate-driven changes are making tropical cyclones more severe and less predictable. Yet, most models still depend on historical data that no longer reflects today’s climate reality.

“You cannot just rely on history or an extrapolation of history to model tomorrow’s risk. There needs to be an understanding of how climate physics drives risk and how that climate is changing.”

— Thomas Loridan, Natural Catastrophes: Flood, Fire & Storm Special Report 2025

This is where Reask’s UTC model comes in. Instead of relying on static, backwards-looking risk assessments, it integrates real-time climate shifts and machine learning to provide a dynamic framework for catastrophe modelling. In today’s changing climate, this isn’t just an advantage, it’s a necessity.

The UTC framework explicitly links global climate data to tropical cyclone activity, improving event set generation, scenario analysis and exposure assessment. This physics-based approach ensures that risk modelling reflects contemporary and future climate conditions, not outdated historical assumptions.

“Extreme weather events are interconnected on a global scale. Climate phenomena like El Niño or La Niña can affect weather patterns around the world. By having a physics-based model that understands these global connections, we provide more accurate assessments of risk across different regions and perils.”

—Thomas Loridan, Natural Catastrophes: Flood, Fire & Storm Special Report 2025

Rather than assuming future storms will mirror the past, UTC simulates hurricanes under evolving climate conditions, adjusting forecasts based on real-time environmental data. This gives risk professionals a more accurate, science-backed view of hurricane behaviour, ensuring risk assessments keep pace with a rapidly changing world.

Why this matters for your business

The cost of maintaining inaccurate risk models far exceeds the investment in upgrading to climate-connected modelling. Mispriced risk leads to capital shortfalls, regulatory penalties and unpredictable reinsurance costs. In 2024, global insured losses from natural catastrophes hit approximately USD 140 billion, making it the third most expensive year on record.

Recent climate conditions are pushing storm severity beyond what traditional models anticipate. Risk curves based on historical baselines rather than real-time environmental data can no longer be trusted to measure capital-at-risk, reinsurance needs or diversification strategies.

This directly impacts insurers and reinsurers in three critical ways:

1. Underestimated Probable Maximum Loss (PML): Traditional models may significantly underprice risk, leaving insurers with insufficient reserves when claims surge beyond expectations.

2. Volatile pricing: Without climate-adjusted models, firms may overpay for coverage or secure too little protection. Both scenarios undermine financial stability and market competitiveness.

3. Regulatory and investor scrutiny: Insurers failing to align models with contemporary climate risks may face higher capital requirements, compliance penalties and investor concerns, directly impacting profitability and long-term stability.

Don’t let outdated risk models leave you exposed

Traditional catastrophe models aren’t built for today’s climate volatility. Underestimating risk puts capital reserves, reinsurance decisions and regulatory compliance at stake.

Reask’s UTC modelling framework replaces outdated statistical assumptions with climate-connected event sets and AI-driven forecasting, giving firms a more dynamic and accurate approach to risk assessment. With better data, risk professionals can anticipate tropical cyclones more confidently and strengthen financial stability.

Ready to protect your business and capital reserves? Reach out directly or request a free demo today.

Sources

Loridan, T., & Bruneau, N. (2024). Reask UTC: A Machine Learning Modelling Framework to Generate Climate-Connected Tropical Cyclone Event Sets Globally. EGUsphere. https://egusphere.copernicus.org/preprints/2024/egusphere-2024-3253/egusphere-2024-3253.pdf
Loridan, T. (2025). Natural Catastrophes: Flood, Fire & Storm Special Report 2025. Reask. https://reask.earth/2025/01/09/insuranceerms-2025-special-report/