A changing view of risk: what the next generation of catastrophe models means for (re)insurers

Catastrophe modelling is entering a new phase

For decades, catastrophe models have relied on historical record and data, but extreme weather and changes in man-made environments have driven the newest generation of models to react to today's climate instead.

For insurers and reinsurers, the question is how this shift in climate and approach could move modelled losses, and what that means for portfolios, underwriting and reinsurance buying. At the same time, advances in catastrophe modelling are creating new opportunities to assess risk through a lens informed by recent catastrophe events.

New model updates from Verisk, alongside anticipated developments from Moody's RMS, reflect this evolution. Newer models are incorporating more current climate and environmental data alongside the historical record.

These developments represent an important step forward. But they also raise a fundamental question for the industry: how can insurers and reinsurers translate increasingly sophisticated model outputs into more informed decisions?

There is broad agreement that the natural catastrophe risk landscape is shifting. How much it is changing, and in what way, is still debated, and confidence varies by peril.

“Each new model release brings a new way of interpreting the historical data,” said Anna Pergerson, Managing Director, Head of Catastrophe R&D at Howden Re. “But models are not all updated at once, so the industry is working with a mix of views across perils. Understanding what is driving shifts in modelled losses, and where uncertainty remains, matters more now than ever.”

How the Latest Models are evolving

Catastrophe models have long used historical events, scientific research and simulated scenarios to estimate potential losses. Increasingly, model vendors are refining these approaches to account more explicitly for the climate conditions influencing natural catastrophe activity.

The clearest example is the move towards a ‘near-present climate’ view, which better reflects current and near-term conditions rather than relying solely on long-term historical averages. Verisk has already applied this approach in its updated hurricane model, following similar updates to its wildfire and severe convective storm models. Moody’s RMS is expected to do the same in its HD hurricane model, due for release next year.

Vendors are also giving users more visibility into the climate assumptions behind their models. This includes climate indices such as the El Niño-Southern Oscillation (ENSO) and sea surface temperatures (SSTs). In Verisk’s updated hurricane model, for example, simulated years can be linked to specific ENSO phases and Atlantic Ocean temperatures. Users can then examine subsets of events tied to particular conditions and build more tailored views of risk. Moody’s RMS plans to include ENSO as a predictor in its new HD model. It will also include the North Atlantic Oscillation (NAO), which it says influences storm tracks.

Separately, a range of climate-specific tools now exists, including several built by the main modelling vendors. These allow users to sensitivity test model results against different climate scenarios.

“These updates give insurers and reinsurers more flexibility in how they assess catastrophe risk,” said Sid Sasanian, Managing Director at Howden Re. “Being able to test how modelled losses change under different ocean and weather conditions gives clients a clearer view of the range of possible outcomes. That supports better decisions on portfolio strategy and reinsurance structure.” 

The ability to assess catastrophe risk under different climate conditions creates opportunities for more tailored portfolio analysis.

Beyond traditional models – building a more complete picture of risk

As catastrophe modelling becomes more sophisticated, the importance of complementary analytical approaches is also growing.

This is particularly relevant for geographically concentrated portfolios, where a smaller set of simulated events may not provide enough information to draw reliable conclusions about localised risk.

“More detailed modelling does not necessarily produce a more reliable view of risk,” said Sid Sasanian. “Independent analysis and specialist judgement are becoming more crucial as the value now lies in knowing where a filtered view adds insight and how much confidence to place in the results.” 

New models provide an essential foundation for understanding catastrophe exposure. However, no single model can capture every dimension of risk, particularly when assessing highly concentrated exposures or the interaction between natural hazards and changing patterns of development.

Howden Re’s proprietary analytics add another perspective. In wildfire, for example, our  Blaze Index capability assesses present-day risk rather than relying on past events.

“Blaze isn’t built on the historical record of wildfires,” said Anna Pergerson. “It gives us the best current estimate of the risk, based on fuels, climate and how communities interact with both. As that underlying data changes, so does our view.” 

Our Portfolio Accumulation Tail Hazard (PATH) capability focuses on wildfire concentration management.

As highlighted in Howden Re's Who dares wins report, PATH analysis conducted in 2024 identified concentrations of high-value properties that later overlapped with the footprints of the January 2025 Palisades and Eaton fires in California.

Looking ahead

The next generation of catastrophe models will give insurers and reinsurers more ways to examine their portfolios. However, changes in modelled losses can reflect many factors. 

A higher modelled loss does not necessarily mean that the underlying risk has risen by the same amount. Equally, a stable modelled view is not proof that exposure is unchanged. The task is to separate changes in the risk itself from the changes in how it is represented. 

“No single model will give the industry the full picture,” said Anna Pergerson. “The firms that test new capabilities carefully, and combine models with their own analysis and specialist expertise, will be the best placed to understand how their exposures are changing.”