BirdsEyeView: Lessons learned from nat cat events in 2025

As hazards manifested in regions once considered low-risk, the industry was forced to confront the widening gap between legacy modelling approaches and a rapidly changing climate reality.

Few events illustrated this more starkly than the 2025 Australian bushfire season. Areas previously classified as marginal fire risk experienced significant and sustained fire activity. Successive wet years across New South Wales, Queensland and Western Australia drove exceptional vegetation growth.

When these conditions were followed by sharp drying and extreme heat, the result was an expansive, highly combustible landscape primed for ignition. Fire behaviour escalated quickly; and overwhelmed assumptions embedded in traditional risk frameworks.

This pattern was not confined to Australia. Similar wet-dry climate cycles played a material role in the severe wildfires experienced in the Los Angeles region during the same year. The parallels between hemispheres underscore a crucial lesson from 2025: wildfire risk is no longer a regional anomaly but a global systemic peril.

A new approach needed

Climate volatility is synchronising hazard behaviour across geographies, eroding the validity of risk classifications based on historical averages.

For reinsurers, these events reinforced a growing concern that ‘non-modelled’ or ‘secondary’ perils are becoming primary drivers of loss. Wildfire, inland flood and convective storms are no longer tail risks; they are central to portfolio performance.

The industry’s experience in 2025 demonstrated that relying on backward-looking data to price forward-looking risk introduces material basis risk at precisely the moment when accuracy matters most.

Looking ahead to 2026, demand for advanced natural hazard risk analytics is set to accelerate. We expect particular growth in the modelling of secondary perils, with wildfire standing out as the most urgent area of focus.

Unlike well-established primary perils such as tropical cyclone or earthquake, wildfire has historically received far less investment in both scientific research and commercial catastrophe modelling. Yet its loss potential is now comparable-and in some regions, increasing faster.

Climate-aware insights

Lloyd’s syndicates underwriting US and Australian wildfire exposures are already seeking more granular, climate-aware insights to support underwriting and portfolio management. Australian managing general agents (MGAs), in particular, are looking for tools that can differentiate risk at a local level rather than relying on broad regional assumptions.

These risk carriers are no longer satisfied with static hazard layers; they want dynamic intelligence that reflects how risk is evolving within a season, not just across decades.

There is also a notable shift in what they expect from risk analytics providers. Beyond pricing and accumulation management, reinsurers and insurers increasingly want decision-support tools that inform capacity deployment, reinsurance structuring and risk mitigation strategies.

Questions around where to grow, where to retrench, and how to design products resilient to climate volatility are driving demand for more actionable, forward-looking insights.

Historical data is history

Despite this momentum, significant challenges remain. The most fundamental is the erosion of historical data as a reliable foundation for modelling. Many legacy models are calibrated on datasets that assume climate stationarity - an assumption that no longer holds.

As hazards such as wildfire and flood break historical bounds, model outputs can lag reality, leaving underwriters exposed to emerging patterns they cannot yet quantify.

Another challenge lies in data resolution and timeliness. Traditional hazard models often operate at coarse spatial and temporal scales, which can mask critical drivers of loss such as fuel load variability, vegetation stress or short-term weather extremes. In a world where hazard conditions can change materially over weeks rather than years, this lack of responsiveness limits model usefulness.

Tech is changing everything

Encouragingly, 2026 is likely to see meaningful progress on these fronts. The convergence of machine learning and satellite-derived data represents a step change in how natural hazards can be understood and modelled. High-frequency Earth observation now provides thousands of daily measurements capturing vegetation growth, moisture stress, fuel conditions and active fire behaviour.

When combined with machine learning techniques, these datasets allow models to learn directly from evolving conditions rather than relying solely on historical analogues.

This approach enables earlier detection of shifts in hazard behaviour, such as the expansion of fire risk into new ecosystems or the intensification of seasonal risk windows. Instead of waiting years for loss data to accumulate, models can adapt in near real time, offering reinsurers a more responsive view of exposure as climate signals emerge.

From retrospective modelling to adaptive intelligence

The lessons of 2025 are therefore as much about opportunity as they are about risk.

Indeed, the events of the year exposed the limitations of established practices, but they also clarified the path forward. Reinsurers that embrace climate-aware, data-driven analytics will be better positioned to navigate volatility, allocate capital efficiently and support cedants facing unprecedented uncertainty.

As climate dynamics continue to evolve, the industry’s ability to learn faster than the hazards themselves will define resilience. The experiences of 2025 should serve as a catalyst - prompting a shift from retrospective modelling to adaptive intelligence.

In doing so, the reinsurance sector can move from reacting to yesterday’s climate to anticipating tomorrow’s risk for the benefit of their bottom-line in 2026, and beyond.

By James Rendell, CEO and founder, BirdsEyeView