Why observed hurricane landfall trends in the U.S. might not tell the whole story

Every year, as hurricane season begins, there’s a familiar rush among modellers, meteorologists and industry professionals to ask:

• What is the hurricane landfall risk this year, or over the next five years?

• Are landfalling hurricanes becoming more common in the U.S.?

• Are they getting stronger?

But here’s the inconvenient truth: using observational data alone for hurricane risk assessment is statistically shaky at best.

In this Research Spotlight, we explore why the observed record of U.S. hurricane landfalls might be misleading, and how Reask’s Unified Tropical Cyclone model (UTC) offers a more complete, climate-aware picture of hurricane risk.

The scarcity of observed hurricane data

Hurricanes making landfall in the U.S. are rare events, fewer than two per year on average over the past century.

With numbers that low, natural variability dominates any trend. One active season like 2005 or 2020 can skew multi-decade averages. Quiet periods can do the same, such as the hurricane drought from 2009–2015, during which only four hurricanes made landfall over seven years.

Figure 1: Observed trends in U.S. hurricane landfalls (Source: NOAA)

We’re left scratching our heads, free to choose any period from the observed record, each yielding a different result. Figure 1 highlights three different trends depending on the period selected:

• 1850–2024: flat

• 1950–2024: slightly positive;

• 2000–2015: strongly negative.

Drawing conclusions from such a small sample is like flipping a coin five times and claiming you’ve found a pattern.

Detecting non-stationary trends

Given the scarcity of data, you might need centuries of observations to detect a clear signal, and we don’t have that.

A simple Poisson model highlights this point (Figure 2). Detecting a positive trend in 80% of simulations, based on a 20% increase in landfall rate (from 2 to 2.4 events per year), would require around 80 years of data, assuming the trend unfolds over a 50-year period.

Figure 2: Detection period, in years, over which a Poisson distribution, with the rate increasing from 2 to 2.4 over a 50-year period, reveals a positive trend.

Looking beyond landfalls

It’s a sobering thought: even if climate change is driving a shift in landfall rate, whether up or down, we don’t have a long enough observational record to detect it reliably. And the truth is, in a dynamic, fast-changing climate, we may never have one.

To make sense of these trends and construct probability distributions that appear less random, we need more information.

Thankfully, we have it. While landfall trends may seem spurious, there are clear, measurable shifts in the physical conditions that influence hurricane behaviour.

Take sea surface temperatures in the North Atlantic, for example. They’ve been rising steadily in recent years, a far more robust and consistent signal to track over time (Figure 3).

Figure 3: Observed trend in Sea Surface Temperatures in the North Atlantic

Reask’s Unified Tropical Cyclone model (UTC)

To make use of this additional climate information, Reask’s global stochastic Unified Tropical Cyclone model (UTC), integrates key climate drivers of tropical cyclone risk, such as sea surface temperatures (SSTs), wind shear, and steering flow, into every statistical component of the model. This includes where storms form and move and how intense they become.

The model is forced with multiple climates, both observed and from global circulation models, to simulate thousands of tropical cyclones under each climate regime. The result is a dataset that captures both trends and variability, without relying on the limited observed record.

Reask’s algorithms have generated over 1.5 million years of synthetic tropical cyclone events, providing a rich sample size that avoids the pitfalls of small-sample randomness. And even if the observational record eventually ‘catches up’, the climate may have already moved on, leaving historical data out of step with present-day probabilities.

The UTC model is peer-reviewed and published in EGUsphere by Dr Thomas Loridan and Dr Nicolas Bruneau, showcasing how our machine learning framework reflects the physical realities of a changing climate.

Putting UTC to work: five real-world use cases

This ability to simulate climate-conditioned storm activity is already helping risk managers make more confident decisions.

1. Long-term climate scenario analysis

In a recently published study by MS Amlin in the Journal of Catastrophe Risk and Resilience, Ed Pope and Sam Phibbs modelled how hurricane losses could evolve under a 2°C warmer climate.

Using Reask’s forward-looking UTC data, they explored the financial implications of warming on U.S. hurricane exposure, showing how climate-conditioned event sets can inform long-range risk planning.

2. Investment risk insight for ILS portfolios in a shifting climate

Schroders Capital explored how climate change is reshaping risk in the insurance-linked securities (ILS) market, highlighting the role of advanced climate-conditioned tropical cyclone models like Reask’s UTC.

By incorporating climate science into catastrophe modelling, asset managers gain a clearer view of potential shifts in hazard frequency and severity, improving capital allocation, portfolio resilience, and investor confidence.

3. Near-term seasonal risk forecasting

Ahead of the 2025 U.S. hurricane season, leading ILS manager Twelve Securis used Reask’s UTC model to quantify the influence of current seasonal conditions on potential insured losses.

By benchmarking 2025 projections against the historical baseline (1985–2024), they were able to identify potential shifts in risk and adjust expectations accordingly.

4. Regional SST signal forecasting

Specialty insurer Inigo used Reask’s UTC model to quantify hurricane activity for the 2025 season, based on forecast SSTs across the Atlantic.

Their seasonal outlook highlighted how current SST conditions in the Main Development Region could influence storm frequency and intensity.

5. Climate variability scenario modelling

In earlier work, Inigo collaborated with Reask to explore how rare combinations, like El Niño events with exceptionally warm SSTs, affect hurricane risk.

Using UTC’s connection to dynamical climate models, they ran simulations to understand how short-term variability (like ENSO) interacts with longer-term trends such as warming oceans.

A tropical cyclone model that reflects a moving climate

Figure 4 presents simulations from Reask’s model. Using observed climate data starting in 1950, the model generates 2,500 simulated storm seasons per year, each representing a plausible version of how storms could evolve under that year’s climate conditions.

In panels (C)–(F), we show selected individual realisations from these simulations, compared to the observed record in panel (A) and the ensemble mean in panel (B).

The mean exhibits an increasing trend in risk, partly driven by the rising sea surface temperatures shown in Figure 3. However, there is significant uncertainty around this trend, as captured by the variability among individual realisations. The observed record can be thought of as just one possible realisation from this broader distribution.

Figure 4: Observations and Reask UTC simulations of the proportion of U.S. storm activity reaching major hurricane strength

Having thousands of realisations enables the model to quantify how likely, or unlikely, the observed record is, given the state of the climate. Some simulations, like Figure 4(E), closely resemble the historical record, while others, like Figure 4(D), show a much stronger upward trend.

Rather than relying solely on abstract statistics and a limited observed record, practitioners can now work with probabilities grounded in physical processes, making them more transparent, easier to explain and better suited to assessing both our current and future reality.

And suddenly, trends in U.S. hurricane landfalls don’t appear so spurious.

A more reliable path forward

With the climate changing dynamically, understanding the past, present and future impacts on hurricane risk becomes a challenging statistical task. This is especially true for extreme events, where observed records are scarce and uncertainties large.

To help quantify these uncertainties, the Reask UTC model represents a significant step forward in catastrophe modelling. It uses climate forcing to generate stochastic event sets based on near-term climate, historical conditions or future warming scenarios.

The key innovation is that every output from the model is tied to a defined climate state, automatically capturing climate change, variability and global correlations. This results in physically grounded statistics that open the door for clear physics-based analysis, while also enabling clear, transparent communication of risk to stakeholders.

Ready to move beyond the limitations of historical data?

Whether you’re forecasting seasonal risk or planning for long-term exposure, Reask’s UTC model gives you the clarity traditional data can’t, supporting more confident hurricane risk assessment and operational tropical storm forecasting.

Get in touch today to learn how Reask’s climate-conditioned insights can support your underwriting, modelling, or investment strategy.