Quantifying the Impact of Recent Climate Trends on North Atlantic Hurricane Activity and Losses

North Atlantic hurricanes rank among the costliest natural catastrophes globally, fuelled by high sea-surface temperatures (SST) in the main development region (MDR) and neutral to positive El Niño Southern Oscillation (ENSO). Record-high SSTs and a predicted shift to positive ENSO ahead of the 2024 season have raised concerns about a “hurricane season from hell”. A key issue is that catastrophe models used to estimate insured loss in practice are calibrated with observations dating far back and may not adequately reflect hurricane risk in today’s climate. Many scientific models focus long term climate change and are thus not fully fit to assess recent climate trends or are not openly accessible for commercial use. Therefore, we built a simplified, physically-based model conditioned on climate variables to quantify changes in hurricane risk from 1980 to today.

The model uses the physical proxies potential intensity (PI) and cyclone genesis index (CGI) calculated from ERA5, as well as hurricane observations. The number of tropical cyclones is modelled as Poisson process with mean equal to the CGI in the MDR. Locations of lifetime maximum intensities (LMI) are drawn from historical observations conditioned on MDR SST and ENSO. LMI is determined based on PI and historical LMI to PI ratios and translated into landfall activity using a statistical method. The model adequately reproduces observed basin and landfall activity when forced with historical climate conditions. By detrending each grid cell using Theil-Sen regression, we project the climate inputs to any specified year to assess climate driven risk changes.

Our results indicate a 17% increase in hurricane landfalls under the 2020 climate compared to historical forcing from 1980 to 2020, with major hurricanes potentially increasing by 22%. Adjusting landfall rates in a vendor catastrophe model accordingly leads to an increase of around 20% in average annual loss. This increase comes mainly from an increased frequency predicted by the CGI, in line with observations. Keeping CGI constant while incorporating PI increases results in fewer lower-category storms, but more categories 4 and 5 storms. Our approach has limitations, notably in translating basin to landfall activity, where we do not simulate the full tracks but rely on historical ratios to determine the landfall intensity. Consequently, shear and steering effects along the track are only implicitly considered, potentially yielding a conservative risk assessment.

Nevertheless, our results highlight a material increase in hurricane risk in the current climate relative to 1980-2020. Given the lag in most catastrophe models, modelled losses may not fully reflect today’s risk. Our methodology can also be used to extrapolate to 2050, to assess climate change impacts, an area of ongoing research.