Every winter, national meteorological agencies across Europe — including the UK Met Office, Météo-France, Agencia Estatal de Meteorología (AEMET), and the Instituto Português do Mar e da Atmosfera (IPMA), among others — monitor, name, and issue warnings for North Atlantic windstorms that threaten life and property. Storm names are assigned by regional groupings of national weather services to identify where the greatest impacts are expected and to support public and civil-protection action. Alongside this public warning infrastructure, sophisticated catastrophe models have become central to how insurers and reinsurers price, aggregate, and manage windstorm risk across much of Europe.
Yet two countries that frequently sit in the path of Atlantic storms — Portugal and Spain — remain largely absent from major vendor wind catastrophe models. This omission is not because Iberia is immune to extreme windstorms, but because historical industry focus, data availability, and model development priorities have centred further north. The result is a significant blind spot in the European risk landscape: a region that can experience damaging storms, but where (re)insurers often lack the standard tools used elsewhere to quantify rarity, aggregate risk, and potential portfolio impacts.
Storm Kristin, named by IPMA and AEMET, brought this issue into focus when it struck Iberia on January 28, 2026. The storm delivered damaging wind gusts above 40 m/s (Figure 1) and intense rainfall, with accumulations up to 100 mm in places, across Portugal and parts of Spain, leading to widespread disruption. Kristin formed part of a sequence of North Atlantic winter storms affecting Portugal and Spain, with nine named storms identified between January and February. The short intervals between successive storms — including just six days between Storms Ingrid, Joseph and Kristin — meant that rainfall accumulated rapidly over already saturated ground, amplifying flooding and associated damage. The Portuguese government has since estimated that direct reconstruction costs from the storm exceed €4 billion, [1] underlining the material economic impact of the event. More importantly for risk professionals, Kristin illustrates the need for alternative, transparent ways to quantify rarity and return periods where standard catastrophe models are unavailable.
Unlike in much of Europe, there is no widely accepted industry model representation of Iberian windstorms that everyone uses as a common starting point for analysis, making it harder to compare impacts across portfolios or with past events in the region. As a result, we cannot easily look up a return period for peak gusts in places like Lisbon, Porto, or inland Portugal. Without a common model catalogue, it is also harder to place an event like Kristin within a shared baseline for how “extreme” it really is. When a high-impact storm strikes Iberia, the market is often forced to judge rarity using a limited historical record, supported by observations and expert judgement, rather than a large event catalogue that simulates extreme tail behaviour.
WTW has a long-standing research partnership with the University of Exeter through the Willis Research Network, aimed at bringing robust atmospheric science into practical risk management. One of the first and most important outcomes of this collaboration was the 2009 paper “Serial clustering of intense European storms”,[2] which showed that severe windstorms tend to cluster in time over northwestern Europe and along the North Atlantic storm track. This work changed industry thinking about aggregate windstorm risk by providing a statistical basis for representing storm sequencing, which was later embedded in standard catastrophe models.
More recently, the partnership has focused on developing large-ensemble European windstorm datasets and improved cyclone-tracking and extreme-wind diagnostics. These ensembles cover the whole of Europe — including Spain and Portugal — and provide an effective storm record of more than 2,500 years, allowing event-based wind footprints to be generated. In well-modelled regions these footprints complement vendor tools; in non-modelled regions like Iberia, they provide a science-based way to estimate hazard and return periods where traditional models are absent.
In our analysis , we use Exeter’s large-ensemble windstorm dataset to place Storm Kristin and its surrounding cluster of storms in both historical and probabilistic context, and to translate that into location-specific estimates of return period that are otherwise unavailable in this region.
We first place Storm Kristin within the historical record using reanalysis, with minimum mean sea level pressure (MSLP) used as a proxy for storm intensity. Of all North Atlantic cyclones that have passed over or near Portugal since 1960, Kristin ranks as the 11th deepest, placing it firmly within the upper tail of observed events — within the top 2% of storms in the historical record — rather than among routine winter storms.
Observed impacts were consistent with this ranking. Inland wind gusts exceeded 40 m/s at multiple weather stations, and at the Monte Real air base — around 150 km north of Lisbon — a maximum gust of 48.9 m/s (175.9 km/h) was recorded. While reanalysis-derived gusts shown in the wind footprint are smoother and lower than point observations due to resolution constraints, they nevertheless indicate a broad swathe of exceptionally strong winds across central and southern Portugal (Figure 1).
Peak rainfall accumulations over the January–early February period reached four to five times the 1980–2021 average in parts of Portugal and Spain (Figure 2). Crucially, this rainfall was concentrated into a very short timeframe: up to 80% of January 2026’s total rainfall fell during Storms Ingrid, Joseph and Kristin. This clustering amplified surface runoff, soil saturation and flood impacts.
To set Kristin’s winds in a broader probabilistic context, we compare its reconstructed wind footprint against the Exeter large-ensemble windstorm dataset, which represents thousands of physically plausible European storms and explicitly includes Spain and Portugal.
This analysis allows us to estimate return periods for peak gusts across Iberia (Figure 3). Kristin’s winds correspond to a 30–40-year event across large parts of Portugal, with return periods approaching 80–90 years locally, particularly in areas affected by the strongest onshore and inland wind maxima. For Lisbon, peak gusts correspond to approximately a one-in-50-year event. Within the available observational and reanalysis datasets beginning in 1980, there is high confidence that this represents the strongest gust event on record for the city. This result confirms that Kristin’s gusts were not merely unusual by historical standards, but also rare when viewed against a much larger ensemble of plausible storms.
Lastly, Exeter’s large-ensemble windstorm dataset enables us to quantify the rarity of the Ingrid-Joseph-Kristin storm sequence itself.
The analysis indicates that a sequence comparable to Ingrid, Joseph and Kristin — defined here as three or more storms with wind gusts of at least 20 m/s over Portugal — has a return period of one-in-12 years in both the observed and ensemble records (Table 1). Although it caused considerable damage, this level of clustering is within the range of what has previously affected the country. The storm sequence surrounding Xynthia in February 2010, for comparison, is placed by the ensemble at closer to a one-in-36-year return period cluster for Portugal. [3]
However, the ensemble also shows that more extreme clustering is plausible. A sequence of three severe windstorms impacting Portugal within a three-day window corresponds to an event with an approximate return period of one-in-150 years — a sequence not yet observed in the historical record. In non-modelled regions such as Iberia, where there is no standard industry catalogue explicitly representing clustering, this aggregate dimension of risk can remain under-recognised until it materialises.
| Window length | Historical return period (years) | Ensemble return period (years) |
|---|---|---|
| 3 days | Unobserved | 149.6 |
| 4 days | 47.0 | 35.8 |
| 5 days | 15.7 | 18.0 |
| 6 days* | 11.8 | 12.2 |
| 1 week | 7.8 | 9.4 |
| 2 weeks | 5.2 | 5.4 |
| 4 weeks | 3.9 | 3.7 |
| Entire winter | 2.6 | 2.5 |
These insights show that Storm Kristin was unusual in the observed record, extreme relative to a much larger ensemble of plausible storms, and locally exceptional for key population and exposure centres such as Lisbon. The associated storm sequence also represents a level of clustering that is uncommon historically and only quantifiable through ensemble analysis.
In combination with the scale of economic losses—likely making Kristin Portugal’s costliest (re)insured windstorm event on record—this analysis highlights both the material nature of Iberian wind risk and the value of alternative approaches for quantifying rarity and return periods in non-modelled regions.
