Of stormy skies and calm horizons: Understanding the recent quiet period in European windstorms

Author: Giovanni Leoncini, Assistant Director - Product Management, Moody's

For the last 20 years or so, windstorm activity has been quiet over Europe. Over this period, many European countries and Europe as a whole have exhibited relatively low levels of losses and hazard. 

The previous most active decade, the 1990s, had large losses that drove up the average annual loss (AAL) and the risk in general, but this was followed by a decrease and then a ‘lull’ in activity.

This decrease in loss activity is linked to a decrease in hazard, both in ‘storm count’ and ‘storm severity.’ As an example of this change, Figure 1 below shows the historical annual losses for the United Kingdom from 1972 to 2022, as represented by the Moody’s RMS European Windstorm HD models (red dots). 

Figure 1: The schematic shows the annual losses for the United Kingdom (red dots) based on the historical event set of the Moody's RMS European Windstorm HD models. The black line is the 10-year running average of the annual losses. Losses have been normalized to the historical Annual Average Loss and are plotted on a logarithmic y-axis.

The schematic above illustrates an initial period of lower losses up to the mid-1980s, a period of high activity until the late 1990s, followed by a decrease and then a plateau. Perhaps these trends are not so apparent due to the large year-to-year variability. From one year to the next, the annual loss easily changes by one or two orders of magnitude.

Such changes are significantly larger than any long-term trend, which must be made visible with a 10-year moving average (black line). The commonly used Storm Severity Index, or the annual count of storms, shows very similar trends. Overall, this is representative of many countries and of Europe as a whole.

The Index is probably the most common hazard measure for windstorms in Europe and was originally developed by Klawa and Ulbirch [11].

These risk variations over time force companies to carefully review their strategy so they can withstand possible large windstorm losses without pricing themselves out of the market and avoid capital burdens. Answers to questions like: “Will the low windstorm activity continue for the next few years?” or “Could a large storm occur regardless of the low activity?” have important consequences.

We have set a goal to address these issues and help our clients tackle them, so we will place the last decades of activity in the wider context provided by longer historical records (see the next section 'Longer historical record for windstorm hazard') as well as future projections (see section entitled 'Future Projections').

Because single strong events are relatively easier to detect and can greatly affect (re)insurers, we will also consider four large storms that occurred in the 1700s and 1800s. Finally, we’ll propose a practical approach for implementation.

Longer historical record for windstorm hazard

To describe European windstorms precisely, accurate and widespread measurements are required. To the best of our knowledge, records available in the years before those shown in Figure 1 do not have the necessary completeness or quality to characterize the risk for the industry. A dataset similar to that for hurricanes does not exist in this case. However, longer records do exist and contain useful information. For example:

• Cusack 2013 [1] examined five Dutch weather stations with high-quality wind measurements and developed the wind-based loss index shown in Figure 2. The data showed two periods of high activity comparable to the 1990s: the 1910s and the 1940s.

Figure 2: Cusack's 2013 figure is reproduced above. The histogram refers to the loss index, while the thin blue line refers to the 10-year moving average.

• Kruger et al. 2019 [2] took advantage of pressure data, which have longer and more reliable measurement records than wind. This allows the estimation of windstorm activity in the northeastern Atlantic, including the North Sea and the British Isles. The figure is reproduced below in Figure 3 below, and shows several peaks of activity; not only the 1990s, but also the 1950s, the 1900s, and the 1880s. 

Figure 3: Kruger et al (2019) figure is reproduced above. Crosses and dots show two different percentiles of a measure of yearly storm hazard over the North Sea and the British Isles. The black lines are obtained by smoothing the same values.

Both these studies show that the windstorm hazard levels of the 1990s are not unusual in the Netherlands and in the northeastern Atlantic region. They also represent limited areas around Europe, but it is reasonable to think that similar levels of storm activity impacted at least the surrounding areas and probably the larger continental Europe as well. Furthermore, all figures show large year-to-year variability, a characteristic of European Windstorms, whether losses or hazard are considered.

Future projections

Future projections are notoriously uncertain (IPCC 6^th Assessment Report [10]). This uncertainty stems, among other things, from the large year-to-year variability of most metrics related to extratropical cyclones as well as from climate modeling uncertainties.

The academic community is actively investigating these issues. Within this vast body of research, we found two particularly insightful papers, from Cusack 2025 [3] and Priestly 2024 [4].

Internal variability and volcanic aerosols have been identified as drivers of the multidecadal variability of the North Atlantic climate, with anthropogenic aerosols possibly playing a significant role in the twentieth century (Zhang et al, 2019 [9]).

Cusack suggests that aerosols likely shaped the high activity period of the 1980s and 1990s. The contemporaneous decrease in anthropogenic aerosol and the absence of volcanic aerosol indicate that the current low activity period may continue in the next few years.

He also highlights the uncertainties in the modeling, which, in our opinion, inhibit any further quantification of the impact of aerosol on windstorm activity. Such uncertainty is an important factor that hinders a detailed description of the interaction between internal dynamics and aerosols, whether anthropogenic or volcanic, all of which is not yet clear.

For example, in Qin’s (2020) [8] study, internal variability also plays an important role for Atlantic Meridional Variability and therefore the windstorm climate. Similarly, Zhang et al (2019) conclude in their review that the Atlantic Multidecadal Variability is not primarily driven by external forcing, i.e., aerosols.

Priestly identifies future trends for the severity and frequency of windstorm activity in different areas of the continent using multimodel ensembles. Some areas reveal aggregate activity increases and average severity, such as in France and Great Britain. Trends for most other regions show decreases in severity and/or frequency.

What is particularly interesting is the confidence associated with the trends, which originate from both internal and model variability. Priestly also highlights that year-to-year variability is larger than any trend, which is also apparent from the three figures shown here. This large year-to-year variability implies that strong storms can occur irrespective of the multidecadal cycle or long-term trends.

Four 'blasts from the past'

Accurate estimations of windstorm activity before the satellite era (i.e., 1979) remain elusive. But some strong extratropical cyclones, similar to Daria in 1990, are large enough to be detected by reanalyses, in some form. Their deepest pressure is likely to be overestimated, with a consequent underestimation of the peak winds, but the presence of an extreme storm is very likely to be detected, even if reanalyses use the rather sparse network of weather stations from the nineteenth century.

The occurrence of an extreme storm can be confirmed with historical damage records, which were already documented in the eighteenth century, by churches and public authorities in Europe. With this in mind, Swiss Re published an interesting paper [5] outlining the impact of three storms that occurred in the late nineteenth century, along with realistic footprints and Swiss Re's own estimates of the damage at the time of publication (2014).

All three storms had gusts in excess of 150 kilometers per hour and are summarized here:

• Lothar’s Big Brother (March 1876): Detailed forestry damage reports indicate the extent of the destruction across continental Europe.
• Daria’s Big Sister (January 1884): Notable for producing the lowest pressure reading ever recorded over the British Isles and continental Europe. It affected mostly the U.K. and France.
• The North Germany Express (February 1894): Caused extensive damage to buildings and forests across northern Germany and Denmark, with strong gusts also in the U.K.

The original Swiss Re estimates for 2014 losses are shown in Table 1. A rough estimate of today’s losses can be obtained by increasing the 2014 estimates by 20-30%, increases mostly driven by the strong inflation of 2022 and 2023 in Europe.

Table 1: Swiss Re's estimation of losses for the storms discussed in the text. Losses are in millions of Euros in 2014.

Notably, all three storms occurred in the late 1800s when the northeast Atlantic was transitioning from a high storm activity period to a low storm activity period, according to Kruger (2019) (Figure 3).

The fourth ‘blast from the past’ is a storm from 1703 documented by Robert Muir-Wood in his “300-year Retrospective” available on the Moody’s website [6]. Record-keeping of damages as well as financial information in the eighteenth century was sufficiently developed to allow for estimating repair costs and damage, and therefore, even an approximated building vulnerability and hazard footprint.

The total damage is estimated to be £6 million at the time. This figure is actually quite large compared to the estimated building stock value of £100-150 million. Therefore, losses were 4% to 6% of the total building stock value. Today’s building stock is certainly less vulnerable, but its value is much higher. So any quantitative estimation of the storm impact on today’s building stock would be very uncertain, but it is safe to say that it would be catastrophic.

A practical approach to return periods

Given the large uncertainties surrounding the loss estimates of those ‘blast from the past’ events, for this discussion, we simply consider them to have incurred losses of the same order of magnitude as Daria, the largest European windstorm in both the U.K. and continental Europe. During the intervening 320 years or so since 1703, Europe has experienced five Daria-like storms corresponding to a return period of around 60 years. Excluding the 1703 storm, that leaves four storms in roughly 150 years (from 1874), or approximately a 35-year return period.

For the U.K., in the 320 years since 1703, the country has been affected by four catastrophic storms: Daria, Daria’s Big Sister, the Northern Germany Express, and the 1703 storm. This corresponds to an 80-year return period. Again, excluding the 1703 storm, it leaves us with three storms in 150 years, or a 50-year return period.

These estimates are certainly rough and should be used carefully for at least two reasons:

1)      There could be other large storms that occurred in the past with potentially catastrophic damage today, given the increases in population and urban areas.

2)      Estimating the probability of losses from a set of single historical events does not account for the location uncertainty, i.e., the possibility that a small displacement of the footprint can have a very large impact on the losses, as shown by Marescot in this blog [7], already in 2014.

Conclusions

While there are still significant uncertainties in the estimation of the past windstorm climate, as well as in the future climate projections, two key facts are very relevant:

a)     It is likely that the level of storm activity of the 1990s is not unusual for the climate system.

b)     Year-to-year variability is larger than trends.

This means that (re)insurance companies must still protect themselves against catastrophic losses even during a period of low activity.

References:

[1] Cusack, S. A 101-year record of windstorms in the Netherlands. Climatic Change 116, 693–704 (2013). https://doi.org/10.1007/s10584-012-0527-0

[2] Priestley, M.D.K., Stephenson, D.B., Scaife, A.A., Bannister, D., Allen, C.J.T., Wilkie, D. (2024) Forced trends and internal variability in climate change projections of extreme European windstorm frequency and severity. Quarterly Journal of the Royal Meteorological Society, 150(765), 4933–4950. Available from: https://doi.org/10.1002/qj.4849

[3] Cusack S: Multidecadal forcing of European windstorm losses in CMIP6-DAMIP models, downloaded on August 15, 2025, from https://stormwise.co.uk/Multidecadal_changes_Europe_windstorms_31July2025.pdf

[4] Krueger O., F. Feser and R. Weisse, 2019, "Northeast Atlantic Storm Activity and Its Uncertainty from the Late Nineteenth to the Twenty-First Century,” Journal of Climate. 1919-1931, DOI: 10.1175/JCLI-D-18-0505.1

[5] Swiss Re, "Winter storms in Europe: messages from forgotten catastrophes" https://www.swissre.com/risk-knowledge/mitigating-climate-risk/winter-storms-in-europe/winter-storms-in-europe-messages.html

[6] Muir Wood 2010 “December 1703 Windstorm: a 300-year Retrospective” available at https://forms2.rms.com/rs/729-DJX-565/images/ws_1703_windstorm_300_retrospective.pdf

[7] Marescot 2014: “Location, location, location: What makes a windstorm memorable?” available at https://www.moodys.com/web/en/us/insights/insurance/location-location-location-what-makes-a-windstorm-memorable.html

[8] Qin, M., Dai, A., & Hua, W. (2020). Quantifying contributions of internal variability and external forcing to Atlantic multidecadal variability since 1870. Geophysical Research Letters, 47, e2020GL089504. https://doi.org/10.1029/2020GL089504

[9] Zhang, R., Sutton, R., Danabasoglu, G., Kwon, Y.‐O., Marsh, R., Yeager, S. G., et al. (2019). A review of the role of the Atlantic Meridional Overturning Circulation in Atlantic Multidecadal Variability and associated climate impacts. Reviews of Geophysics, 57. https://doi.org/10.1029/2019RG000644

[10] Intergovernmental Panel on Climate Change, Sixth Assessment Report, https://www.ipcc.ch

[11] Klawa, M., & Ulbrich, U. (2003). A model for the estimation of storm losses and the identification of severe winter storms in Germany. Natural Hazards and Earth System Sciences, 3:725-732. https://doi.org/10.5194/nhess-3-725-2003