logo image

Insurance Solutions

Formerly Moody’s RMS

With annual windstorm losses in Europe ranging from a couple of hundred million to tens of billions of Euros, it is no wonder the insurance industry is interested in forecasting winter storminess. However, we cannot let the potential of good returns distract from a full understanding of what winter forecasts really say about future wind losses.

Over the past few years, RMS have been distilling the vast amount of research in this field into key insights for the insurance industry, with a series of annual blogs on the outlook backed up by a more detailed research paper (available to RMS licensing clients). Before we discuss the forecast for this year, we look back to last year’s forecast.

How Good was the Forecast Last Year?

In last year’s blog, we noted how the climate predictors indicated a stormier winter than in recent times, and the 2017/18 season turned out to be probably the costliest windstorm season since 2006/07, with about three billion Euros (US$3.45 billion) of insured wind losses in Europe. While this headline is very encouraging for research efforts, a detailed assessment is more sobering:

  • The forecast contained no spatial information, e.g. on how Germany would bear the brunt of the losses.
  • The forecasted stormier conditions in the later part of the 2017/18 winter did not validate, instead freeze and snow were the top perils.
  • Luck plays a big part when the outcome is close to the expected value of a wide distribution of possibilities.

In brief, the 2017/18 forecast was successful, but provided little information on regional loss, and had some good luck.

Forecast for the 2018/19 Winter Season

The forecast summary is:

Meteorological indicators point to a less stormy North Atlantic this winter compared to last

Two major caveats are attached to this forecast. First, forecasts contain a wide distribution of possible outcomes, therefore this winter could be very different from the expected value.  Second, windstorm losses in Europe have a weak relation to the forecasted meteorological variables.

How Do We Arrive at This Forecast?

This year, we will conduct a more detailed analysis of some aspects of the science underlying the forecast, and emphasize the main messages below in bold font. We describe the status of the main indicators of North Atlantic storminess, namely the El Niño-Southern Oscillation (ENSO), the Quasi-Biennial Oscillation (QBO), solar flux, North Atlantic sea-surface temperature (SST) and Arctic sea-ice; and how they collectively point to a less stormy winter.

1. An El Niño isforecast to occur this winterand this tends to favor negative phases of theNorth Atlantic Oscillation(NAO), meaning fewer storms in the North Atlantic and northern Europe.The left plot in Figure 1 below shows the ENSO impacts on the NAO Index. The spread of possible outcomes is large, due to different patterns of sea surface temperature (SST) anomalies in the equatorial Pacific, a variety of other modulating factors as well as internal noise in the system.

Additionally, researchers found the opposite signal in early winter: the El Niño favors more Atlantic cyclones in the Nov-Dec period.

Of special note this year are larger anomalies in the central than eastern Pacific, and the right panel of Figure 1 illustrates how these central Pacific El Niños tend to produce cooler January-March periods (fewer expected cyclones) in northern Europe.

EUWS activity
Figure 1: Left: histogram of January-March NAO index anomalies for ENSO phases beyond one standard deviation between 1706 and 2000; Right: Temperature at Uppsala in January–March as a function of September-to-February averages of NINO3.4, for central Pacific (right) ENSO type, using data from 1870 to 1995. (From Figures 7 and 13 of Brönnimann, respectively).

2. The current easterly phase of the QBOin the stratosphere also favorsweaker westerly winds over Europeor fewer storms in winter. As with ENSO, the signal is of a much smaller amplitude than the spread. It is notable how the current winds in the equatorial upper stratosphere have westerly anomalies and these have recently been linked to stronger westerly winds over northern Europe in December.

3. The solar driver indicates more westerly winds over the North Atlantic in early winter and cold, non-stormy easterly anomalies in later winter. Gray et al. did a controlled study of the effects of the eleven-year solar cycle on North Atlantic winter weather using long observational datasets and described how the winter-mean impact hides significant monthly variability. Next winter is the fifth since the peak in early 2014 of Solar Cycle 24, and Figure 2 below shows the observed monthly anomalies in mean sea-level pressure in such a solar cycle phase, based on data from the past fourteen solar cycles. The pressure anomalies point to weak anomalous westerlies over northern Europe in December, which flips to quite strong anomalous easterlies (much less storminess) in February.

EUWS activity
Figure 2: the anomalies in monthly mean sea-level pressure (hPa) in the fifth winter after a solar cycle peak, based on observed data from 1870 to 2010. From Figure 5 of Gray et al. (2016).

Labitzke and Kunze analyzed the non-linear interactions between ENSO, QBO, and solar flux on the late winter at high latitudes. Their Figure 4 contains four different Februaries (1966, ’73, ’77 and ‘87) similar to current conditions — easterly QBO, low solar flux and El Niño — and all four had a warm polar vortex in February indicating a less stormy late winter, and consistent with the signals from each driver in isolation. However, the significant storms on April 2, 1973, and March 27, 1987, are reminders of the limited influence of the polar vortex state on European storm losses.

4. The North Atlantic SST anomalies for Septembercontain a horseshoe pattern thatis related to more westerly airflow over northern Europe (or more storms) in the following November to January. This is quite consistent with the early winter signal from the previous three drivers, though not consistent in January.

5. Arctic sea-ice is the last factor we consider for this winter’s outlook. There are two aspects here, and we begin with the second-order effect as it is clearer, then discuss major issues concerning the big picture of sea-ice decline.

Sea-ice extents in September fromNSIDCshow more sea-ice on the Canadian side (especially Beaufort Sea) and less off the Russian coastline (especially Laptev Sea) compared to last year.This change favors a negative NAO, though its impact is small, and much smaller again when converting NAO to loss impacts in Europe.

The big picture is one of long-term decline of Arctic sea-ice, and researchers have shown how declines, especially in the Kara and Barents Seas, are strongly linked to negative NAO, e.g. Honda et al.Yang and Christensen and many others. However, observations contain a puzzle: sea-ice extent in September in the Kara Sea over the past 15 years is about one third of the value from 1980 to 1999, yet the winter NAO over the past 15 years has been on average slightly positive. Where is the sea-ice forcing of NAO?

A plausible explanation is that the winter NAO responds to transient, year-to-year change in sea-ice, and this fits with some results (e.g. the Wang et al. model de-trend all time-series, and Honda et al. do seasonal experiments).

The central question for insurance is whether wind losses are following the total sea-ice extent, or those interannual transients captured by the NAO. Recent December to February periods with significant positive NAO and low sea-ice (2011/12, 2013/14, 2014/15, 2015/16, 2016/17) have produced Europe-wide wind losses less than the median of the Dec-Feb loss distribution. This indicates long-term losses are more influenced by the total sea-ice extent rather than NAO in recent times. Of course, the low losses in those positive-NAO winters could occur by chance from a small sample, however, there is a compelling argument to indicate this is a signal:

  • Windstorm losses are caused by the tiny fraction of most severe storms
  • And the severities of the most extreme storms are limited by the heat contrast between the sub-polar and sub-tropical air masses
  • And the sub-polar air mass is warming much faster due to Arctic Amplification
  • Hence sea-ice loss has more impact on those extreme storms driving loss, compared to the everyday cyclones reflected in the winter NAO

The evidence suggests the new, lower Arctic sea-ice climate regime could be shifting the balance towards extreme storms becoming rarer, and the NAO is not reflecting this longer-term change. If true, then this winter could be in a low quantile of the long-term climate of winter losses. Besides being an untested hypothesis of current risk, the near-term outlooks for sea-ice extent and sub-tropical temperatures are uncertain, and more research is needed before changing views of windstorm risk.


This year, the climate drivers point to more cyclones earlier in the season, and fewer after the New Year. Given the relative importance of the Jan-Mar period, we expect a less stormy North Atlantic sector this winter compared to last. The usual caveats apply: the forecast contains a wide distribution of possible outcomes hence winter could be very different from the expected value, and windstorm losses in Europe have a weak relation to the forecasted variables.

You May Also Like
Storm Formation
November 06, 2020
Examining the European Windstorm Outlook This Winter
March 05, 2020
Update on Multidecadal Variability of European Windstorms

At this moment, you might expect a blog about European windstorms to be about recent Storms Ciara-Sabine, Dennis and Jorge causing wind and flood losses of a couple of billion euros in Europe. However, the losses this winter are modest in a longer-term context. Instead, I think the recent insights into longer-term variations in wind losses could have much more impact on how we price windstorm risk. We first noticed multidecadal variability of European windstorm activity ten years ago, with 50 percent lower frequencies of damaging storms in the new millennium than in the eighties and nineties. This variability is important: a company’s length of loss experience is unlikely to match the model calibration period, which impacts model validation. It also held the promise of improved risk management, if the storminess changes could be anticipated. We needed to know more about it. Hundred-year records of wind data at several stations from the Dutch weather service, KNMI, showed pronounced multidecadal variability of storminess throughout the period. An RMS review of published work found: A wide variety of observational evidence of multidecadal variability of storminess in Europe An incomplete understanding of multidecadal drivers, including mixed results from earlier climate models What has the past decade of wind data and research taught us about these slow variations in storm activity? The next few sections will show we have learned a lot over the past ten years. Observed Storminess: What are the Changes? The Dutch study was recently updated with nine additional years of wind observations from KNMI, to span 1910-2019. The wind data was homogenized using the same procedures to ensure changes in storminess reflected real meteorological causes rather than observational changes. Figure 1 below shows the main result: the twenty-first century lull has extended throughout the 2010s in the Netherlands. Figure 1: Time series of annual storm loss in the Netherlands, with 10-year running meansStorminess on the continental-scale was measured using a preliminary dataset of European windstorm footprints spanning 1972 to 2018. Each footprint consists of winds at 25 kilometer cells, and a Europe-wide loss index was computed for each historical event. The aggregate loss was computed for each year, then transformed into a standard normal distribution. The Europe-wide annual losses are shown in Figure 2 below, together with annual rates of occurrence of damaging windstorms, also standardized for comparison. Figure 2: Time series of standardized annual aggregate loss index and storm occurrence rates, and their five-year running means, for the whole of EuropeStorm Climate Drivers What could be the drivers of the multidecadal storm variation? Over the past ten years, researchers have made huge advancements in understanding European winter climate variability through analysis of observations and experiments with better climate models. Specifically, they have identified heat anomalies in the North Atlantic Ocean and the Arctic as two main drivers of our winter wind climate at decadal and longer timescales. North Atlantic Ocean Forcing Gulev et al. (2013) examined observations over the 1880-2007 period and identified an ocean region in the central northern North Atlantic which forces the atmosphere at decadal and longer timescales, as shown in Figure 3 below. The resonance in this ocean area is caused by the co-located storm track. Figure 3: The observed correlation, at decadal timescales, of surface heat fluxes and sea surface temperatures in the period 1880-2007; Figure 1b of Gulev et al. (2013). Positive correlations indicate ocean temperature anomalies produce surface flux changes of same sign, to drive the atmosphere.Researchers had been reporting confusing, mixed results from climate models. The situation has since been clarified by Scaife et al. (2012), who explained the need for a high model top and vertical resolution to simulate mid-latitude winter signals. There are few tests with appropriate models. Omrani et al. (2014) found a significant ocean forcing of atmosphere winds, about one half of the corresponding observed anomalies over Europe, from a high-top model. Peings and Magnusdottir (2014) used a more modern climate model and found roughly the same result, and indicated that the same area as Figure 3 was the key to ocean forcing of atmosphere on long timescales. A cooling in this key area raises storminess, and vice versa. Arctic Forcings There have been some remarkable changes in northern hemisphere winters over the past 20 years or so: the Arctic winter warmed at a rapid rate, while mid-latitude winters have warmed slower than the global trend, and even cooled in some areas. Observations reveal a strong link between declining sea ice and a stronger Siberian High. Mechanisms to explain this were reviewed in Cohen et al. (2020), and they viewed the process depicted in Figure 4 as most robust. Figure 4: A schematic of the process linking Barents-Kara sea ice to European winter climateThere has been a long debate on the size of the Siberian High anomaly forced by Arctic warming, fanned by mixed climate model results. This is getting resolved by recent research indicating climate models have special requirements to simulate mid-latitude winter signals, such as high model top and fine vertical resolution and fully coupled ocean-atmosphere models.  Six modern climate models meeting most requirements had a consistent signal of sea ice loss causing a stronger Siberian High, and a weakening of the westerlies carrying storms into Europe. Their modeled signal is about half of the observed circulation change over the past few decades. The Outlook How are the two main drivers likely to evolve over the next several years, and what does this entail for European windstorm activity? North Atlantic Ocean Outlook Figure 5 shows the mean temperature anomalies (lower plot) for the key ocean area (upper plot). Will the recent cooling continue, or reverse? There are no published forecasts of ocean heat in the specific key region for the next decade. Instead, we estimate changes based on the known drivers of heat in this area, and conclude the likeliest outcome is for persistence of these cooler anomalies in the next few years, implying the stormier North Atlantic of recent years will continue. Will they raise windstorm loss? That depends more on the second driver, and what it does to the Siberian High. Figure 5: Time series of mean temperature anomaly in the top 400 meters of the ocean in November to April (lower plot), for the region off Newfoundland indicated by red box (upper plot). Ocean temperatures from EN4 were de-trended to remove global warming signal.Arctic Outlook Anthropogenic forcing is the main cause of Arctic warming and sea ice decline, and the IPCC indicate this is very likely to continue in the future. On shorter timescales, Årthun et al. (2017) describe how North Atlantic Ocean heat anomalies are carried north to modulate sea ice in the Barents and Kara Seas. The northern Atlantic has been cool since 2015, and Figure 6 below shows winter sea ice area in this region has risen since its nadir in 2016. The northern North Atlantic waters are expected to remain cooler over the next few years, and this could stabilize sea ice, or perhaps even reverse its longer-term decline. Raised sea ice in this region would lift windstorm losses. Figure 6: The sum of sea ice area in Barents and Kara Seas for December and January. Source: NSIDCUncertainties in the Outlook There are significant uncertainties in the multiannual forecast: Could the known drivers evolve unpredictably, in this time of changing climate? Could an unexpected climate process, such as a mode of tropical variability, or ongoing anthropogenic forcing of the stratosphere, or simply random variability, emerge to dominate? Will an explosive volcanic eruption occur, to raise windstorm risk for the following few years? Summary The continental-scale windstorm lull in the first decade of this century continued through the past ten years, with notable regional variability. There have been several advances in understanding mid-latitude storm climate variability: Heat anomalies in the North Atlantic Ocean, underneath the storm-track, affect winter storminess in the Atlantic sector Heat anomalies in the northern North Atlantic modulate Arctic sea ice, notably in the Barents and Kara Seas Barents and Kara sea ice modulate the Siberian High to affect European storminess Both observation-based analyses and fit for purpose climate models support these processes In brief, heat anomalies in the North Atlantic Ocean and Arctic regions explain at least half of the recent multidecadal change in storminess. The best estimate for the next few years is a slight upward trend in windstorm activity, with significant uncertainty. The new challenge for insurance is how to benefit from climate forecasting skill, while maintaining safe management of European windstorm risk. We explore this in an upcoming RMS white paper. Debate is positively encouraged, please contact our Product Manager Michele.Lai@rms.com in the first instance.…

Stephen Cusack
Stephen Cusack
Senior Director, Model Development

Stephen is a Senior Director in the Hazard Climate team. After joining RMS in 2009, most of Stephen’s focus has been on researching and developing the Europe Windstorm (EUWS) model, with particular focus on the hazard. Stephen also spent 18 months leading the recalibration of the U.S. and Canada Severe Convective Storm model, released in January 2014. Before RMS, Stephen worked in a wide variety of research and development posts during 13 years at the U.K. Meteorological Office.

cta image

Need Help Managing Your Portfolio?

close button
Overlay Image
Video Title

Thank You

You’ll be contacted by an Moody's RMS specialist shortly.