Known indicators point to stormier conditions in the North Atlantic this winter. However, what this means for Europe windstorm losses is much less certain.
Our ability to understand and forecast variability of North Atlantic winter storminess continues to improve year-on-year. Research highlights in 2017 include:
- A new, and skillful, empirical forecast model for winter climate in the North Atlantic revealed that sea ice concentrations in the Kara and Barents Seas are the main source of predictable winter climate variations over the past three decades. Interestingly, a separate 2017 study supports earlier forecasts of either a slowing or reversal of the sea ice reductions in the Barents and Kara Seas between now and 2020, implying an uptick in storminess over the next few years.
- An innovative tool to analyze sources of predictability in a numerical forecast model revealed strong links between tropical climate anomalies and winter climate in the North Atlantic in that model.
Twelve months ago, the forecasting indicators for the windstorm season broadly pointed to a 2016/17 season characterized by below average storminess — a forecast borne out by subsequent observations. We have already had a fairly active start to the 2017/18 season, with Windstorms Xavier, Herwart, and ex-Hurricane Ophelia causing local damage, but what is the outlook for the rest of the season?
The rapid cooling in the eastern tropical Pacific over the past three months is pointing to a La Niña phase of the El Niño Southern Oscillation (ENSO) this winter. The quasi-biennial oscillation (QBO) is in a westerly phase which could persist through the winter, and the solar phase is near the minimum of its 11-year cycle. This configuration of drivers is typically associated with a deeper polar vortex in late winter, producing a more westerly circulation over the Atlantic and favoring increased storminess in the North Atlantic.
We must also consider the state of sea ice in the Arctic. In September 2017, the extent of sea ice in the Kara and Barents Seas was much lower than the 1981-2010 average. However, the 2017 extent was greater than observed during the previous two Septembers, and far above the record low seen in September 2012. This evidence suggests a slight increase in storminess in the North Atlantic compared to recent years. Furthermore, we speculate that the relatively less severe sea ice declines above Siberia this autumn, compared to near-record declines above northern Canada and Alaska, could be an additional indicator of raised North Atlantic storm activity, via hemisphere-scale changes in upper-level air flow.
Figure 1: Average monthly Arctic sea ice extent for September 1979-2017 (Source: NSIDC)
Together, these signs suggest a more active North Atlantic basin in the upcoming season relative to the recent quiet period, particularly later in the winter. This is especially the case if the westerly phase of the QBO persists and La Niña conditions continue to develop over the next few months.
But what does this mean for insured losses?
RMS recently reviewed the growing evidence of predictability of winter weather in the North Atlantic, with the aim of identifying opportunities for the insurance industry. Our findings are published in a detailed research paper available to RMS clients.
We find that most research to date has focused on predicting the North Atlantic Oscillation (NAO) index for the upcoming winter period, defined as “December-January-February” (DJF). However, this forecast parameter has both a spatial and temporal mismatch with annual windstorm losses:
- The winter NAO describes time-mean weather over the whole North Atlantic, whereas severe windstorm losses arise mainly from winds hitting major European cities which represent small areas of dense exposure much further to the east. (In this regard, we note with interest recent research suggesting European windstorms are better explained using the Scandinavian Pattern, which has centers of action directly over Europe.)
- Major windstorm losses can occur outside DJF, including two of the top six historical windstorm losses of the past five decades (Lower Saxony in November 1972 and 87J in October 1987). More recently, the major loss in the 2013-14 season occurred in October (Storm Christian), and the major loss in 2014-15 hit on the last day of March (Storm Niklas).
These dislocations mean that although the winter-mean NAO has some correlation with windstorm losses in the same DJF period, especially in northwestern Europe, the correlation with damage over annual periods is poor. This suggests that even a perfectly skillful forecast of the winter-mean NAO index would have limited utility for insurance purposes.
Thus, we conclude that although there are clear signs indicating the potential for a more active North Atlantic basin during the forthcoming winter, the signs for European windstorm losses over the whole storm season are much more uncertain.