This may seem like a confession only ever heard at a gathering of Meteorologists Anonymous, but I have a favourite windstorm. It's possible you've never heard of "The Greenhouse Low", but in terms of its severity, it's one from which we can learn things, especially with regards the type of event we might see when The Big One finally happens in Europe.
Over the course of two days in early February 1991, an area of low pressure in the central Atlantic accelerated, deepening extremely rapidly with the result that it generated winds in southern Iceland equivalent to those of a moderate category 4 hurricane. It was dubbed the "Greenhouse Low" because of widespread damage to greenhouses in Iceland during this event.
The chart here, taken from a paper in 1995 by Kristjannson and Thorsteinsson, highlights one of the key elements of the storm that made it so damaging: how quickly the storm developed. The dynamics of rapidly deepening windstorms mean they can pack a stronger punch than more "common-or-garden" windstorms. As a way of classifying rapid development of windstorms, a low pressure system will sometimes be called a "bomb" when the central pressure of the low deepens by more than 24 millibars (mb) in 24 hours. The Greenhouse low's central pressure (see chart) dropped by 55mb in 24 hours.
But what did the storm look like on the ground? Iceland has quite complex topography which can act locally to accentuate the impact of the wind. That said, the reported mean wind speed of 56.6 m/s at one location is remarkable as windstorms go. Convert this to a 1-minute sustained windspeed and you get 63 m/s, or about 142 mph: a moderate Cat 4 hurricane. This is a rough-and-ready calculation, but you get the impression here than when the atmosphere is primed just how strong windstorms can potentially become.
The picture here shows the peak gust across the simulation. The actual simulation may not have produced the strongest of winds but you can see the length of the track where notably damaging winds (the black contour showing peak gusts over 40 m/s [90 mph] over the sea) span about 16 degrees of latitude: around 1800km.
There is the potential for a storm like this to cause quite widespread damage if such an event occurred over Europe. The length of the storm footprint is quite startling: by way of geographical comparison, the straight-line distance from Galway in western Ireland to Munich in Southern Germany is 1600km.
There is also a likeness in this event with the (in)famous October Storm that caught the UK unawares in 1987, both in the structure of the event but also the extent of the winds:
The chart above shows the footprint of both events are fairly similar in their extent and length. The mechanisms that led to the most damaging winds in the two storms turned out to be very similar - more about this in a future blog on the what makes the "Perfect Windstorm".
The other item of note is how similar they looked in satellite imagery as they underwent rapid development. The large, curved "cloud head" seen in centre of the charts here is a now well-recognised by forecasters as a signifier for rapid developing of a extratropical cyclones.
(As an aside - the forecasters of the October Storm in 1987 would not have known to recognise the "cloud head" feature in the imagery above. If they had known what we know now about the hallmarks of developing windstorms in satellite imagery, the bad forecast the UK got on that night likely wouldn't have happened...!)
But what does all this mean in the context of catastrophe modelling?
We have here two examples of historical windstorms that had the potential to cause widespread damage. In the case of the October Storm in 1987, it did cause significant damage in SE England and NW France, but its potential, thankfully was not fully realised by way of damage over land. For the most part, the Greenhouse Low had damaging winds over the sea until it hit Iceland.
These are two historical examples of what I like to call a "Tangible Disaster Scenario" - an historical event that, if its track or orientation had been slightly different, could have been a lot worse in terms of the damage it had caused. A recent UCL Natural Hazards for Insurers project by Matthew Raywood at Sompo Canopius tackled this "what if" topic well - I've included a link to the presentation and document that is on his LinkedIn profile.
We can also use the properties of these historical events to stack up alongside the sort of events that cause 50,100,200 year losses in our catastrophe models to see if these type of historical events - specifically their size and shape - exist in the stochastic models. Do these stochastic catalogues contain these very long, fairly narrow, intense windstorms in their tails? Clearly length of track that causes damage is an important factor when discussing cross-country loss correlation (often referred to as 'clash') in the tail of the curve. Knowing that long-track storms such as these have existed in the past can help guide the confidence we have in the tail of catastrophe models.
Much as UK weather forecasters have learnt from the 1987 October Storm to recognise rapid windstorm development from satellite imagery, learning from the past can also be a powerful tool in our understanding and validation of catastrophe models.
A very brief hello! I intend to use this blog to write on anything from a topical subject to a distillation of research that will appear on the...