An eye on the skies

Cyclone_Catarina_480p

How accurate are hurricane forecasts and do global insurance premiums reflect the real risk?

On the weekend of 15-16 October 1987, hurricane-force winds struck the UK, killing 19 people, downing an estimated 15 million trees and causing widespread damage to homes and property.

It was a night that BBC weatherman Michael Fish will never live down, failing to forecast the worst storm to hit South East England for 300 years.

This was an isolated incident, but how often do forecasters get it wrong when it comes to extreme meteorological events, and what are the repercussions for the insurance industry worldwide?

Alex Jarman, a postdoctoral researcher in the Department of Statistics|, has spent three years critiquing and improving the techniques used to assess the quality of hurricane forecasts.

His PhD study was funded by Munich Re|, one of the world’s leading reinsurers, whose gross premiums surpass €27 billion, and who rely on accurate seasonal forecasts to set yearly premiums in advance.

The scientific consensus, backed up by the Intergovernmental Panel on Climate Change (IPCC), is that global warming is leading to an increase in hurricane intensity around the world.

 “The fact they can wreak so much havoc is causing concern for the insurance industry and it is crucial they have more reliable forecasts to set premiums which reflect the risk,” he adds.

His PhD thesis involved using statistical models and a different approach to scientists in the meteorological industry in predicting hurricanes.

Part of the problem, he says, is that current annual hurricane data analysis is based on a sample of historical storm observations which is reliable over a relatively short time period – the past 40 years – and does not take into account a changing climate.

Recent evidence shows that Atlantic basin hurricane forecasting has been patchy. In April 2012, US meteorologists predicted a low likelihood of Atlantic basin hurricanes for the June-December season but these forecasts proved wildly inaccurate. The season actually recorded the third-highest number of tropical storms on record, including Hurricane Sandy, which killed 286 people and left a damage bill of US$68 billion.

Well known US meteorologist Joe Bastardi overestimated the number of hurricane landfalls by five in 2006, 2007, 2010 and 2011, and by three in 2009, although his predictions were spot on in 2008.

Two other prominent industry forecasters, Philip Klotzbach and Bill Gray, have also fallen short of the mark with their predictions for major hurricanes, missing Hurricane Sandy completely.

A study by Ria Persad of StatWeather|, a probability-based weather predictions system for the risk management industry, shows that conventional forecasting science is “ahead of the curve” when it comes to tropical storms but far less accurate in predicting whether those storms will develop into major hurricanes.

Persad notes that “when it comes to predicting whether a season will have more or less hurricanes, the public’s guess is as good as the weather forecaster’s.” 

Jarman is quick to point out that weather is not an exact science and notoriously difficult to predict with any certainty.

“There is arguably no such thing as a perfect model for any physical system in the real world, but existing forecasting methods certainly need improving,” he says.

In his thesis, Jarman recommended ‘best-practice’ techniques which, he suggests, could also be used to evaluate financial market forecasts.

His thesis, ‘On the provision, reliability and use of hurricane forecasts on various timescales’, is available at http://etheses.lse.ac.uk/943/|

Useful notes

For more details about the IPCC’s outlook on tropical storms, go to page 216 of the following report|: http://www.climatechange2013.org/images/report/WG1AR5_ALL_FINAL.pdf|

Lloyds has also produced a report on the value of long-range weather forecasting for the insurance industry: http://www.lloyds.com/the-market/tools-and-resources/research/exposure-management/emerging-risks/emerging-risk-reports/climate/forecasting|

17 October 2014

Share:Facebook|Twitter|LinkedIn|