Notes/commentary on hurricane season forecasts
August 14, 2007
(Inspired by / in response to Brendan Loy’s commentary on hurricane season forecasts)
Background / history of hurricane season forecasts
A pair of papers published in the September 1984 issue of the American Meteorological Society’s Monthly Weather review outlined the basis of a forecast scheme devised by Colorado State University’s William Gray that could be used to predict the number of tropical and hurricanes that would occur in the season ahead. Under the original scheme a forecast would be released in June, with an update in August. Further research conducted in ensuing years gave Gray and his graduate students confidence that a good forecast could be put out even earlier, in December and such forecasts started being released in 1992.
At first, this was a fairly unnoticed experimental research exercise, somewhere along the way insurance companies realized an interest in the research and became a source of funding for it. Somewhere along the way, due to early success in the forecasts, they were released to the media. (I can recall them being reported in the early-mid 1990’s. (1994 Virginian Pilot article on the Gray forecast, for example.)
Until 1998, the Colorado State team was the only player in the seasonal forecasting game. The next effort was borne out of an effort “to assist the competitiveness of the UK insurance industry by using the UK science effort to improve the assessment of risk”. It is now known as Tropical Storm Risk. To my knowledge they don’t do much to push their forecasts to the public, they just make them available.
In 1999, a specialized agency of the National Weather Service, the Climate Prediction Service quietly issued a brief outlook on the 199 hurricane season. In 2001, this outlook was expanded to give a range of numbers for forecast activity, rather than a general ‘below average’ or ‘above average’. By 2003 this outlook was being pushed out to the media via press release and press conferences were being held by 2005.
In 2006 a team from NC State University issued a seasonal forecast at the AMS hurricane conference. It was little noticed at the time, but became much more realized after their forecast of a quieter season than anyone else anticpated was verified (NCSU team called mild hurricane season).
My thoughts on seasonal forecasts
I don’t take too much interest in them personally and don’t like how they are being pushed to the general public. They are a experimental works in progress and should be treated at such. I am most displeased with NOAA’s trumpeting of their forecasts. It gives the public the sense that these are operational forecasts that are on par with the other forecasts of the National Weather Service and that is definitely not the case. Brendan’s post included a question he was posed, “Where are all the hurricanes the NHC had forecast for the last 2 years? just curious as to why we should panic over predictions that have little or no accuracy?”) This shows the confusion that the hurricane season forecasts cause because the National Hurricane Center is not the agency that puts out the seasonal forecast and, as I just said, the seasonal forecasts do not have the same accuracy as the operational forecasts put out by the NHC.
As one can discern by examining who backed the earlier forecasting ventures, these forecasts are most useful for people who have a stake in the macro-scale, namely insurance companies. They are of little value to individuals. National Hurricane Center forecaster Richard Pasch said it best: “An active year is the year when you get hit.” These forecasts as they are know don’t tell people that and may instead give them a false sense of security when an inactive year is forecast. (1992, the year of Andrew was a quiet season if you didn’t live in South Florida).
On ‘crying wolf’, overforecasting seasons
This Excel file contains all of the Colorado State forecasts from 1984 to to 2006 and include what the actual numbers ended up being. There have been three seasons that were significantly overforecast: 1993, 1997, 2006. The common thread between those three years: El Niño conditions that were either totally unexpected by the forecast team or were stronger than expected. On the flip side, there have been an equal number of seasons that were significantly under-forecast: 1995, 2001, and 2005. Both TSR and the CPC had similar issues for the seasons that they were doing forecasts.
On ‘cooking the books’ , naming storms that wouldn’t have been named in the past and/or are undeserving of a name
One of the comments to Brendan’s post raised this issue, which has come up in recent years as an allegation that the number of storms are being inflated to help meet a forecast number.
There is a bit of truth to it, there are some storms in recent years that were classified as tropical storms that wouldn’t have been in the past. This is because meteorologists have many more tools in the toolbox than they used to. There is a mountain of data available from remote sensing that wasn’t available ten years ago and is that data that enables forecasters to realize that some of those storms that once would have been considered non-tropical do in fact have tropical characteristics (Chantal being a prime example of this, based solely on appearance it would have been considered non-tropical).
This concern to make sure that every tropical storm gets classified is not restricted to the present. The Re-Analysis Project of the Hurricane Research Division is devoted to reviewing as much as possible on past seasons to find storms that were missed or were stronger than originally thought.