FLHurricane.com: The Seasonal Forecast Report Card

The April forecasts for the total number of storms for the 2011 Atlantic Hurricane Season have been made by various agencies/individuals and the input is as follows:

CSU 16/9/5
TSR 14/8/4 (the TSR update on 5/24 maintains the same forecast)
Accwx 15/8/3
JB 13-15 Named Storms

These forecasts all reflect a narrow range, with the numbers down considerably from last year – primarily due to an outlook for a much cooler Atlantic basin with SSTs about 1C cooler, which is indeed a significant loss of thermal energy. Mid-level upper air temperatures are also lower than last season – globally.

How well do the April seasonal forecasts verify? The answer is ‘not very well at all’. CSU has archived forecasts available back to the 1999 season. In the past 12 seasons they have made 6 good forecasts and 6 bad forecasts – which translates to ‘zero’ skill, i.e., no better than even odds for any given season. The statistics for each season are as follows:

APRIL FORECAST ACTUAL SCORE
1999….14/9/4……….12/8/5…...4 (good)
2000….11/7/3……….15/8/3……5 (good)
2001….10/6/2……….15/9/4……10 (bad)
2002….12/7/3……….12/4/2……4 (good)
2003….12/8/3……….16/7/3……5 (good)
2004….14/8/3……….15/9/6……5 (good)
2005….13/7/3……….28/15/7…..27 (poor)
2006….17/9/5……….10/5/2……14 (bad)
2007….17/9/5……….15/6/2……8 (bad)
2008….15/8/4……….16/8/5……2 (excellent)
2009….12/6/2……….9/3/2……..6 (bad)
2010….15/8/4……….19/11/5…..8 (bad)

In the first six years they made 5 good forecasts and one bad forecast and in the next six years they made one good forecast and 5 bad forecasts which again suggest the lack of any refinement in the development of forecast skill for this type of a forecast. Perhaps part of the problem was associated with some changes in forecast methodology in the past few years, or perhaps, more realistically, seasonal forecasts (including my own attempts based primarily on forecasts) still have a long road ahead before a better reliability is achieved.

The seasonal forecasts were originally developed for ‘Insurance’ purposes – and they are probably still used by various insurance companies for planning potential claims payments. Note that the range of named storm forecasts in the 12-year period has been from 10 to 17 and the actual totals ranged from 9 to 28. Two hyperactive seasons were forecasted but did not verify (2006 & 2007) but one did come close (2007). On the other side of the coin, four hyperactive seasons did occur that were not forecasted in April (2003, 2005, 2008 & 2010).
ED

Edited by Ed Dunham (Fri May 27 2011 12:29 PM)