Musings: Hurricane Forecast Calls 2011 Potentially Worse Than 2010

The forecasting team at Colorado State University (CSU) led by Phil Klotzback and Bill Gray say in their first forecast for the 2011 hurricane season that they expect it to be an above-average Atlantic basin tropical storm season along with having an above-average probability of a major hurricane landing on the U.S. coastline and in the Caribbean.  The team acknowledges that for the past 19 years of issuing early December forecasts, they have yet to demonstrate real-time forecast skill.  They have, on the other hand, demonstrated significant real-time forecasting skill with their early June and early August predictions. 

In an interesting discussion, the team explains that the early December forecasting scheme they initially developed in 1991 demonstrated good hindcast skill for the period 1950-1990, but did not give skillful results when utilized for 10 real-time forecasts between 1992 and 2001.  The problem was due to the discontinuation of the strong relationships they had earlier found between West African rainfall and the stratospheric quasi-biennial oscillation (QBO) with Atlantic basin hurricane activity 6-11 months in the future.  At this point they remain at a loss to understand why after 41 years of a good predictive relationship it all changed.

In 2002, they developed a new forecasting scheme, which did not utilize the West African rains and relied less on the QBO.  The newer forecasting scheme gave better hindcast skill but did not demonstrate real-time forecast skill for the four years from 2003-2006 that it was used.  They admit that four years is a short time to demonstrate success, so the next step was to further modify the scheme and try again.  In 2007 they made modifications for the 2008 forecast season.  They continue to use the new system that now relies on only three predictors rather than six as before.  This scheme has demonstrated a 75% success in forecasting above- or below-average tropical storm seasons for the 61-year period of 1950-2010.  Importantly, this new forecast scheme has had a smaller error than climatology in 41 out of the 61 years for a 67% success.  They believe the new forecasting scheme is well-tuned to the multi-decadal active hurricane periods from 1950-1969 and 1995-2010 versus the inactive hurricane period from 1970-1994.

The CSU December forecast calls for 17 named storms, nine hurricanes and five major hurricanes (Category 3-4-5).  There should be 85 named storm days, 40 hurricane days and 10 major hurricane days.  The 2011 forecast begins to look very similar to the 2010 season’s actual results, especially in terms of the number of days for each forecasted category of storm.  The total number of named storms is slightly below the recent past season (17 versus 19) and the number of hurricanes should be less with nine rather than the 12 experienced in 2010.  The forecast does call for a similar number of major hurricanes to last season with five.

Exhibit 8.  CSU 2011 Early Hurricane Forecast
CSU 2011 Early Hurricane Forecast
Source:  Colorado State Univ., PPHB

In preparing its forecast, the CSU team also looks for those analog years that have similar dynamics for the formation of tropical storms and their development into hurricanes and major hurricanes.  The analog years the CSU team has selected for this upcoming season include 1956, 1961, 1989, 1999 and 2008.  It is interesting that most of these analog years have fewer named storms, hurricanes and major hurricanes than the CSU team is predicting in its early December forecast.  It will be interesting to see how, and if, the list of analog years changes as the CSU team begins preparing its later forecasts, which we know have demonstrated greater hindcast and predictive skill.

Exhibit 9.  Analog Years For 2011 Hurricane Forecast
Analog Years For 2011 Hurricane Forecast
Source:  Colorado State Univ., PPHB

Importantly, the CSU team expects that the probability of a major hurricane making landfall on the entire U.S. coastline is 73% compared to the average for the last century of only 52%.  The average for a storm making landfall along the U.S. East Coast including the Peninsula of Florida is estimated at 49% compared to the historical rate of 31%.  Along the important Gulf Coast from the Florida Panhandle to Brownsville, Texas, they are predicting a 48% probability of a major hurricane making landfall versus the past rate of 30%.  The higher probability of a strong hurricane barreling through the Gulf of Mexico and hitting the Gulf Coast with attendant damage to the offshore oil and gas industry will keep energy companies focused on preparing for next year’s hurricane season.