The failure of preseason predictions for the hurricane season of 2006

The preseason predictions of an extremely active hurricane season were spectacularly wrong. Only nine named storms and five hurricanes formed in the Atlantic, one below the average of ten named storms and six hurricanes. We ended up with the quietest hurricane season since 1997, much to the relief of regions ravaged by the unprecedented activity of 2004 and 2005. What happened to make the prognostications such a abysmal failure?



Dry air and dust
A significant reason for the failure appears to be the unusual amount of dry air laden with African dust that came off the Sahara Desert during July and August. Hurricanes need moist air at mid-levels of the atmosphere in order to form, and recent research suggests that the dust within the dry air may act as an inhibiting factor as well, through some not well-understood process. In addition, these dry air outbreaks are frequently accompanied by a strong jet of easterly winds that brings hostile wind shear over the Atlantic. As seen in a plot of the relative humidity at 500 mb (roughly 18,000 feet altitude) in August of 2006 (Figure 1), there was much drier air than usual over a large portion of the tropical Atlantic where hurricanes like to form.

Why was there so much dry air and dust? During the early part of the rainy season (May-July) in the southern Sahel region of Africa, precipitation was well below average (Figure 2). Rainfall was also below average in this region in 2005, and these factors could have contributed to more Saharan dust being blown out over the tropical Atlantic in 2006.

Figure 1. Departure from normal of relative humidity for the month of August, at 500 mb (about 18,000 feet altitude).

Figure 2.Departure of precipitation from normal for May-July 2006 for Africa. Note that the southern Sahel region (approximately 10-15�N, 0-15�W) had much below average precipitation, and this likely contributed to the dry air and widespread Saharan dust outbreaks observed over the tropical Atlantic this year. Image credit: Bill Gray and Phil Klotzbach's Summary of 2006 Atlantic Tropical Cyclone Activity and Verification of Author's Seasonal and Monthly Forecasts at Colorado State University.

El Ni�o
The other reason the hurricane season of 2006 was so mild is probably due to the arrival of El Ni�o conditions in September. It is well known that when the warming of the Equatorial Pacific waters off the coast of South America brings about an El Ni�o event, the number and intensity of Atlantic hurricanes is sharply reduced. Conversely, action in the Eastern Pacific is enhanced, and we saw both of these effects in 2006. The reason usually given for this lack of activity in the Atlantic is an increase in wind shear. The warm waters of the eastern Pacific lead to more rising air than usual there, and when that rising air hits the top of the troposphere (the lower atmosphere), it spreads out and creates strong upper-level winds that blow from east to west towards the Western Pacific, and west to east over the Atlantic. These strong upper-level winds create hostile wind shear that tears apart developing hurricanes. However, in 2006, it appears that El Ni�o-induced wind shear was not a serious impediment to Atlantic hurricane formation. Wind shear was near average over the Atlantic during most of hurricane season (Figure 3). There are additional reasons El Ni�o suppresses hurricane activity, and foremost among these is the introduction of stable, sinking air over the Atlantic. It is likely that El Ni�o brought such conditions to the Atlantic during large portions of the 2006 hurricane season, significantly inhibiting hurricane formation.


Figure 3. Wind shear over the tropical Atlantic. Except for a big above-normal spike in October, wind shear in 2006 (blue line) was near normal (black line) over the tropical Atlantic during hurricane season. Image credit: NOAA/CIRA.

So why did the long range forecasts fail?
The long range seasonal hurricane predictions are statistical in nature--they look for early season patterns in winds, pressures, and ocean temperatures that occurred in years past that one can combine to make a skillful prediction of the hurricane season. One of the variables these prediction schemes typically do not include is the early season rainfall in Africa. If it is an unusually dry and dusty year over the Atlantic like 2006 was, then the forecast is going to be wrong. The other problem was the unusual nature of the El Ni�o event that developed this year. We went from La Nina conditions in March to a full-fledged El Ni�o in September. This was by far the largest percentage warming of SST anomalies between June-July and August-September in the tropical Pacific for a year that had El Ni�o conditions in August-September. In addition, the the timing was unusual--it is uncommon for El Ni�o events to start in the Fall. Since the historical record had very few cases mimicking the behavior of this year's El Ni�o event, it is no wonder that the statistical models which rely on past years' data to come up with forecasts of hurricane activity failed. El Ni�o behaved too strangely this year to anticipate, and the computer models had no idea it was coming until about March or April.

Jeff Masters