On the irregular nature of hurricane seasons
July 17, 2010
A recent post by Dr. Masters, which noted a few more hurricane season predictions inspired this post. He points out a prognostication by Penn State forecasters of 19-28 named storms and the UK Meteorological Office’s forecast (July-November) of 20 storms. These forecasts are a bit above the 17 storm consensus derived from the Colorado State, Florida State, N.C. State, NOAA, and Tropical Storm Risk forecasts (compiled in a table on my post regarding 2005-2009 predictions). Given that there’s only been one named storm so far with none on the horizon, the Penn State and UKMET forecasts may be overdone, but the others are not necessarily busts.
There are two of the problems associated with public expectation settings caused by forecasts released in May and early June . One is that even though the Atlantic hurricane season begins on June 1, the bulk of activity occurs from the end of July to the end of October. Similarly, as freshly minted Dr. Ryan Maue (congratulations!) mentioned in a note expressing skepticism over the high end of the season forecast numbers, activity is not evenly distributed. It just doesn’t work to say “OK, 20 named storms over 5 months… 20 weeks, 20 storms, we’ll get a storm a week.” Intra-season oscillations cause periods of inactivity along with ones featuring multiple storms. Even relentless seasons like 2005 have spots of downtime (I think of the period from after Emily (July 21) through the death of Tropical Depression 10 (on August 18) as the “2005 phony season”**. The two week lull in mid-August caused 2005 to briefly fall behind 1995’s pace.)
To amplify this point, I went back to the Unisys hurricane season archives and looked at tropical cyclone formation dates for eventual tropical storms during the 1995-2009 time-frame, grouping them by 10 day periods starting on June 1. Only seven seasons had cyclone formation during half of the periods during hurricane season. 2005 lead the way, of course, with formation occurring in 78% of the time-steps. But still, it had gaps in peak season; one being the aforementioned Aug 10-19 period (when TD 10 formed but failed to develop further) and Sep 19-28 (the tropical depression that became Rita formed on the 18 and nothing else formed until October 1). Grouped this way 1995 had quite a streak going from July 1-August 29, but fell quiet in the beginning of September. That was the last time the first August 30-September 8 dekad didn’t have any tropical depressions that became tropical storms; every season’s since has had at least one while 2002’s, 2003’s, and 2005’s had three.
The August 30-September 8 period is more or less centered around the time-steps during which tropical cyclone formation is probable, starting on July 31 and extending to October 18. Our shallow record of seasons with 20+ named storms (2005 and 1933 being the only examples) indicates that having exceptional activity (4-5 storms) before then would be necessary to rach such high numbers. 15 named storms is quite possible despite lack of tropical cyclogenesis before the end of July, as 2004 and 1998 aptly demonstrate.
That past history, combined with current large-scale conditions, indicate that forecast consensus of 17 named storms remains on track. Unlike the last time a truly extreme season was forecast, 2006, the El Niño Seasonal Oscillation has not given any surprises. Back then, it offered a surprise when El Niño conditions unexpectedly formed. This caused most season forecasts to be epic failures. In 2010, however, the expected transition from El Niño to La Niña has occurred. Also, expectations of exceptionally warm sea surface temperatures in the main development region of the tropical Atlantic continue to be met. Below is a chart of the departure from 1970-2000 norms:
Another item demonstrating this point is on Dr Maue’s weather page. Along with a multitude of other useful charts, one can find of comparison of current SST conditions to those at this point in past seasons, such as 2010 compared to 2005 ***. Not only is the eastern end of the MDR significantly warmer than normal, it is warmer than in 2005 as well. The upshot is an increased number of storms. A possible benefit relative to 2005 is that storms will form further east and at least have a chance to recurve at sea rather than making landfall.
So, even though it is quiet at the moment, I do not expect that to be the case two weeks from now. While a casual glance of the season forecasts gives appearances that all are in trouble, the 17 storm consensus will at least come close to verifying.
**Taking after the October 1939-April 1940 “phony war” pause between Germany’s invasion of Poland the later attack on western Europe
*** I chose to link to this chart because it’s focused on the Atlantic. The downside of this is that it’s only a 3 day average, which causes it to be affected by short term events. The eastern Gulf of Mexico is currently significantly warmer than it was at this point in 2005 due to the then recent passage of Hurricane Dennis. Comparing 2005 charts for that day and different days later on do not show the eastern MDR difference to be caused by a similar short-term event, as far as I can tell.