23 July 2014

Nature’s roadblock to hurricane prediction

Posted by Nanci Bompey

By Bob Henson

Hurricane Mitch, the strongest storm observed in 1998, is the second deadliest Atlantic hurricane on record. Mitch caused more than 10,000 deaths, mainly due to torrential rainfall across Central America. Credit: Wikimedia Commons/NOAA satellite image.

Hurricane Mitch, the strongest storm observed in 1998, is the second deadliest Atlantic hurricane on record. Mitch caused more than 10,000 deaths, mainly due to torrential rainfall across Central America.
Credit: Wikimedia Commons/NOAA satellite image.

The quiet Atlantic hurricane season of 2013 came as a surprise to many, as seasonal forecasts had consistently predicted an unusually large crop of named storms. A new study published in the Journal of Geophysical Research: Atmospheres, a journal of the American Geophysical Union, finds that internal variability—processes that unfold without being dictated by larger-scale features—can make one season twice as active as another, even when El Niño and other large-scale hurricane-shaping elements are unchanged. The results suggest that seasonal hurricane forecasts could be improved by conveying the amount of unavoidable uncertainty in the outlook.

The new study by scientists at the National Center for Atmospheric Research in Boulder, Colorado, and North-West University in South Africa focused on the busy 1998 Atlantic season, which produced 14 tropical storms (10 of which became hurricanes). Half of the 14 named storms took shape over the deep North Atlantic tropics, the area south of about 23°N and east of 65°W. About 85 percent of major hurricanes originate from tropical waves that move off Africa into this region.

The study team recreated the 1998 season multiple times using an NCAR-based version of the Weather Research and Forecasting model (WRF). They incorporated the role of internal variability by introducing minor variations in the atmosphere, too small for seasonal forecast models to capture, at the beginning of each model run. This resulted in 16 simulations that had different but equally likely atmospheric features at the start of the key study period (May 1). Ocean temperatures were identical through the season in all simulations.

Despite the similarity of their starting points, the 16 model runs produced strikingly varied results for the season. The total number of named storms in the deep North Atlantic tropics varied from as few as six to as many as 12, depending on the simulation. In the actual 1998 season, a total of seven named storms formed in that region.

Most of the internal variability produced in the simulations was found in two stages of hurricane formation. At the very earliest point, showers and thunderstorms congeal into a tropical wave. At a later point, a fully developed tropical low with warm, moist air at its core goes on to intensify into a tropical storm. At both of these stages, small-scale features such as a blossoming cluster of thunderstorms or a pocket of strong upper-level wind can either nurture or crimp a potential hurricane. The specific timing, locations, and magnitudes of these small-scale features are dependent on internal processes of the climate system and thus cannot be known in advance of the season.

The study did not examine variations in El Niño and La Niña, which are among the main factors used in seasonal hurricane prediction. It’s well established that El Niño years tend to produce fewer Atlantic hurricanes than average, while La Niña years tend to produce more than average.

Although El Niño/La Niña conditions do help in seasonal forecasting, the new study implies that internal variability needs to be considered as well. Over a recent 30-year period (1981-2010), the number of named storms in the study area was as low as zero and as high as 10 from season to season. However, natural variability alone produced a range of storm counts as low as six and as high as 12 in the study’s reproductions of the 1998 season. Thus, natural variability adds a significant component of uncertainty to seasonal forecasts.

“Due to the inherent natural variability, it appears there is an upper limit on how well we can predict hurricane frequency in advance of a season,” said James Done, a scientist at NCAR who led the new study. With more research, he added, it’s possible that tools that explicitly include the amount of natural uncertainty could be developed for use in seasonal hurricane prediction. This would help give stakeholders and the public a better idea of each outlook’s margin of error.

Guest blogger Bob Henson is a writer/editor at the University Corporation for Atmospheric Research, which manages the National Center for Atmospheric Research in Boulder, Colorado. He is the lead writer for the NCAR/UCAR AtmosNews website, where this post also appears.