NSF: New forecasting algorithm helps predict hurricane intensity and wind speed

The National Science Foundation has covered the hurricane modeling research of SMU engineers Yu Su, Michael Hahsler and Margaret Dunham in a Dec. 5 “Discoveries” article on its web site.

Su, Hahsler and Dunham have written a white paper on their method for predicting hurricanes: “Learning a Prediction Interval Model for Hurricane Intensities.” The three scientists are in the SMU Lyle School‘s Department of Computer Science and Engineering.

Read the full story.

EXCERPT:

National Science Foundation
Each year, hurricanes cause tremendous destruction across the globe. It is not a coincidence that the word “hurricane” derives from Huracán, Hunraken or Jurakan, the evil god of winds and destruction in Mayan civilizations of Central America and the Tainos of the Caribbean.

But what makes them so menacing and powerful to deserve such mythos?

“A hurricane’s destructive power is directly related to the hurricane’s intensity–its maximum sustained wind speed,” said Yu Su, a Ph.D. student at the Department of Computer Science and Engineering, Lyle School of Engineering, at Southern Methodist University (SMU).

Yet, predicting the intensity of hurricanes is a difficult challenge.

A team of National Science Foundation- (NSF) funded scientists at SMU’s Intelligent Data Analysis Lab (IDA) developed a new forecasting algorithm called the Prediction Intensity Interval model for Hurricanes (PIIH), to help better predict hurricane intensity.

PIIH also predicts the potential ranges, from high to low, of maximum hurricane wind speeds, specifying the likelihood of wind speeds in varying ranges.

“Accurately predicting intensity means vastly improving hurricane readiness and reducing the risk to property and human life,” said Michael Hahsler, visiting assistant professor for Computer Science and Engineering at SMU. “With more accurate predicting of intensity, governments and the communities they serve will be able to make better decisions on the extent of an evacuation and when to evacuate. This will result in real dollar savings as well as saving lives.”

The PIIH algorithm is based on an aggregate hurricane model that uses previous data, including current maximum intensity, potential for increase in intensity, time of year, various temperature measurements, direction of storm movement and wind shear–the difference in wind speed and direction over a relatively short distance in the atmosphere. PIIH applies this model of past hurricane behavior to predict the intensity of current hurricanes up to five days from any given time point.

“When a future intensity is to be predicted for a current storm, similar states in the life cycle model are found,” said Margaret Dunham, Computer Science and Engineering professor at SMU. “A forecast is created by constructing a weighted average of forecasts from similar storm states found in previous storms. Confidence bands are constructed based upon observing the frequency distributions of intensity values found in previous storms. Based on these and the current intensity value, confidence intervals for future predictions are created.”

By analyzing 2011 storms, through Hurricane Nate, which struck in September 2011, researchers observed that just over 96 percent of the PIIH observations fell within the 95 percent confidence band, which is a very high probability that the PIIH prediction confidence bands were accurate.

Read the full story.

SMU is a nationally ranked private university in Dallas founded 100 years ago. Today, SMU enrolls nearly 11,000 students who benefit from the academic opportunities and international reach of seven degree-granting schools. For more information see www.smu.edu.

SMU has an uplink facility located on campus for live TV, radio, or online interviews. To book an interview in the SMU studio, call SMU News & Communications at 214-768-7650.

This entry was posted in Earth & Climate, Economics & Statistics, Researcher news, SMU In The News and tagged , , , . Bookmark the permalink.

About Margaret Allen

EA-PubAffairs(Periodicals)