Quantcast

Researchers Develop New Model To Predict Seasonal Hurricane Activity

September 11, 2012
Image Credit: Mechanik / Shutterstock

Brett Smith for redOrbit.com — Your Universe Online

Researchers from North Carolina State University have developed a new method for predicting the activity in an upcoming hurricane season that is 15% more accurate than previous models, according to their new report in the journal Data Mining and Knowledge Discovery.

“This approach should give policymakers more reliable information than current state-of-the-art methods,” said co-author Nagiza Samatova, an associate professor of computer science at North Carolina State. “This will hopefully give them more confidence in planning for the hurricane season.”

Current methods used to predict seasonal hurricanes utilize classical statistical methods based on historical trends. These models have become quite complicated because of the large amount of data and other factors that need to be entered for different places and times. All influence storm development.

One of the most important determinations in using these models is weighting certain factors with more significance than others. With only 60 years of historical data to plug into the models, meteorologists are making a series of highly complex calculations with a relatively small data set.

The new “network motif-based model”, developed at North Carolina State, evaluates historical data for all of the variables in all of the necessary places and times in order to accurately weight factors that have the greatest influence on seasonal hurricane activity. For example, while some combinations of factors may correlate to low activity, others may only be influential during high activity hurricane seasons.

After the factors are properly weighed and categorized by the network motif-based model, they are plugged into a program that creates a suite of statistical models that present the probabilities for hurricane activity for the forthcoming season. A typical output of this program might show a 75% chance of high activity, a 15% chance of average activity and a 10% chance of low activity.

In addition to the complexity associated with different factors and their influence during different types of hurricane seasons, the classifications of these activity levels differ from region to region. For example, the North Atlantic high activity level is defined as eight or more hurricanes during the season. Normal activity in the same region is defined as five to seven hurricanes, and low activity is four or fewer.

To check the validity of their new model, the researchers used historical data to compare the new method´s results to subsequent historical events. They found that their model has an 80% accuracy rate of predicting the level of hurricane activity, compared to a 65% accuracy rate for previously used methods.

In addition to confirming the importance of previously identified factors, the model also identified several new predictive groups of factors. The researchers plan to use these newly identified relevant factors to further the study of hurricane variability and behavioral mechanisms. A statement from North Carolina State said the new model could ultimately improve our ability “to predict the track of hurricanes, their severity and how global climate change may affect hurricane activity well into the future.”

The research team´s paper, “Discovery of extreme events-related communities in contrasting groups of physical system networks” was published in this month´s edition of the journal Data Mining and Knowledge Discover.


Source: Brett Smith for redOrbit.com – Your Universe Online



comments powered by Disqus