HR: 0800h
AN: NG31A-0848    [Abstracts]
TI: Nonlinear Prediction of Atlantic Hurricane Abundance in 2004 Season
AU: * Tang, J
EM: jtang@gmu.edu
AF: CESOR, School of Computational Sciences, George Mason University, 4400 Univeristy Drive, MS5C3, Fairfax, VA 22030 United States
AU: Barbara, D
EM: dbarbara@gmu.edu
AF: ISE Dept., School of Information Technology and Engineering, George Mason University, 4400 University Drive,, Fairfax, VA 22030 United States
AU: Yang, R
EM: ryang@gmu.edu
AF: CESOR, School of Computational Sciences, George Mason University, 4400 Univeristy Drive, MS5C3, Fairfax, VA 22030 United States
AU: Kafatos, M
EM: mkafatos@gmu.edu
AF: CESOR, School of Computational Sciences, George Mason University, 4400 Univeristy Drive, MS5C3, Fairfax, VA 22030 United States
AB: Hurricanes are an eddy phenomenon in the weather system. Each year hurricanes form out of the global atmosphere circulation in a couple of months and then disappear. One of the directions in hurricane research is to discuss the characteristics of annual hurricane behavior in order to answer questions such as how many hurricanes may occur in a coming hurricane season (i.e., the hurricane annual abundance or frequency) and how long they will last (i.e., the hurricane duration) [1]. Usually the historical best track data are utilized for analysis of the climatology of hurricane seasons. The time series of the annual abundance of North Atlantic hurricanes over the period 1886-2003 [2] shows a series of bursty data with self similarity, which suggests that a nonlinear prediction method based on fractal dimension may be good to describe the abundance of hurricane season in 2004. This paper tries to use F4 [3] - a nonlinear prediction method based on fractal dimension to forecast the number of hurricanes in 2004 season. The hurricanes annual frequency in the Atlantic basin from 1886 to 2003 is extracted from the NHC best track data [2]. The whole series contains 118 data points. A set of experiments is designed as followed. A 2004 estimate is computed by F4 using all the 118 data points. That is, the whole data series from 1886 to 2003 is used to forecast the number in 2004. A second estimate is computed by F4 using the series from 1886 to 2002, totally 117 data points. Of course F4 will produce estimates of the number of hurricanes in both 2003 and 2004 seasons. The idea is to test F4 estimate with the actual observation in the 2003 season at the mean time to make a forecast of the 2004 season. This idea can be applied where the series from 1886 to 2001, to 2000, to 1999, until to 1994 are used respectively to make estimates till 2004. Overall, F4 states that there will be 6-7 hurricanes in the Atlantic basin for this 2004 season. References: [1] James B. Elsner and A. Birol Kara, Hurricanes of the North Atlantic: Climate and Society, Oxford University Press, June 1, 1999, ISBN: 0195125088. [2] NHC/TPC archive, http://www.nhc.noaa.gov/pastall.shtml, last accessed on Sep 09 2004. [3] Deepay Chakrabarti and Christos Faloutsos, 2002, F4: Large-scale Automated Forecasting Using Fractals, in Proceedings of the 2002 ACM CIKM International Conference on Information and Knowledge Management (CIKM 2002), McLean, VA, USA, November 4-9, 2002, pp 2-9.
DE: 3250 Fractals and multifractals
SC: Nonlinear Geophysics [NG]
MN: 2004 AGU Fall Meeting