Extended Forecast of Atlantic Basin Seasonal Hurricane Activity

摘要

Meteorologists have studied about genesis, intensification and cycle life of hurricanes in tropic zones. A fundamental element considered to evaluate is hurricane activity. During last decades, Dr. Gray had developed a statistical model to estimate seasonal hurricane activity in the North Atlantic basin. His forecast scheme is based on a variety of climate-related global and regional indicators. Our main objective was to improve hurricane seasonal activity forecast. We used climatological indices obtain from atmosphere, surface and ocean layers. Moreover, we applied optimization techniques like variable selection scheme, and multicollinearity elimination to select better meteorological indicators related to hurricane activity. We generated seasonal lineal regression forecast equations. We developed a second scheme that estimates seasonal hurricane activity probability. Given that hurricane generation can be represented by a Poisson probability density function, we selected better meteorological indicators to obtain hurricane seasonal probability. Depending season probability we estimated hurricane activity forecast. Finally, we compare lineal regression scheme with Poisson probability model to determinate which scheme provide us better forecast in Atlantic Basin, based on forecast mean square error.