P5.16

**Prediction of trends of tropical storms in the North Atlantic basin**

**Nazario D. Ramirez**, University of Puerto Rico, Mayaguez, Puerto Rico; and J. M. Castro, J. Gonzalez, and M. Angeles

A statistical model is developed to estimate the annual seasonal hurricane activity in the North Atlantic basin. The purpose of this work is to estimate what would be the expected hurricane activity during 2006 – 2030. The suggested methodology consists on four major steps: identify the years with similar meteorological characteristics, develop a statistical model, perform data adjustment and predict the hurricane activity. The first step consists on identifying the years that exhibit similar meteorological characteristics to the predicted year. The areas of the Earth that are likely to exhibit significant correlation with the number of tropical storms are selected and dimensionality reduction is applied into the vertical levels. The empirical orthogonal functions (EOF) are input into a self organized neural network to determine the years with similar characteristics to the predicted year. The second step includes implementing an optimal variable selection algorithm to develop a parsimonious regression model without multicolinearity problem and with the largest coefficient of multiple determination. The variable selection algorithm is based on a combination of a stepwise regression with a parsimonious-multicoliniarty algorithm to select the best regression predictors. The variable selection algorithm identifies the horizontal location, the vertical level and the specific variables that are highly correlated with the number of tropical storms and hurricanes for a specific year. The third step is to adjust the outputs from the Global Numerical Model (GCM) to the NCEP/NCAR reanalysis data. The GCM simulates the atmospheric conditions by using the business as usual scenario, which was defined by the Intergovernmental Panel on Climate Change (IPCC). This simulation shows the possibility of controlling the CO2 emissions in the next years using renewable energy, by the commitment of the highly industrialized countries in reducing the dependency of the petroleum. According to the results of this simulation, the sea surface temperature will stay an increasing tendency, given to greenhouse effect. In addition, these results indicate that rainfall average in the planet did not change significantly; however, the periods of drought of high rain levels will be more intense. Therefore, the variability of the synoptic conditions, in the surface and the atmosphere in the future will be different from the past. The statistical model adjustment is accomplished by using a regression model that includes two additional correction factors. The first factor is additive and the second factor is multiplicative to correct the first and the second moments of the process, respectively. The correction factor is applied only during an overlapping period (2000-2005), i.e., when observations from NCEP and estimation from GCM are available. The fourth step consists on using the corrected GCM outputs to evaluate the developed regression models. The final output presents the ensemble prediction and the confidence interval for the expected number of tropical storms and hurricanes during the period of 2006-2030.

Poster Session 5, Tropical Cyclone Modeling and Prediction

**Tuesday, 25 April 2006, 1:30 PM-5:00 PM**, Monterey Grand Ballroom** Previous paper Next paper
**