Atlantic Hurricane Forecast and Climatological Impacts: A Statistical Analysis

A reliable forecast of hurricane activity in Atlantic Basin has the potential to help mitigate the economic losses caused by hurricanes. In general, the insurance industry accepts large risks due to the frequency and severity of extreme events. Catastrophe loss models have become wildly used by insurance/reinsurance industries in order to determine annualized and return period-based losses for a given exposure. One of the difficult problems is to make reasonable forecasts of catastrophe losses based on short record of historical observations. Elevated Ocean and atmospheric conditions influence tropical cyclones development. Considering complex interactions among climatological factors, prediction of future hurricane activity is a challenging course of study. This paper is broken into two parts. In part I, the historical time series of hurricane activity from National Hurricane Center (NHC) are used to generate a forecast set for 1990 through 2010. An autoregressive integrated moving average (ARIMA) is used to model a long-run behavior of Atlantic hurricane frequency. The average number of hurricane per year increases slightly over time and indicates a nonstationary behavior in hurricane frequency time series. Analysis of autocorrelation functions suggests a first-order moving average model, first differences and first-order autoregressive model ARIMA (1, 1, 1) and provides a reasonable technical analysis. We present a comparison of Dr. Gray's forecast with ARIMA model. Results suggest that Dr. Gray's forecast model, in general, is superior to results obtained by simple forecast analysis. Part II focuses on the relationship between the climate signals and hurricane activity. Several climate factors are known to influence hurricane activity: El Niño-southern oscillation (ENSO), North Atlantic oscillation (NAO), sea surface temperature (SST), etc. In this study, principal components analysis (PCA) is used to identifying possible patterns in historical data based on six climate indices, prior and during hurricane season. The objective is to reduce the data set to a smaller set while most of the variability observed in the real data exists. The eigenvectors and eigenvalues are identified to define patterns that characterize the data. The variances observed in an orthogonal system indicate the order of the contribution of each mode shape. The correlation between each shape mode and the number of Atlantic hurricanes per year is examined and the resulting relationships are analyzed.