Long-Range Forecasting of Iceberg Season Severity

The development of statistical schemes for the long-range prediction of Grand Banks iceberg season severity is described. Empirical Orthogonal Function (EOF) Analysis is employed to isolate the dominant modes of variability in atmospheric pressure, geopotential height, thickness and air temperature predictor fields over a broad region of the North Atlantic. Correlations between eigenfunction coefficients and an iceberg severity index are employed to isolate predictors for input to a multiple regression model. Results of two hindcasts for the interval 1951-1980 are presented to show the achievement of high skill in predicting the occurrence of extreme iceberg seasons. One of the demonstrated models correctly predicts the numerical rank (from a population of 29 years) to within
\pm 1
unit in seven of the eight most extreme seasons.