Predictability of ice concentration anomalies in the high latitudes of the North Atlantic using a statistical approach

Abstract

Based on a 27 year data record from the COADS and SEIC data sets, a statistical analysis of ice concentration, sea surface temperature (SST), air temperature, U and V wind components, and sea level pressure anomaly data was conducted for five locations in the ice-covered waters of the North Atlantic. Spectral densities and autocorrelations of the time series for each variable were calculated to establish a measure of persistence and periodicity. Regression equations were formulated based on the above data sets to forecast both the winter and summer ice concentration anomalies for each location. The differing effects of land and ice boundaries, currents, storm passages and wind velocity anomalies on the ice concentration anomalies at each location were reflected by the parameters retained by each of the regression equations. In addition to ice concentration anomalies at various lags, the inclusion of meteorological and oceanographic parameters was shown to increase the total explained model variance, which should improve the accuracy of an ice concentration anomaly forecast at lead times of at least one season over a forecast based on ice concentration anomaly persistence alone.