Long-Range Prediction of Regional Sea Ice Anomalies in the Arctic

Abstract

Gridded fields of sea ice concentration are used to evaluate weekly and monthly anomalies of sea ice coverage in 22 Arctic subregions. The primary period of study is 1972?1988, although statistical comparisons are made with data of lesser quality from 1953?1971. The various time series of regional ice coverage permit the evaluation of ice anomaly persistence as a function of region, season, and lag (forecast range). The fractions of variance explained by anomaly persistence in most regions are considerably larger than corresponding atmospheric values. The fractions typically decrease from 50% to 10% as the forecast range increases from several weeks to several months. Anomaly persistence from the winter months is generally largest, although the regions of greatest persistence-derived forecast skill tend to migrate seasonally with the marginal ice zone. Biases in the regional analyses of the 1950s and 1960s inflate the apparent persistences in the North Atlantic during 1953?1971, but the persistences in most other regions are generally similar in the pre-1972 and post-1972 data. The inclusion of lagged regional cross-correlations provides little increment of forecast skill over persistence at the 1-month range, but this strategy appears to have the potential to enhance the usefulness of ice forecasts at ranges of several months. Analog-based forecasts show statistically significant skill but are generally unable to outperform persistence at the 1-month range.