An Analog Method for Seasonal Forecasting in Northern High Latitudes

Abstract

An analog forecast method designed for monthly and seasonal outlooks is applied to the Arctic. The analog selection process uses pattern matches based on agreement with historical data to identify past years with similar distributions of sea level pressure, upper-air geopotential height, surface and upper-air temperatures, precipitation, and sea surface temperatures. The evolution of the atmosphere in the analog years is then the basis of a prediction for the target year. Users can choose the predictor domain, the predictand domain, the variable to be predicted, and the number of antecedent months on which the analog selection is based. We provide an example of a monthly forecast generated by the analog forecast tool. In comparisons with operational dynamical model forecasts over the period 2012-2019, the analog system underperforms the dynamical models in middle latitudes but generally outperforms the dynamical models in monthly forecasts of surface air temperatures in the Arctic. The improvement over the dynamical models is especially apparent in the late summer and early autumn (August-October).