A long‐range forecast of Arctic summer sea‐ice minimum extent

This paper discusses the development of simple multiple linear regression (MLR) models for predicting the annual pan‐Arctic minimum sea‐ice extent at monthly intervals from February through August. The predictor data is based on mean monthly weighted indices of sea‐ice concentration (WIC), surface skin temperature (WST), surface albedo (WAL), and downwelling longwave flux at the surface (WDL). The final regression equations retain either one or two sea ice and surface energy and radiation balance predictors, and each of the MLR models is superior to climatology, persistence, or random chance models. The mean absolute error (MAE) for the MLR models decreases from 0.36 × 106 km2 in February to 0.15 × 106 km2 in August; the corresponding r2 increases from 0.46 in February to 0.90 in August. In addition to improving long‐range predictions, the models provide insight into the physical mechanisms affecting recent large reductions in sea‐ice extent.