Understanding the limits of the seasonal prediction of sea ice in the Arctic

Abstract

The Community Earth System Model’s Large Ensemble is used to test the hypothesis that spring melt pond area is a robust predictor of September sea ice minimum extent in the Arctic. Melt pond area is examined in the context of a number of plausible predictors, focusing on the thermodynamic mechanisms that control ice growth and melt at the pan-Arctic level. Most of the variables individually perform poorer than a persistence forecast until the latter half of the 21st century, when snow and ice thickness become more effective predictors. MWhile melt pond areas are is not the most effective predictors of pan-Arctic September sea ice extent when examined in the context of multiple predictors (a claim which is static throughout a two century perioddoes not change when tested over two centuries of model simulation);, however, itthey does provide improved skill when handled at the regional levelused in regional forecasting. Maximum covariance analysis provides an effective mechanism to enhance regional forecasting.