Decreasing summer ice cover in the Arctic has led to an extension of the Arctic open water season, with implications for resource extraction, shipping, tourism and scientific research. Short- and long-term forecasts of summer sea ice cover are useful planning and preparation aids for stakeholders hoping to access the Arctic Ocean during the summer season. Current forecast models, however, struggle to offer useful forecasts earlier than three months out. Researchers supported by CPO’s Climate Variability & Predictability (CVP) program have developed a new forecasting method which can provide, up to seven months early, a probabilistic spatial map of the location of sea ice during the September sea ice minimum. The method, described in the Earth and Space Science journal, uses a stasitical approach, Bayesian logistic regression, to provide uncertainty information about the predictions given by the model which allows users to gauge how reliable the model forecasts are. The forecast model debuted by the researchers relies on a combination of atmospheric, oceanic, and sea ice predictors, and was built and tested with September minimum sea ice cover data from 1980 through 2018. The researchers used ‘skill scores’ to evaluate how well their forecasts perform in a variety of circumstances. While extreme events can cause uncertainty and seasonal variability will always limit predictive accuracy, the new forecast model can skillfully predict sea ice cover from one to seven months before the September sea ice minimum.
Photo credit: Jeremy Potter NOAA/OAR/OER