A Method for Forecasting Sea Ice Concentration for Arctic Shipping Utilizing eXtreme Gradient Boosting and Gaussian Process Regression

Le résumé fourni par la source

The ongoing global climate change and the progressive melting of Arctic sea ice have considerably improved the navigability of Arctic shipping routes. Nevertheless, navigation safety is influenced by sea ice conditions. Thus, detailed forecasting of these circumstances is crucial. This article proposes a method for long-term daily forecasting of the sea ice concentration (SIC) in the Northern Sea Route. The monthly trend of SIC is predicted with the eXtreme Gradient Boosting (XGBoost) machine learning algorithm. Moreover, the daily change of SIC in each month is regressed from the gaussian process given the covariance information. The proposed approach decouples the forecasting of long term SIC into a global trend portion and a local fluctuation part, and remit the problem of error accumulation. The method offers the dramatic accuracy of long term daily SIC forecasting for the Arctic shipping, hence giving substantial assistance for the Arctic navigation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.