A machine learning model of Arctic sea ice motions

TL;DR

This paper introduces a deep learning CNN model that predicts Arctic sea ice motions from surface wind data, outperforming traditional models and highlighting the importance of spatial patterns in predictions.

Contribution

A novel CNN-based data-driven approach for modeling Arctic sea ice motions, surpassing existing thermodynamic-dynamical models in predictive accuracy.

Findings

CNN achieves an average correlation of 0.82 with actual sea ice motions.

CNN outperforms local point-wise predictions and CICE5 model.

Connective spatial patterns are crucial for accurate sea ice motion prediction.

Abstract

Sea ice motions play an important role in the polar climate system by transporting pollutants, heat, water and salt as well as changing the ice cover. Numerous physics-based models have been constructed to represent the sea ice dynamical interaction with the atmosphere and ocean. In this study, we propose a new data-driven deep-learning approach that utilizes a convolutional neural network (CNN) to model how Arctic sea ice moves in response to surface winds given its initial ice velocity and concentration a day earlier. Results show that CNN computes the sea ice response with a correlation of 0.82 on average with respect to reality, which surpasses a set of local point-wise predictions and a leading thermodynamic-dynamical model, CICE5. The superior predictive skill of CNN suggests the important role played by the connective patterns of the predictors of the sea ice motion.