Author(s): Changhong Hu; Gang Wu; Meng Cui; Christo Pimpirev; Songtao Ai
Linked Author(s):
Keywords: Sea ice concentration; Arctic; Time series; Forecast
Abstract: This dataset provides daily Arctic sea ice concentration (SIC) forecasts from January 1, 2020, to December 31, 2029. The forecasting methodology integrates least squares trend modeling with an autoregressive model to correct residuals, enabling accurate daily SIC forecasts over extended periods. The model demonstrates robust performance, with the mean absolute error of SIC forecasts remaining below 13% across the entire five-year forecast horizon (2020-2024) from a single initialization. Additionally, evaluation based on binary accuracy between forecasted and observed sea ice margins shows an overall accuracy exceeding 85% for most months, with peak performance surpassing 90% during the winter months. This highlights the framework's ability to effectively capture both seasonal and interannual SIC variability. This dataset provides a valuable resource for monitoring and forecasting sea ice conditions, with significant implications for climate research, Arctic navigation, and resource management.
Year: 2026