Improving Arctic Sea Ice Predictions of the Norwegian Climate Prediction Model Through Dynamical Downscaling

Abstract Recent declines in Arctic sea ice and increasing shipping activity demand more precise sea ice predictions. However, the coarse spatial resolution of Earth system models, such as the Norwegian Climate Prediction Model (NorCPM), limits their ability to resolve fine‐scale ice features that are critical for safe Arctic navigation. To address the limitation, we implement a dynamical downscaling approach in a case study covering the 2023 autumn freeze‐up season. Specifically, a NorCPM hindcast provides atmospheric forcing for the regional coupled ocean‐sea ice model, the Norwegian High‐resolution pan‐Arctic ocean, and sea ice Prediction System (NorHAPS), which produces high‐resolution (3–5 km) hindcasts of Arctic sea ice concentration (SIC). The downscaled SIC predictions show improved performance throughout the prediction period with particularly notable reductions in the overestimation bias along the Northeast and Northwest Passages prior to mid‐to‐late October especially in marginal ice zones. Furthermore, NorHAPS provides a more accurate representation of local discontinuities and fine‐scale sea ice structures in key regions of the Arctic passages, such as the Laptev Sea, Canadian Arctic Archipelago, and Beaufort Sea. These improvements are associated with a more realistic simulation of sea ice freeze‐up processes, which mitigates the premature freezing found in NorCPM outputs. Overall, our results demonstrate that dynamical downscaling is a viable method for refining the outputs of coarse‐resolution climate models. This approach generates detailed sea ice predictions, which can support safe Arctic maritime operations.