A Local Analytical Optimal Nudging for assimilating AMSR2 sea ice concentration in a high-resolution pan-Arctic coupled ocean (HYCOM 2.2.98) and sea ice (CICE 5.1.2) model

Abstract. A Local Analytical Optimal Nudging (LAON) is introduced and thoroughly evaluated for assimilating the Advanced Microwave Scanning Radiometer 2 (AMSR2) sea ice concentration (SIC) in the Norwegian High-resolution pan-Arctic ocean and sea ice Prediction System (NorHAPS). NorHAPS is a developing high-resolution (3–5 km) pan-Arctic coupled ocean and sea ice modeling and prediction system based on the HYbrid Coordinate Ocean Model (HYCOM Version 2.2.98) and the Los Alamos multi-category sea ice model (CICE Version 5.1.2), with the LAON for data assimilation. In this study, our focus is on the LAON assimilation of AMSR2 SIC, which is designed to update the model SIC in every time step such that the analysis will eventually reach the optimal estimate. The SIC innovation (model minus observation) is designed to be proportionally distributed to the multiple sea ice categories.

A twin experiment is performed with and without the LAON assimilation for the period 1 January 2021 to 30 April 2022. The results show that the LAON assimilation greatly improves the simulated sea ice concentration, extent, area, thickness and volume, as well as the sea surface temperature (SST). It also produces significantly more accurate sea ice edge and marginal zone (MIZ) than the observed AMSR2 SIC that is assimilated when evaluated against the Norwegian Ice Service (NIS) ice chart. The results are also compared with the Copernicus Marine Environment Monitoring Service (CMEMS) operational SIC analyses from NEMO, TOPAZ4 and neXtSIM which use ensemble Kalman filters and direct insertion for data assimilation. It is shown that the LAON assimilation produces significantly lower integrated ice edge error (IIEE) and integrated MIZ error (IME) than the CMEMS SIC analyses when evaluated against the NIS ice chart. The LAON also produces a continuous evolution of sub-daily SIC, which avoids abrupt jumps often seen in other assimilated products. This efficient and accurate method is promising for data assimilation in global and high-resolution models.