A New Sea Ice Type Concentration Retrieval Algorithm from Microwave Remote Sensing Data

Sea ice types, e.g., first-year ice (FYI) and multi-year ice (MYI), can be discriminated based on their radiometric and scattering signatures. However, changes in ice surfaces caused by factors such as ice deformation and melt-refreeze events can lead to extensive ice type misclassification. To solve this problem, a new sea ice type concentration (SITC) algorithm from microwave observations (SITCAM) is proposed in this study. It builds upon a previous algorithm, namely ECICE, but improves from two perspectives. Firstly, a new cost function is employed, with weights indicating the separation efficiencies of microwave parameters. Secondly, a pre-classification scheme is incorporated to account for the bimodal distributions in microwave characteristics. With SITCAM, daily Arctic SITCs are retrieved for the winters of 2002–2011 using passive (AMSR-E) and active (QuikSCAT and ASCAT) microwave data. The results are compared with a sea ice age product (SIA) and evaluated with ice type samples and SAR images. Overall, SITCAM performs well on mitigating the misclassifications induced by the aforementioned factors. The Arctic MYI area agrees well with that from SIA. Compared to ECICE, the retrieval accuracy for MYI and FYI samples increases to 96% and 90%, respectively (increasing by 5% and 15%, respectively), in SITCAM. The bias in MYI concentration between the SITC retrievals and SAR-based results has reduced from 15% to 4%. Furthermore, instead of being limited to specific observations (e.g., Ku-band scatterometer data), SITCAM performs well with various combinations of microwave data, even solely passive microwave data. This universality allows for a long-term record of SITC, which enables the potential of dating SITC back to late 1970s.