A New Normalized Difference Ice Index (NDII) Based on GOCI Imagery for Accurate Sea Ice Extraction: Application in Peter the Great Bay, Sea of Japan

The rapidly changing sea ice conditions in seasonal ice zones highlight the necessity for effective sea ice monitoring. However, distinguishing between ice and water, particularly in regions with large areas of thin sea ice, remains a challenging task. To address this, a new normalized difference ice index (NDII) has been developed to accurately retrieve sea ice areas in Peter the Great Bay, Sea of Japan, using Geostationary Ocean Color Imager (GOCI) satellite imagery under clean atmospheric conditions. The NDII algorithm utilizes the violet (412 nm) and near-infrared (NIR) (745 nm) bands, exploiting reflectance differences for ice–water differentiation. The accuracy of the NDII, along with four other commonly used algorithms, was evaluated against a ground truth dataset from 2020 and intercompared. The NDII algorithm outperformed all others, achieving the highest overall statistical accuracy in distinguishing sea ice from seawater and exhibiting the lowest mean absolute percentage bias (8.26%), demonstrating a higher agreement with ground truth data. Additionally, the NDII classification accuracy was significantly higher than that of the other algorithms ($p < 0.01$). A total of 421 GOCI images, captured under clean atmospheric conditions from 2012 to 2021, were further used to extract sea ice areas based on the NDII algorithm and analyze the spatiotemporal variations. The analysis reveals that both the annual average and maximum sea ice areas have decreased, though the changes were not statistically significant. Overall, our findings highlight the potential of the NDII algorithm for sea ice observations in seasonal ice zones, offering valuable insights for researchers to advance this field in future studies.