Retrieval of sea ice thickness from FY-3E data using Random Forest method

In this study, we employ a Random Forest approach to estimate sea ice thickness (SIT) using Fengyun-3E (FY -3E) and Soil Moisture Ocean Salinity (SMOS) data. This method relies on four input parameters: incidence angle ( h), reflectivity ( C ), sea ice salinity ( S ), and sea ice temperature ( T ). In addition, FY -3E can receive both Global Positioning System (GPS) and Beidou Navigation Satellite System (BDS) reflected signals. Evaluation for the Arctic region based on data spanning from October 2022 to April 2023 reveals that the proposed models trained on GPS and BDS signals from FY -3E achieve high consistency and low error. Take GPS signals as an example, coefficients of determination are 0.97 and 0.91 and mean absolute errors are 0.019 m and 0.032 m for the training and test sets, respectively. In general, SIT inversion based on GPS signals slightly exhibits a higher accuracy than that based on BDS signals, but both approaches display high performances. The areas with the highest accuracy of SIT estimation based on GPS and BDS signals are the Shelikhov Bay and the Okhotsk Sea, followed by the Bering Sea and the Bering Strait. We conclude that machine learning and data fusion are effective for SIT estimation. (c) 2024 COSPAR. Published by Elsevier B.V. All rights reserved.