Refining Arctic Marine Gravity Field in Ice-Covered Regions using ICESat-2

The accurate determination of the Arctic marine gravity field is crucial for applications such as marine resources exploration, geological structure detection, and underwater navigation. However, the extensive sea-ice cover makes the inversion of Arctic marine gravity field challenging using traditional radar altimetry due to waveform contamination and limited lead detections, resulting in lower precision compared to mid-latitude oceans. While ICESat-2 offers smaller footprints (~11 m) ideal for lead detection and a theoretical advantage for sea surface height measurement, the official ATL07 product, which provide surface heights and surface-type classifications in the Arctic, suffers from inaccurate classification errors and high uncertainty in sea surface height estimation over leads. This study presents a refined processing chain for ICESat-2 ATL03 data to enhance sea surface height (SSH) and marine gravity field retrieval in the Arctic. A two-step denoising algorithm was developed to remove noise photons and mitigate after-pulse effects, improving the surface height precision of the strong beams from 0.12 m to 0.08 m as validated against airborne measurements. Furthermore, this study integrated Sentinel-2 imagery with a combined unsupervised and supervised machine learning approach to achieve high-accuracy classification of sea ice, leads, and open water. Validation with Sentinel-2 imagery demonstrated that this refined classification increased lead identification accuracy from 46.6% to 98.6%. By integrating the processed ICESat-2 data with multi-mission radar altimeter data (Cryosat-2, SARAL, Sentinel-3a), a new Arctic marine gravity field model was developed. When compared with the models SIO V32.1 and DTU21, the standard deviations of the discrepancies of our model are 3.76 mGal and 3.15 mGal, respectively. Comparison with the ArcGP gravity dataset indicated an improvement of approximately 0.5 mGal in gravity anomaly accuracy north of 80°N after incorporating ICESat-2 data. Further comparison with the GEBCO_2024 bathymetric model revealed that the inclusion of ICESat-2 data also resulted in an enhancement in the resolution of marine gravity model over ice-covered oceans. This study demonstrates that incorporation of high-precision ICESat-2 data might enhance the accuracy of marine gravity field in sea ice-covered regions.