Abstract
The accuracy of pan-Arctic sea ice thickness (SIT) derived from the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) primarily depends on the representation of snow depth during the conversion from ICESat-2 laser freeboard to SIT. Here, we systematically assess ICESat-2 Arctic SIT retrievals by combining multiple snow depth products for the freezing seasons (October-April) during 2018-2020, including those obtained from satellite radiometry (FY3/MWRI and ASD_(M)R), reanalysis (TOPAZ4b), a merged climatology/passive-microwave product (MW99/AMSR2), a Lagrangian snow evolution model (SMLG), and dual-satellite altimetry (KuLa). We adopt the National Snow and Ice Data Center's ICESat-2 SIT product (IS2SITMOGR4, v4) as the reference, in which snow depth is prescribed by the NASA Eulerian Snow on Sea Ice Model (NESOSIM v1.1). The bulk densities of sea ice, snow, and seawater are set to the default values used in IS2SITMOGR4. The results show substantial uncertainty in ICESat-2 SIT associated with differences among snow depth products. During the study period, spatially averaged SIT ranges from 0.75 m ± 0.62 m (MW99/AMSR2) to 1.37 m ± 0.86 m (TOPAZ4b) in October and from 1.68 m ± 1.08 m (NESOSIM) to 2.23 m ± 1.16 m (FY3/MWRI) in April. The basin-scale mean SIT growth rate ranges from 0.09 m month⁻¹ (TOPAZ4b; P < 0.05) to 0.20 m month⁻¹ (FY3/MWRI; P < 0.05). Inter-scheme uncertainty in ICESat-2 SIT is largest over thick-ice regions and the Atlantic sector, with ranges (defined as the maximum minus the minimum) of 0.8-1.4 m and coefficients of variation (CVs) of 0.1-0.4, whereas the Pacific sector shows comparatively low uncertainty, with ranges of 0.2-0.4 m and CVs below 0.1. Evaluation against airborne, mooring, shipborne, and ground-based observations indicates that the ICESat-2 SITs, derived using multi-source snow depth data, exhibit mean correlation coefficients ranging from 0.71 to 0.82 and root-mean-square errors between 0.52 and 0.80 m. The best-performing snow depth scheme depends on the observational dataset used for evaluation. Furthermore, the synergistic effects of sea ice bulk density parameterization schemes and multi-source snow depth data on the estimation accuracy of ICESat-2 SIT are also discussed. This study provides new insights into parameter optimization for the operational ICESat-2 Arctic SIT product; however, it continues to encounter challenges arising from inherent uncertainties in observational references and the redistribution of snow cover at a kilometer-scale resolution.