River ice monitoring in the Yellow River using harmonized Landsat and Sentinel data

River ice strongly influences hydrodynamics and ice-flood risk along the Inner Mongolia section of the Yellow River. This study applied the Harmonized Landsat and Sentinel-2 (HLS) dataset in Google Earth Engine to monitor river ice in this region from 2020 to 2025. Three indices—NDSI, NDUWI, and RDRI—were assessed for class separability. Among them, NDSI showed strong ice–soil separability, RDRI provided robust ice–water discrimination, while NDUWI exhibited limited effectiveness. Validation showed that the combined use of NDSI and RDRI effectively distinguished ice from both soil and water, achieving high accuracy (OA ⪆0.93, Kappa ⪆0.90). Time-series analysis revealed consistent seasonal cycles: ice formed in November–December, peaked in January–February, and decayed rapidly in March–April. Spatial differences among three channel types were analyzed. Results showed gradual freeze–melt in the braided reach, asymmetrical patterns under curvature control in the meandering reach, and abrupt changes in the canyon reach due to topography and reservoir effects. These findings confirm the value of HLS data for river ice monitoring and highlight spatial heterogeneity of ice dynamics in the Inner Mongolia reach.