Work

HLS-Based Assessment of Suspended Sediment Concentration in the Middle and Lower Reaches of Liaohe River

Year: 2026

Type: article

Abstract: High-frequency monitoring of suspended sediment concentration (SSC) is important for sustainable river management, but conventional cross-sectional measurements and single-sensor satellite observations cannot simultaneously provide fine spatial detail and sufficient temporal continuity. This study evaluated whether Harmonized Landsat and Sentinel-2 (HLS) observations can improve SSC monitoring in the middle and lower Liaohe River, China. Daily mean cross-sectional SSC records from five hydrological stations were matched with same-day HLS L30 and S30 surface reflectance observations during the ice-free months from 2016 to 2022. Three Ridge regression schemes, including L30-only, S30-only and L30 + S30 fusion models, were assessed using date-grouped cross-validation, temporal extrapolation, station extrapolation and quality-based reliability labeling. The fusion model achieved an out-of-fold RMSE of 0.329, MAE of 0.245 and R2 of 0.454 on the log10(SSC) scale, with negligible bias and no clear accuracy degradation relative to the best single-sensor schemes within sensor-specific subsets. The L30–S30 union increased valid observation days by 64.6% compared with the single sensor with a greater number of valid observation days at each station and raised station-month availability to 87.9%. Reliability labels based on clear-water pixel number and proportion helped identify lower-error observations, while spatial transferability varied more substantially among stations than temporal extrapolation. These results show that HLS fusion improves the temporal availability of quality-screened SSC observations in medium-width rivers, although uncertainty remains greater at the extremes of the SSC range and when the model is transferred to new cross-sections.

Source: Water

DOI: 10.3390/w18151830