Assessment of land adjacency effect correction on deriving aquatic reflectance over river waters from Landsat-8/9 and Sentinel-2 imagery

Land adjacency effects (LAEs) significantly bias satellite-derived aquatic remote sensing reflectance ( R rs ), particularly in small water bodies and narrow rivers. Although several algorithms have been developed for oceans and lakes, their performance in fluvial environments remains poorly understood. Here, we evaluated six atmospheric correction (AC) algorithms using Landsat-8/9 OLI and Sentinel-2 MSI data, supported by in situ measurements from three Yellow River sections with varying turbidity and terrain. For all OLI and MSI matchups (N = 54), without LAE correction, the mean absolute percentage error (MAPE) averaged 57.8%, with red and near-infrared bands performing better than blue bands. Incorporating LAE correction substantially improved accuracy (MAPE = 27.6%), and the TSDSF-RAdCor method achieved the best results (MAPE = 17.4%), especially in visible bands. Further analysis reveals that performance varied with river morphology and turbidity: riffle-dominated reaches showed higher accuracy than canyon sections, and highly turbid waters improved correction efficiency. Suspended particulate matter estimates from LAE-corrected R rs were 16-30% more accurate than those from uncorrected R rs . These findings highlight the necessity of LAE correction in fluvial remote sensing to enhance the reliability of riverine water quality retrievals and support large-scale environmental monitoring and management.