A fusion retrieval approach for monitoring upstream suspended sediment fluctuations using Sentinel-2 imagery

Remote sense data are increasingly being developed also to accomplish environmental monitoring activities. In the context of climate change, a growing number of extreme events are being observed worldwide, leading to significant changes in surface hydrological processes. This shift is a crucial factor in the alteration of suspended sediment concentration (SSC) in large rivers. Particularly, large rivers originating in the high mountainous regions of Asia and flowing into densely populated areas of Southeast Asia are more susceptible to erosion due to their geomorphic characteristics and intensive human activities. Faced with grand and complex geomorphic conditions, there is a need to adopt a basin-wide perspective by employing a broader array of monitoring methods. This is particularly important for the Pearl River, a major river in southern China. Retrieval of SSC using remote sensing is one of the popular monitoring methods in the past decades. Unfortunately, its application has mainly focused on the estuary and coastal open water. In this study, we place a stronger emphasis on the basin-wide scale, specifically focusing on the upstream and major tributaries. We recalibrated model parameters using a general index model (G_index) and a regional high-precision model (C_SSC). These recalibration results were combined with Sentinel-2 imagery and field data to establish a basin-wide suspended sediment monitoring program. The results of the integrated model fit (n = 29), R² = 0.95, RMSE = 14.15 mg/L. The modeling results indicated that the spatial distribution of SSC in the upstream and tributaries of the Pearl River showed a clear concentration pattern, with markedly different concentrations in the upstream and downstream reaches. In the upstream, the SSC distribution was clearly divided into two parts, with concentrations of 104.85 mg/L and 13.43 mg/L, respectively, reflecting a substantial difference. It is worth noting that the monthly SSC statistics were clearly seasonal related to the precipitation. May and June were two months with high SSC concentrations in the whole river, with median values of 85.51 mg/L and 106.48 mg/L, respectively. In addition, we observed an abrupt change in suspended sediment downstream of the large reservoirs. This difference is likely caused by the streamflow resulting from the high drop of the dam or channel narrowing. Consequently, we have analyzed the suspended sediment dynamics of both the mainstem and major tributaries of the entire Pearl River. The results will enhance the understanding of suspended sediment changing in large rivers, serving as a valuable complement to water resource management and soil erosion risk assessment.