Assessment of the decreasing suspended sediment concentrations in the Rhine River using the Google Earth Engine

Abstract

Suspended sediment transport plays a crucial role in river systems. Human influences have strongly altered river systems around the world causing a strong decline in the suspended sediment concentrations (SSC) and loads. This decline has led or may lead to insufficient sediment supply to compensate for sea level rise and, consequently, to increased land subsidence or loss of coastal land. Knowing the origin of this decline is important for the implementation of effective management strategies. This study explores the capabilities of remote sensing tools to identify the decline of the SSC in the Rhine River and its major tributaries from 1990-present. Using the Google Earth Engine (GEE), an SSC model for both the Landsat and Sentinel image collection was developed. These models were calibrated using SSC measurements at in-situ locations. Based on the available observations of the Sentinel and Landsat image collections, no significant trend could be observed. Observations from the Sentinel model showed a gradual downstream increasing trend in SSC. It also showed a higher resemblance to the in-situ measurements in comparison to the Landsat model. The period (2017-2022) of data availability from the Sentinel model was too short to observe any clear temporal trend. The Landsat model had a longer period (1990-2022) of data availability, but it performed poorly in observing low sediment concentration events and underestimated the high events. Therefore, the Landsat model was unable to show any clear SSC trends. It is advisable to use fitted models for the Upper-Rhine, Mid-Rhine, Lower-Rhine, and each tributary separately. This improves the reliability of the SSC predictions at each section of the Rhine. To do so, more consistent and up to date in-situ data is needed for the calibration. Future SSC studies of the Rhine River would benefit most from the use of the Sentinel image collection as the increased resolution of 10mx10m and the use of the Red Edge1 band resulted in a better prediction of the SSC in comparison to the Landsat model.