EVENT-SCALE QUANTIFICATION OF STREAMBANK EROSION WITH PLANET SATELLITE IMAGERY

Streambank erosion is a significant contributor to sediment load in intensively managed agricultural landscapes. Channel-forming events with near-bankfull discharge can act as preparatory processes and initiator of channel incision and eventual streambank slumping. More research is needed to quantify the contribution of these large events to watershed sediment loads. Due to the spatial limitations associated with traditional methods used to quantify streambank erosion (e.g., bank pins), remotely sensed imagery (aerial, drone, and satellite) and advances in geospatial computational tools provide unprecedented opportunities to analyze longer river reaches and multiple watersheds. In this study, we utilized Planets Labs daily multispectral imagery, coupled with LiDAR-derived DEMs, and Google Earth Engine (GEE) platform to quantify event-scale streambank sediment mass losses (Mg/meter reach length/year), over a seven-year period (2016-2022) for six HUC-8 watersheds. A random forest machine learning algorithm was applied to the complete set of data and used to identify geospatial characteristics that influence streambanks. The results of this analysis could inform the partitioning of sediment sources in watershed models. Our research provides important insight into the impact of flood events on streambank erosion and highlights the need for better management of the riparian environment.