In the last 30 years, over 1,500 dams have been removed as part of river restoration efforts in the United States. This trend is expected to continue in the future due to aging infrastructure, high rehabilitation costs, and the impacts of climate change on the severity of hydrologic events. However, only less than 10% of past dam removals have been closely monitored and documented, severely limiting quantitative and predictive understanding of river response. As a result, there is a pressing need to develop innovative approaches that allow to reconstruct the propagation of past fluvial sediment pulses and expand scientific knowledge on how these sediment-supply disturbances impact rivers. This study presents the development and validation of an integrated modeling approach that couples machine learning, remote sensing, numerical modeling, and parameter optimization to reconstruct the propagation of fluvial sediment pulses generated after past dam removals. The performance of the modeling approach was critically assessed using the well-documented case of the Elwha River dam removals, which took place in Washington state between 2011 and 2014, becoming the largest dam removal in the United States to date. Results indicated that the integrated modeling approach was able to successfully reconstruct the initial phase of the fluvial sediment pulse from 2012 to 2014, with its accuracy decreasing thereafter. The efficacy of the integrated modeling approach was primarily constrained by the availability of cloud-free USGS Landsat imagery used to estimate turbidity-based suspended sediment concentrations, which were in turn used as input for the numerical modeling and parameter optimization when reconstructing the spatiotemporal propagation of the fluvial sediment pulse.
