Using dynamic regression and a process-based watershed erosion model to evaluate suspended sediment reductions from stream restoration projects

Abstract:

Elevated turbidity levels from suspended-sediment (SS) flux during and following flood events can degrade water supply quality and aquatic ecosystem integrity. Streams draining glacially conditioned mountainous terrain, such as those in the Catskill Mountains of New York, are particularly susceptible to chronic and acute high levels of turbidity from SS sourced from erosional contact with glacial related sediment. Implementing stream restoration projects at the SS-sourced headwater basins can be effective to reduce turbidity in water supplies. However, it is often difficult to assess the effectiveness of these restoration projects due to changing hydrologic conditions (Q), which are anticipated under a changing climate. After Hurricane Irene in August 2011, New York City (NYC) and the federal government invested millions of dollars in 8 stream sediment and turbidity reduction projects (STRPs) in the Stony Clove Creek catchment, an 83 km² watershed in the Ashokan Reservoir basin and part of the NYC water supply system. In 2016, the USGS started a 10-year SS source monitoring research project with NYC using 29 monitoring stations in the Ashokan basin. Reduced SS concentrations (SSC) in the watershed have been observed. However, flood hydrology since 2012 has been abnormally low, potentially obscuring restoration impacts. This study builds on the ongoing NYC-USGS research initiative to advance a framework to identify restoration project effectiveness in the presence of hydrologic trends. We use Dynamic Linear Models (DLMs) to statistically characterize daily variations in the SSC-Q rating curve prior to and following STRP installation, in order to isolate the timing of SS yield changes independent of flow. We compare the observed rating curve dynamics against those in nearby catchments with limited or no STRP installation, and rating curve dynamics that are simulated by a process-based River Erosion Model (REM) calibrated to the Stony Clove Creek catchment and parameterized without influences from STRPs. The comparative analysis provides a way to estimate the amount of SSC reduction attributable to restoration projects and hydrologic trends, respectively. The proposed framework can expedite the assessment of projects in order to generate more rapid feedback that can guide additional investment decisions.