Predicting Reach-scale Sediment Regimes to Prioritize River Restoration and Conservation

Vermont rivers have been brought into focus by changes in climate, water quality concerns, and extreme events like Tropical Storm Irene. Extreme precipitation events have become more frequent, which has led to an increased interest in flood mitigation to address excessive sediment transport. The latter, along with streambank erosion has been linked to water quality issues such as algal blooms in Lake Champlain. Restoration and conservation practices along river segments helps reduce the severity of floods, decreasing sediment erosion, and allowing the river to return to a more balanced state. It has thus become important to identify and prioritize river segments that provide the most benefit from restoration and conservation practices. The state of Vermont has been collecting and compiling river data at the reach scale via the VTANR’s the Rapid Geomorphic Assessment (RGA) program and has assessed more than 1,371 stream miles to date (Kline and Cahoon 2010). This research leverages the RGA data to provide a framework for prioritizing the stream reach locations best suited for conservation or restoration, which is challenging as more data come online.

This research looks to advance the data processing into an applied tool that produces a map of prioritized river segments using a Multi-Objective Genetic Algorithm (MOGA). The prioritization is weighted using an objection function that minimizes project costs while maximizing the returned benefit of those projects as defined by a group of stakeholders. Prioritization is focused on several key factors, one of which is the sediment transport regime at the individual river reach scale. The regime type provides an indicator for the river reach sensitivity to adjustment and erosion. Over the years, this regime type classification has been determined by local experts using the VTANR RGA data as well as their field expertise. This work has provided a unique dataset for training machine-learning algorithms that can help alleviate the labor intensity and expert bias in future classifications. Recent work by Underwood, et al, (in review, 2018) has developed a Self Organizing Map (SOM) that classifies regime types that are comparable to expert classification. We build on this earlier work to provide a framework for automating regime classification statewide. Identifying these regime types is important, but is only one factor in determining which subset of river segments to target for conservation and restoration. The costs associated with projects in a given location as well as a variety of other factors are included in the multi-objective optimization function for prioritization. The SOM and MOGA are packaged in a desktop application that provides management groups (watershed groups, Regional Planning Commissions, municipalities, state and federal agencies) a means to visualize tradeoffs between project benefits and costs.