Document Type

Thesis - Open Access

Award Date

2024

Degree Name

Master of Science (MS)

Department / School

Natural Resource Management

First Advisor

Christopher Cheek

Abstract

Stream connectivity is crucial for fish movement and genetic diversity in fragmented landscapes. Road-crossings often act as barriers, disrupting hydrology and isolating fish populations. The Southeast Aquatic Resources Partnership (SARP) has developed rapid road-crossing assessment protocols and tools to prioritize restoration projects. However, managers lack tools for planning, directing, and initiating roadcrossing assessments. Managers also require modeling approaches to effectively identify and monitor road-crossings across dynamic stream networks. Presented here is a case study offering a roadmap for effective implementation of collaborative road-crossing assessments and prioritization of remediation projects, alongside an exploration of predictive modeling approaches to identify problematic crossings across large, dynamic stream networks. A mobile decision-support tool was created to help plan and direct assessment efforts using datasets of roads, streams, road-crossing locations, completed assessments, and species of greatest conservation need (SGCN). In 2022 and 2023, a total of 528 road-crossings were assessed, and a training workshop in August 2023 increased assessment capacity, resulting in a contribution of 91 assessments in just 1.5 days. Top remediation candidates produced through prioritization, such as Willow Creek near Watertown, South Dakota, could provide ~94 miles of quality upstream habitat. Plans to expand the tool nationwide include refining map layers and functionality to enhance agency goals and foster collaborations. Two modeling frameworks were developed and evaluated for remotely estimating road-crossing barrier severity (the degree to which fish passage is impeded): Boosted Regression Trees (BRT) using landscape-level covariates and Image Scene Classification (ISC) of aerial imagery. Results from both frameworks provided encouraging training accuracies (BRT: training correlation of 78.3%, ISC: training accuracy of ~78%) but fell short of precise predictions. However, this effort provided valuable insights into specific watersheds and landscape characteristics such as scour pool size, stream discharge, and amount of pasture at the catchment level contributing to barrier severities in eastern South Dakota. Ultimately, additional barrier assessments are needed in South Dakota to improve predictive power and accurately identify problematic road-crossing locations.

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright