Authors: Ethan Shavers*, United States Geological Survey, Larry Stanislawski, United States Geological Survey
Topics: Water Resources and Hydrology, Geomorphology, Earth Science
Keywords: Headwater streams, hydrology, UAV, lidar
Session Type: Paper
Start / End Time: 12:40 PM / 2:20 PM
Room: Marshall East, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Headwater streams make up more than 50 percent of streams by length in the conterminous United States and play an important role in hydrologic, nutrient, and ecologic systems. The mutable nature of headwater streams makes automated and efficient mapping of stream features necessary for modeling and monitoring. This work applies optical and lidar data collected from a UAV to mapping headwater streams in the Meramec River Watershed, south-central Missouri. The study area is a roughly 10 square kilometer forested and hilly region within a state conservation area. Here we present a remote mapping strategy that employs point cloud and optical data analysis. A 3-dimensional (3D) vegetation model identifies structural variation relative to streams. Increases in low vegetation density and canopy height are shown to correlate with the presence of surface water and are used to identify areas of interest. A pixel-based classification tree algorithm is used to filter features identified by the 3D model. The input layers for the classification tree include RGB and NIR data as well as lidar-derived high-resolution digital elevation models. Resulting features are converted to vectors and match lines are generated by comparing pixel and 3D-generated features to derived flow accumulation lines. Results are validated using field surveys. The methods presented here uses open source tools, which can potentially be fully automated. The correlation of return point density from vegetation below 1 meter is stronger than that of tree height, yet these results likely vary in other regions with differences in forest management practices.