Document Type

Open Access Thesis

Embargo Period


Degree Program


Degree Type

Master of Science (M.S.)

Year Degree Awarded


Month Degree Awarded


Advisor Name


Advisor Last Name



Recent advances in remote-sensing technologies and analysis methods, specifically airborne-LiDAR elevation data and corresponding geographical information system (GIS) tools, present new opportunities for automated and rapid fluvial geomorphic (FGM) assessments that can cover entire watersheds. In this thesis, semi-automated GIS tools are used to extract channel centerlines and bankfull width values from digital elevation models (DEM) for five New England watersheds. For each study site, four centerlines are mapped. LiDAR and NED lines are delineated using ArcGIS spatial analyst tools with high-resolution (1-m to 2-m) LiDAR DEMs or USGS National Elevation Dataset (NED) DEMs, respectively. Resampled LiDAR decreases LiDAR DEM resolution and then runs spatial analyst tools. National Hydrography Dataset (NHD) lines are mapped by the USGS. All mapped lines are compared to centerlines delineated from photography and LiDAR DEMs. Bankfull widths at each site are determined through three methods. Regional regression equations are applied using variables derived from LiDAR and NED DEMs separately, producing two sets of width results. Additionally, the Hydrogeomorphological Geoprocessing Toolset (HGM) is used to extract widths from LiDAR data. Widths are also estimated visually from aerial photos and LiDAR DEMs. Widths measured directly in the field or derived from field-data are used as a baseline for comparison.

I find that with a minimal amount of preprocessing, specifically through DEM resampling, LiDAR data can be used to model a channel that is highly correlated with the shape and location of the mapped channel. NED-derived channels model the mapped channel shape with even greater accuracy, and model the channel location only minimally less accurately. No tool used in this study accurately extracted bankfull width values, but analysis of LiDAR data by the HGM toolset did capture details that could not be resolved using regression equations. Overall, I conclude that automated, computerized LiDAR interpretation needs to improve significantly for the expense of data collection to be cost-effective at a watershed scale.