A Comparison of DEM-based methods for fluvial terrace mapping and sediment volume calculation

Abstract

Fluvial terraces form in both erosional and depositional landscapes and are important recorders of land-use, climate, and tectonic history. Terrace morphology consists of a flat surface bounded by valley walls and a steep-sloping scarp adjacent to the river channel. Combining these defining characteristics with high-resolution digital elevation models (DEMs) derived from airborne light detection and ranging (lidar) surveys, several methods have been developed to identify and map terraces. This research introduces a newly developed objective terrace mapping method and compares it with three existing DEM-based techniques to determine which is most applicable over entire watersheds. This work also tests multiple methods that use lidar DEMs to quantify the thickness and volume of fill terrace deposits identified upstream of dam sites. The preliminary application is to the Sheepscot River watershed, Maine, where strath and fill terraces are present and record Pleistocene deglaciation, Holocene eustatic forcing, and Anthropocene land-use change. Terraces were mapped at four former dam sites along the river using four separate methodologies and compared to manually delineated area. The methods tested were: (1) edge detection using MATLAB, (2) feature classification algorithms developed by Wood (1996), (3) spatial relationships between interpreted terraces and surrounding natural topography (Walter et al., 2007), and (4) the TerEx terrace mapping toolbox developed by Stout and Belmont (2013). Thickness and volume estimates of fill sediment were calculated at two of the study sites using three DEM-based models and compared to in situ data collected from soil pits, cut bank exposures, and ground penetrating radar surveys. The results from these comparisons served as the basis for selecting methods to map terraces throughout the watershed and quantify fill sediment upstream of current and historic dam sites. Along the main stem and West Branch of the Sheepscot River, terraces were identified along the longitudinal profile of the river using an algorithm developed by Finnegan and Balco (2013), which computes the elevation frequency distribution at regularly spaced cross-sections normal to the channel, and then mapped using the feature classification (Wood, 1996) method. For terraces upstream of current or historic dam sites, thickness and volume estimates were calculated using the two best performing datum surfaces. If all analyzed terraces are composed of impounded sediment, these DEM-based results suggest that terraces along the main stem and West Branch of the Sheepscot River potentially contain up to 1.5 x 106 m3 of fill. These findings suggest powerful new ways to quickly analyze landscape history over large regions using high-resolution, LiDAR DEMs while relying less heavily on the need for detailed and costly field data collection.