Abstract

The overall objective of this study is to utilize high resolution lidar-derived digital elevation models (DEMs) to improve classification and understanding of forested watersheds. Since geographic information systems technology became broadly used in natural resource fields in the 1980s, scientists have used digital elevation models to study aspects of forested ecosystems including the delineation of drainage networks, geomorphic modeling, and ecological classification for forest management and ecosystem management. With recently available lidar elevation data, we have improved our ability to ―see features on the landscape by orders of magnitude. Existing methodologies for assessing geomorphometry and hydrologic network delineation across the landscape may not suffice for all tasks. By taking a multi-scale, multidisciplinary approach, we can improve our understanding of headwater ecosystems and how to assess and predict the relationship between terrain and vegetation. This research was performed in the Leading Ridge experimental watersheds, the site of a long-term study analyzing the impact of forest management practices on stream water quality. The Leading Ridge experimental watersheds are also located within the Susquehanna/Shale Hills Critical Zone Observatory.

In order to assess the ability of lidar-derived DEM to improve stream network modeling, the stream network for Leading Ridge watershed number one was recorded using a GPS unit during base flow conditions. The stream network was then modeled using lidar-derived 1 m, 3 m, and 10 m resolution DEMs as well as photogrammetrically-derived NED (National Elevation Dataset) DEM. All of the lidar-derived DEMs resulted in a relatively accurate stream network model, with the 3 m DEM providing the most accurate model. There was no significant difference between any of the lidar-derived modeled stream networks, but they were all significantly different from the NED DEM-derived stream network, which was much less accurate. Topographic index (TI) was modeled using multiple DEM products and presented very different statistical distributions and spatial patterns. The distribution of TI could have an impact on hydrologic models, while the improvements in network delineation could substantially improve our knowledge of headwater streams on the landscape. This could in part impact forest management, site planning, and ecosystem modeling.

Surface roughness was calculated for Leading Ridge using several algorithms on two different lidar-derived DEMs to evaluate patterns of roughness on the watershed. Roughness metrics included standard deviation of slope, value of pittedness in cells, standard deviation of curvature, and the difference between the original DEM and a splined surface. Micro-plots and transects were surveyed to ground truth roughness metrics. Although the scale of the 1 m DEM was too coarse to assess micro-topography at the same scale as the ground survey, unique patterns were identified on different landforms and soil types. There was also substantial interaction between the roughness algorithm and the DEM creation algorithm. The results suggest that although there are many complicating factors when assessing surface roughness using a lidar-derived DEM, there is information about soils and topography that can be obtained. Also, DEMs studied here had slightly higher elevation values (about 0.3 m) on average than the field-surveyed elevations.

In order to relate topography to vegetation, curvature was chosen to model landforms based on its importance to water transport on an ecosystem. There was evidence of curvature being reflective of underlying geology and predictive for soil properties that may affect vegetation. Leading Ridge watershed was delineated into nine curvature classes using a 10 m DEM, and patterns of curvature were used to construct four recurring formations: hidden hollows, rock ridgelets, scalloped slopes, and rounded ridges. Based on a vegetation analysis of these formations, there was a difference in both vegetation community and structure based on formation. Similar formations were calculated for a broader region of the Ridge and Valley Province and vegetation communities on formations were identified. There was association between the identified vegetation community and the delineated formation.

Overall, methodologies were developed to explore properties of forested ecosystems in the Ridge and Valley Province. Using lidar elevation data, delineation of the stream network and characterization of terrain and micro-topography were all improved, and curvature was utilized to help classify the landforms in watershed. Further research should attempt to validate these results across a broader area, as well as work to develop techniques to use together to create a multi-scale, hierarchical classification system incorporating hydrologic data, surface roughness, and landscape level terrain data.

Citation

Kristen M. Brubaker (2011): MULTI-SCALE LIDAR-BASED APPROACHES TO CHARACTERIZING STREAM NETWORKS, SURFACE ROUGHNESS AND LANDFORMS OF FOREST WATERSHEDS. Doctor of Philosophy, Forest Resources, The Pennsylvania State University, p. 162.