A Relatively Simple Method for Extracting Wet Channels under Canopy from Spatial and Spectral LiDAR Data

Abstract:

Headwater channels are vital to ecological health, water quality, and watershed connectivity. However, the geographic extent and temporal dynamics of headwater drainage networks are not thoroughly understood due to mapping limitations, namely due to the occlusion of channels from overhead imagery via canopy cover. This inhibits management decision-making and appropriate designation, protection, and conservation of these channels. The utilization of LiDAR holds great promise in this regard, as LiDAR has demonstrated capacity in penetrating canopy cover and generating data on the ground beneath with exceptional accuracy. In this study, we present a novel masking workflow that employs LiDAR elevation and return-intensity data, and implements LiDAR’s ability to act as a spatial and spectral segmentation tool for wet channel delineation, with a focus on Yellowstone National Park. As our data shows, we observed a phenomenon whereby LiDAR return-intensity is further reduced when encountering both a canopy obstruction and liquid water at ground level, relative to either scenario individually. This quantifiable LiDAR signature was leveraged to locate water bodies otherwise undetectable by optical imagery.

In our automated workflow, we first delineate a drainage network from a DEM, mask an intensity raster to neglect pixels non-coincident with the network, and then hone in on regions with potential channel morphology that exhibit a high-density of low-intensity returns, outputting a classified data product. An accuracy assessment of our method was performed on each study site using stratified random sampling, with an overall accuracy as high as 85%. Though future work would be beneficial prior to operationalization, this study demonstrates a proof-of-concept for additional research and field validation. Supplementary work might include a study aligned with a robust ground-based data campaign, an examination of effectiveness across a broader range of land cover types, and an investigation on how to optimize relevant thresholds for areas of interest dissimilar from the Greater Yellowstone ecosystem.