Objective 2: Develop a process for identifying furrows from a LiDAR dataset using the tools available in WhiteboxTools
As established in Objective 1, the LiDAR data were converted into the raster format in order to be used with the MaxAnisotropyDev tool. The tool works at multiple scales and assigns a unitless anisotropic value to a raster cell based upon the topographic position of itself and its surrounding cells using an iterative process in a range of user-defined kernels (D. Newman, personal communication, Feb 12th, 2018). The tool was designed to automatically identify geographic features that are anisotropic such as drumlins or barchan dunes, but works well on identifying linear features as well, such as roads or stream channels (J. Lindsay, personal communication, Dec 7th, 2017). Since the tool is multiscale, it may be able to identify micro-topographic features such as furrows if the grid resolution of the DEM is fine enough. The design process therefore require that the LiDAR be converted to raster format so that it could be entered into the MaxAnisotropyDev tool.
The first step of the process was to prepare the LiDAR data for interpolation into a DEM. The Peterborough LiDAR data is partially classified, so the ground points of the dataset were filtered before being interpolated into a raster. To isolate the furrows and ridges from the rest of the ground data within the DEM, the FeaturePreservingDenoise tool within WhiteboxTools was used. As noted by Turner et al. (2014), bare soil has a unique surface roughness when imaged using remote sensing technologies. The FeaturePreservingDenoise tool uses a modified algorithm created by Sun et al. (2007) which calculates the normal vector for each cell before smoothing the cells based upon whether or not they exceed a user defined threshold. The process runs in a kernel that is set by the user and runs through as many iterations as the user desires. The smooth and non-smooth DEMs can be put into the Subtract tool, with the result being a raster that just contains the surface roughness. The results can then be entered into the MaxAnisotropyDev tool. Given the variation in width of furrow-ridge systems noted by Beaudoin et al. (1990), it was necessary to test the process at multiple grid resolutions. During discussion with OMAFRA officials, it was indicated that an accurate DEM with a grid resolution as low as 0.20 metres could be made using the Peterborough data. A conceptual model of this process can be seen below in Figure 3.
Figure 3 - Proposed Conceptual Model for Identification Process