Smooth High Resolution DEM

Description

Smoothing is used to blur DEMs to remove the changes in elevation that are too small to indicate features of interest (i.e., microtopographic noise), which are ubiquitous in high-resolution DEMs.

Usage

• Microtopographic noise can be the product of both erroneous data or rue variations in the elevation of the vegetated ground surface.
• Identifying and filtering noisy data are challenging as this risks modifying the true land surface or degrading features of interest.
• Although many smoothing methods have been proposed, just a few are selected for their common use in hydrology-related applications. These include mean, median, gaussian, and Perona-Malik smoothing.
• In addition to the smoothing algorithm, the rate of smoothing is an important parameter in this DEM preprocessing step. The rate of smoothing controls the scale of features that are preserved. Users should choose this parameter based on the scale of the hydrologic parameter that is being extracted.

Parameters

Parameter Name	Type	Direction	Data Type	Dialog Reference
Input High Resolution DEM (*.tif)	Required	Input	Raster Layer	Input high-resolution (~ 2m resolution or finer) DEM to be smoothed. Must be in TIF format.
Smoothing Method	Required	Input	String	Mean smoothing performs a linear convolution using the user-defined smoothing window size. This smoothing window slides across the input DEM and the center cell of each window is replaced with the mean of all cells within the window. The rate of smoothing is determined by the smoothing window size. Median smoothing performs a nonlinear convolution using the user-defined smoothing window size. This smoothing window slides across the input DEM and the center cell of each window is replaced with the median of all cells within the window. Median smoothing is less affected by outliers in windows and is typically well-suited to salt-and-pepper type noise. The rate of smoothing is determined by the smoothing window size. Gaussian smoothing performs a linear convolution using a Gaussian function. Gaussian smoothing ensures causality, meaning it will no result in spurious features because any features at a coarse resolution must have a cause at a finer resolution. The rate of smoothing is determined by the Gaussian kernel, where the standard deviation parameter asks as a proxy for a smoothing window size. Warning: Gaussian smoothing results in boundary effects that cause erroneous values at the edge of the input DEM. The smoothed DEM will be returned with smaller extents to remove these values. The larger the smoothing window, the more boundary values that need to be removed. Perona-Malik smoothing performs a nonlinear, anisotropic diffusion. This method estimates geomorphic feature boundaries to be where the slope is steeper than 90% of all slopes within the DEM. Then, smoothing is promoted only between estimated feature boundaries to preserve feature edges and achieve preferential smoothing. The rate of smoothing is determined by the number of smoothing iterations.
Output Smoothed DEM (*.tif)	Required	Output	Raster Dataset	Name of the resulting smoothed DEM. Must be in TIF format. If the directory does not exist, it will be created.
Smoothing Width (m)	Optional	Input	Double	For mean and median smoothing, the smoothing width refers to the X and Y length of the smoothing window in meters. If linear units of the DEM are not in meters, the conversion is executed automatically. Features smaller than the area of the smoothing window will be blurred. For Gaussian smoothing, the smoothing width is used to approximate the standard deviation used in the Gaussian kernel, according to Lashermes et al. (2007). If linear units of the DEM are not in meters, the conversion is executed automatically. Default value given are starting points, but users should adjust these based on their specific application.
Perona Malik Iterations	Optional	Input	Double	For Perona-Malik smoothing, the number of iterations determines the scale of features preserved and smoothed, however this parameter has no unique and uniform equivalent spatial scale. A higher number of iterations will result in coarser output landscapes. Default value given are starting points, but users should adjust these based on their specific application.