A New Automated Technique Within ArcGIS to Compute the Attitudes of Planar Topographic Features

Introduction Common methods of visualizing topographic data such as contour maps or shaded relief do not yield a clear and quantitative representation of the full 3D orientation of a surface. ArcGIS® contains built-in tools (the Surface toolset in Spatial Analyst® toolbox) that can determine the Aspect and the Slope via nearest neighbor algorithms [1], but lacks the capacity to use kernel sizes larger than 3x3 pixels, or to provide goodness-of-fit estimates (Fig. 1: AD). A New ArcGIS Tool Herein, we present our in-house developed ArcGIS add-on that computes surface derivatives using Least Squares Regression (LSR) which is similar to the approach used in the manual planar attitude calculations used within the software Orion [2, 3]. The new approach allows more flexibility by enabling the use of larger, user-defined kernel sizes, and the ability to calculate Mean Squared Error (MSE) (Fig. 1: E, G). Unlike ArcGIS built-in tools, our software has a capacity to generate parameters for Cartesian plane equations, thus making it possible to distinguish parallel surfaces from coplanar. We also demonstrate a new, Augmented Visualization of Attitude (AVA) presentation for displaying attitude data in RGB format, wherein strike is encoded as hue and dip as saturation (Fig. 1: F, H). Advantages of the New Tool The ability to use larger kernel sizes becomes useful where surface roughness is an issue, or where the Digital Elevation Model (DEM) is subject to other noise. Increasing the kernel size minimizes the effect of small (e.g., single pixel size) surface undulations, without sacrificing data quality. As long as the kernel size is markedly smaller than the features of interest, the quality of fit will increase as more points are available for regression analysis. The MSE parameter provides a cleaner, less noisy representation of surface irregularities then can be obtained using ArcGIS® Curvature tool. Points with high values of MSE are indicative of either invalid pixels in the parent DEM, exceptional roughness, or bends. Estimating surface roughness using MSE is superior to doing the same via the neighborhood standard deviation, as the latter does not take into account the slope of the surface.