A METHODOLOGY FOR ENGINEERING GEOLOGY TERRAIN ANALYSIS WITHIN THE BOREAL FOREST REGION OF THE CANADIAN SHIELD
Landform (roughness) and relief are derived automatically from an analysis of a detailed hydrological-conditioned DEM. For roughness, a nearest-neighborhood analysis of the variety of elevation is performed on an integer version of the DEM. The results are grouped into 3 classes: 2 types of plains with relief averaging 1 m and 3 m and rough areas with relief greater than 5 m. The rough areas are further subdivided following an analysis of the variety of aspect and application of a majority filter. This enables the rough areas to be divided into areas of escarpments (low variety of aspect) and areas of irregular terrain (high variety of aspect), thus producing 4 roughness classes.
Relief is derived from the DEM using a nearest-neighborhood analysis of the range of elevation. The circular neighborhood used has a radius greater than half the width of the largest landforms in the area studied. The results are classified into 6 classes and then the radius of the neighborhood used internally buffers each class. They are then added sequentially and clipped by the roughness polygons. The plains from the roughness process are assigned relief values of less than 5 m and the rough area polygons are subdivided into polygons with the classified values of relief. The spatial distribution of the various roughness-relief polygons assists in defining landforms, such as a ridge, terrace etc.
Regional drainage conditions are estimated from RADARSAT imagery with a mask based on aspect and slope being used to limit the effects of topography on backscatter. Various types of analysis of Landsat imagery are then combined with all the above digital data layers to interpret material type.
An automated method that first determines the relationships of the various layers of digital information in areas of known terrain conditions, then predicts terrain conditions in adjacent unmapped areas, is being evaluated.