| Airborne SAR for Geological Mapping
By D.F. Graham and D.R. Grant Introduction
Conventional surficial geological mapping is based on ground observation, which serves to establish the composition of distinctive landforms. This is aided by the interpretation of standard panchromatic airphoto stereo-pairs, which permits the extent of the landform associations to be sharply delimited. In general, 1:20,000-scale photos are used, and these permit differentiation of surficial units, which typically range in area from 0.01 to 1 km2. Contacts vary from gradational to sharp, where widths are as little as a few meters. With the experience backed by field observations, photo-interpretation has become the common means for reliably delineating surficial materials and landforms.
With the advent of digital imaging systems, however, there is interest in evaluating their utility as mapping tools relative to the aerial photo method, and specifically to determine whether digital images reveal additional features. Previous work by Singhroy et al. (1989, 1993) and Kenny et al. (1991) also reported on the use of SAR for surficial mapping in Canada. In an initial test that compared SAR data and aerial photo interpretation, Graham and Grant (1991) used photographic prints of 16 meter (wide swath) SAR data from the Canada Centre for Remote Sensing (CCRS) and focused only on lineaments of bedrock and surficial origin over an area of central newfoundland. Lineaments interpreted from radar were found to correlate well with those obtained from aerial photos, and some new features were noted. One reason for the extra information is that SAR is able to image the ground surface at a relatively low angular range, thereby enhancing subtle and large-scale features that are either not evident or are only weakly expressed on vertical aerial photos.
That work prompted a second test using the 6 meter (narrow swath) CCRS SAR data for the same general region. This test differs from the first in that the radar data were of higher resolution, custom-enhanced, geographically corrected, and output digitally by software as a continuous-tone photographic print. The surficial geology contacts, overlain digitally on the SAR image, allowed a direct visual comparison to be made. Method SAR Data
The data were processed in several operations. The original SAR data were acquired as high-resolution (6 m) narrow-swath, X-band SAR data. Imaging was accomplished on board CCRS's Convair 580 aircraft at an altitude of 6 km, a depression angle of between 14 and 45 degrees, and a ground swath of 18 km. The raw pixel dimensions were calculated as 4.67 m azimuth by 4.41 m range and resampled to a 5 m2 pixel size. Next, a correction for antenna pattern in the ground range direction was made. Directional and median filters did enhance major lineaments but removed subtle lineaments that were important for feature interpretation. Finally, to reduce the foreshortening that is inherent in radar data, a third-order geometric correction was made, thereby registering the SAR image geographically to UTM coordinates. The image was saved in a TIFF format and imported into an image processing program (Adobe PhotoShop), then the de-speckle filter was applied to reduce image noise (radar speckle), increasing the clarity of all radar lineaments. Aerial Photographs and Geology
The SAR image was compared to a portion of two typical Geological Survey of Canada surficial geology maps. The maps show surface materials and geomorphic features based on ground observation and delineated by the interpretation of 1:50,000-scale black and white aerial photo stereo-pairs. Of the 10 to 15 classes of surficial material that are usually identified, 11 are present in the test area: three classes of bedrock exposure, three of varying till thickness, three of alluvial origin, one colluvial, and one organic. Landform elements (shown by symbols) include bedrock structural lineaments (fault and fold traces), longitudinal glacial lineations (drumlins, fluting, crag-and-tail hills), and side-hill glacial meltwater channels. Areas of individual units range from 0.01 km2 to 1 km2, and the widths of their boundaries range from typically 50 m to 100 m, but for bedrock, alluvial, and organic terrains may be as narrow as 1 to 2 m.
The surficial geology contacts and features from portions of five aerial photo stereo-pairs were compiled optically onto the SAR image, using lakes for registration. The resulting composite image (Figure 1) provides the basis for an evaluation of the degree to which radar data can differentiate the terrain characteristics that are seen on aerial photos and traditionally used to classify surface materials as to their origin, composition, and thickness. Results Information Extraction
The correspondence between the tones and textures on the SAR image and the units interpreted from aerial photos varied from good to poor, depending on the unit but irrespective of depression angle. First, however, it must be acknowledged that mapping with aerial photos is dependent on stereoscopy. Using a single aerial photo, without the benefit of stereographic aerial photo interpretation, much less could be mapped - a limitation that also applies to SAR imagery. Radar stereoscopy increases the ability to interpret terrain relief variations. Allowing for this disadvantage, the correspondence can be considered to be generally good, as outlined below.
The SAR image (Figure 1) shows variations in textural roughness and tonal brightness, which generally correspond to surficial units mapped from aerial photos and substantiated by ground-checking. Within till veneer, three distinct drumlin sets (numbered 1,2, and 3 on Figure 1) of different ages that range in trend from northwest, through north, to northeast are present. These drumlin trends record the changing flow regime as the ice sheet shrank from its maximum extent offshore to a remnant cap situated south of the study area.
Within the till blanket, meltwater channels record the final recession of an ice front towards the north. Application to Mineral Exploration
Some practical applications of the additional terrain data acquired by SAR that are of value to mineral exploration, as well as to a variety of other land use activities, can be cited. The ability to image larger areas and discern more subtle relief features than is possible with conventional aerial photos and standard stereoscopic examination enhances any interpretation based on aerial photos. Of particular value to mineral exploration is the clarity with which bedrock structural lineaments and drift lineations related to ice flow are expressed. Additionally, radar provides a convenient means of assessing the relative depth to bedrock surfaces in that its roughness signature is used as an indicator of relative till thickness. The ability of radar to detect variations in the stoniness of till surfaces provides insights into till composition that can only be guessed at from aerial photos. Its ability to detect small-scale widespread evidence of reworking by glacial meltwater is an indirect indication that the surface has been winnowed of fines and rendered more stony, a factor to consider in a mineral exploration program based on geochemical analysis of overburden.
Some advantages of working mainly with digital files are noted. Of prime value are the overall efficiency, the flexibility to examine various subsets of data, and the reduction of information loss, which inevitably occurs in the generation of photographic prints. Developing the best image by on-screen iteration, thereby limiting the hard-copy output to one print, greatly reduces costs and is more environmentally responsible. Conclusion
This test further underlines the value, in terms of various geoscience applications, of interpreting SAR data through the collaboration of a radar specialist and a terrain scientist with experience in the area. Given sufficiently extensive 6 m coverage, the imagery is particularly well suited for mapping the extent of broad patterns and the continuity of large linear features, both of structural and glacial origin. However, before radar can achieve greater potential as a geoscience mapping tool, it is necessary to have stereo coverage and an improved accuracy in geometric correction by the introduction of a Digital Elevation Model (DEM). Acknowledgements
Airborne SAR data were acquired by the Data Acquisition Division of the Canada Centre for Remote Sensing (CCRS). David Graham is employed by MIR Teledetection Inc., Longueuil, Quebec, under joint contract with the Geological Survey of Canada and CCRS. This article was published with the permission of the Canadian Journal of Remote Sensing. About the Authors:
D.F. Graham is with the Remote Sensing Office, Mineral Resources Division and may be reached at 613-943-0205. D.R. Grant is with the Terrain Sciences Division of the Geological Survey of Canada. Back |
