Southeastern Section - 66th Annual Meeting - 2017

Paper No. 11-9
Presentation Time: 4:00 PM

USING A GIS-BASED MODEL TO IDENTIFY OUTCROPS ALONG ROADWAYS


LANNOM, Michael F., Earth and Planetary Sciences, The University of Tennessee, 306 EPS Building, 1412 Circle Drive, Knoxville, TN 37996-1410, BAUER, Jennifer E., Earth and Planetary Sciences, The University of Tennessee, 1621 Cumberland Ave, 602 Strong Hall, Knoxville, TN 37996-1410 and SUMRALL, Colin D., Department of Earth and Planetary Sciences, University of Tennessee, 1621 Cumberland Ave, 602 Strong Hall, Knoxville, TN 37996-1410, mlannom1@vols.utk.edu

A significant issue for geologists is georeferencing old locality data. Old localities often lack precise coordinates and are described in reference to landmarks that no longer exist making them difficult to resample. Here, we developed a model where GIS is utilized to identify outcrops of interest along roads. This model, in concert with general knowledge of bedrock geology, is a beneficial tool for geologists attempting to discover localities containing data relevant to their research.

The model classifies areas of interest as bare rock or soil from Landsat imagery and combines these data with an analysis of the slope angle from Digital Elevation Models (DEM), where higher slope angles suggest higher probability of exposed strata. These data are then evaluated by proximity to roads and outliers are removed from the dataset. The resulting data highlight areas with a high probability of rock being exposed along roadways. The initial parameters used were supervised classification of rock in the visible spectrum (0.4 - 0.7µm) where we selected the spectra assigned to rock, >30° slope in 10-meter resolution DEM, and 30-meter road proximity. We can further test the accuracy of each parameter (spectral response, elevation change, and road proximity) to provide the highest probability of the model predicting exposed rock while minimizing false positives. Using a wider spectral range and higher resolution datasets will also increase accuracy.

We evaluated this model with known outcrop localities, in western and central Kentucky. We targeted upper Mississippian mixed carbonate and shale strata that were being studied for a faunal analysis. This area contains an abundance of topographic variation clearly seen in the DEM images. At each of the outcrops variation was noted in exposed rock face and slope of the outcrop. Each of these previously visited outcrops suggest that the model is viable but performs best in regions with variation in slope compared to flat lying areas.