Spectral Enhancement and Automated Extraction of Potential Sinkhole Features from NAIP Imagery - Initial Investigations

Abstract

In eastern USA, karst sinkholes are responsible for millions of dollars of damage to buildings, infrastructure, agribusiness, and land availability. Drilling and geophysical techniques have been employed to locate developing sinkholes that have not yet expressed themselves at land surface through cover collapse, but these techniques are expensive and time consuming if not targeted to specific, small areas. This paper presents initial research into locating cover-collapse sinkholes through the use of aerial remote sensing. National Agricultural Imagery Program imagery, captured in the summer of 2004 with 1-m ground resolution, in 3, 8-bit bands (visible, RGB), was found to be effective in highlighting circular features that appear to be related to sinkhole occurrence on two farm sites. In some instances, the features were sinkholes up to 5 m in depth. In other instances, the features did not represent existing sinkholes but potential sites of incipient cover collapse. Differences in agriculturally related vegetative cover at this time affected the efficiency of determining the extent of karst development. Surface geophysical techniques and drilling will be used to characterize the subsurface when cropping practices allow. The goal is to correlate imagery features to subsurface characteristics so that remotely sensed imagery can be used as a means of delineating potential cover-collapse sinkholes before they express themselves at the surface. Development of such a tool will provide a time- and cost-effective approach to land-use planning and, therefore, substantially reduce damage done by cover collapse.

Keywords: Cost, cover collapse, feature extraction, karst, linear, remote sensing, spectral enhancement