HYPERSPECTRAL SIGNATURE OF THE SALT SCAR AND OTHER OILFIELD DISTURBED AREAS AT USGS OSAGE-SKIATOOK PETROLEUM ENVIRONMENTAL RESEARCH (OSPER) SITE A, NORTHEAST OKLAHOMA - A PRELIMINARY REPORT

James K. Otton*
U.S. Geological Survey
MS 939 Box 25046
Lakewood, CO 80225
Voice: 303-236-8020
Fax: 303-236-0459
E-mail: jkotton@usgs.gov

David B. Reister
Oak Ridge National Laboratory
Oak Ridge, TN

On October 12, 2003, selected areas of Osage County, including the USGS' OSPER site A, were surveyed with an airborne, 126-channel spectrometer for wavelengths from 450-2500 microns under a DOE-funded contract to Oak Ridge National Laboratory (ORNL). The survey was designed to characterize areas disturbed by oil production to determine if remote-sensing data can be used to map such disturbance and its intensity. The survey was flown at a nominal altitude of 1400 m with a 3-m ground resolution. Site A is an abandoned oil production site active from about 1913-1973. The site consists of variably vegetated open areas in oak trees. A partly revegetated salt scar about 0.2 hectares in size has formed in the north part of the 1.5 hectare site downslope from two production pits. Other smaller salt scars are present elsewhere. At the time of the flight, adjacent Skiatook Lake was down about 1.5 m from its normal pool elevation exposing lake bottom sediments along the lake margin.

The 23,100 pixels at Site A have been classified using a new clustering algorithm developed at ORNL. Many distinct root pixels were found. The members of a cluster related to a root pixel are all of the pixels with a spectral correlation coefficient above a specified limit. Varying the cluster correlation limit from 99% to 99.6% controls the number of clusters that are generated. These pixel clusters have been mapped using distinct colors. The resulting cluster map can be related to features on the ground, including salt scars, roads, pits, exposed lake bottom, other disturbed areas of varying vegetation type and density (mostly grasses and forbs), and little-disturbed oak forest. For the 99% case, the method found 46 clusters for Site A; the cluster with the most members corresponding to the dense oak forest. Based on ground mapping and aerial photos, a salt scar area was defined that contains 261 3x3 m pixels. Two clusters, containing 18 pixels uniquely associated with the salt scar, are located in the core area of the scar, which is nearly devoid of vegetation. Other clusters occur on the scar but they also occur elsewhere including lightly vegetated areas on the exposed lake bottom, an area adjacent to the nearby recently active tank battery, old well sites and associated drill pads, and roads, and lightly vegetated natural areas where thin sandy soils occur over shallow sandstone bedrock. For the 99.6% case, the method found 100 clusters for Site A. The number of clusters uniquely associated with the salt scar increases from 2 to 3 but the total number of unique pixels found on the salt scar decreases from 18 to 9. However, the number of clusters where more than half the pixels occur on the salt scar increases from 2 to 10 with 54 pixels occurring on the salt scar. At either level of analysis, the method provides a unique signature for only a small part of the salt scar, however the 100 cluster case provides greater discrimination. The signature of most of the partly revegetated salt scar is thus similar to other oilfield disturbed areas at Site A and to the exposed lake bottom. The degree of correlation between the dominant cluster (the oak tree canopy) and all other pixel clusters was calculated and the degree of correlation mapped to show the relative intensity of site disturbance.