Searching for Oil Seeps & Oil-impacted Soil with Hyperspectral Imagery
By James Ellis, Ph.D.

For quite some time, specialists have considered using remote sensing technology to detect oil seeping out from subsurface petroleum reservoirs. Until recently, these efforts have been hampered by the fact that the spatial and spectral resolution of most sensors is too limiting for use in onshore applications, especially in remote terrain. In addition, environmentalists and alternative land-use proponents continue to put pressure on operators of refineries, tank farms, pipelines and other industrial sites, asking them to look for alternative ways to more efficiently detect and prioritize oil-impacted sites.
      The remote sensing team from HJW GeoSpatial Inc. (HJW) - formerly known as Hammon, Jensen, Wallen & Associates Inc. - wondered if the relatively new airborne hyperspectral technology that they were using for mineral exploration could be utilized for petroleum applications as well. In 1998, and with this goal in mind, HJW joined an airborne hyperspectral group shoot initiated by The Geosat Committee Inc. (Geosat).
      Earlier this year, HJW initiated a follow-up study of the 1998 Geosat shoot. Revisiting areas known to contain oil seeps and oil-impacted soils, hand held hyperspectral sensors that had been imaged during the group shoot were put to use. The primary mission of this ongoing project was to document spectral characteristics that are typically associated with oil-impacted soils and seeps. These findings were then used to construct a spectral library that will make detection of onshore oil seeps and oil-impacted soil more rapid and reliable than traditional methods.

Evaluating Airborne Imagery
As part of Geosat's cooperative R&D; project, airborne hyperspectral flight strips were acquired over the West Sulphur Mountain in Southern California in 1998. Eight sponsors participated - Chevron, Shell and Exxon-Mobil among them - as did seven technology providers, including HJW. Hyperspectral data cubes were acquired by a sophisticated sensor built by Integrated Spectronics, owned and operated as Probe-1 by Earth Search Sciences Inc. Each of these data cubes contained 128 bands that covered the entire VNIR-SWIR wavelength spectrum.
      HJW came to this project already familiar with the prolific oil production history of the West Sulphur anticline because, during the 1970s, a member of their remote sensing team had mapped its geology and onshore oil seeps. The company consulted geologic maps and databases to jumpstart the evaluation as to whether airborne hyperspectral imagery could detect onshore oil seeps and oil-impacted soils. Oil-seep locations were found on a geologic map that had been published by the Thomas Dibble Foundation in 1990. This map, along with the hyperspectral imagery, was draped over a USGS Digital Elevation Model (DEM) to facilitate the location of "Dibble oil seeps" on the flight strip (Figures 1a-1c).
      The pixels in the vicinity of the oil seeps were individually analyzed. HJW discovered a family of hyperspectral signatures that was unique and appeared to correlate spatially with pixels that were identified as being within or adjacent to the published "Dibble oil seeps" (Figures 1d-1f). However, the value of this spatial correlation was questioned for a number of reasons. First, the geologic map was published at a scale of 1:24,000 providing, at best, x/y accuracy of plus-or-minus 12 meters for "well-defined points," which do not include oil-seep locations as mapped by a field geologist. Second, the hyperspectral imagery lacked camera orientation data and ground control points. Finally, there was no way to know how various materials such as liquid oil, tar, oil-impacted soil, clean soil, vegetation, etc., were interacting within the airborne 5x5-meter pixels, and what effect this interaction had on the resultant spectral signatures.
      Twenty kilometers east of the "Dibble seeps," HJW observed clearings on the airborne flight strips that were situated within the heavily vegetated north flank of the Sulphur Mountain anticline. These clearings were interpreted to be oil well sites. Technicians evaluated individual pixels within these clearings, discovering that these clearings provided the same spectral signatures as those found within the "Dibble oil seeps" (Figure 1).
      In both areas, the airborne hyperspectral signature presumed to identify oil seeps or oil-impacted soil had some vegetation associated with it. Also, there was evidence that the oil seep or oil-impacted soil was associated with more sparsely vegetated terrain that could have been clearings, roads, well sites, etc. On the airborne imagery, the signature was possibly associated with road shoulders and unpaved access-road surfaces on the anticline that had either been built with materials mixed with oil residue or else were subjected to sheetwash from the upslope areas of oil seeps. Based upon this airborne imagery, HJW hypothesized that oil seeps and oil-impacted soil may foster development of an open landscape, giving rise to a unique vegetation assemblage that would alter soil characteristics such as color, amounts and type of iron, etc.

Field Measurements and Observations
In August 2000, HJW sent an experienced crew into the field with a handheld hyperspectral sensor and a GPS receiver. Their objective was to confirm that the family of airborne spectral signatures found earlier was caused directly by oil seep and oil-impacted soils (Figure 2). An ASD spectrometer capable of measuring VNIR-SWIR spectral responses was used to confirm the "Dibble Seep Complex" (Figures 2b and 2c). HJW discovered that, along with numerous active oil seeps, a very large tar deposit was sited in this area. There were areas in or near the deposit where vegetation appeared stressed, but there were also areas where vegetation grew vigorously. Botanical observations were collected in addition to seep samples.
      To determine the effect that different mixtures had on hyperspectral signatures, a systematic measurement program was designed for the handheld spectrometer. In particular, areas were chosen for measurement that had liquid oil or tar associated with different amounts of vegetation or soil. These data were then processed to determine the spectral characteristics of different mixtures, and also the difference between impacted and non-impacted sites.
      Observations and measurements made in the field established that the airborne signatures were valid. The airborne sensor detected the signal from soils impacted by oil, and HJW's newly designed work processes successfully extracted this subtle signal from the data cube. Soils impacted with hydrocarbons - as well as onshore oil seeps - were determined to have physical properties and composition different from soils not impacted. Sophisticated hyperspectral sensors with narrow bands and high signal-to-noise characteristics were capable of detecting these differences, provided that these differences were not masked or overly subtle. HJW's research further indicated that the full wavelength spectrum, from visible through shortwave-infrared, was needed so as to differentiate oil-impacted pixels from those that were not. The company ultimately documented changes along the spectrum that were indicative of oil-impacted soils and oil seeps.
      A number of sites characterized by oil seeps were visited during the August 2000 field session. These included a spot along the Southern California coast at More Mesa, located west of Santa Barbara, that is a prominent outcropping of tar possessing live oil seepage zones (Figure 2). Various materials were measured in the field using the handheld spectrometer, which were then compared directly with pixels in the airborne flight strip (Figures 2b and 2c). These field and airborne measurements will ultimately form the basis for using and understanding the new spectral library of oil seeps and oil-impacted soils.
      An industrial facility on the hyperspectral flight strip was analyzed for oil-impacted surfaces (Figure 3) to evaluate the effectiveness of the preliminary spectral library. Results of the classification were not field-checked due to inaccessibility, but the spatial distribution of the impacted surfaces was reasonable and included a circular site within the facility, plus access roads and turnouts. Storm runoff might possibly have carried hydrocarbons from nearby natural oil seeps onto the road surface. HJW found no hyperspectral evidence of stressed vegetation along the perimeter of the facility (lower right portion of Figure 3).

The Spectral Library
Hyperspectral sensors are unique in that they have enough spectral resolution to identify individual surface materials based solely on respective spectral signatures. Spectral libraries contain a group of hyperspectral signatures that have been positively matched with specific materials at the Earth's surface. Spectral libraries have been constructed for minerals, plant communities, man-made materials, vegetation stress, and so on.
      The spectral library of oil-contaminated soils and onshore oil seeps built by HJW may be the first of its kind anywhere in the commercial sector. This spectral library will include pixels with varying amounts of oil-impacted surfaces, and can be used with handheld, airborne or satellite sensors that span the spectrum from visible through shortwave-infrared. The library will enable skilled analysts to detect oil-impacted surfaces across remote geologic structures and within such dense urban sites as refineries, tank farms, pipelines, and brownfields. When combined with HJW's integrated work processes, this spectral library will open up the use of hyperspectral technology for oil exploration, establishing baselines, and monitoring changes at both active and abandoned industrial sites.
      HJW's advancement of hyperspectral imaging for oil-seep detection expands the technology's functionality throughout the oil industry. In addition to facilitating geologic exploration for onshore oil seeps, operators of industrial facilities can also use the technology to plan detection, prioritization, and remediation programs. Hyperspectral imaging also has wide and proven uses for mapping vegetation communities, vegetation stress, water areas that might include industrial settling ponds, waterways, wetlands, etc., and infrastructure areas within these respective industrial sites. Geosat and HJW ultimately expect the hyperspectral library to become a valuable tool that will step up the commercialization of airborne hyperspectral imaging for even greater vertical applications.