Home ]


SWLGS Luncheon Topics

Updated Dec. 6, 2008

Dec. 9, 2008

Unexploded Oil: Lessons Learned from the Search for Unexploded Ordnance Applied to the Search for Hydrocarbons Using CSEM


Jack Stalnaker, Geology Department at the University of Louisiana at Lafayette


Abstract Summary

Controlled source electromagnetic (CSEM) geophysical techniques have long been considered nearly ideal for the detection of unexploded ordnance (UXO) given the compact, metallic (and therefore highly conductive) nature of an unexploded bomb. However, the problem is complicated by the presence of metallic debris, or “frag” comprised of debris from exploded bombs, which is generally found in the same location as UXO. The low resolution of CSEM precludes any direct imaging of the buried targets, and the geophysicist is forced to resort to inversion by way of a physical model of induction in a metal target. Realistically, the complex governing physics and the sheer size of the areas to be surveyed dictate that the model must be a simple approximation. Classification of the target is accomplished by comparing estimated model parameters with a library of known targets. The acts of library description and comparison, though easy by human standards, are delicate processes, and many approaches have been proposed to accomplish the comparison.. 

Hydrocarbon reservoirs of the kind typically found off the Louisiana coast may likewise be described as compact—albeit resistive—targets. It stands to reason, therefore, that a similar processing approach may be undertaken. In order to apply the UXO approach, a simplified physical model is desirable (though not necessary), and some description of known hydrocarbon reservoirs in terms of the model parameters is needed. Probability distribution functions can be non-parametrically derived from statistical descriptions of map and log data using kernel density estimation (KDE). An a posteriori probability of reservoir detection can then be estimated using a Bayesian approach. A related application of Bayes’s Theorem can also be used to classify a fresh data set as hydrocarbon bearing or non-hydrocarbon bearing, and the Chernoff bound can be computed to determine the level of confidence one should place in the decision. The utility of a model-based probabilistic description of hydrocarbon reservoirs in combination with a rapid approximate model is expansive. For instance, Monte Carlo simulations can be used to describe the range of expected CSEM responses from a suite of known reservoirs. Consequently, the effectiveness of such ubiquitous CSEM reservoir detection approaches as magnitude and phase variation with offset (MVO and PVO) can be evaluated rigorously.


Biographical Sketch



Jack Stalnaker is an assistant professor in the Geology Department at University of Louisiana at Lafayette. His areas of research include near surface geophysics, controlled-source electromagnetic geophysics, statistical signal processing, physics-based modeling, archaeology, computer science, and environmental studies. Jack received a Ph.D. in geophysics from Texas A&M University in 2004, and Bachelor’s degrees in geology and in anthropology from College of Charleston in 1998. Before coming to Lafayette, Jack was a postdoctoral researcher in the Electrical Engineering Department at Northeastern University.