It seems like the seismic acquisition side of the exploration business has been exploding at the seams over the past few years, and one of the most exciting areas has been in sampling studies. Some companies are pushing the boundaries of channel count on land, others are looking at node technology in marine environments to provide superior imaging and yet others are relying on new broadband techniques to better sample the frequency spectrum.

Attitudes toward sampling geometry are also changing, and ConocoPhillips is driving a push toward what it calls “compressive seismic imaging” (CSI) to both optimize source and receiver location and reduce the total number of sources and receivers needed. The concept is based on compressive sensing sampling theory, which was developed to reduce aliasing in sensing.

According to an article on aliasing by Bruno A. Olshausen at the University of California-Berkeley, aliasing occurs when signals are sampled at rates that can’t capture the changes in the signals. To reduce or avoid aliasing, Harry Nyquist proposed a theorem stating that the sampling frequency should be at least twice the highest frequency contained in the signal. But in an article published in the April 2014 issue of The Leading Edge, several ConocoPhillips authors note, “The reality of limited access and funding requires us to make do with orders of magnitude fewer sampling points than Nyquist theory would dictate.”

Enter compressive sensing. The new sampling theory, according to the article, provides a mathematical basis for designing non-uniform sampling systems. This new mathematical basis can be used to determine the degree to which signal can be separated from sampling noise for a given sensor layout. The most common technique used in the industry to date is to use randomized sampling locations, which allows coherent signals to be separated from random sampling noise using well-known signal processing techniques.

ConocoPhillips has dubbed its particular brand of compressive sensing non-uniform optimal sampling (NUOS). The newer method uses an optimization loop to establish optimal source and receiver locations rather than relying on random techniques.

The system was further tweaked to allow for simultaneous sources. Two realizations of shooting patterns for dual-vessel operations, one random and one optimal, showed very similar characteristics and had comparable values of coherence. But the optimal plot also minimized both coherence within each pattern and cross-coherence between patterns. This results in 50% lower cross-coherence in the optimal realization.

The method was field-tested in the North Sea at the end of a production survey to provide a baseline for comparison. After constructing a detailed geologic model, NUOS designs for source spacing were built based on minimal cross-coherence criteria. Continuous receiver records were created by interpolating shots from the model and then blending them into continuous records, the authors noted.

After processing, it was concluded that this technique enables the design of simultaneous source surveys that maximize the fidelity of de-blended source records.

Comments