Interpreting seismic data may be a science, but it is also an art
Acquiring the seismic data is essentially taking acoustic pictures through the earth’s strata – rather like a camera that, for all the effort to create a precision machine capable of generating pin-sharp images, always delivers blurred results that then need correcting. In that regard it’s little like correcting the Hubble telescope’s early life aberration to enable it to deliver workable results pending more permanent re-engineering only, for all the efforts to date, precision seismic eludes.
However, substantial strides have been made over the past 15 years especially, though there are interpretation techniques that have been available for significantly longer but which only now are gaining recognition for the vale they can bring.
A prime current in this regard is perhaps ‘Gabor deconvolution’, which is founded on pioneering work carried out by British physicist Dennis Gabor, who won the Nobel Prize in 1971 for inventing the hologram.
In a nutshell, deconvolution of data is a fundamental part of the construction of
high-resolution seismic images. This approach handles the difficult task of separating a seismic trace (sensor recording) into two components, both of which are unknown. One represents the seismic waveform, the other reflectivity, which is the cause of image ‘blurring’.
But it is the massive advance in computing power that has been the main enabler for the advances of recent years. Bear in mind that, prior to the mid 1990s, seismic survey vessels capable of on-board real-time preliminary processing were very few and far between. This capability essentially grew out of the rapid take-up of 3D seismic during the 90s and, today, is commonplace.
Coupled with that is increasingly powerful, faster data processing and interpretation
shore-side, including setting up visualisation suites – often referred to as ‘hives’.
A further catalyst to the rapid acceleration is commoditisation of seismc. Until the mid 1990s and certainly the late 1997 through 1999 oil price slump, oil company preference was for exclusive arrangements with survey companies.
However, with companies like BP in the vanguard of change at that time, this approach was scrapped in favour of buying data from the market. That had much to do with precipitating the large seismic vessel-building programme of the past decade or so, with a significant number of high capacity units joining the global fleet and yet more on the way. It also has much to do with the significant corporate consolidations that have taken place too including, of late, the snapping up of small boutique companies offering interpretation services with a twist.
Deals just closed or in-train at the close of 2007 include the merger of Norwegian TGS_NOPEC and Wavefield Inseis to create TGS Wavefield; PGS has tabled a mandatory offer for Arrow; while Eastern Echo of Dubai was fighting off Schlumberger.
Upstream petroleum currently spends around
US$4–7billion a year on data acquisition and analysis. Companies like WesternGeco, Fugro, TGS-Nopec, PGS and CGG-Veritas are the backbone of the sector and all essentially rely on what are best described as seismic processing farms, comprising tens of thousands of CPUs (central processing units), to continuously process data.
The demand for real-time seismic analysis is growing rapidly and, as both power and space become issues at data processing centres on the beach, the industry is hungry for yet faster, more efficient processing techniques capable of delivering supercharged seismic processing continues.
This is why there is growing interest in coprocessor acceleration, which uses FPGAs (field programmable gate array), GPUs (graphics processing unit), Cell processors, and ClearSpeed boards to enable the most
computer-intensive parts of seismic processing. But still it's not enough and the search for the ultimate technology continues.
Consider the application of GPU technology. It was towards the end of 2005 that Houston company Headwave set up an offshoot to offer the upstream sector 3D visualisation software designed specifically for viewing and interacting with pre-stack data.
According to the group, this marked the first commercially available solution in the industry to make use of GPUs on graphics cards for raw processing power.
By utilising the GPUs on a cluster of PC nodes, Headwave offered the ability to visualise enormous amounts of data, and can apply geophysical algorithms to the datasets with real-time results for true interactive visualisation.
Currently, much of the data gathering by upstream seismic practitioners use up to 18bits (14bits of data and 4bits of amplifier gain). This means that the current 32-bit implementations of GPUs are more than adequate for the job and the 64bit device is in the pipeline, according to Headwave.
It should be remembered that 3D visualisation has traditionally been a tool for the interpreter and the reservoir analyst, yet the very features that make it such a powerful implement for post-stack analysis are directly applicable to pre-stack workflows … the ability to view huge amounts of data very quickly, to move through a volume (of data) and get a sense of the quality of that data in seconds, to calculate and display attribute information over the seismic data, and to interpret geological events. But finding and developing new approaches is expensive and, although the free-market prevails in seismic, nonetheless, the sector does occasionally attract government money. A prime example is a joint initiative implemented under the European Union’s BEINGRID programme, which falls under the EU’s sixth research framework scheme.
As stated already, modern seismic data processing and geophysical simulations require immense amounts of computing power, data storage and sophisticated software. The research community hardly keeps pace with this evolution, resulting in difficulties for small or medium research centres to exploit their innovative algorithms.
Grid Computing is a opportunity to foster sharing of computer resources and give access to large computing power for a limited period of time at an ‘affordable cost’, as well as sharing data and sophisticated software.
The capability to solve new complex problems and validate innovative algorithms on real scale problems is also a way to attract and keep the brightest researchers for the benefit of both the academic and industrial R&D geosciences communities.
Business Experiment 18, as it is called, brings together three upstream petroleum supply chain partners: CGGVeritas, NICE and Petrosoft, plus agency TNO. Of the partners, CGG-Veritas is a major player in the seismic market; TNO is a Dutch applied sciences organisation; Petrosoft, is a SME with long-time experience in a research and development of seismic data processing software; while NICE, is a private,
self-financed European company which offers software products and IT solutions and is versed in Grid.
The experiment offers access to computing resources and various state of the art technology services for seismic processing and reservoir simulation using Grid technology.
Whether or not this work will make a contribution to the need for cross-industry standards is not known. Indeed, the issue of establishing American Petroleum Institute standards for seismic processing is a very real one, otherwise there is a danger that this will impact the sector's efficiency and therefore value.
