By John Kemp
LONDON (Reuters) - No one has ever seen an oil field.
Typically buried thousands of feet below the surface, oil fields are like a sponge saturated with a mixture of oil, water and gas, rather than the underground cavern most people imagine when they think about oil and gas reservoirs.
Many rock formations contain limited amounts of oil and gas, especially in sedimentary basins, but only a few contain enough to be worth the costs of drilling. Finding substantial amounts of oil is therefore akin to seeking a needle in a haystack.
Before deciding whether to sink a well, exploration and production firms must estimate how much oil the reservoir contains, and how much might be technically and economically recoverable, usually based on interpreting relatively limited amounts of data on what is actually down there.
As the easy oil runs out, and exploration and production companies delve deeper below the surface, in tougher environments like deepwater and the Arctic, the costs of drilling and field development rise sharply, and the risks of getting it wrong are correspondingly greater.
Oil companies have resorted to more and more computing power to "de-risk" the process.
Rather than drill wildcat wells and hope for the best, which is basically what happened in the 19th and early 20th centuries, most exploration and production companies now rely on super-computers to process and combine vast amounts of data gathered from seismic, gravity and magnetic surveys, in addition to well logs, to produce a three-dimensional "map" of the subsurface.
The aim is to identify drilling locations that maximize the chance of finding an exploitable amount of oil and gas, minimize the danger of drilling a "dry hole", and target the most-productive "sweet spots" in a formation.
The most modern models often employ four-dimensional visualizations to identify how oil, gas and water flow through the reservoir during production. They enable field engineers to plan the optimum layout for producing and injection wells, and to scrape some of the oil and gas left behind from primary production through waterflooding and carbon dioxide injection.
WILDCATS AND SUPER-COMPUTERS
Oil majors like BP and Shell own and operate some of the world's biggest super-computers to help process the vast amounts of information gathered in the exploration process.
BP's computer needs have grown 10,000 times since 1999, according to the company, as a result of its increasingly data-driven exploration and production process.
Its High Performance Computing Center, in Houston, Texas, has maxed out its computing power and cooling capacity. The company is building a new facility on the same site that will roughly double its computing power.
The new three-story facility at Westlake Campus, scheduled to be operational by the middle of 2013, will have more than 67,000 central processing units, and data storage equivalent to 147,000 Apple iPods with 160GB of memory each. It will be able to perform more than 2,000 trillion operations per second (2 petaflops).
"This is not just about building a bigger and better computer. BP's new high performance computing centre will be as important to our global search for new energy resources as any piece of equipment we employ today," the company said in December.
In the Middle East, Saudi Aramco's Exploration and Production Engineering Centre (EXPEC) operates a legendary control room that enables engineers to visualize Ghawar, the kingdom's super-giant oil field, as if they were walking through it, using special eye-pieces and screens, backed by a super-computer.
Aramco's latest and most powerful simulator is GigaPOWERS, which can break a giant oilfield up into billions of separate cells to analyze it better.
"GigaPOWERS is an innovative reservoir simulation technique (that) helps us to better analyze and predict the production rate of our oil and gas reservoirs over time so we can manage reserves well into the future," Aramco says.
"By combining relevant physics, chemistry and thermodynamic relationships, we create highly complicated mathematical equations. The solutions to these equations are then presented in grid blocks that create a detailed visual picture of an oil field," Aramco explains on its website.
"The increased resolution of GigaPOWERS has allowed us to identify bypassed oil zones and additional oil zones, leading us to drill new wells and recover more oil," the company adds.
Oilfield service companies like Schlumberger and Baker Hughes now market their own sophisticated reservoir visualization software making it available to even small and midsize exploration and production companies.
PROCESSING AND DATA LIMITATIONS
Despite the enormous strides in computing technology, super-computers and geologists still rely on indirect measurements such as gravity anomalies, seismic patterns and cores brought back to the surface from test wells to estimate the amount of oil in place and potential recovery rates.
Petroleum geology remains an art, relying on good judgment and interpretation, as much as a science.
Data remains expensive to acquire and comparatively scarce.
Geologists have to estimate the thickness and extent of petroleum-generating source rocks as well as the reservoir formations from which the oil and gas is recovered.
The total organic content of the source rock, and its temperature-pressure history, all determine how much of the organic matter has been converted into oil or gas, and have to be estimated (or intelligently guessed).
The porosity of the reservoir, the connectedness of its pores, and its relative saturation with oil, gas and water all determine how much can ultimately be recovered, how many wells may be needed, and how far apart, as well as whether special techniques like horizontal drilling and hydraulic fracturing will be needed.
In most cases, initial estimates of the amount of recoverable oil and gas across comparatively large areas must be made based on surface seismic surveys and rock samples and well logs from just a small number of wells.
Even small changes to estimates of total organic content, porosity, connectedness and other factors can result in enormous changes in estimated resources and reserves. For this reason, resources and reserve estimates are subject to huge uncertainty.
CONSERVATISM IN RESOURCE ESTIMATES
It is normally good practice to adopt a conservative approach, employing the least realistically feasible estimates for the various parameters to produce a cautious estimate, and hope any surprises will be on the upside.
As the field is developed, more wells are drilled, and more data becomes available, it is possible to update the initial estimates. If good practice has been followed, reserves and resource estimates will normally be revised upwards.
This caution is one reason why the recent U.S. Geological Survey estimates for the ultimate technically recoverable resources from the Bakken shale deposits in North Dakota remain low compared with estimates published by some exploration companies and consultants involved in the play.
Technology is the other. Most of the oil originally contained in a reservoir is never recovered (though exactly what percentage remains is left behind is also subject to uncertainty because no one knows for certain how much was actually there in the first place). But as technology improves it has usually become possible to produce more of the oil that was originally left behind.
In many cases, it has been possible to identify pools of stranded oil, or fractured the reservoir rock and pump in water and chemicals to drive extra oil towards the wells.
Super-giant fields like Saudi Arabia's Ghawar and California's Wilmington have already produced many times more oil than geologists originally thought possible, and are still thought to be capable of producing billions of barrels more.
North Dakota's Bakken is no exception. Once thought to be essentially impossible to produce because there was not sufficient connectivity between the pores, it has already produced more than 500 million barrels and could ultimately yield at least eight times as much, according to USGS.
Given the conservative estimating methods employed by USGS it may eventually yield far more.
(John Kemp is a Reuters market analyst. The views expressed are his own)
(Editing by Keiron Henderson)