When scientist David Valentine and colleagues published results of a study in early 2011 reporting that bacterial blooms had consumed almost all the deepwater methane plumes after the 2010 Gulf of Mexico Deepwater Horizon oil spill, some were sceptical.
How, they asked the University of California at Santa Barbara (UCSB) geochemist, could almost all the gas emitted disappear?
In new results published this week in the journal Proceedings of the National Academy of Sciences (PNAS), Valentine; Igor Mezic, a mechanical engineer at UCSB; and coauthors report that they used an innovative computer model to demonstrate the respective roles of underwater topography, currents and bacteria in the Gulf of Mexico.
This confluence led to the disappearance of methane and other chemicals that spewed from the well after it erupted on 20 April 2010.
The National Science Foundation (NSF) funded the research.
"As scientists continue to peel apart the layers of the Deepwater Horizon microbial story," said Don Rice, director of NSF's chemical oceanography programme, "we're learning a great deal about how the ocean's biogeochemical system interacts with petroleum - every day, everywhere, twenty-four / seven. "
The results are an extension of a 2011 study, also funded by NSF, in which Valentine and other researchers explained the role of bacteria in consuming more than 200,000 metric tons of dissolved methane.
"It seemed that we were putting together a lot of pieces," Valentine said. "We would go out, take some samples, and study what was happening in those samples, both during and after the spill.
"There was a transition of the microorganisms and a transition of the biodegradation, and it became clear that we needed to incorporate the movement of the water."
The scientists believed that there was an important component of the physics of the water motion - of where the water went.
Valentine turned to Mezic, who had published results in 2011 forecasting where the oil slick would spread.
"Our work was on the side of: here's where the oil leaked and here's where it went," Mezic said. "We agreed that it would be beautiful if we could put a detailed hydrodynamic model together with a detailed bacterial model."
The resulting computer model has data on the chemical composition of hydrocarbons flowing into the Gulf of Mexico, and is seeded with 52 types of bacteria that consumed the hydrocarbons.
The physical characteristics were based on the US Navy's model of the gulf's ocean currents and on observations of water movements immediately after the spill and for several months after it ended.
The scientists then sought the help of Mezic's former colleagues - engineers at the University of Rijeka in Croatia.
"We needed somebody to build the software," Mezic said. "It was a big task, a mad rush, but they did it.
"The power behind this is a tour de force. A typical study of this kind would take a year, at least. We found a way that led us to answers in three or four months."
The model revealed that one of the key factors in the disappearance of the hydrocarbon plumes was the physical structure of the Gulf of Mexico.
"It's the geography of the gulf," Valentine said. "It's almost like a box canyon. As you go northward, it comes to a head.
"As a result, it's not a river down there; it's more of a bay. And the spill happened in a fairly enclosed area, particularly at the depths where hydrocarbons were dissolving."
When the hydrocarbons were released from the well, bacteria bloomed. In other locations outside the gulf, those blooms would be swept away by prevailing ocean currents.
But in the Gulf of Mexico, they swirled around as if they were in a washing machine, and often circled back over the leaking well, sometimes two or three times.
"What we see is that some of the water that already had been exposed to hydrocarbons at the well and had experienced bacterial blooms, then came back over the well," Valentine said.
"So these waters already had a bacterial community in them, then they got a second input of hydrocarbons."
As the water came back over, he explained, the organisms that had already bloomed and eaten their preferred hydrocarbons immediately attacked and went after certain compounds.
Then they were fed a new influx of hydrocarbons.
"When you have these developed communities coming back over the wellhead, they consume the hydrocarbons much more quickly," Valentine said, "and the bacterial composition and hydrocarbon composition behaves differently. It changes at a different rate than when the waters were first exposed."
The model allowed the scientists to test this hypothesis and to look at some of the factors that had been measured: oxygen deficits and microbial community structure.
"What we found was very good agreement between the two," Valentine said.
"We have about a 70-per cent success rate of hitting where those oxygen declines were. It means that not only is the physics model doing a good job of moving the water in the right place, but also that the biology and chemistry results are doing a good job, because you need those to get the oxygen declines. It's really a holistic view of what's going on."
There are valuable lessons to be learned from the study, the scientists believe.
"It tells us that the motion of the water is an important component in determining how rapidly different hydrocarbons are broken down," Valentine said. "It gives us concepts that we can now apply to other situations, if we understand the physics."
Mezic said that this should be a wake-up call for anyone thinking of drilling for oil.
"The general perspective is that we need to pay more attention to where the currents are flowing around the places where we have spills," he said.
"We don't have models for most of those. Why not mandate a model?
"This one worked - three-quarters of the predictions were correct. For almost everything, you can build a model. You build an airplane, you have a model. But you can drill without having a model. It's possible we can predict this. That's what a model is for."
The US Department of Energy and the US Office of Naval Research also supported the research.
In addition to Valentine and Mezic, co-authors of the paper are Senka Macesic, Nelida Crnjaric-Zic, and Stefan Ivic, of the University of Rijeka in Croatia; Patrick J. Hogan of the Naval Research Laboratory; Sophie Loire of the Department of Mechanical Engineering at UCSB; and Vladimir A. Fonoberov of Aimdyn, Inc. of Santa Barbara.