Science News
from research organizations
Modeling the probability of methane hydrate deposits on the seafloor
Sandia scientists use machine learning
Date:
March 17, 2021
Source:
DOE/Sandia National Laboratories
Summary:
Researchers have developed a new system to model the likelihood of finding methane hydrate and methane gas that was tested in a region of seafloor off the coast of North Carolina.
Methane hydrate, an ice-like material made of compressed natural gas, burns when lit and can be found in some regions of the seafloor and in Arctic permafrost.
Thought to be the world’s largest source of natural gas, methane hydrate is a potential fuel source, and if it “melts” and methane gas is released into the atmosphere, it is a potent greenhouse gas. For these reasons, knowing where methane hydrate might be located, and how much is likely there, is important.
A team of researchers from Sandia National Laboratories and the U.S. Naval Research Laboratory have developed a new system to model the likelihood of finding methane hydrate and methane gas that was tested in a region of seafloor off the coast of North Carolina.
While methane hydrate deposits have been found in a variety of locations, there are significant unknowns in terms of how much methane hydrate exists on the seafloor and where. It is challenging to collect samples from the seafloor to find methane hydrate deposits. This is where Sandia’s computer modeling expertise comes in.
“This is the first time someone has been able to approach methane hydrate distribution in the same way we approach weather forecasting,” said Jennifer Frederick, a computational geoscientist and lead researcher on the project. “When you hear a weather forecast for a 60% chance of two inches of rain, you don’t necessarily expect exactly two inches. You understand that there is uncertainty in that forecast, but it is still quite useful. In most places on the seafloor we don’t have enough information to produce an exact answer, but we still need to know something about methane and its distribution. By using a probabilistic approach, similar to modern weather forecasting, we can provide useful answers.”
The new system combines Sandia’s longstanding expertise in probabilistic modeling with machine learning algorithms from the Naval Research Laboratory. The system was tested and refined by modeling the area around Blake Ridge, a hill on the seafloor 90 to 230 miles southeast of North Carolina’s Outer Banks with known deposits of methane hydrate and methane gas.
The team shared their model for Blake Ridge and compared it with previous empirical data in a paper published on March 14 in the scientific journal Geochemistry, Geophysics, Geosystems.
‘Forecasting’ methane by combining uncertainty modeling with machine learning
The Naval Research Laboratory’s Global Predictive Seafloor Model provides site-specific details on seafloor properties, such as temperature, overall carbon concentration and pressure. If data is missing for a certain region, the Naval Research Laboratory’s model uses advanced machine-learning algorithms to estimate the missing value based on information about another area that may be geographically distant but similar geologically.
The research team imported the data from the Naval Research Laboratory’s model into Sandia software that specializes in statistical sampling and analysis, called Dakota. Using Dakota, they determined the most likely value for influential seafloor properties, as well as the natural variation for the values. Then, in a statistical manner, they inserted a value from this expected range for each property into PFLOTRAN, another software maintained and developed at Sandia. PFLOTRAN models how chemicals react and materials move underground or under the seafloor. The team conducted thousands of methane production simulations of the Blake Ridge region. All the software involved in the system is open source and will be available for other oceanographic researchers to use.
“One of the biggest things we found is that there is almost no formation of methane hydrates shallower than 500 meters, which is to be expected given the temperature and pressure needed to form methane hydrate,” said William Eymold, a postdoctoral fellow at Sandia and primary author of the paper. Solid methane hydrate is known to form in low-temperature, high-pressure environments where molecules of methane are trapped within well-organized water molecules.
The team also found methane gas formed closer to shore. They were able to compare their model to methane hydrate values calculated by past studies and samples collected a few decades ago by the National Science Foundation’s Ocean Drilling Program, he said. For example, methane hydrate was detected in a seafloor sample collected from a hole drilled on Blake Ridge called Site 997.
“The fact that we predicted methane hydrate formation in similar amounts to past studies and observations really showed that the system appears to be working pretty well, and we will be able to apply it to other geographic locations that may have less data,” Eymold said.