Preventing Human-Induced Earthquakes

The most common source of such fluid injections is from the oil and gas industry’s disposal of wastewater that is brought up along with oil. Field operators dispose of this water through injection wells that continuously pump the water back into the ground at high pressures.

“There’s a lot of water produced with the oil, and that water is injected into the ground, which has caused a large number of quakes,” Hager notes. “So, for a while, oil-producing regions in Oklahoma had more magnitude 3 quakes than California, because of all this wastewater that was being injected.”

In recent years, a similar problem arose in southern Italy, where injection wells on oil fields operated by Eni triggered microseisms in an area where large naturally occurring earthquakes had previously occurred. The company, looking for ways to address the problem, sought consulation from Hager and Juanes, both leading experts in seismicity and subsurface flows.

“This was an opportunity for us to get access to high-quality seismic data about the subsurface, and learn how to do these injections safely,” Juanes says.

Seismic Blueprint
The team made use of detailed information, accumulated by the oil company over years of operation in the Val D’Agri oil field, a region of southern Italy that lies in a tectonically active basin. The data included information about the region’s earthquake record, dating back to the 1600s, as well as the structure of rocks and faults, and the state of the subsurface corresponding to the various injection rates of each well.

The researchers integrated these data into a coupled subsurface flow and geomechanical model, which predicts how the stresses and strains of underground structures evolve as the volume of pore fluid, such as from the injection of water, changes. They connected this model to an earthquake mechanics model in order to translate the changes in underground stress and fluid pressure into a likelihood of triggering earthquakes. They then quantified the rate of earthquakes associated with various rates of water injection, and identified scenarios that were unlikely to trigger large quakes.

When they ran the models using data from 1993 through 2016, the predictions of seismic activity matched with the earthquake record during this period, validating their approach. They then ran the models forward in time, through the year 2025, to predict the region’s seismic response to three different injection rates: 2,000, 2,500, and 3,000 cubic meters per day. The simulations showed that large earthquakes could be avoided if operators kept injection rates at 2,000 cubic meters per day — a flow rate comparable to a small public fire hydrant.

Eni field operators implemented the team’s recommended rate at the oil field’s single water injection well over a 30-month period between January 2017 and June 2019. In this time, the team observed only a few tiny seismic events, which coincided with brief periods when operators went above the recommended injection rate.

“The seismicity in the region has been very low in these two-and-a-half years, with around four quakes of 0.5 magnitude, as opposed to hundreds of quakes, of up to 3 magnitude, that were happening between 2006 and 2016,” Hager says. 

The results demonstrate that operators can successfully manage earthquakes by adjusting injection rates, based on the underlying geology. Juanes says the team’s modeling approach may help to prevent earthquakes related to other processes, such as the building of water reservoirs and the sequestration of carbon dioxide — as long as there is detailed information about a region’s subsurface.

“A lot of effort needs to go into understanding the geologic setting,” says Juanes, who notes that, if carbon sequestration were carried out on depleted oil fields, “such reservoirs could have this type of history, seismic information, and geologic interpretation that you could use to build similar models for carbon sequestration. We show it’s at least possible to manage seismicity in an operational setting. And we offer a blueprint for how to do it.”

This research was supported, in part, by Eni.

Jennifer Chu is a writer for the MIT News Office. Reprinted with permission of MIT News.