It's not a huge mystery why Los Angeles experiences earthquakes. The city sits near a boundary between two tectonic plates - they shift, we shake. But what about places that aren't along tectonic plate boundaries?

For example, seismicity on the North American plate occurs as far afield as southern Missouri, where earthquakes between 1811 and 1812 estimated at around magnitude 7 caused the Mississippi River to flow backward for hours.

Until now, the cause of that seismicity has remained unclear.

While earthquakes along tectonic plate boundaries are caused by motion between the plates, earthquakes away from fault lines are primarily driven by motion beneath the plates, according to a new study published by USC scientist Thorsten Becker in Nature on Aug. 27.

Just beneath the Earth's crust is a layer of hot, semi-liquid rock that is continually flowing - heating up and rising, then cooling and sinking. That convective process, interacting with the ever-changing motion of the plates at the surface, is driving intraplate seismicity and determining in large part where those earthquakes occur. To a lesser extent, the structure of the crust above also influences the location, according to their models.

"This will not be the last word on the origin of strange earthquakes. However, our work shows how imaging advances in seismology can be combined with mantle flow modeling to probe the links between seismicity and mantle convection," said Becker, lead author of the study and professor of Earth sciences at the USC Dornsife College of Letters, Arts and Sciences.

Becker and his team used an updated mantle flow model to study the motion beneath the mountain belt that cuts north to south through the interior of the Western United States.

The area is seismically active - the reason Yellowstone has geysers is that it sits atop a volcanic hotspot. Previously, scientists had suggested that the varying density of the plates was the main cause. (Imagine a mountain's own weight causing it to want to flow apart and thin out.)

Instead, the team found that the small-scale convective currents beneath the plate correlated with seismic events above in a predictable way. They also tried using the varying plate density or "gravitational potential energy variations" to predict seismic events and found a much poorer correlation.

"This study shows a direct link between deep convection and shallow earthquakes that we didn't anticipate, and it charts a course for improved seismic hazard mapping in plate interiors," said Tony Lowry, co-author of the paper and associate professor of geophysics and geodynamics at Utah State University.

Becker and Lowry collaborated with researchers from University Roma Tre in Italy, the University of New Mexico, Scripps Institution of Oceanography and the Massachusetts Institute of Technology. The full study can be found online here