Michele Masterson

Voice Recognition Captures Yellowstone Geyser Bubbles

Voice recognition is not just for humans. Scientists have found a way to use the technology to measure seismic activity in geyser basins at Yellowstone National Park.

Norris Geyser Basin

U.S. Geological Survey in California scientist Phil Dawson and colleagues recently used voice recognition software and found that background seismic activity in geyser basins can be closely linked to daily cycles of heating and cooling.Dawson published a paper about his findings in the January 2012 Geophysical Research Letters.

The authors looked at data from 2003, when colleagues from the Yellowstone Volcano Observatory placed a seismic and other monitoring stations in the Norris Geyser Basin to try and understand the cause for a “disturbance” of increased surface temperatures, according to the U.S. Geological Survey. There were few measurable earthquakes originating in the geyser basin during that time period. However, a closer look at the constant background seismic noise revealed the presence of millions of tiny events. These events represent the constant bubbling, boiling and crackling of the geyser basin.

According to the U.S. Geological Survey, earthquakes in volcanic and geothermal areas often happen in the same place in the same way, so what’s recorded at the seismic sensors often produces the same data-output patterns. The continuous seismic record from a station on a volcano can be considered a song and the individual events as the repetition of specific words in the song.

Examples of: one minute of continuous seismicity for four of the seismic stations (the song).

“The cool thing about volcanoes is very often there are repeated signals, indicating that the source process is the same,” Dawson told The Casper Star Tribune. “What that means is, in the context of speech recognition, it’s very easy to see them as a sentence or a song.”


Dawson said that for the last five or six years, the data showed undecipherable “seismic noise.” Dawson tried using voice recognition software supplied by his colleague Carmen Benítez, a signal-processing expert at the University of Granada in Spain, to identify the small seismic events. The software was able to detect specific types of small seismic events embedded in the background noise. These events were interpreted to be due to the violent collapse of small steam bubbles during their rise toward the surface.

Dawson then trained the software to recognize these events and count their abundance, and tracked a persistent tremor clearly linked to movement of water in shallow sub-surface cracks. He found that the region of the disturbance, in Norris’ Back Basin had more active and consistent noise than surrounding areas at Norris.

“It’s very exciting science,” Hank Heasler, geologist for Yellowstone National Park, told the Tribune. “Who would have thought that voice recognition software could be applied to this kind of problem?”