Tuesday, June 16, 2009

What's the signal, and what's the noise?

As anyone who listens to the (non-satellite) radio knows, signal-to-noise ratio is an important consideration when analyzing a data set. If the ratio is too low, all you get is static. But what if that static actually contained its own signal?

The idea of useful information being "hidden" in apparent noise is nothing new--after all, cosmic background radiation was once thought to be just noise (and for many applications it still is). But in the May 22 issue of Science, Peter Bromirski outlines a rather unusual case of noise-becoming-signal: seismological evidence for climate change.

Geologists use seismographs to record the movements of the crust. Most of the time, the crust doesn't move much, aside from a background "hum" that results from Earth's natural oscillations. That hum can actually show up on seismograms; it has a period of 1-8 minutes or so. Occasionally, though, an earthquake--geologists also sometimes call it a "seism"--causes the crust to move much more emphatically.

During an earthquake, the movements of the crust cause the seismograph needle (or the digital analogue) to move in a specific way. The speed, amplitude, and duration of that motion are related to the motion that occurred to cause the earthquake, as well as to the composition and structure of the materials the resulting seismic waves passed through to get to the seismograph. By studying seismographs from around the world, geologists can infer where and when the earthquake occurred, what caused it, and how the waves it produced traveled. The background hum is just noise, and it's generally ignored.

The thing about seismographs is that, for the most part, they're terrifically sensitive. It's not unusual for them to detect trains and traffic. And, as Bromirski points out, under the right conditions they can also detect ocean waves, particularly those produced by big storms.

During a large storm over the ocean, high winds blow over the ocean's surface. The wind transfers energy to the water, which is where the big ocean waves come from. That energy can generate "microseisms" in the ocean crust. (A microseism is exactly what you'd guess from the name: a very low-amplitude vibration in the crust.) The vibrations produced by wave energy travel through Earth, just like those from an earthquake, and they can be detected on seismographs, too. Therefore, hypothetically, one could use seismogram records to determine the average storminess of the oceans over time.

The use of seismograms to study storminess has a few advantages over more common methods. For one thing, there are accurate seismograms that go back to the early 20th century--as far back as 1930, in some areas. These seismograms were all collected using pretty much the same technology and have similar precision, so they're readily comparable. This is unusual in climate science; many of the techniques commonly used to study recent climate change are fairly...well...recent, so the records don't go very far back or, if they do, they're much less precise.

Another advantage to using seismograms is that the global seismograph network (which has become more and more widespread over time) allows for comparisons between signals from different areas. This can allow scientists to infer the approximate paths and durations of storms in a region. In some cases, microseisms can give information about wave frequency and duration along specific coastline regions, data that may be hard to obtain otherwise.

Some studies using these long-term seismic records do suggest that Earth is becoming stormier: the ambient noise on the seismograms has increased over time.

Some researchers are also studying ways to use storm-driven microseisms to study more than storms. An important use of earthquake seismogram data is the study of Earth's interior. It's similar to the use of ultrasound to see inside the body: just as the path of a sound wave through your body depends on the density and structure of the organs below the skin, so the path of a seismic wave depends on the composition, temperature, and structure of the rock within Earth. Typically, seismologists use earthquake-generated seismic waves to study Earth's interior, because they're very high amplitude and generate strong signals. However, earthquakes are relatively rare and unpredictable. "Background" microseisms produced by storms and wave activity may provide a more long-term and consistent energy source for the study of Earth's interior.

Bromirski, Peter D., 2009. "Earth Vibrations." Science 324: 1026-1027. doi: 10.1126/science.1171839.

No comments: