This evening Silicon Valley did a little dance with a 5.6 magnitude earthquake. While there wasn’t much damage here in the land of quake-resistant building codes, it was, as
ValleyWag notes, this was “the largest quake to hit the Bay Area since Loma Prieta in October ’89.”

After the house stopped shaking, and then swaying, the first thing we did was to get on the USGS web site and try and find out what we could about the quake. From the main information page, we quickly found the “Did you feel it?” questionnaire for the event, which we filled out. The DYFI program collates the data from responses like ours and uses them to create a map of the earthquake intensity by geographic region, as well as some additional data. The map shown above represents the intensity felt by 60,000 respondents in the first few hours after the quake. It is by far the finest example that I have ever seen of “citizen science” in action– apparently objective data collected over a wide area in a short period of time with (in many cases) good statistics.

Of course, it could be better. Much better. Like other Mac laptops (and many others), my computer contains a “sudden motion sensor” to protect the hard drive in case the computer is dropped. Fundamentally this is just an accelerometer and can be used for any arbitrary purpose besides just waiting for the computer to fall. A program called SeisMac has already been developed that can turn a Mac laptop into a makeshift seismometer. (SeisMac is freeware and based is on open source libraries for the sensor.)

The logical leap that has to happen is to turn this into a distributed system. Here is how it could work. You go to a website and download the application– let’s call it “Quake@Home.” When you run the installer, it checks against a list known computers that have an appropriate sensor, and walks through an appropriate calibration routine. You enter your address, which is then geocoded. If you commute with your computer, you can enter multiple addresses, which it can be set to identify by changes in the network environment. Once the Quake@Home client is installed on your computer, you never see it again (except perhaps as part of a screen saver), while it runs quietly in the background. It keeps a continuous record of the last six hours of seismic data. When a quake happens, as detected by the traditional monitoring system, the Quake@Home system will gather data from participating laptop users. If requested to do so, your client sends a highly compressed version of its time-stamped record back to the central server. Naturally, it would be helpful if the clock on your computer is correct. In order to facilitate this, it may be easiest if clients periodically (once per hour) poll the central server via HTTP to see if their data is requested.

A system like this could yield very interesting and sensitive seismographic data, even in many cases where people do not notice the quake. Correlation between the different waveforms returned would make it possible to discriminate between local disturbances and the quake event. With automated data returns, the participation ratio could be very high for any given event. With gradual improvement, it could provide a useful research tool on its own. (Is it a substitute for real research grade seismometers? Not by a long shot. Accelerometers– more or less– measure the sorts of accelerations that humans can also feel, while research-grade seismometers measure actual displacements, down to much lower frequencies.)

A separate, and arguably more difficult, proposition would be to actually use the data as part of a detection and early warning system. If your computer detected what might be the effect of a weak quake (or of being set down on the table), it would need to compare this to other local seismographic records, and quickly, to be sure. Rather than have each laptop talk to the central server every time that it is moved, it would be much less burdensome to use a direct peer-to-peer system for this stage. Naturally, we are not the first to think of it. An earlier proposal for tsunami warnings is based on the use of hard drive feedback signals to generate the motion data. However, reading out the accelerometer is a straightforward method that is easy to calibrate and can operate even when the hard drive is powered down.