The Sumatra earthquake was recorded by the first of five NSF-funded long-baseline tiltmeters in Washington State designed to monitor slow earthquakes associated with deep subduction zone activity. Each tiltmeter consists of two horizontal, half-km-long water pipes buried at right angles, and half filled with water. The sensors are referenced to 10-m-deep steel piles driven to refusal in glacial moraine.

The raw data are recorded by three sensors on each water pipe , the data from the two pipes are processed to form four tilt signals that provide a measure of fidelity of the measured tilt. Because the instrument is lightly damped, data of shorter period than its resonant frequency (25 minutes) are incoherent, however, signals with duration longer than 30 minutes are 90% coherent during the passage of surface waves from the earthquake. The processed data (body tides removed, but load tides and thermoelastic effects remain) reveal a 4.5 nanoradian transient tilt down to the NE with a half period of 50-70 minutes. Note that it is not possible to distinguish between tilt down to the NE, or horizontal acceleration to the NE. If the signal were entirely horizontal acceleration its amplitude would be ≈4.5 µGal.

The figure below shows the path of the tilt vector with the time indicated at approximately 1 minute increments. In this view the tilt data have been filtered with a 60 minute binomial filter using IGOR (™ Wavemetrics). The slow drift of the instrument in the 23 minutes before the arrival of the seismic waves is caused by local noise in the instrument (load tides and thermal effects on the 1-km-long tiltmeter have not been removed). The tilt (or acceleration) is approximately tangential to the Sumatra/Andaman trench, with a relatively minor radial signal.