Rapid Magnitude Scaling with GPS measurements
Brendan Crowell, University of Washington
March 24, 2014
Without reliable earthquake prediction, seismologists have developed methods to rapidly
detect earthquakes as soon as they happen and then predict the amount of shaking as well as the
arrival of shear (S) and surface waves at given points further away from the event. This system is
known as Earthquake Early Warning and it predicates itself on the fact that the compressional
(P) waves travel much faster than the S waves and are much less damaging. Currently, the
western United States is building a prototype Earthquake Early Warning system as a joint project
between the California Institute of Technology, University of California – Berkeley, and the
University of Washington (UW), funded through the private Moore Foundation. The system at
UW includes both a seismic and geodetic early warning system that work together to form the
most robust estimate of earthquake magnitude. One of the problems for seismic early warning is
the saturation of parameters for large earthquakes (M > 8). The UW system uses the ElarmS
module which estimates the peak displacement of the P wave in the first 4 seconds, known as Pd.
Pd decays linearly on a log-log plot as a function of distance, and at a given distance for larger
and larger earthquakes, Pd gets larger and larger. However, this simple relationship does not hold
for M > 8 earthquakes using just seismic data. One issue with seismic (acceleration) data is that
in order to integrate into displacement, one must perform a high-pass filter to prevent the
integration from being unstable due to sensor rotations and tilts. This filtering in effect limits the
low frequency part of the P wave from being measured and causes the Pd measurement to
saturate. In Crowell et al. [2013], they showed using seismogeodetic data that there is no
saturation of Pd and it can be used to determine the size of a large earthquake within seconds of
being detected with an uncertainty of 0.4 magnitude units. Furthermore, tracking the peak ground
displacement (PGD) at each station over the course of strong shaking can provide an even more
accurate estimate of earthquake magnitude, and is available within a minute or two of the start of
the event.
For the Alaska Shield exercise, we tested the Pd and PGD algorithms using the predicted
ground motions from the source model. The Pd initially overestimates the magnitude at 10.0 15
seconds after the start of the event using 4 stations. The Pd magnitude decreases constantly to 9.4
at 44 seconds and 9.1 at 60 seconds. The final Pd magnitude estimate using all stations is 8.75.
One thing to note is the Pd attenuation in the Alaska Shield exercise is roughly -2.5 whereas
what we have seen from previous earthquakes is around -1.7. Even given this discrepancy, the Pd
magnitude estimates from the GPS data indicate a very large earthquake in the magnitude 9
range. The initial 4 station PGD magnitude estimate is 9.1 at 85 seconds after the start of the
earthquake. The PGD magnitude estimate stays between 9.0 and 9.3 for the remainder of the
seismic wave propagation. This revised magnitude estimate is useful for places 200 km or further
from the epicenter.
Figure 1: Pd and PGD magnitude estimates from the Alaska Shield exercise and other large earthquakes from Crowell et al.
[2013]
Crowell, B. W., D. Melgar, Y. Bock, J. S. Haase, and J. Geng (2013), Earthquake magnitude
scaling using seismogeodetic data, Geophys. Res. Lett., 40, 6089–6094,
doi:10.1002/2013GL058391.