Attempting to Constrain Upper Bounds to Preseismic Velocity Changes

We investigate the existence of preseismic velocity changes before the 24 September 2004 Parkfield and 20 January 2007 Odaesan, Korea, earthquakes using two different techniques. For Parkfield we test the limits of ambient noise monitoring. For Korea we use a sequence of nine repeating events and the doublet method. Brenguier et al. (2008) measured a 0.1% coseismic decrease in velocity for the 2004 Parkfield event using the borehole array and ambient noise monitoring with 30 day stacks. For temporal monitoring it is desirable to have finer time resolution. We perform a similar analysis using the same array with 30 day stacks down to 1 day stacks. All three components are considered to improve the signal-to-noise ratio giving up to nine pair combinations as compared to the vertical alone. We find that the coseismic change is still measureable down to 1 day and remains at -0.1%. It does not increase in magnitude like expected if averaging over longer time would have damped out the signal. These results are in agreement with a paper that just came out by Hadziioznnou et al. (2011). For constraining an upper bound on a preseismic signal we use the median absolute value of the days before the mainshock assuming there is no change in the three months prior. This estimate includes any effects due to changes in noise source distribution. So far 72 pairs have been considered. There is a tradeoff with upper bound constrained and days in the stack as expected. To measure a preseismic change that is greater than 0.05% (half the coseismic) it takes at least three days in the stack. We find that adding additional components doesn't reduce the scatter in the velocity change measurement as much as additional stations presumably due to correlated errors. For the Odaesan earthquake we analyze a very unique foreshock sequence in that it was nine repeating events that occurred within three days and 1 km of the hypocenter of the mainshock offering a rare opportunity to measure any possible velocity changes near the source at depth. The repeating event hypocenters were within 100 m of each other. Measured velocity changes ranged from -0.2% to -1% depending on station and component and phase analyzed. A correlation of the fluctuations of velocity changes with depth and along strike distance was observed implying that biases were introduced by the slight position differences of the events, albeit small. An attempt was made to correct the delay measurements for these position differences. After correction the measured velocity changes have maximum range (upper bound of scatter) from 0.006% to 0.1%. This is over a factor of ten improvement in measurement precision using the doublet method and is able to extend its usefulness to less than perfect repeats.