HR: 0800h
AN: S31A-0268    [Abstracts]
TI: Absolute Time Measurements From Seismic Noise: Monitoring Temporal Velocity Changes and Correction of Station Clock Drifts.
AU: * Campillo, M
EM: campillo@obs.ujf-grenoble.fr
AF: LGIT - Universite Jospeh Fourier - CNRS, 1381, rue de la Pisicne BP 53 , Grenoble, 38041 France
AU: Stehly, L
EM: lstehly@obs.ujf-grenoble.fr
AF: LGIT - Universite Jospeh Fourier - CNRS, 1381, rue de la Pisicne BP 53 , Grenoble, 38041 France
AU: Shapiro, N
EM: nshapiro@anquetil.colorado.edu
AF: Laboratoire de Sismologie - IPGP, 4,place Jussieu, Paris, 75005 France
AB: Recent works (Shapiro and Campillo, 2004) have shown that the Rayleigh wave part of the Green function between two stations A and B can be reconstructed by cross correlating the vertical component of seismic ambient noise recorded at these two stations. The cross-correlation (CC) contains waves travelling from A to B and from B to A for positive and negative times, respectively. In principle, because of time symmetry of the wave equation, the travel time of Rayleigh wave should be the same for the positive and negative correlation times and should be independent of the time window correlated, as far as the medium is not changing. However this is not always the case for three reasons: 1) The medium changes with time (for example, because of change of stress or of fluid content due to a strong earthquake). In this case both part of the CC should be affected symmetrically: we should either measure faster or slower velocity in both part of the CC. 2) One of the clocks of the two stations is not stable. This would produce a drift in the correlations: The causal part should accelerate while the anti-causal part should decelerate or vice-versa. 3) The reconstruction of the Green function is not perfect because of a insufficiently homogeneous distribution of sources of noise. Because the causal and anti-causal part are not sensitive to the same sources of noise, this would affect separately the causal and the anti-causal part of the CC. Changes in the incomplete distribution of noise sources yield to changes in the reconstructed signal although the apparent velocity is always larger than the true one. We studied the evolution of the symmetry in time of the cross-correlation of seismic ambient noise at three stations located in southern California. We computed the one bit cross-correlation month by month for 13 years of continuous vertical records of seismic noise. By studying the evolution of the arrival time of the Rayleigh wave reconstructed in the positive and negative time of the CC with the time window cross-correlated, we are able to identify and to evaluate the three effects discussed above. Our results shows that the largest fluctuations of the arrival time of Rayleigh waves can be of the order of 2 seconds for stations separated by 190 Km. Most of the delay is unambiguously related to uncorrected clock drift at one of the stations. A clear seasonal effect can be observed that we interpret as the mark of the incomplete reconstruction leading to an overestimation of the velocity. Because of the dispersion of surface waves, slight changes in the spectral content produce changes in the group velocity measurement. Correcting for these effects, we found residual fluctuations of the order of less than 1/20 of the wave period. This indicates that the use of the correlation method can lead to precise measurements on long time series.
DE: 7255 Surface waves and free oscillations
DE: 7299 General or miscellaneous
SC: Seismology [S]
MN: Fall Meeting 2005