A Search for Temporal Variations in Station Terms inSouthern California from 1984 to 2002by Guoqing Lin,*Peter M. Shearer, and Egill HaukssonAbstract We use relative arrival times and locations for similar earthquake pairsthat are found using a cross-correlation method to analyze the time dependence of Pand S station terms in southern California from 1984 to 2002. We examine 494 similarevent clusters recorded by Southern California Seismic Network (SCSN) stationsand compute absolute arrival-time variations from the differential arrival-time resi-duals obtained following event relocation. We compute station terms from the robustmeans of the absolute arrival-time residuals from all events recorded by each stationat 3-month intervals. We observe nine stations with abrupt offsets in timing of 20–70 msec, which are likely caused by equipment changes during our study period.Taking these changes into account could improve the relative location accuracyfor some of the event clusters. For other stations, we generally do not see systematictemporal variations greater than about 10 msec. Analysis of residuals along individualray paths does not reveal any clear localized regions of apparent velocity changes atdepth. These results limit large-scale, long-lasting temporal variations in P and S ve-locities across southern California during this time period to less than about 0:2%.However, there is an increased fraction of individual travel-time residuals exceeding20 msec immediately following major earthquakes from source regions near the main-shock rupture.IntroductionRecently, waveform cross correlation has become anincreasingly important tool for improving relative earth-quake locations, characterizing event similarity, and study-ing earthquake source properties (e.g., Nakamura, 1978;Poupinet et al., 1984; Got et al., 1994; Dodge et al.,1995; Nadeau et al., 1995; Gillard et al., 1996; Shearer,1997, 1998; Rubin et al., 1999; Waldhauser et al., 1999;Astiz and Shearer, 2000; Astiz et al., 2000; Shearer,2002; Shearer et al., 2003; Rowe et al., 2004; Haukssonand Shearer, 2005; Schaff and Waldhauser, 2005; Sheareret al., 2005; Lin and Shearer, 2007; Lin et al., 2007). Relativelocations can be greatly improved compared to standardmethods because the differential times for events within sim-ilar event clust ers are much more precise than individualarrival-time picks, and differential location methods canlargely remove the biasing effects of a large-scale3D ve-locity structure between the source and receiver. Most ofthese methods implicitly assume stability of seismic veloci-ties and the relative timing among the stations within therecording network. However, it is also possible to use wave-form cross-correlation data to search for time dependence inobserved travel times, and a number of studies have used thisapproach to identify (or place limits on) small crustal veloc-ity changes or changes in scattering behavior associated withlarge earthquakes (e.g., Poupinet et al., 1984; Ellsworth et al.,1987, 1992; Haase et al., 1995; Dodge and Beroza, 1997;Baisch and Bokelmann, 2001; Nakamura et al., 2002; Rubin-stein and Beroza, 2004; Peng and Ben-Zion, 2006; Sawazakiet al., 2006; Li et al., 2007). In addition, Rubin (2002) dem-onstrated that differential times provide precise calibrationinformation for detecting station and network timing dis-crepancies. He observed abrupt changes in station timingof up to 20 msec, which are associated with changes in sta-tion electronics, and found that correcting for these effectscan reduce the scatter within clusters of repeating earth-quakes and the apparent seismogenic thickness of faults innorthern California.Here, we perform a comprehensive search for possibletemporal variations in station timing or seismic velocitiesacross southern California from 1984 to 2002, using relativearrival times and locations from similar event clusters identi-fied in our recent comprehensive relocation of southern Cal-ifornia seismicity using waveform cross correlation (Sheareret al., 2005; Lin et al., 2007). We observe nine stationswith abrupt offsets in timing of 20–70 msec, which are likely*Present address: Department of Geology and Geophysics, University ofWisconsin-Madison, Madison, Wisconsin 53706.2118Bulletin of the Seismological Society of America, Vol. 98, No. 5, pp. 2118–2132, October 2008, doi: 10.1785/0120070243caused by equipment changes during our study period. Forother stations, we generally do not s ee systematic temporalvariations greater than about 10 msec, although there is anincreased fraction of individual travel-time residuals exceed-ing 20 msec immediately following major earthquakes fromsource regions near the mainshock rupture.MethodOur differential times and earthquake locations arefrom the waveform cross-correlation projects of Sheareret al. (2005) and Lin et al. (2007), in which events are re-located separately within individual similar event clusters.The similar event clusters are obtained by first searching forcorrelated event pairs that have eight or more P or S mea-surements with correlation coefficients above 0.65 for sta-tions within 80 km of the events. A cluster analysis methodis then applied to the set of correlated event pairs to identifyrobust distinct sets of linked events. These similar event clus-ters are typically between 1 and 2 km in size and containfrom 40 to 1000 events. The relocations are performe d as-suming that seismic velocities and station timing are constantduring the entire 1984–2002 period. Our goal here is to lookfor possible time dependence in the residuals from the relo-cations that would suggest some changes in velocity and/ortiming have occurred. However, the problem is complicatedby the fact that the precise times provided by waveform crosscorrelation are differential times between a pair of events oc-curring at different times. Thus, our first step is to estimateabsolute time residuals for each event from the set of differ-ential time residuals.For a similar event cluster with n events recorded by agiven station, we may express the differential arrival-time re-siduals, drij, for a pair of events, i and j, after relocation asdrij dToij dTpij dt0ij (1a)Toi Toj Tpi Tpj t0i t0j (1b)Toi Tpi t0i Toj Tpj t0j (1c) ri rj; (1d)where dToijis the observed differential arrival time betweenevent i and event j measured using waveform cross correla-tion; dTpijis the