The Berkeley Seismological Laboratory has accumulated a catalog of regional moment tensor (MT) solutions which were determined using an automated system as part of REDI project over several years [Gee et al., 1996]. The automated system uses two independent methods of MT determination, a complete waveform (CW) approach and a method using surface waves (SW) in the frequency band typically from 0.02 to 0.1 Hz [Dreger and Helmberger, 1993; Romanowicz at al., 1993; Pasyanos, 1996; Pasyanos et al., 1996]. Having two independent methods of determining MTs is an advantage of the BDSN system. In general the solutions determined by these methods are consistent with each other for events with M4.5 except in the Mendocino and Mammoth Lake regions, and solutions by the CW method tend to show better variance reduction than those by the SW method. For smaller events (M < 4.5) effects of emphasized structural noise is sometimes comparable to source signal in the pass-band between 0.02 and 0.1 Hz where MT solutions are reasonably stable [Mégnin et al., 2000].
At present the CW method uses either one of two 1-D models,
So.Cal [Dreger and Helmberger, 1993] or Gil7 [Dreger and Romanowicz, 1994]
to calculate Green's functions on the basis that the lateral variations
concentrate in the upper several kilometers of the crust for which
the body waves are not very sensitive [Dreger and Helmberger, 1991].
The selection of model, So.Cal or Gil7, is determined by
the earthquake source location (see the regional boundary of the two models
in Fig. 19.1). The SW method uses six to eight different 1-D
for the entire northern and central California region, and is strongly
affected by the unmodeled lateral heterogeneity in the uppermost crustal
layers (shallower than several km).
The improvement of structural models is critical for the
Mendocino/Gorda plate to northern California region where the oceanic
structure transitions to continental structure and no optimal velocity models
have been established to our knowledge.
To identify areas where existing velocity models may not represent the actual propagation paths, we examined an assembled catalog of MT solutions by evaluating consistency between SW and CW solutions for the period from 05/29/1997 to 12/31/1999, that includes 236 events (M3.5). There are 120 events with M4, out of which 51 events show a large value of (0.5) in the automated solutions. Here is a measure to assess the difference between SW and CW solutions [Pasyanos, 1996], and if 0.5, the two solutions are considered to be discrepant. The percentage of events with discrepant solutions, i.e., is 42.5% in the entire region. However, the rate of events with is higher in the Mendocino (west of 124oW) and Mammoth Lake (east of 120oW and south of 38.5oN) regions than 42.5%.
There are 28 events (M4) in Mendocino and 14 of them show . The occurrence rate with is 50%. However, after revision by analysts was reduced to under 0.5 for six events, i.e., the occurrence rate with is 29% (see the locations in Fig. 19.1). In Mammoth 24 events out of 55 show yielding the discrepant event occurrence of 44% or 19 events (35%) after revision. If the events in Mendocino and Mammoth are excluded, 11 events (30%) (or 6 after revision, 16%) out of 37 show 0.5 in the rest of the entire region. All of these events with 0.5 fall in the magnitude range below 4.5. After revision the average occurrence rate with is 27.5% in the entire region, 29% in Mendocino, 35% in Mammoth, and 16% in the rest of the region. While a large value of does not necessarily mean poorly constrained structural models, we investigated path effects in the regions with a higher value.
Dreger et al.  studied the source processes and near structural effects for events in the Mammoth Lake region, which have shown substantial compensated linear vector dipole (CLVD) and concluded the primary cause of the limited variance reduction or the discrepancy between the CW and SW solutions is the source itself in this region. On the other hand the coastal region of northern California from 39oN to 42oN represents a region in which the active tectonic regime changes from subduction to transform motion at the Mendocino Triple Junction (MTJ) [Oppenheimer et al., 1993; Trehu et al, 1995]. The layered model Gil7 currently used to calculate Green's functions for this region does not necessarily account for the transitional structure from the off-shore region to the continent. The large difference between SW and CW solutions ( 0.5) could be due to the unrepresented velocity structures.
There are six events for which Harvard CMT solutions are also available in this region. Figure 19.1 compares the CW and CMT solutions. The two solutions are more or less similar to each other except for the July 24, 1996 event which is located 200 km to the northwest of MTJ, and the rays to stations ORV, WDC, and YBH (used for CW determination) sample the complex tectonic structure of the Gorda plate and north America plate underlain by the Pacific plate.
Given the information on the MT solutions, we are calibrating propagation paths in the Mendocino region using high quality broadband waveform data recorded by BDSN. The procedure of this analysis involves forward modeling of waveforms in which Green's functions are calculated using FK code, and source time-function (duration and complexity) and source depth are evaluated. Figure 19.2 compares velocity models, So.Cal, Gil7, and two models conventionally used for routine analysis for off-shore Oregon region, one with a typical thin oceanic crust (JOCH.W), and the other one (JOCH.E) with a thicker crust than JOCH.W and Gil7, but also higher crustal velocity than Gil7 at shallower depths (<17 km) (provided by J. Braunmiller).
A velocity model used for Cape Mendocino region (east of 125.75o) by USGS (USGS.men) and a global average model iasp91 are also shown for comparison. USGS.men has a similar Moho depth to but slower P wave velocity at shallower depths (<20 km) than Gil7. We tested several different 1-D models including modified versions of the models in Figure 19.2 (typically between JOCH.E and JOCH.W) for each event data set of waveforms.
While an optimal 1-D model can be determined for a specific path, it is not possible to determine a 1-D model that represents the entire Mendocino region. With the relatively spotty event distribution, it is not practical to identify the boundary of an area that can be represented by an optimal model using a 1-D modeling approach, either. However, in the frequency band between 0.02 and 0.05 Hz a new model (denoted as New1 in Fig. 19.2) that is characterized by a similar Moho depth to, and higher P-wave velocity above Moho than Gil7, represents the transitional structure in Mendocino fairly well. The velocity structure in the upper several km of the crust, that strongly affects surface wave generation, vary from path to path reflecting the tectonic background in this region. Thus, at present we are using a 3-D finite difference code by Larsen to model the complex structure from the offshore west of MTJ to the coastal region, that traverses the primarily oceanic structure to the coast where it transitions to continental structure.
L. Gee provided us with the list of events with CW and SW MT solutions, J. Braunmiller with two velocity models for off-shore Oregon region, D. Oppenheimer with the USGS velocity models and A. Tréhu with her reprints. We thank C. Mégnin for some technical help during the course of this project, and S. Larsen for the use of his 3-D finite difference code.
Dreger, D. S., and D. V. Helmberger, Complex faulting deduced from broadband modeling of the 28 February 190 Upland earthquake (Ml =5.2), Bull. Seism. Soc. Am., 81, 1129-1144, 1991.
Dreger, D. S., and D. V. Helmberger, Determination of source parameters at regional distances with single station or sparse network data, J. Geophys. Res., 98, 8107-8125, 1993.
Dreger, D. S., and B. Romanowicz, Source characteristics of events in the San Francisco Bay region, U. S. Geol. Surv. Open-File Rept. 94-176, 301-309, 1994.
Dreger, D. S., H. Tkalcic, and M. Johnson, Dilatational processes accompanying earthquakes in the Long Valley Caldera, Science, 288, 122-125, 2000.
Gee, L., D. Dreger, D. Neuhauser, R. Uhrhammer, and B. Romanowicz, The REDI program, Bull. Seism. Soc. Am., 86, 936-945, 1996.
Mégnin, C., F. Tajima, D. Dreger, and B. Romanowicz, Near real-time determination of moment tensor and centroid location for northern California earthquakes using broadband waveforms from a sparse seismic network, submitted to J. Geophys. Res., 2000.
Oppenheimer, D., G. Beroza, G. Carver, L. Dengler, J. Eaton, L. Gee, F. Gonzalez, A. Jayko, W. H. Li, M. Lisowski, M. Magee, G. Marshall, M. Murray, R. McPherson, B. Romanowicz, K. Satake, R. Simpson, P. Sommerville, R. Stein, D. Valentine, The Cape Mendocino, California, earthquakes of April 1992: subduction at the triple junction, Science, 261, 433-438, 1993.
Pasyanos, M. E., D. S. Dreger, and B. A. Romanowicz, Toward real-time estimation of regional moment tensors, Bull. Seism. Soc. Am.,86, 1255-1269, 1996.
Romanowicz, B. A., D. S. Dreger, M. E. Pasyanos, and R. Uhrhammer, Monitoring of strain release in central and northern California using broadband data, Geophys. Res. Lett., 20, 643-1646, 1993.
Tréhu, A. M., and Mendocino Working Group, Pulling the rug out from under California: seismic images of the Mendocino Triple Junction region, EOS Trans. A.G.U., 76, pages, 369, 380-381, 1995.