DCL-11-005, Report on the Analysis of the Shoreline Fault Zone, Appendix C, Seismicity Reviews Through Appendix D, Potential Field Data - Magnetic Surveys

Report on the Analysis of the Shoreline Fault Zone, Appendix C, Seismicity Reviews Through Appendix D, Potential Field Data - Magnetic Surveys

ML110140428
Person / Time
Site:	Diablo Canyon
Issue date:	01/07/2011
From:	Pacific Gas & Electric Co
To:	Office of Nuclear Reactor Regulation
References
DCL-11-005
Download: ML110140428 (91)

Text

0 Appendix C Seismicity Reviews C-i: Report: Central California Offshore Earthquake Assessment Phases 1-3, Dr. Clifford Thurber C-2: Central California Offshore Earthquake Location Assessment, Dr. Felix Waldhauser Shoreline Fault Zone Report, Appendix C Seismicity Reviews

Appendix C-I Report: Central California Offshore Earthquake Assessment Phases 1-3 Dr. Clifford Thurber Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-1 of 35

Phases 1 and 2 Report: Central California Offshore Earthquake Assessment PG&E Contract No. 2500174779 Clifford Thurber March 18, 2009 I report on my analysis of the locations of events offshore California, 35.1 to 35.3' N, with focus on an apparent NW-SE lineation identified from the tomographic inversion and relocation results of Dr. Jeanne Hardebeck, USGS (the "target events"). The tasks for Phases 1 and 2 are as follows:

Phase 1 - assessment of Jeanne Hardebeck's location/tomography results a) Evaluate the parameters and other aspects used in the completed tomographic inversion work Set up and run tomoDD inversion using original parameters p Assess sensitivity of results to parameter changes: relative weighting of data types, inversion sequence, event linkage b) Assess suitability of tomographic velocity model grid c) Evaluate potential for S minus P time moveout analysis to test reality of offshore lineations d) Provide report including my preliminary professional assessment of the off-shore lineation Phase 2 - assessment of offshore earthquake location uncertainties a) Carry out relocation analysis using two different, existing regional 3D velocity models b) Repeat tomographic inversions using these two different models as starting models c) Delete-one tests of sensitivity of locations to observing stations d) Evaluation of realistic location uncertainty estimates e) Provide report including my detailed professional assessment of the robustness of the location results for the off-shore events My first step in Phase 1.was an evaluation of the parameters used in and other aspects of Hardebeck's tomographic inversion work using tomoDD (Zhang and Thurber, 2003). In my opinion, the inversion grid adopted by Hardebeck is too finely spaced, with 2 km grid intervals in Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-2 of 35

o 2 all three Cartesian directions throughout most of the model volume. The available data cannot adequately constrain such a spatially fine model. Hardebeck appears to have compensated adequately for the excessively fine grid by using relatively high model perturbation smoothing weights. I was unable to replicate her tomographic inversion due to her dense grid parameterization (nearly 25,000 nodes) combined with undetermined compiler differences. The results I obtained using a decimated version of Hardebeck's model, however, show moderately small location differences for the target events, but unstable locations for a number of other events (Figure 1a, b, and c). The location results are insensitive to the details of the tomoDD parameter choices, as tests on a broad range of key parameter values for the decimated model do not result in significant location differences for the target events. The parameters tested are the model perturbation smoothing weight (20, 200, 2000) and linking distance (60 to 1 for cross-correlation data, 60 to 2 for catalogrdata). A smoothing weight of 2000 yields locations that are indistinguishable from those for 200 (the reference value, used by Hardebeck). For a smoothing weight of 20, some target events shift on the order of a kilometer, but most move less than 500

m. Changes in linking distance make at most minorchanges to the locations, except for a small number of unstable events.

My summary evaluation of Hardebeck's tomography work and the associated'location results is that a more conservative (i.e. coarser) velocity model grid would have been more appropriate, but with the heavy smoothing applied, stable tomography results were obtained. My tests using a decimated version of Hardebeck's starting model show very little sensitivity to a broad range of parameter values. The use of different models and different data subsets results in substantially larger hypocenter changes for the target events, however, as discussed below.

To complete the phase 1 tasks, I made a preliminary evaluation of the potential for using an S minus P time moveout analysis to test the reality of the offshore lineation. Unfortunately, the number of earthquakes with a significant number of S picks is quite small. For the earthquakes associated with the lineation, there is only one event with 7 S picks, and there are 7 events with 5 and 19 events with 4 S picks. A simple moveout analysis also requires events at comparable depths. I have not been able to identify a suitable subset or even a pair of events within the lineation that could be used for a meaningful moveout analysis. There is a potential for determining regional (average) Vp/Vs values from the use of Wadati diagrams on the available S picks. I determined single-event Vp/Vs' values ranging between 1.72 and 1.79.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-3 of 35

0

-121 120.75" km 0 5 10 Figure la NW SE NW SE

-20 -15 -10 -5 0 5 10 15 20 -20 10 -5 0 5 10 15 20 0 0 0 0 5 -5

-10 -10 -10 -10

-15 15 -15 -15

-20 .20 -20 -20

-20 a e15 10 -5 0 5 10 15 20 - ,2 0 15 .10 -5 0 5 10 15 20 Figure lb Figure Ilc Figure la compares the epicenters from Hardebeck's tomographic inversion (red stars) to epicenters derived from a tomographic inversion using a decimated version of Hardebeck's model as a starting model with all of the available data (blue circles). Note that there are numerous events with unstable locations, but the vast majority of the events move very little, especially those in the offshore lineation (the target events). The relatively linear character of the offshore events is preserved quite well compared to Hardebeck's results. In subsequent relocation maps, Hardebeck's locations are also used as the reference events (red stars). Figure lb shows a depth section for the target events. Events within 2 km of a great circle through the target events (center point at 35.210, -120.86', azimuth 1600 CCW from North) are shown. The locations in section view are comparable to those provided to us by Hardebeck (Figure I c), although the events that fall within the 2-km-wide zone used for the projection differ slightly between the two plots.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1l-4 of 35

0 4 Next I present the results from Phase 2, evaluating the sensitivity of the offshore epicenters to the velocity model and data subsets used to locate them, and delete-one-station tests to derive estimates of relative location uncertainty. In Figures 2 through 5, I show a series of epicenter maps and cross-sections illustrating the relative stability (or instability) of the hypocenters in the region relative to the use of different velocity models and combinations of data.

Figures 2 and 3 show the effect on the locations of using two other regional 3D models as starting models for the tomographic inversion. For Figure 2, the initial model is extracted from the statewide 3D Vp model of Lin et al. (2009), with a constant Vp/Vs value of 1.78 used to generate the starting Vs model. More events show unstable location perturbations compared to the decimated Hardebeck model. The model of Thurber et al. (2006) for the greater Parkfield region is used as the starting model for the results in Figure 3. Significant location shifts are obtained that are largely systematic in character, with most events moving to the northeast by 2 to 4 km. These results must be viewed with caution, however, because the region of interest is near the corner of the Thurber et al. (2006) study region, so the model gridding may not be adequate to represent the 3D structure faithfully.

The above tests largely reflect uncertainties in absolute locations for the earthquakes. An extreme example is Figure 3a, in which a tomographic solution is obtained that results in a very systematic and relatively large location shift. Although we can identify a likely cause for the presumably poor performance of this solution, in general it would be difficult to state with authority whether one set of tomographic results is superior to another.

Previous experience shows that including active-source data, especially for shots recorded at network stations, helps remove some of the nonuniqueness that is inherent in passive-source (earthquake) tomographic inversions, and can help lead to reasonably accurate relative locations (Thurber et al., 2003, 2004). There have been previous active-source experiments in the study region, including off-shore/on-shore work. Such data should be incorporated in future tomographic modeling work. If network picks can be obtained for these shots from archived waveforms or catalog records, or alternatively for known quarries in the region, such data could be quite valuable in constraining absolute locations better than the present earthquake-only dataset can.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-5 of 35

5 U

1 5.25'

• 4 km

.125' Figure 2a 0 5 10 I  !

NW SE NW SE 15 10 -5 0 5 10 15 20 .20 10 -5 0 5 10 15 20 0 0 0 0 0 0 -

-5 -5 -5 -5 0

0

-10 .10 -10 -10

-15 -15 -15 -15

-20 -20 -20 -20

-20 -15 -10 -5 0 5 10 15 20 15 5 0 5 10 15 20 Figure 2b Figure 2c Figure 2a compares the epicenters from Hardebeck's tomographic inversion (red stars) to epicenters derived from a tomographic inversion using an initial model extracted from the statewide 3D Vp model of Lin et al. (2009) (blue circles), with a constant Vp/Vs value of 1.78 used to generate the starting Vs model, with all of the available data. There are a surprisingly large number of events with unstable locations, but the majority move less than a kilometer, especially those in the offshore lineation (the target events). The linear character of the offshore events is again preserved reasonably well. Figure 2b shows a depth section for the target events from their Figure 2a locations. As before, events within 2 km of a great circle through the target events (center point at 35.21', -120.86', azimuth 1600 CCW from North) are shown. The locations in section view are similar to those provided by Hardebeck (Figure 2c), although the events that fall within the 2-km-wide zone used for the projection differ slightly between the two plots.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-6 of 35

0

-121 12075 5]25 , >- - 1 125!o 05 km . -oI 10 Figure 3a NW SE NW SE 15 -10 -5 0 5 10 15 20 -20 *15 5 0 5 10 15 20 0 0 0 0

-5 .5 .0

-5 -5

-10 10 -10 -10

-15 15 -15

.20 .20 -20 -20 15 -10 -5 0 5 10 15 20 15 5 0 5 10 15 20 Figure 3b Figure 3c Figure 3a compares the epicenters from Hardebeck's tomographic inversion (red stars) to epicenters derived from a tomographic inversion using an initial 3D Vp model from Thurber et al. (2006) for the greater Parkfield region (blue circles), with a constant Vp/Vs value of 1.78 used to generate the starting Vs model, and with all of the available data. There is a systematic northeastward shift of 2 to 4 km for most events. Figure 3b shows a depth section for the target events from their Figure 3a locations. As before, events within 2 km of a great circle through the target events (center point at 35.21 0, -120.86', azimuth 1600 CCW from North) are shown. The locations in section view are quite different those provided by Hardebeck (Figure 3c), with the events being systematically shallower in Figure 3b.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-7 of 35

0 q" 7 Withholding of either the cross-correlation data or the S data also leads to some location scatter, although much less than for the results shown in Figures 2 and 3. In Figure 4, we repeat the tomographic inversion using the decimated version of Hardebeck's model as a starting model, but remove the cross-correlation data from the inversion. Similarly, in Figure 5, we repeat the tomographic inversion using the decimated version of Hardebeck's model, but remove the S-wave data from the inversion. In both cases, the locations are much more stable than I expected.

The location shifts for the target events nearest the coast in particular are quite small, less than 500 m in epicenter in many cases. In depth, the location differences are relatively subtle.

It is quite important to assess the level of relative location uncertainty in addition to characterizing the absolute location uncertainty. One way to approach this is to selectively and repeatedly remove data from the inversion and use the resulting variability in the locations to provide an empirical estimate of the relative location uncertainty. I used the strategy of deleting individual stations from the inversion, one at a time, and evaluating the standard deviations of the location coordinates (inthe Cartesian reference frame of the cross-sections shown above). I sequentially deleted one of the five coastal stations ARHV, DCHV, LMHV, PBHV, and PSHV, chosen because of their locations and their provision of numerous S-wave picks. Thus, removal of these stations should have the largest possible impact on the stability of the relative locations.

This jackknife-like estimate of relative location uncertainty yields values of 140 m in the direction parallel to the lineation, 190 m perpendicular to that, and 280 m in depth. This can be compared to the hypoDD-style (Waldhauser, 2001) location uncertainty estimates provided by tomoDD that are typically 3 to 10 m.

On the basis of the above tests, my firm conclusion is that the offshore seismic lineation feature discovered by Hardebeck is real. Joint tomographic inversions with different starting models yield locations that show systematic shifts but tend to preserve the lineation relatively clearly. The locations are robust to the removal of cross-correlation data or S-wave data from the inversion. An empirical estimate of relative location uncertainty yields values that are much higher than the typical optimistic values produced from the double-difference approach, but they are still sufficiently small to give sufficient confidence to Hardebeck's results. Further

,geophysical investigation of this feature is warranted, in my opinion.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-8 of 35

8

-120.75' 2

5.25"

/

km 0 5 10 Figure 4a NW SE NW SE 15 -10 -5 0 5 10 15 20 15 5 0 5 10 15 20 0 0 v0 0 0

-5 1-5 -5

-10

-10 -10 -10 15

-15 -15 -15

-20 20 -20 -20 15 -10 -5 0 5 10 15 15 5 0 5 10 15 20 Figure 4b Figure 4c Figure 4a compares the epicenters from Hardebeck's tomographic inversion (red stars) to epicenters derived from a tomographic inversion using the same input model but excluding the waveform cross-correlation data (blue circles). The location shifts are generally small, especially for the target earthquakes. This indicates that the catalog absolute and differential times by themselves are adequate to yield well-constrained locations. Figure 4b shows a depth section for the target events from their Figure 4a locations. Again, events within 2 km of a great circle through the target events (center point at 35.21', -120.86', azimuth 160' CCW from North) are shown. The locations in section view are quite similar to those provided by Hardebeck.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1-9 of 35

9 S

• - N

~'

.125* 0 k 5 . 10  !

5. 0 ....... . L Figure 5a NW SE NW SE 15 -10 -5 0 5 10 15 20 15 -10 -5 0 5 10 15 20 o0 u 0 0

-5 -5 .5 -5

-10 10 -10 -10

-15 15 -15 -15

-20 -20 -20 -20 15 -10 -5 0 5 10 15 20 15 -10 -5 0 5 10 15 20 Figure 5b Figure 5c Figure 5a compares the epicenters from Hardebeck's tomographic inversion (red stars) to epicenters derived from a tomographic inversion using the same input model but excluding the S-wave data (blue circles). The location shifts are generally small, especially for the target earthquakes. This indicates that the catalog absolute and differential times by themselves are adequate to yield well-constrained locations. Figure 4b shows a depth section for the target events from their Figure 4a locations. Again, events within 2 km of a great circle through the target events (center point at 35.21', -120.86', azimuth 1600 CCW from North) are shown. The locations in section view are extremely similar to those provided by Hardebeck.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1 -10 of 35

10 Appendix by I include here two early relocation plots that have been superseded For completeness, the results in the main body of the report.

PGE_1 A

1 '9 9.

9

• 9999 35.375' *9

• .~.  :.*.

j9 .o4.

9..

• -** 99

.9 35.25' p p t I

S 9

.9 9,

.9 9 9 I

.9

~q~9 I

0 35125' "

0 km 5 10

-120.75'

-121' PGE_2 A.

j9

• 99

9. 1.

9 9 9 9 9 . 9 9*. 9

~99 35.375

• 4

~9. A 44~

9 99 4 9 I I*

99 9 9 9 9 9 9. 9 4

• 9 4,9 94 9 35.25' 9 9 .A 9 0*

9 9 9 9 9.

%, ~ .9.

9-.

.9.4 e' . C,

• .9 A

• 9, 9

~ '9...

• 9 km 35.125' 10 0 5

-120.75' Page C1 -11 of 35 C-1, Thurber review Shoreline Fault Zone Report, Appendix

References Lin, G., C. H. Thurber, H. Zhang, E. Hauksson, P. M. Shearer, F. Waldhauser, T. M. Brocher, and J. Hardebeck, A California statewide three-dimensional seismic velocity model from both absolute and differential times, Bull. Seism. Soc. Am., submitted, 2009.

Thurber, C., S. Roecker, K. Roberts, M. Gold, L. Powell, and K. Rittger, Earthquake locations and three-dimensional fault zone structure along the creeping section of the San Andreas Fault near Parkfield, CA: preparing for SAFOD, Geophys. Res. Lett., 30, .10.1029/

2002GL016004, 2003.

Thurber, C., S. Roecker, H. Zhang, S. Baher, and W. Ellsworth, Fine-scale structure of the San Andreas fault and location of the SAFOD target earthquakes, Geophys. Res. Lett., 31, L12S02, doi 10.1029/2003GL019398, 2004.

Thurber, C., H. Zhang, F. Waldhauser, J., Hardebeck, A. Michael, and D. Eberhart-Phillips, Three-dimensional compressional wavespeed model, earthquake relocations, and focal mechanisms for the Parkfield, California, region, Bull. Seism. Soc. Am., 96, S38-S49, 2006.

Waldhauser, F., hypoDD: A computer program to compute double-difference hypocenter locations, U.S. Geol. Surv. Open File Rep.,01-113, 25 pp., 2001.

Zhang, H., and C. H. Thurber, Double-difference tomography: the method and its application to the Hayward fault, California, Bull. Seism. Soc. Am., 93 1875-1889, 2003.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-12 of 35

0 Phase 3 Report: Central CalifOrnia Offshore Earthquake Assessment PG&E Contract No. 2500174779 Clifford H. Thurber Department of Geoscience University of Wisconsin-Madison 1215 W. Dayton St.

Madison, WI 53706 August 18, 2009 Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-13 of 35

0 Phase 3 Report: Central California Offshore Earthquake Assessment PG&E Contract No. 2500174779 Clifford Thurber August 18, 2009 I report on my extended analysis of the locations of events offshore California, 35.10 to 35.30 N, focusing on an apparent NW-SE-trending seismic lineation identified from the tomographic inversion and relocation results of Dr. Jeanne Hardebeck, USGS (the "target" earthquakes). The tasks for Phase 3 are as follows:

Task 1 - Waveform cross-correlation analysis.

a) Redo the waveform cross-correlation analysis with the same method as Jeanne Hardebeck used.

b) Redo the waveform cross-correlation analysis with a more sophisticated method.

Task 2 - Replicate Jeanne Hardebeck's exact tomoDD results.

Task 3 - Redo Jeanne Hardebeck's tomoDD analysis with improved cross-correlation data and adjusted velocity model and parameters.

Task 4 - Write a fully documented report providing all the important input and output files.

Task 1 - Waveform cross-correlation analysis a) Redo the waveform cross-correlationanalysis with the same method as JeanneHardebeck used.

The first step in Task 1a was the reconstruction of the waveform database used by Hardebeck, who provided me with the raw waveforms in multiple compressed tar files. Once the necessary directory structure was established, and the byte order of the waveform data was determined, I started to work with Hardebeck's cross-correlation code (xcor) and found that it did not function properly on our computer system. Debugging required changing to a different Shoreline Fault Zone Report, Appendix C-1, Thurber review Page CI-14 of 35

0 computer platform and swapping the data byte order. Eventually, debugging succeeded and I was able to replicate the cross-correlation results for 3 stations chosen at random by Hardebeck and provided to me, up to an occasional round-off difference in the sixth decimal place (e.g.

0.670000 instead of 0.669999). Thus, I was able to confirm the validity of the original cross-correlation data.

b) Redo the waveform cross-correlationanalysis with a more sophisticatedmethod.

The first step in Task lb was reorganization of the waveform database for compatibility with an independent cross-correlation code, corr-sac (Rowe et al., 2002), along with setting up the necessary auxiliary files. One of corrsac's advantages over standard cross-correlation computations (such as xcor) is the use of dynamic coherency-weighted spectral filtering. The goal is to down-weight incoherent frequency bands while avoiding removal of useful signal, which occurs in standard cross-correlation analysis when a fixed frequency band is used for data filtering. Uncertainties for the differential times are determined using multiple narrow-band estimates. For reliably high correlation values (typically 0.9 and above), the cross-spectral phase method is used to determine a subsample differential time (Poupinet et al., 1984), using multi-taper spectral estimation (Thomson, 1982). Based on the quality of the relocation results shown below, I believe I succeeded in obtaining a very high quality set of cross-correlation data from my analysis with corr sac. I also obtained substantially more high-correlation differential times:

Hardebeck obtained 17,271 differential times above her specified acceptance threshold of 0.8 for the PG&E stations (listed in Table 1) using xcor, whereas I was able to obtain 41,614 differential times above that same threshold for the same set of stations and events with corr sac. Spot checks of individual differential times values show that differences, where present, are typically on the order of one to a few samples.

Task 2 - Replicate Jeanne Hardebeck's exact tomoDD results In Phases 1 and 2 of my work, I attempted to replicate the original double-difference (DD) tomography results of Hardebeck using tomoDD (Zhang and Thurber, 2003). Initial problems involved compilation issues, but even when those issues were overcome previously, I was unable to replicate the original results precisely. This problem continued in the current Phase 3 work, until I determined from a detailed comparison of Hardebeck's tomoDD log files Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C1 -15 of 35

0 versus mine that Hardebeck's inversion used a different number of stations than my mine did despite my use of the exact files provided to me by her. With the station list identified as the likely source of the discrepancies between the results, I discovered that the station file provided to me post-dated the actual inversion run I was trying to match. Using the matching station file (which contained an error that led to Hardebeck modifying it before sending it to me) and the tomoDD executable compiled on Hardebeck's c6mputer system, I was finally able to exactly replicate her inversion and relocation results. A map view and cross-section of my relocations that match the original are shown in Figure 1. Specifically, my locations in Figure 1 exactly match those of Hardebeck's February 26, 2009, tomoDD results. Thus, I was ultimately able to confirm the validity of Hardebeck's original tomoDD results.

Task 3 - Redo Jeanne Hardebeck's tomoDD analysis with improved cross-correlation data and adjusted velocity model and parameters My main concern about Hardebeck's original tomoDD results was the extremely fine model grid she employed in her inversions, with 2 km node spacing in all directions throughout most of the inversion model. For a large study region (100 km by 100 km) with a coarse station spacing (generally 10 km and greater), it is my opinion that use of a very fine grid (in this case, 2 km) is potentially pushing the limits for achieving a stable inversion result. Hardebeck used a high smoothing weight in her inversions, which appears to have compensated adequately for the fine gridding. In my Phase 1 and 2 work, I tested the use of two different initial models, one based on Subregion 3 of the statewide tomography study of Lin et al. (2009) with 10 km horizontal gridding, and another from the Parkfield regional model of Thurber et al. (2006) with variable gridding (3 to 20 km). The results I obtained with the former initial model were more stable and the velocity model grid covered the present study area more uniformly, so I ultimately adopted a modified version of the Lin et al. (2009) model grid, with 10 km horizontal gridding covering a smaller region, to use for most of my Phase 3 tomoDD inversions - see Figure 2 a and

b. For additional comparisons, I also decimated Hardebeck's model to 4 km and 10 km horizontal gridding (and with coarser vertical gridding as well) and carried out a pair of inversions with each decimated model.

I carried out 24 inversions using tomoDD with varying cross-correlation datasets, smoothing weights, data weighting, and solution step sequences. Key aspects of the inversions Shoreline Fault Zone Report, Appendix C-1, Thurber review, Page C1 -16 of 35

are listed in Table 2. The locations of the "target" earthquakes vary in detail from one inversion to another, but significantly less so than observed in the inversions for my Phase 1 and 2 report.

In the present work, however, the use of what I consider to be more appropriate initial models (two with a 10-km grid-spacing that is more compatible with the station spacing, and one with a grid interval double that of Hardebeck's) combined with an increased cross-correlation dataset for the PG&E stations resulted in smaller location variations among inversions compared to my previous results.

My preferred results are shown in Figure 3, and the control parameters for that inversion are provided in the Appendix. My preference for this result is based on (1) my opinion that my cross-correlation dataset is of higher quality than that of Hardebeck's and (2) the slightly better data fit for the results shown in Figure 3 compared to that for the results shown in Figures 4 through 7 (the residuals are a few percent smaller on average). The relocations from several other runs with different data and/or initial models are shown in Figures 4 through 7. For Figure 3, absolute and catalog differential P and S waves were used, and the only cross-correlation data included were from my reanalysis of the PG&E station data. For comparison, the location results from a P-wave-only inversion are shown in Figure 4. The differences are subtle. The relocations in Figure 5 were obtained by adding Hardebeck's NCSN and SCSN cross-correlation data, with P and S absolute and catalog differential times included. The differences from Figures 3 and 4 are quite minor. For the relocations in Figures 6 and 7, decimated versions of Hardebeck's initial model with 4 and 10 km gridding were used, and the inversions were otherwise identical to the one producing the results in Figure 4. In the case of Figure 6, the target lineation lies slightly closer to the coastline than the others. Overall, the results are remarkably stable to a variety of significant changes to the inversion.

Comparing my results to those from my replication of Hardebeck's inversion (Figure 1),

the "target" earthquake lineation appears to be slightly more sharply defined in Figures 3 through 7 compared to the replicated Hardebeck results, except for the seismic activity off the tip of San Luis Point. Similarly, the seismicity associated with the Hosgri Fault farther offshore also appears somewhat better aligned in the new results compared to Hardebeck's. In depth, the "target" earthquakes in my inversion results appear slightly more clustered into limited depth ranges (2-3 km, 5-7 km, 10-12 km) than in Hardebeck's results, something that is a characteristic of numerous California strike-slip faults (Rubin et al., 1998; Waldhauser et al., 1999; Schaff et Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-17 of 35

a., 2002; Waldhauser et al., 2004). There is a general theme in these papers that the repeating events indicate relatively steady state loading due to creep surrounding these stuck patches.

Nadeau and coworkers have papers that estimate creep rate from the moments and frequency of repeaters, for example Schmidt et al. (2005). I attribute the increased sharpening and clustering to the quality of the new cross-correlation data. As discussed below, it is not possible to state with certainty whether the original locations in Figure 1 or the new locations in Figure 3 are superior to the other.

In my previous report, I discussed delete-one-station estimates of location uncertainty, which provides robust estimates of relative locations, not absolute locations, because the same velocity model is used for all realizations. Those estimates were on the order of 150-300 meters.

The theory for estimating absolute location uncertainties where the uncertainty in the velocity model is considered explicitly (Pavlis, 1986) is complex, and such an analysis is beyond the scope of this project. A rough estimate of absolute uncertainty can be obtained by considering the variations in absolute locations resulting from the use of different starting velocity models and different control parameter settings. Figures 8 a and b show examples of location differences between Hardebeck's results and two of my solutions (Run 1 Is, from Figure 3, and Run 2 1s, from Figure 7). Shifts of up to about 500 meters are observed for the target earthquakes, with larger shifts for the events along the Hosgri Fault system farther offshore. I consider 500 meters to be a reasonable estimate of the absolute location uncertainty for the target earthquake lineation. The fact that the locations hardly change whether or not S waves are included in an inversion (c.f. Figures 3 and 4) supports this relatively small absolute location uncertainty estimate.

Task 4 - Write a fully documented report providing all the important input and output files This report contains the description of the key results that have emerged from my work.

In addition, I am providing a compressed tar file of the input parameter files and the final hypocenter location files for the inversion results shown in the figures, as well as the cross-correlation results I obtained for the PG&E stations. Upon request, I can provide any other files that are deemed useful.

..Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-18 of 35

Conclusions and Recommendations On the basis of my research in this third phase of work, my previous conclusion is confirmed: that the coastal seismic lineation feature discovered by Hardebeck is real. Joint tomographic inversions with different datasets and control parameters yield locations that show systematic shifts but tend to preserve the lineation quite clearly, and in some cases sharpen its appearance. A similar sharpening of the seismicity trend associated with the Hosgri Fault situated even farther from the coast, adds confidence to the reality of the coastal lineation.

Further geophysical investigation of this feature is clearly warranted.

There are a number of reasons why seismic tomography by itself cannot provide definitive results on absolute locations of earthquakes. An obvious reason is data noise, which generally will be worse for S waves. Data noise can lead to both relative and absolute location errors - random pick errors produce location scatter and systematic pick errors cause systematic location shifts. It is also not possible to represent the Earth's structure exactly. The seismic velocity structure of the Earth varies over a wide range of scales, and seismic tomography can only capture part of that range. First-arrival tomography cannot be used to detect and model sharp discontinuities, and it inevitably produces a blurred model. A more subtle problem is that because the seismic velocity model is not perfectly known, first-arrival ray paths will be incorrect in general and sometimes can be seriously inerror (e.g., an up-going wave instead of a down-going wave). Finally, the lack of strong constraint on shallow structure due to the absence of very shallow earthquake sources and the sparse distribution of stations makes the well-known depth-origin time trade-off a serious problem for obtaining absolute depths, especially when the earthquakes lie outside the network.

My previous experience with local earthquake tomography shows that including active-source data, especially for shots recorded at network stations, helps remove some of the non-uniqueness that is inherent in passive-source (earthquake) tomographic inversions, and can help lead to reasonably accurate absolute locations (Thurber et al., 2003, 2004). There have been previous active-source experiments in this study region, including offshore/onshore work. The active-source data available through the California earthquake data centers and USGS colleagues was incorporated in my tomographic modeling work (Figure 2a), which included picks from the PG&E (Lines 1 and 3) and Morro Bay refraction lines, but the data I was provided did not include PG&E network picks for these refraction lines. If network picks can be made available Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-19 of 35

for these shots from archived waveforms or catalog records, they could be quite valuable in constraining absolute locations better than the present dataset can. Alternatively, if it is possible to identify events that can be unambiguously associated, with known quarries in the region, as was done by Lin et al. (2007) for-southern California, that could provide constraints that are nearly as strong as active-source data. In additional, there may be other active-source data not available through the California earthquake data centers that could be instrumental in improving absolute location constraints, or perhaps a new offshore-onshore dataset could be acquired.

I also strongly recommend an effort to determine whether the location of the lineation derived from my work or Hardebeck's work can be associated with features suggestive of faulting in other sources of data, although the likelihood that this feature represents strike-slip faulting with possibly limited total offset may make such work difficult. Marine seismic reflection data from offshore of Buchon Point might reveal features that could constrain the absolute location of the causative fault. Onshore high-resolution seismic profiling, trenching, or other paleoseismic work might reveal if the fault comes onshore, and if so where. High-resolution potential field data may provide key clues to the absolute position of the fault. There also may be results already published in the gray literature that could yield critical information.

Shoreline Fault Zone Report, Appendix. C-1, Thurber review a Cl-20 of 35 Page

0 Table 1. List of PG&E stations for which xcor cross-correlation processing was done by Hardebeck and corrsac cross-correlation processing was done by Thurber.

STATION NAME LATITUDE LONGITUDE AR 35.4632 -120.9783 BL 35.5338 -120.9067 BP 35.7537 -121.1432 CS 35.6005 -121.0940 DC 35.2122 -120.8408 DC2 35.2138 -120.8503 DP 35.2330 -120.7817 EC 35.3333 -120.7182 LM 35.3803 -120.8247 LQ 35.6688 -120.9907 LS 35.2973 -120.8430 ML 35.3225 -120.6025 NP 35.0863 -120.4725 OC 34.8500 -120.4772 OF 35.0160 -120.6055 PB 35.7090 -121.2715 PS 34.9022 -120.6217 RP 35.7772 -121.3058 SH 35.1682 -120.7613 WR 35.4588 -120.8817 Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-21 of 35

Table 2. Information on selected key inversion control parameters and input data for the tomoDD inversions carried out in this project. Modified L is model modified from the Lin et al.

Subregion 3 model; Dec. H is decimated version of Hardebeck's model at 4 or 10 km grid spacing. Cross-correlation datasets are as follows: H, Hardebeck's original xcor; TO.7, Thurber's corr sac with 0.7 threshold; TO.8, Thurber's corrsac with 0.8 threshold, P+S; T0.8P, Thurber's corr sac with 0.8 threshold, P only. Inversion sequences included 3, 4, or 5 steps: 3 =

hypocenter only, joint inversion, and hypocenter only; 4 = joint, hypocenter, joint, hypocenter; 5

= hypocenter, joint, hypocenter, joint, hypocenter. Data weighting combinations: h = high, e equal; A = absolute, CT=catalog, CC=cross-correlation (for example, hA means high weight on absolute data, eACT means equal weight on absolute and catalog data, etc.).

Run

• Initial V P or P+S? CC data Smooth Wt Inversion sequence 0 Modified L P+S H 20 3hA, 3eACTCC 1 Modified L P+S H 20 3eACT, 3hCT, 3hCC 2 Modified L P+S H+TO.7 20 3hA, 3hCT, 3hCC 3 Modified L P+S H+TO.7 200 3hA, 3eACT, 3hCC 4a Modified L P+S H+TO.7 100 4eACT, 4hCC 4b Modified L P+S H+TO.8 100 4eACT, 4hCC 5a Modified L P+S H 20 4eACT, 4hCC 5b Modified L P+S TO.8P 20 4eACT, 4hCC 6 Modified L P+S H+T07 200 5eACT, 5hCC 7a Modified L P+S TO.8 200 5eACT, 5hCC 7b Modified L P+S T0.8P 200 5eACT, 5hCC 8a Modified L P+S H+TO.7 200 5hA, 3eACT, 5hCC 8b Modified L P+S TO.8P 200 5hA, 3eACT, 5hCC 9 Modified L P+S H 200 5hA, 3eACT, 5hCC 10a Modified L P+S TO.8 100 ShA, 3eACT, 5eCTCC 10b Modified L P+S T0.8P 100 5hA, 3eACT, 5eCTCC llp Modified L P T0.8P 100 5eACT, 3hCT, 5hCC lls Modified L P+S TO.8P 100 5eACT, 3hCT, 5hCC 12p Modified L P H+TO.8P 100 5eACT, 3hCT, 5hCC 12s Modified L P+S H+TO.8P 100 5eACT, 3hCT, 5hCC 20p Dec. H 4 P TO.8P 100 5eACT, 3hCT, 5hCC 20s Dec. H 4 P+S TO.8P 100 5eACT, 3hCT, 5hCC 21p Dec. H 10 P T0.8P 100 5eACT, 3hCT, 5hCC 21s Dec. H 10 "P+S T0.8P 100 5eACT, 3hCT, 5hCC Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-22 of 35

35.2*,

a)-121" -120.8*

oNW 10 o SE 0 0 0, 0 00 a):: -52o o108 NW SE(b 1~8 0 0o CL o 0 0 0 ~ 00 0 Q-io 00 0

0

-15 , 1 , ,

-10 -5 0 5 10 b) Horizontal Distance (kin)

Figure 1. (a) Map view and (b) cross-section (center point at 35.21', -120.86', azimuth 1290 CW from North, half-width 1 km) of replicated locations from Hardebeck's tomographic inversion.

The box in (a) indicates the earthquakes plotted in (b).

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-23 of 35

42" 40O 38*

36" 34*

32*

-16*

-14 -22 -2 *,* -11*:. -1_6' -114*

Figure 2a. Map showing the distribution of all the earthquakes (colored circles), explosions (red stars), seismic stations (black triangles) and model grid nodes (colored diamonds) included for the DD tomography inversions for Subregion 3 of Lin et al. (2009), from which the initial velocity model used here was extracted.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-24 of 35

0 37.50 0 A A.

370 *

• 04 A.* 4

• 4 , . - . ,,.

• LA 0A, "0

37

• A' A-*

365 365 " 0.

'A

"- A*AAA A**~ *A . * .~

0A+A .4- A'A 36* , . A o.

-A ,

A ..

-A A A . A A' AL.* "

A_.

AA~

A.. AAA AA At 36 A A kmA 0 AA0 A.

0 po5 A.A. *A 0 A 0 AA

• . **, -. 0 3500 " 0 "" " 0' * "

- A

'A A *A A AA' saA.A 350 A 34.5° &

km A 20,27 0 50 3,and45kmdeth A 34.50 Figure 2b. Map showing the distribution of seismic stations (red triangles) and model grid nodes (black dots) for my tomoDD inversions, modified from the work of Lin et al. (2009). The grid has a uniform spacing of 10 km in the horizontal directions, and has nodes at -1, 1, 4, 8, 14, 20, 27, 35, and 45 km depth.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-25 of 35

35.2* 00 '~ sos~*

km" 0 10 0 5C1 a)a 0 0 0 0~0 0 E: 0 C6 o0 o CL 0 CD 000 0"0000 000 0:3-10 0 p 0

o o

-15 , ,1 ,

-10 -5 0 5 10 b) Horizontal Distance (km)

Figure 3. (a) Map view and (b) cross-section (center point at 35.2 10, _120.86', azimuth 1290 CW from North, half-width 1 km) of locations from a tomoDD inversion using P and S waves, the modified initial model from Lin et al. (2009), and cross-correlation data from my reanalysis of PG&E data. The box in (a) indicates the earthquakes plotted in (b). Run 11 s in Table 2.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-26 of 35

0

• 1

• es 0 1..0

• 5

'00.

352 41'0*ie 35.20 0.

".S* ' . 3 l~ \ - -" i 0*

• S 00 0,, a 0 NW S 0 0 0 *) 00 0 0 0 0 S S

-1210 CL 00 0 0 6- 0 "10 ",

0 00 0,00.

o 0 00 C}-0 o oo o 10 0O 5 10 oc Fiur 4.21 (a a iwad()cosscin(etrpita 520 120.86' zmuh19

-15 , ,,0 10 --0 05051 b) Horizontal Distance (kin)

Figure 4. (a) Map view and (b) cross-section (center point at 35.2 10, _120.86o, azimuth 129° CW from North, half-width 1 km) of locations from a tomoDD inversion using the modified initial model from Lin et al. (2009), cross-correlation data from my reanalysis of PG&E data, and P waves only. The box in (a) indicates the earthquakes plotted in (b). Run lp in Table 2.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-27 of 35

35.26 * -. . -.

km "

0 5 1'0

-121 ° -120.8*

a)

NW SE 0

0 0 co E -5 0 C0

- -1 0 CPo o 0 001 0 000 00 W) 00 0 0

0 CLo 00

-15 , , I1

-10 -5 0 5 10 b) Horizontal Distance (km)

Figure 5. (a) Map view and (b) cross-section (center point at 35.2 10, _120.86o, azimuth 1290 CW from North, half-width 1 km) of locations from a tomoDD inversion using P and S waves, the modified initial model from Lin et al. (2009), and cross-correlation data from my reanalysis of PG&E data plus Hardebeck's cross-correlation data for NCSN and SCSN stations. The box in (a) indicates the earthquakes plotted in (b). Run 12s in Table 2.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-28 of 35

O 35.21 4b 4%

0 0 k

4%; rz* au'0

• 52 5% s 0

00 0 10

-121 -10*

0 5 0 0 00 -

35.2 o oo 'S 0~*

0 000 0 0 00 00 tE -5 0 0 r* 0 0 0 0

0 0

-15 ,1 1

-10 5 10 b) Horizontal Distance (kmn)

Figure 6. (a) Map view and (b) cross-section (center point at 35.210, _120.860, azimuth 129' CW from North, half-width I km) of locations from a tomoDD inversion using P and S waves, F the 4-km grid decimated Hardebeck model, and cross-correlation data from my reanalysis of PG&E data. The box in (a) indicates the earthquakes plotted in (b). Run 20s in Table 2.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page CI-29 of 35

35.2 *N_"Z -:.

0 km 5 1 ,"

a)-121° -120.80 NW SE 00 0

0 5 10 Q 0 208 0 0 0-10 0 00 0

0 00 0-10 0 0

0 0

0 000 00

-10 -5 0 5 10 b) Horizontal Distance (km)

Figure 7. (a) Map view and (b) cross-section (center point at 35.21', -120.86', azimuth 1290 CW from North, half-width 1 km) of locations from a tomoDD inversion using P and S waves, the 10-km grid decimated Hardebeck model, and cross-correlation data from my reanalysis of PG&E data. The box in (a) indicates the earthquakes plotted in (b). Run 21 s in Table 2.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-30 of 35

35.25- °

• 00

• 1*

km 35.1255

-121" -120.875' -120.75*

Figure 8. (a) Comparison of the epicenters from Hardebeck's tomographic inversion (red stars) to the epicenters shown in Figure 3 (blue circles; Run IIs in Table 2). The location shifts are relatively small for the target earthquakes, generally a few hundred meters seaward. The location shifts for events along the Hosgri Fault system are substantially larger, ending up on the order of a kilometer farther away from the coast.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-31 of 35

35.25 I 35.125 , km 0 5

-121 -120.875' -120.75' Figure 8. (b) Comparison of the epicenters from Hardebeck's tomographic inversion (red stars) to the epicenters shown in Figure 7 (blue circles; Run 21s in Table 2). The location shifts are slightly larger for the target earthquakes, with some moving up to about a half kilometer seaward. The location shifts for events along the Hosgri Fault system are not as large as in Figure 8a, ending up on the order of half a kilometer farther away from the coast.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page C 1-32 of 35

Appendix List of parameter settings used for the tomoDD inversion shown in Figure 3.

• --- data type selection:

• IDAT: 0 = synthetics; 1= cross corr; 2= catalog; 3= cross & cat

• IPHA: 1= P; 2= S; 3= P&S'

• DIST:max dist [km] between cluster centroid and station

• IDAT IPHA DIST 3 3 200

• --- event clustering:

• OBSCC: min # of obs/pair for crosstime data (0= no clustering)

• OBSCT: min # of obs/pair for network data (0= no clustering)

• OBSCC OBSCT CC-format 0 0 2

- solution control:

• ISTART: 1 = from single source; 2 = from network sources

• ISOLV: 1 = SVD, 2=lsqr

• NSET: number of sets of iteration with specifications following

• ISTART ISOLV NSET weightl weight weight3 air depth 2 2 13 100 100 100 -1.7

• i3D delti ndip iskip scalel scale2 iuses 2 0 9 1 0.5 1.00 2

• xfac tlim - nitpb(1) nitpb(2) stepl 1.3 0.0005 50 50 0.5

• latOrig lonOrig Z Orig iorig rota 36.5000 -120.0000 0 1 -36.0

"--- data weighting and re-weighting:

• NITER: last iteration to used the following weights

• WTCCP, WTCCS: weight cross P, S

• WTCTP, WTCTS: weight catalog P, S

• WRCC, WRCT: residual threshold in sec for cross, catalog data

• WDCC, WDCT: max dist [km] between cross, catalog linked pairs

• DAMP: damping (for isqr only)

• --- CROSS DATA ---------CATALOG DATA

• NITER WTCCP WTCCS WRCC WDCC WTCTP WTCTS WRCT WDCT WTCD DAMP JOINT THRES 2 0.01 0.01 -9 -9 1.0 0.5 -9 -9 1 200 0 2 2 0.01 0.01 -9 -9 1.0 0.5 ' -9 -9 1 200 1 2 2 0.01 0.01 -9 -9 i'0 0.5 -9 -9 1 200 0 2 2 0.01 0.01 -9 -9 1.0 0.5 -9 10 1 200 1 2 2 0.01 0.01 -9 -9 1.0 0.5 -9 10 1 200 0 2 2 0.1 0.05 -9 -9 1.0 0.5 -9 10 .1 200 0 2 2 0.1 0.05 -9 -9 1.0 0.5 -9 10 .1 200 1 2 2 0.1 0.05 -9 -9 1.0 0.5 -9 10 .1 200 0 2 2 1 0.5 8 -9 1.0 0.50 6 10 .01 200 0 2 2 1 0.5 8 -9 1.0 0.50 6 10 .01 200 1 2 2 1 0.5 8 5 1.0 0.50 6 10 .01 200 0 2 2 1 0.5 8 5 1.0 0.50 6 10 .01 200 1 2 2 1 0.5 8 5 1.0 0.50 6 10 .01 200 0 2 Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-33 of 35

References Du, W., C. H. Thurber, and D. Eberhart-Phillips, Earthquake relocation using cross-correlation time delay estimates verified with the bispectrum method, Bull. Seism. Soc. Am., 94, 856-866, 2004.

Lin, G., P. Shearer, E. Hauksson, and C. Thurber, A three-dimensional crustal seismic velocity model for Southern California from a composite event method, J. Geophys. Res., 112, B11306, doi: 10.1029/ 2007JB004977, 2007.

Lin, G., C. H. Thurber, H. Zhang, E. Hauksson, P. M. Shearer, F. Waldhauser, T. M. Brocher, and J. Hardebeck, A California statewide three-dimensional seismic velocity model from both absolute and differential times, Bull. Seism. Soc. Am., revised and resubmitted, 2009.

Pavlis, G. L., Appraising earthquake hypocenter location errors: a complete, practical approach

. for single-event locations, Bull. Seismol. Soc. Am. 76, 1699-1717, 1986.

Poupinet, G., Ellsworth, W.L., Frechet, J., Monitoring velocity variations in the crust using earthquake doublets: An application to the Calaveras fault, California, J. Geophys. Res., 89, 5719-5731, 1984.

Rowe, C. A., R. C. Aster, B. Borchers and C. J. Young, An automatic, adaptive algorithm for refining phase picks in large seismic data sets, Bull. Seism. Soc. Am., 92, 1660-1674, 2002.

Rubin, A. M., D. Gillard, and J,-L. Got, Streaks of microearthquakes along creeping faults, Nature, 400, 635-641, 1999.

Schaff, D. P., G. H. R. Bokelmann, G. C. Beroza, F. Waldhauser, and W. L. Ellsworth, High resolution image of Calaveras Fault seismicity, J. Geophys. Res. 107, no. B9, 2186, doi 10.1029/2001JB000633, 2002.

Schmidt, D. A., R. Biirgmann, R. M. Nadeau, and M. D'Alessio, Distribution of aseismic slip rate on the Hayward fault inferred from seismic and geodetic data, J. Geophys. Res. 110, B08406, doi 10.1029/2004JB003397, 2005.

Thomson, D. J., Spectrum estimation and harmonic analysis, Proc. IEEE, 70, 1055-1096, 1982.

Thurber, C., S. Roecker, K..Roberts, M. Gold, L. Powell, and K. Rittger, Earthquake locations and three-dimensional fault zone structure along the creeping section of the San Andreas Fault near Parkfield, CA: preparing for SAFOD, Geophys. Res. Lett., 30, 10.1029/

2002GL016004, 2003.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-34 of 35

22 Thurber, C., S. Roecker,-H. Zhang, S. Baher, and W. Ellsworth, Fine-scale structure of the San Andreas fault and location of the SAFOD target earthquakes, Geophys. Res. Lett., 31, L12S02, doi 10.1029/2003GL019398, 2004.

Thurber, C., H. Zhang, F. Waldhauser, J. Hardebeck, A. Michael, and D. Eberhart-Phillips, Three-dimensional compressional wavespeed model, earthquake relocations, and focal mechanisms for the Parkfield, California, region, Bull. Seism. Soc. Am., 96, S38-S49, 2006.

Waldhauser, F., W. L. Ellsworth, and A. Cole, Slip-parallel seismic lineartions along the northern Hayward fault, California, Geophys. Res. Lett. 26, 3525-3528, 1999.

Waldhauser, F., W. L. Ellsworth, D. P. Schaff, and A. Cole; Streaks, multiplets, and holes: High-resolution spatio-temporal behavior of Parkfield seismicity, Geophys. Res. Lett., 31, L 18608, doi: 10. 1029/2004GL020649, 2004.

Zhang, H., and C. H. Thurber, Double-difference tomography: the method and its application to the Hayward fault, California, Bull. Seism. Soc. Am., 93, 1875-1889, 2003.

Shoreline Fault Zone Report, Appendix C-1, Thurber review Page Cl-35 of 35

0 Appendix C-2 Central California Offshore Earthquake Location Assessment Dr. Felix Waldhauser Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-1 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 P M Project Report Project Name: Central California Offshore Earthquake Location Assessment Project Number: 2500179119 Prepared by:

Dr. Felix Waldhauser 423 W 1 2 0 th Street, Apt 88 New York, NY 10027' Tel: 212 678 4804 Email: felixw(,ldeo.columbia.edu Prepared for:

Dr. Marcia McLaren, Geosciences, PG&E Summary A series of double-difference relocations is carried out using the hypoDD algorithm of Waldhauser and Ellsworth (2000), together with different velocity models and hypoDD parameters, to evaluate the robustness of earthquake locations along the Shoreline Fault Zone (SFZ). The SFZ is a linear seismicity feature off-shore and southeast of Point Buchon, Central California, between latitudes 35.1°iN and 35.3' N. Existing seismic arrival time picks and cross-correlation delay times, as well as new synthetic data, were used to relocate the events and estimate their location uncertainty and robustness. This study finds that the resolution in relative hypocenter locations, given the available data and network configuration, is on average better than 0.2 km in horizontal and 0.7 km in vertical directions. This study confirms the spatial distribution of the seismic activity along the Shoreline Fault Zone as imaged by J. Hardebeck, USGS.

1. Objectives The focus of this study is to test the robustness of earthquake locations estimated by Jeanne Hardebeck (USGS) with the tomoDD program (Zhang and Thurber, 2003) within the linear seismicity feature offshore and southeast of Point Buchon, Central California, between latitudes 35.1°N and 35.3' N (Figure 1), referred to in the following as the Shoreline Fault Zone (SFZ). This is done by re-analyzing existing pick and cross-correlation based differential times (Jeanne Hardebeck, pers. Comm.), as well as synthetic travel times. The goal is to estimate hypocenter locations and associated location uncertainties using the double-difference algorithm hypoDD (Waldhauser and Ellsworth, 2000; Waldhauser, 2001) together with ID and 3D seismic velocity models.

The various location estimates that result from this study are then compared with the locations obtained by J. Hardebeck.

1 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-2 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 A 9/09 8:47 P M Figure 1 Study area in Central California. Epicenters color-coded by depth. Coastline and mapped surface traces of faults are shown. Polygon includes events analyzed in this study. Elongated ellipse includes events along the Shoreline Fault Zone (SFZ). DCPP:

Diablo Canyon Power Plan. Figure courtesy of J. Hardebeck, USGS.

2. Double-Difference Relocation Analysis The hypoDD algorithm minimizes, in a weighted least squares sense, residuals between observed and predicted travel time differences for pairs of earthquakes to solve for adjustments in the vectors connecting the hypocenters through the partial derivatives of the travel times for each event with respect to the unknown. The method is described in detail in Waldhauser and Ellsworth (2000) and Waldhauser (2001). We apply the 1-D model of McLaren and Savage (2001) and the 3-D tomoDD model provided by Jeanne Hardebeck to the original data set of 1418 earthquakes included in the polygon shown in Figure 1, and two smaller subsets of the data that focus on the Shoreline Fault Zone. A 2

Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-3 of 23

M r7) J l~tlf .' MRIA 4 001UU I I9 - %aeL*,l .*a.k-sltm CAld ' IIImeIIL 'IUJeLCL I"0JUIL Vr. I I It I I* Z .7-I-4 IVI version of the hypoDD program that uses 3-D velocity models has previously been used in Rietbrock and Waldhauser (2004), and has also been applied in Thurber et al. (2005),

and Waldhauser and Schaff (2008).

In all hypoDD runs for which results are shown in the following, we require each event pair to have at least four delay time links. Only stations within 150 km distance from each event pair are used. The SimulPS locations of Jeanne Hardebeck are used as starting locations in all cases. The system of double-difference equations is solved using the conjugate gradient solver LSQR (Paige and Saunders, 1982). In the following figures the various hypoDD location results are always compared to the final tomoDD hypocenter locations provided by Jeanne Hardebeck (in the following also referred to as JH locations).

2.1 Delay time data The analysis includes -1418 earthquakes in the central coastal California region within the polygon provided by Jeanne Hardebeck (USGS), located approximately between latitudes 34.8°N and 35.5°N. In addition to the nearly 21,000 cross-correlation differential times measured by Jeanne Hardebeck, over one million delay times are formed from the phase pick data using the program ph2dt (Waldhauser, 2001).

1.

Parameters used to run ph2dt are listed in Table The accuracy and consistency of the phase arrival time picks and the cross-correlation delay times provided by Jeanne Hardebeck is evaluated by comparing the two data sets for all pair of events for which both data types are available. The distributions of differences in P-wave delay times are shown in Figure 2a for pairs of events with correlatibn coefficients Cf - 0.7 (2588 pairs) and Cf - 0.9 (471 pairs). The distributions have standard deviations of 0.150 s for delay times with Cf > 0.7 and 0.117 s for those with Cf? 0.9. They exhibit long but thin tails. The median absolute value is 0.015 s for both Cf thresholds, indicating both the consistency of the phase picks as well as the precision of the correlation data down to the cutoff threshold of Cf = 0.7. The distribution characteristics are similar for the 22 correlated pairs of earthquakes along the SFZ for which both pick and cross-correlation delay times were available (Figure 2b). The relatively small number of correlation measurements is likely due to the sparseness of the seismicity along the SFZ (i.e., the low number of event pairs with short inter-event distances), and faulting complexities that cause seismograms at common stations to be dissimilar even for nearby events.

Table 1 Parameters used to generate network of delay time links from phase pick data (programph2dt, see Waldhauser 2001 for parameter description):

MINWGHT MAXDIST MAXSEP MAXNGH MINLNK MINOBS MAXOBS 0.0 150 5 100 6 1 60 3

Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-4 of 23

OCAA4 7ni 4n rý + I t' A C ý+k L, A + - D ^+ 0ý ^r+wo 1 111 l QI Q f&A_7 DA P delay times P delay times 5

mf>07 b Cf>0.7 500 (a) m C0.9) 4 400-o3

'E 300 'E 2002 100 I1 o" 0

-0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2 dtpk - dtxor dtl - dtxcorr Figure 2 Histograms of differences between P-wave cross-correlation delay-times obtained by Jeanne Hardebeck and corresponding delay-times formed from the phase picks. a) Distribution shown for 2588 pairs of earthquakes within the polygon in Figure 1 with cross-correlation coefficients Cf a- 0.7 and 471 pairs with Cf ; 0.9. Standard deviations are 0.150 s for Cfa 0.7 and 0.1 17s for Cfa 0.9. Median absolute values are 0.015 s for both Cf thresholds. b) Distribution shown for 22 pairs of earthquakes along the Shoreline Fault Zone with Cf> 0.7 and 4 pairs with Cf> 0.9. Standard deviations are 0.070 s for Cfz 0.7 and 0.022 s for Cf> 0.9. Median absolute values are 0.033 s (Cfa I 0.7) and 0.014 (Cfa 0.9).

2.2 Relocation with 1D model (DD-1D)

The combined pick and correlation data have been inverted using the layered 1-D model of McLaren and Savage (2001) (MS2001, blue line in Figure 3). The velocity depth function of the MS2001 model has been re-sampled for its use with hypoDD to avoid strong velocity contrasts (red line in Figure 3). S-wave velocities are obtained by scaling the P-velocity model by a factor of 1.73. For comparison, we also show the model used for relocating the NCSN earthquake catalog in the Central Coast region (Waldhauser and Schaff, 2008) (green line in Figure 3). The model is based on the 1-D model used by the NCSN for routine earthquake location (their model CST)

(Oppenheimer et al., 1993). Notable differences between the two models are the slower velocities at shallow depths in the NCSN model.

The double-difference results for 1380 events, included in the polygon (DDIDa) are shown in map view and cross sections in Figure 4, and compared to the JH tomoDD locations (gray circles). The map view is centered at 350 12.9769' N and 1200 52.3128' E. The root mean square (RMS) of the weighted pick delay time residuals of the relocated events is 0.071 s (down from 0.18 s before relocation), and that of the correlation time residuals is 0.005 s. The hypoDD parameters that were used to control the iterations and data weighting are listed in Table 2.

Because in the DD approach all events are connected through a network of weighted differential-time links, each event may influence the locations of its neighboring events.

Bias can be introduced when the distribution and quality of the differential times is heterogeneous across the network, as is the case here. While the onshore events are generally well recorded by nearby well-distributed stations, most offshore events have 4

Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-5 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM 0

large azimuthal station gaps. Furthermore, events along the SFZ, for example, connect to onshore events over comparably large distances. In order to evaluate the effect of the (well-constrained) on-shore seismicity on the location of the off-shore locations, two additional DD inversions using the 1-D model have been carried out: the first one uses a smaller (zoomed) box including 349 events and centered on the SFZ (DDlDz) (Figure 5) and one using only the 65 events along the SFZ (DD 1Df) (Figure 6). The RMS values for the former DD relocations are 0.064 s for pick and 0.003 s for correlation data, and for the latter locations these values are 0.048 s and 0.003 s respectively. Except for the damping value, the weighting parameters for these additional runs were kept the same as listed in Table 2.

Table 2 hypoDD iteration and weighting parameters used in this study (see Waldhauser, 2001 for parameter description):

NITER WTCCP WTCCS WRCC WDCC WTCTP WTCTS WRCT WDCT 6 0.01 0.01 -999 -999 1.0 0.5 14 5 6 0.01 0.01 20 -999 1.0 0.5 7 5 6 0.01 0.01 20 -999 1.0 0.5 7 4 6 1.0 0.5 20 5 0.01 0.005 7 4 6 1.0 0.5 10 3 0.001 0.0005 7 4 0

-5

-10 E

-. -15 N

-20

-25

-30 2 3 4 5 6 7 8 9 Vp (km/s)

Figure 3 Velocity-depth functions for the study region: M&S2001: McLaren and Savage (2001); M&S DD: re-sampled McLaren and Savage (2001) model; USGS CST: model used by the NCSN for routine location purposes for the Central Coast (CST) region (Oppenheimer et al., 1993).

5 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-6 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM

../reloc.1 d.ctcc. 150km.5km.MSmod/hypoDD.reloc C:

0 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 0-000 00 -5)

T 0, so 0%

.c -10 ,-b -c -10 >

01

-o -15

-15

-20 -10 0 10 20 distance [km] -2 0 2 distance [krn]

Figure 4 Map view (top panels) and cross-sections (bottom panels) of double-difference solutions (DD1Da) for 1380 earthquakes (out of 1418 selected) within the polygon (see Figure 1) computed using the ID model shown in Figure 3 (red) and the combined data set of 938,000 P- and 112,000 S-wave picks and 8,440 P- and 4,757 S-correlation delay times. In the top left panel, blue squares denote stations that have pick data only and orange squares those that have additional cross-correlation data. Red dots are epicenter locations, gray lines fault traces, and black line coastal line. In the top right panel, color scale indicates hypocenter depths (in km). Box outlines events shown in cross-sections.

Gray circles are the tomoDD locations by J. Hardebeck, with gray lines connecting to the DD locations. Note: In the top right panel, additional events outside the shown area are also included in the relocation. The top left panel shows all events relocated (in red). The map view (top right) is centered on the SFZ at 350 12.9769' N and 1200 52.3128' E.

6 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-7 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM

../reloc.1 d.ctcc.zoom.150km.5km.MSmod/hypoDD.reloc 121.0 W 120.6 W

=a 35.5 N a.E

!flu r-35.0 N

-101

-15'1

-15 -10 -5 0 5 10 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 -q-0

-5 o go 0o

-1 V 0 "5 '

" -10o*

-1

-o o 00 -15

)nI

-20 -10 0 10 20 distance [km] -2 0 2 distance [km]

Figure 5 Same as Figure 4, but showing double-difference solutions (DD1Dz) computed for 349 earthquakes (out of 363 selected) in a smaller area centered on the Shoreline Fault Zone, using the ID model and a combined data set of 203'000 P- and 24'000 S-wave picks and 847 P- and 653 S-wave correlation delay times.

7 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-8 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 0 11 /19/09 8:47 PM 121.0" W 120.6° W ../reloc.1d.ctcc.fault. 150km.5km.MSmod/hypoDD.reloc 15 35.5 N DUU [

a~g 10 35.0 N 5

Ra C,

C 0

-5

-10

-15 `

-15 -10 -5 0 5 10 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 0 a

-5

-10 S10 0~

a,

-15

-15

-9A

-20 -10 0 10 20 on, distance [km] -2 0 2 distance [km]

Figure 6 Same as Figure 4, but showing double-difference solutions (DD1Df) computed for 65 earthquakes along the Shoreline Fault Zone, using the ID model and a combined data set of 14'800 P- and 2'0O0 S-wave picks and 137 P- and 115 S-wave correlation delay times.

8 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-9 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM 2.3 Relocation with 3D model (DD-3D)

In a second step, the combined pick and correlation data have been relocated using hypoDD and the 3D P- and S-wave tomography model provided by Jeanne Hardebeck.

The double-difference results for 1375 events (out of 1418 total) (DD3Da) within the region outlined by the polygon in Figure 1 are shown in map view and cross sections in Figure 7, together with the JH locations (gray circles). The RMS of the final pick delay time residuals for the relocated events is 0.074 s (down from 0.18 s in the initial data),

and that of the correlation time residuals is 0.005 s. These values are similar to the DD1Da results. The weighting parameters are the same as used in the DD1D inversions (see Table 2).

Similar to the 1D cases, two additional 3D DD inversions were performed to evaluate the dependency of the offshore. event locations on the onshore seismicity during the DD relocation procedure. Relocation results from events in the zoomed area (DD3Dz) and events along the SFZ (DD3Df) are shown in Figure 8 and Figure 9. RMS values for the

.DD3Dz relocations are 0.067 s for pick and 0.004 s for correlation data. For the DD3Df relocations these values are 0.059 s and 0.004 s.

9 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-1 0 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/1 9/09 8:47 PM

../reloc.3d.ctcc.150km.Skm/hypoDD.reloc 121.0 W120.6 W Um m n 15 35.5 N 10 S'I;"

t. U 35.0 N A"Qi 5

0

'0

-5

-10

-15 -10 -5 0 5 10 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 m 0

0 0

°ho00 0° 0 -5

~-10 64o8"o 10 $0 C

. k0 CL 0

v-15

-15 o1

~~~If1 '.

-20 -10 0 10 20 -20 L distance [km] -2 0 2 distance [km]

Figure 7 Same as Figure 4, but showing double-difference solutions (DD3Da) for 1375 earthquakes (out of 1418 selected) using the 3D model from Jeanne Hardebeck. The top right panel shows a subset of the events that went into the DD inversion.

10 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-11 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM

../reloc.3d.ctcc.zoom. 150km.5km/hypoDD.reloc 121.0 W 120.6 W a W on 35.5 N 13 III?

M 35.0 N Cf 85 W

-15' 1 1 1 1

-15 -10 -5 0 5 10 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 F-0

_-5 T -5 O 0o 0.. -10 0 0 .c-10

", a)

-15 0

-15 Q

-20 -10 0 10 20 -20' distance [km] -2 0 2 distance [km]

Figure 8 Same as Figure 5, but showing double-difference solutions (DD3Dz) computed for 351 earthquakes (out of 363 selected) using the 3D model from Jeanne Hardebeck.

All events used in this inversion are shown in the map view.

11 Shoreline Fault Zone, Appendix C-2 Waldhauser .Page C2-12 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM

../reloc.3d.ctcc.fault.1 50km.5km/hypoDD.reloc 121.0 W 120.6 W aD 35.5 N 130 ti Mi 35.0 N U

C (U

(I,

5 0 15 distance [km]

Cross Section: B-B' Cross Section: A-A' 0 m v

-5

• "-5 E 00N 0 00;

-10 0 00 .- 10 CL

-- 15 0

-15 0

-20

-2(0 -10 0 10 20 -20 J distance [km] -2 0 2 distance [km]

Figure 9 Same as Figure 6, but showing double-difference solutions (DD3Df) computed for 65 earthquakes using the 3D model from Jeanne Hardebeck.

12 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-13 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM 2.4 Estimates of relative location uncertainty Relative location errors are estimated for each of the 1375 3-D relocated events by bootstrapping, with replacement, the final un-weighted double-difference residual vector (Waldhauser and Ellsworth, 2000). Error ellipsoids are obtained at the 90% confidence level for 200 bootstrap samples. The distributions of the major and minor axes of the horizontal and vertical projections of these ellipsoids are shown in Figure 10. Note i) the strongly elongated lateral ellipses due to the one-side station distribution, and ii) the long tails of the error distributions associated with weakly linked events.

Figure 11 shows locations and error ellipses for the 65 events along the Shoreline Fault Zone computed with the 3D model (DD3Df, see Figure 9). The large lateral errors for two events indicate little control of their location through the delay times that remain after 30 iterations (see Table 2). However, control of these locations may have been better during the first iterations. Formal least squares errors derived from the full covariance matrix for selected number of events show error estimates that are similar to the bootstrap results.

The bootstrap error analysis is also carried out for the DD3Dz case, which includes events surrounding the SFZ (Figure 12). We observe large elongated ellipses for some (mostly onshore) events that indicate loss of control due to weak links to neighboring events. The error ellipses of the events along the SFZ, now derived by including the onshore events, are comparable to the ellipses derived from inverting the SFZ events alone (Figure 11). However, they are overall slightly smaller because of the additional data linking the events along the SFZ to events away from the fault. Events in the center of a cluster are typically best constrained, events at the edge of a cluster worst.

90

- lateral (major) 80 ..... lateral (minor) 70 1 vertical 60 50 CD Cr:340 u- 30 20 10 0

0 0.5 1 1.5 Relative location error [km]

Figure 10 Histograms of lateral and vertical relative location errors of 3D double-difference solutions for 1375 events using the combined pick and correlation data (see Figure 7 for event locations). Errors are computed from the mayor axes of the horizontal and vertical projection of the 90% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector based on 200 samples with replacement.

Percentage values are computed within bins of 0.1 km.

13 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-14 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM 121.0" W 120.6° W ../reloc.3d.ctcc.fault.150km.5km/hypoDD.reloc IID

-'U 35.5 N 10 Qa 35.0 N 5 U

00 0

0)

"o

-5

-10

-15

-10 -5 0 5 10 15 distance [km]

Cross Section: B-B' 0

Cross Section: A-A' r 1 0

T* -5 -5

• - -10 -c -10

-s. o,"0 -.

-o

-15 0

-20

-2 0 -10 0 10 20 -20' distance [km] 2 0 2 4 distance [km]

Figure 11 Relative location errors for the 65 earthquakes along the Shoreline Fault Zone using the 3D model (DD3Df, see also Figure 9). Relative horizontal and vertical error ellipses are shown at the 90% confidence level. Ellipses are computed from the mayor axes of the horizontal and vertical projection of the 90% confidence ellipsoids obtained from a bootstrap analysis of the final double-difference vector based on 200 samples with replacement. Gray lines connect to corresponding JH locations.

14 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-15 of 23

11 /19109 8:47 PM Assessment - Project Report v2.1 2500179119 - Central CA Earthquake

../reloc.3d.ctcc.zoom.1 50km.5km/hypoDD.reloc 121.0 W120.6 W 10 35.5 N UC lII i, 5 35.0 N an U-o 0

CO

-5

-10

-15 0 5 10 15

-10 -5 distance [kin]

Cross Section: B-B' 0, 1 Cross Section: A-A' 0 -5

-5 x

-10

-10[ a)

-o

-15 0 -15f

-20' 10 20 -20 w J

-20 -10 0 2 0 2 4 distance [km] distance [km]

12 Same as Figure 11, but showing the double-difference locations with their Figure for the error ellipses at the 90% confidence level relative horizontal and vertical bootstrap 3D DD solutions shown in Figure 8.

15 Page C2-16 of 23 Waldhauser Shoreline Fault Zone, Appendix C-2

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM

3. Synthetic tests Tests with synthetic travel times are carried out to evaluate the resolution capabilities of the available seismic stations. Travel times are computed from 140 sources through the 3D model to 33 stations that recorded the real earthquakes within 100 km. The sources are evenly distributed (spacing = 2 kin) on a vertical plane between 1 and 13 km depth (gray circles in Figure 13), mimicking the fault defined by the earthquakes along the Shoreline Fault Zone.

The sources are then randomly mislocated between -4 and 4 km in all three directions (color coded in the top right panel and blue circles in the bottom panels of Figure 13)

(Table 3), and Gaussian noise with a standard deviation of 0.1 s is added to the perfect travel times. These values are conservative estimates of the uncertainties in location and pick accuracy of the initial (SimulPS) solutions. From these synthetic travel times we form delay times between nearby events and relocate the events using hypoDD with parameters similar to the ones chosen for relocating the real data. The relocation results for these synthetic data are shown in Figure 14. Average mislocations in each of the three directions are less than 200 m, and the maxium mislocation is less than 500 m (Table 3).

The mislocations are somewhat smaller when Gaussian noise with a standard deviation of 0.01 s, similar to the precision of the cross-correlation data, is added to the perfect data before relocation (Table 3). Note,. however, that events along the SFZ are predominately constrained by delay times from phase picks (see Figure 2b).

Table 3 Horizontal (DX, DY) and vertical (DZ) deviations of shifted starting and relocated synthetic sources from true locations, with Gaussian noise (standard deviation =

0.1 s and 0.01 s) added to the perfect data before relocation (in km):

DX DY DZ Mean Max Mean Max Mean Max Initial 1.9 4.0 1.9 3.9 2.0 4.0 (shifted)

Relocated 0.11 0.32 0.19 0.42 0.13 0.42 noise=0.1 s Relocated 0.09 0.18 0.18 0.32 0.15 0.32 noise=0.01s 16 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-17 of 23

0 2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM 121.0 W 120.6 W ../syn/relocthypoDD.Ioc U

• U U

• U U

35.5 N T EU U

• U U r-U U

U U

35.0 N U -10 0 10 U

distance [kin]

Cross Section: B-B' 0 0 o

-2 ý% o o Cross Section: A-A' 000 OwS to,* *^

• b!b o 0 0 0 O- 00% 09 0 oc 0ow" * "-8 0ý 0p 0oO

-5

-10 0,IB6 0o@ 0, o 0(t, 10.

-10 0 o 0) 0 OD 0 b 0

-15 0 -12 8 o 0

-10 -5 0 5 10 distance [kin] -14 00 0 OCO 0

-16 0 0

-2 0 2 distance [km]

Figure 13 Test with synthetic data. Map view (top panels) and cross sections (lower panels) of synthetic sources (gray circles) for which perfect travel times are calculated through the 3D model to the stations within 100 km (blue squares in the top left panel).

Depth (in km) color-coded (top right panel) and blue (bottom panel) circles show starting locations used in the DD inversion. Starting locations are obtained by randomly shifting the true locations within 4 km in each direction.

17 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-18 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19109 8:47 PM 121.0 W 120.6 W ../syn/reloc/hypoDD.reloc U

U

• U 15 U

• U 10 U

35.5 N 5 U.

U T,

• U 0

U U *

• a N * -5 sq *

• U

-10 U

35.0 N -15 S -10 0 10 U

II distance [km]

Cross Section: B-B' 0

-2 Cross Section: A-A' -4 0

060000,0060000000 0 000 00000000006000000000 -6

-5 oOo0 oooo 0000 oo0ooooo E o00GOOOQQQO Ooo 066000 -8 C-0ooooooo0oooooooooo00

-10 00000000000000000000

-15 -12F U=C

-10 -5 0 5 10 distance [kin] -141

-2 0 2 distance [km]

Figure 14 Same as Figure 13, but with the blue circles now representing the DD relocated sources obtained from the starting locations shown in Figure 13. Gray circles are true locations. Gaussian noise with a standard deviation of 0.1 s was added to the perfect data before relocation.

18 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-19 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM

4. Conclusions A comparison between the six different sets of earthquake locations (1D vs. 3D, for each of the large, local, and fault scale event sets) is given in Figure 15 and Figure 16.

Figure 15 shows epicenter locations from all six data sets superimposed and color coded for comparison. Figure 16 shows, for each event along the SFZ, the deviation in hypocenter locations in horizontal and vertical direction from the mean of all six solutions. Two main trends can be identified. Absolute epicenter locations shift west with increasing area of relocation, and events locate deeper in 3D solutions (DD3D) compared to 1D solutions (DD1D). The west shift from fault scale DD relocations (DDf) to full area relocations (DDa) is about 500 m (Figure 16, left panel). This is presumably due to the larger distances over which the off-shore events are linked to the on-shore events in large-scale relocations, and thus the off-shore events become more sensitive to the apparent faster velocities of the ID model compared to the true structure. That shift is smaller for 3D solutions than it is for ID solutions, indicating that the 3D model is a better representation of the true structure. 3D solutions are on average about 1 km deeper than ID solutions (Figure 16, right panel). Again, this points to the differences between the two velocity models used, in particular at upper crustal depths. Depths are similar for different scales of relocation areas within the 1D and 3D solutions, indicating that relative depths are less sensitive to the larger distances over which offshore events are linked to onshore events.

Table 4 summarizes the differences between the various DD locations obtained in this study and the locations obtained by Jeanne Hardebeck (JH locations). Only the events along the SFZ are included in these statistics. The median absolute deviations are typically less than a few hundred meters from the JH locations. They are highest (-700 m) in vertical directions for hypoDD solutions obtained in the ID model. For comparison, median vertical deviations of the 3D solutions from the JH locations are

-200 m. Maximum deviations range from several hundred of meters (north-south-direction) to less than 3 km in vertical direction. They are smallest for the 3D DD solutions. The SimulPS locations that were used as starting locations in all DD inversions show the largest deviations from the JH locations (-300 in). Relative location errors for events along the SFZ, estimated from a bootstrap analysis of the final double-difference vector, include, in most cases, the locations determined y J. Hardebeck (Figure 11).

The synthetic experiment demonstrates that with the available seismic stations and data accuracy the relative locations of the SFZ events can be constrained within less than 200 on average. The synthetic results indicate no systematic bias in the source locations and no distortion of the general geometry of the simulated vertical fault plane.

In summary, this study confirms the spatial distribution of the seismic activity along the Shoreline Fault Zone as imaged by J. Hardebeck. It is pointed out again here that the underlying seismic data used in this study is the same as used by J. Hardebeck to generate her solutions, as are the fundamental equations to solve for relative event locations. This study focused on testing the robustness of Hardebeck's solutions using different velocity models and various parameters that control the network of delay time links and the behavior of these data during their inversion for relative hypocenter locations.

19 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-20 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM 35.33 N W

35.25 N All lJ 35.18 N " "

• 0DD3Df 5km 6) 00 a o DD3Da 5kinOm¢ O; DD1If 5kinm 35.10° N o DD1Dz5km DD3Df5KM O DO1a 5k 35.10'n¶ Np ~lzk 0 5km 120.92 W 120.84 W 120.750 W 120.66 W Figure 15 Superposition of epicenters from the six different inversions carried out in this study and shown in Figure 4 - Figure 9. Gray circles are S-I locations. DD1Da, DD3Da:

ID and 3D solutions for all events in the polygon shown in Figure 1; DD1Dz, DD3Dz:

ID and 3D solutions for all events in rectangular area centered on the Shoreline Fault Zone; DDlDf, DD3Df: ID and 3D solutions for events along the Shoreline Fault Zone.

20 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-21 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11/19/09 8:47 PM 1

X

-2 [

X X X 1ýqxX 0.5 X *<

X X X

XX X

XX -1 L X X X X V 0 X 01 xXIX

-0.5 [ X xX N

'0 1

X X

X)t XX 2[

-1

-0.5 0 0.5 1 -1 0 1 dX (km) dX (kin)

Figure 16 Relative hypocenter locations (in km) in horizontal (left panel) and vertical (right panel) direction after subtracting the mean of all six solutions for each event along the SFZ. Color-coded as in Figure 15.

Table 4 Summary of absolute differences between Jeanne Hardebeck's tomoDD locations (JH) and the various relocation runs carried out in this study, for the 65 events along the Shoreline Fault Zone (in km):

JH vs. IDX IDY IDZI Median Max Median Max Median Max SimulPS 0.39 1.39 0.23 1.67 0.32 2.84 DD1Da 0.18 1.47 0.21 0.58 0.73 2.97 DDIDz 0.29 1.50 0.18 0.63 0.71. 2.91 DD1Df 0.41 1.64 0.15 0.69 0.72 2.82 DD3Da 0.12 1.25 0.12 0.95 0.20 1.73 DD3Dz 0.18 1.03 0.18 0.89 0.22 1.81 DD3Df 0.19 1.17 0.20 1.12 0.21 2.09

5. Electronic Files 0 ASCII files with double-difference hypocenter locations for all case studies. The files can be downloaded from www.ldeo.columbia.edu/-felixw/PGE/

21 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-22 of 23

2500179119 - Central CA Earthquake Assessment - Project Report v2.1 11 /19/09 8:47 PM

6. References McLaren, M. K. and W. U. Savage (2001), Seismicity of South-Central Coastal California:

October 1987 through January 1997, Bull. Seism. Soc. Am. 91, 1629-1658.

Oppenheimer, D. H., F. W. Klein, J. P. Eaton, F. W. Lester (1993), The Northern California Seismic Network bulletin, January-December 1992, U.S. Geol. Surv. Open-file Report,93-578, Menlo Park, California.

Paige, C. C. and M. A. Saunders (1982). LSQR: Sparse linear equations and least squares problems, ACM Transactionson MathematicalSoftware 8/2, 195-209.

Rietbrock, A. and F. Waldhauser (20014), A narrowly spaced double-seismic zone in the subducting Nazca plate, Geophys. Res. Lett., 31, L10608, doi:10.1029/2004GL019610.

Thurber, C., H. Zhang, F. Waldhauser, J. Hardebeck, A. Michael, and D. Eberhart-Phillips (2006), Three-Dimensional Compressional Wavespeed Model, Earthquake Relocations, and Focal Mechanisms for the Parkfield, California, Region, Bull. Seism. Soc. Am. 96, 38-49.

Waldhauser, F. and W.L. Ellsworth (2000), A double-difference earthquake location algorithm:

Method and application to the northern Hayward fault, Bull. Seism. Soc. Am. 90, 1353-1368.

• Waldhauser, F. (2001), HypoDD: A computer program to compute double-difference earthquake locations, U.S. Geol. Surv. Open-file Report,01-113, Menlo Park, California.

Waidhauser, F. and D.P. Schaff (2008), Large-scale relocation of two decades of Northern California seismicity using cross-correlation and doible-difference methods, J. Geophys. Res.

113, B0831 1, doi:10.1029/2007JB005479.

Zhang, H. and C.H. Thurber (2003),. Double-Difference Tomography: The Method and Its Application to the Hayward Fault, California, Bull. Seism. Soc. Am, 93, 1875-1889, DOI:

10.1785/0120020190.

22 Shoreline Fault Zone, Appendix C-2 Waldhauser Page C2-23 of 23

Appendix D Potential Field Data - Magnetic Surveys Shoreline Fault Zone Report, Appendix D Magnetics Page D-1 of 32

Introduction Potential field (gravity and magnetic) data available at the time of the original LTSP, (e.g. McCulloch and Chapman, 1977; Beyer and McCulloch, 1988; Chapman et al.,

1989) had insufficient resolution for detailed characterization of the Hosgri fault zone.

Residual magnetic intensity data offshore DCPP, presented in LTSP Plate Q43m-2 (PG&E, 1989a), were flown in 1976 with a flight line spacing of 1 mile (1.6 km) at an altitude of 2000 feet (610m) above sea level. Their primary utilitywas in the identification of changes in basement rock characteristics and morphology and in confirming seismic data interpretation of basementstructures.

Modem high resolution geophysical data have significantly improved the imaging of basement structure. In 2008 and 2009 new marine magnetic and aeromagnetic data were collected in the study area. Aeromagnetic data were collected at a lower altitude (305 m v. 610 m) with half of the line spacing (800 m v 1.6 km) using differential GPS navigation. Marine magnetic data were collected at 400 m line spacing. Additional aeromagnetic data at - 50 m elevation and 150 m line spacing were collected along the coastline between Pt. Buchon and Pt. San Luis in late 2009. Figure D-1 compares the earlier LTSP magnetic data with the high resolution data that was collected in 2009.

Aeromagnetic Data Total-field aeromagnetic data were-collected by Eon Geosciences, Inc. from July to September 2008 along the central California Coast Ranges from north of Cape San Martin, near San Simeon, to Point Concepcion. A total of 20,508 line-kilometers of data were collected using a Geometrics G-822A cesium-vapor magnetometer along flight lines spaced 800 m apart at a nominal terrain clearance of 305 meters. Tie lines were flown 8,000 m apart. Figure D- 2 shows the survey area. Two areas within the survey area were not flown because of proximity to condor nesting sites. Aeromagnetic data were collected at a 10 Hz sampling rate, and were adjusted for tail sensor lag and diurnal field variations.

Further data processing included micro leveling using the tie lines and subtraction of the reference field defined by IGRF2005 extrapolated to August 1, 2008. Data were transformed to a Universal Transverse Mercator Projection (Base Latitude 0', Central Meridian -123' W.). and interpolated to a square grid with a grid interval of 200 m using the principle of minimum curvature (Briggs, 1974). A magnetic base station was established near the San Luis Obispo Airport. Base-station readings were recorded every second using a GEM Systems Overhauser GSM-19 magnetometer. A contour map of the aeromagnetic data was published as USGS Open File Report 2009- 1044 (Langenheim et al., 2009a) and is shown in Figure D-3.

An additional 933 line-kilometers of total field aeromagnetic data were collected by New-Sense Geophysics in October 2009 between Pt. Buchon and Pt. San Luis using a Scintrex CS-3 Cesium vapor magnetometer mounted on Bell 206 B3 helicopter with a fixed stinger assembly (New Sense Geophysics, 2010). A three-axis Bartington fluxgate magnetometer was used to measure the orientation and rates of change of the aircraft to compensate for magnetic fields generated by the aircraft itself. Flight lines were spaced 150 m apart at a nominal terrain clearance of 50 to 100 m. Tie lines were flown 1500 m apart. These data were collected by helicopter to fill the gap between the marine and Shoreline Fault Zone Report, Appendix D Magnetics Page D-2 of 32

fixed wing surveys along the coastal zone to provide higher resolution of features near the shoreline. Figure.D-4 shows the area of the 2009 survey and the corrected Total Magnetic Intensity map for the survey area. The sampling rate for the cesium and fluxgate magnetometers and radar altimeter was at 10 Hz. Data processing included micro leveling using the tie lines and subtraction of the reference field defined by IGRF2005. A magnetic base station was established at DCPP in an area of low magnetic gradient free from cultural electrical and magnetic noise sources (Latitude 35.211513, Longitude -120.845162). Base station readings were recorded every second using an Overhauser GSM- 19 magnetometer.

Marine Magnetic Data Marine magnetic data were collected by the USGS in 2008 and 2009 (Sliter et al., 2009, revised 2010) using a Geometrics G882 cesium-vapor magnetometer that was towed approximately 30 m behind the research vessel R/VParke Snavely in 2008 and approximately 50 m behind the vessel in 2009. Magnetic data were collected at a 10-Hz sampling rate on a line spacing of 800 m along shore-perpendicular transects simultaneously with the mini sparker seismic-reflection data (see Appendix H).

Additional magnetic transects were run across segments of the Hosgri Fault Zone in 2008, resulting in a localized line spacing of 400 m. In 2009, an additional series of marine magnetics transects were conducted closer to shore between Pt. Buchon and Pt.

San Luis, resulting in a uniform line spacing of 400 m for the survey area. Seismic-reflection and marine magnetic data were also collected along three shore-parallel tie lines. See Figure D-5 for survey area. A magnetic base station was set up onshore in Morro Bay State Park on the south flank of Black Hill (Lat 35.35360 N., Long 120.8314' W.) in 2008 to continuously measure the local magnetic field in order to remove diurnal field variations from the offshore survey. In 2009, two base stations were established onshore during different portions of the survey, one in Morro Bay State Park and one.

near Avila beach in Wild Cherry Canyon (Lat 35.18700 N, Long 120.7467' W). See Figure D-6 for base station locations. Base-station readings were recorded every minute, using a Geometrics G856 proton-precession magnetometer.

Initial processing of the magnetic data involved smoothing, mainly to remove noise from the mini sparker source. The smoothed data resulted in approximate along-track data spacing of 10 m for lines where seismic-reflection data were also collected (boat speed, 4 knots), and 25 m along the magnetometer-only lines (boat speed, 10 knots). Further processing included removal of the diurnal field variation by using the magnetic-base-station readings and subtracting the reference field defined by the IGRF2005. Data were imported into the Geosoft Oasis montajTM geophysical software package, where survey lines were leveled by using the tie lines and then gridded at 200 m (Figure D-6). Marine magnetic data were subtracted from the fixed wing aeromagnetic data (Langenheim et al.,

2009a) to produce a residual magnetic, anomaly map dominated by magnetic sources in the shallow subsurface (< 500m below the seafloor (see Figure D-7)). Figure D-8 shows

.the merged onshore and offshore magnetic data. The marine data were filtered to effectively place those data at the same height as the regional aeromagnetic survey (Langenheim et al., 2009b). Figures D-7 and D-8 both show the dominant NW Shoreline Fault Zone Report, Appendix D Magnetics Page D-3 of 32

magnetic grain that is truncated by the Hosgri fault zone on the west and bounded by the Oceanic fault on the east.

2D Processing of Magnetic Field Data The magnetic intensity data from the helicopter survey were processed by Fugro Magnetic and Gravity Surveys, Inc. (Houston, Texas) with assistance from FWLA to aid in the interpretation of shallow magnetic features along the Shoreline fault zone (Fugro Gravity and Magnetics Services, 2010) . Since the magnetic inclination in this area is approximately 590, the total magnetic intensity anomaly field has anomalies that are not symmetrical and generally not centered over the causative magnetic source bodies or contacts between bodies of different magnetic susceptibilities. However, the inclination of 590 is considered to be fairly high latitude and therefore the total magnetic intensity anomalies for simple tabular shaped sources will be dipolar but with a larger positive peak than negative trough. This also means the differences in line azimuth in 2D modeling will not be as important as for low magnetic latitudes and inclinations. The objectives of the processing included positioning anomalies over their sources using reduction to pole (RTP) (Li, 2008) and highlighting lateral boundaries using tilt angle and horizontal gradients (Miller and Singh, 1994; Verduzco etal., 2004). To facilitate calculation of these derivative magnetic anomaly maps, the helicopter - total magnetic intensity anomaly data (Figure D-4) were re-gridded with a 30 m cell size (Figure D-9).

The 30-m cell size is about one fifth the line spacing and about as small as recommended.

The first processing step was the reduction-to-pole (RTP) transformation of the re-gridded data, which removes first-order magnetic field inclination effects to better position anomalies over their sources (Figure D-10). Reduction to the pole makes the simplifying assumption that the rocks in the survey area are all magnetized parallel to Earth's magnetic field. This is true in the case of rocks with an induced magnetization only, however remnant magnetization will not be correctly dealt with if the remnant direction is different to the direction of Earth's magnetic field. Thus the RTP transformation is associated with varying degrees of error and introduction of noise, such that only first-order features are robust. The primary change is that many of the RTP magnetic anomalies (Figure D-10) are narrower than the TMI anomalies (Figure D-9) and sometimes the position of peak anomaly amplitudes have shifted slightly compared to the total magnetic intensity anomalies.

There are factor of 10 differences in water depth within the span of the helicopter magnetic survey. The helicopter data acquisition elevation was small (50 m) relative to water depths that range from 10 m near the coast to over 100 m in the southwestern end of the helicopter survey area (Figure D- 11). Thus, it is important to consider the effects of bathymetry on the amplitude and width of the RTP magnetic anomalies and derivative maps. Figure D- 11 superimposes bathymetry contours on the RTP magnetic field map.

The Shoreline fault RTP anomaly is very narrow in the shallow (-10 m) water nearest to shore, but the peak anomaly amplitude decreases and the anomaly width increases by about a factor of three in 40 m of water (Figure D- 11). All the large-amplitude offshore RTP anomalies decrease in amplitude and broaden to the west as they trend toward water depths > 40 m (Figure D- 11) consistent with first-order effects water depth being on the Shoreline Fault Zone Report, Appendix D Magnetics Page D-4 of 32

I/

order of the helicopter flight elevation of 50 m. Calculations using a first-order vertical dipole approximation indicate that for the flight elevation of 50 m, RTP anomalies observed in 40 m of water should have widths about three times wider and amplitudes about 40% of RTP anomalies in areas with 10 m water depths. Hence, while the peaks of the RTP anomalies are useful to provide first-order estimates of the position of source bodies, the widths and amplitudes of the RTP anomalies as they trend toward deeper water should not be inferred to mean that the tops of source bodies are moving significantly below the seafloor, are becoming broader, or are necessarily starting to dip significantly away from vertical.

The RTP image was used to calculate vertical and horizontal gradients to produce a tilt angle anomaly map based on the arc tangent of the vertical-to-horizontal gradient ratio (Figure D-12). The tilt angle highlights the lateral extent of magnetic source bodies.

From Miller and Singh (1994), "The tilt angle may be used to detect the presence of sources of potential field anomalies and to provide information about their horizontal extent. It has the attractive property of being positive over the source, crosses through zero at, or near, the edge of a vertical sided source, and it is negative outside the source region. The tilt angle has the additional attribute, unique among the various potential field edge detectors, of responding equally as well to shallow and to deep sources and is, therefore, able to resolve the presence of subtle deeper sources which are often swamped in the larger responses of shallower sources."While these attributes of tilt-angle are generally true for level ground, the abrupt change in the seafloor elevations and slopes between the north and south sides of the Shoreline fault and the N40W fault have different effects on the vertical versus horizontal components of RTP total magnetic intensity. In particular, the bathymetry contours intersect the western Shoreline and N40W fault segments at shallow angles (Figure D-1 1). Most of the Shoreline fault closely follows a series of narrow prominent tilt-angle anomalies near the shore (Figure D- 12). However, the tilt-angle anomaly along the southwestern portions of the Shoreline and N40W faults may in part be smeared out laterally to the southwest (compare Figure D- 11 and D-12) due to quite different water depth characteristics across the deeper water portions of these faults.

The horizontal gradient of RTP results in positive peaks along the edges of blocks.

However, for thin steeply dipping sources a double positive anomaly results. The horizontal gradient results are important because the consistent double positive peaks along all the Shoreline fault segments where water depth is less than 40 m indicate that these portions of the source body along the Shoreline fault are near-vertical (Figure D-13). As the water depth becomes about 40 m or more, the horizontal gradient becomes more diffuse and lower amplitude (Figure D-13). The greater distance from the flight elevation of 50 m to the top of the source bodies as water depth increases broadens the horizontal anomaly and reduces its amplitude, resulting in a lower amplitude, broader horizontal gradient in the deeper water portions of Figure D- 13. This result confirms that the apparent broadening of the tilt-angle anomalies in regions with water depths > 40 m (Figure D- 14) is at least in part an artifact of the increased water depth. Thus, both the horizontal gradient (Figure D-13) and narrow tilt-angle anomaly regions in water shallower than 40 m (Figure D-14) suggest a source body that has a steep dip and is on Shoreline Fault Zone Report, Appendix D Magnetics Page D-5 of 32

the order of 200 m thick or less. There is a one-to-one spatial correlation of the position of the double positive peaks of horizontal gradient with the narrow tilt-angle anomalies (Figure D- 14) and a first-order one-to-one spatial correlation of the/tilt-angle anomalies with the Shoreline fault segments (Figure D-15).

The RTP, horizontal gradient, and tilt-angle maps are all consistent with a near-vertical source body associated with the Shoreline fault along all the near-shore segments of the fault (Figures D-10, D-14, and D-15.). The possible dips for the source bodies along portions of the Shoreline and N40W faults located in regions with water depths of 40 m or more are not as well constrained due to possible effects of deep water relative to the helicopter flight elevation of 50 m. However, even in the regions with water depths of 40-50 m, the horizontal gradient map shows double positive peak anomalies along the Shoreline and N40W faults (Figure D- 13), suggesting that the deeper water portions of these faults are steeply dipping. 2D modeling described below is used to further investigate magnetic constraints on Shoreline fault dip.

2D Modeling of Magnetic Anomaly Data 2D forward modeling of the magnetic anomaly data is used to evaluate alternative magnetic models to determine the first-order magnetic anomaly constraints on fault dips and magnetic basement depth. Modeling of the magnetic-intensity data from the combined marine, fixed wing, and high-resolution helicopter surveys were evaluated by Fugro Magnetic and Gravity Services (Houston, Texas) and Fugro William Lettis &

Associates to aid in the interpretation of prominent magnetic features. A single profile location was selected to incorporate the maximum amount of geologic and geophysical constraints available in the area while crossing the Hosgri and Shoreline faults and the central portion of the Irish Hills. Geologic mapping constrains surface geology and formation dips along the profile. Constraints on the structure of the Pismo syncline and Tertiary sediments are provided by proprietary seismic reflection profiles in the central Irish Hills. These seismic reflection profiles and the Honolulu-Tidewater well located in the central Irish Hills (PG&E, 1989b) provide constraints on the structure of the Pismo syncline at depth, including the minimum possible thickness of Tertiary sediments within the syncline along the 2D magnetic anomaly profile. The position of the profile was adjusted to incorporate the highest concentration of gravity measurements in the offshore and onshore sections of the profile.

The helicopter magnetic data being flown at lower elevation of 50 m than the aeromagnetic data (305 m) provide better resolution of shallow magnetic anomalies than the aeromagnetic data. Therefore, the higher resolution helicopter magnetic data was spliced-in into the portions of the profile where it was available for use in the 2D modeling. Total Magnetic Intensity (TMI) values were used for the modeling of the helicopter magnetic data and Residual Magnetic Intensity (RMI) values were used by Langenheim et al. (2009b) to model the marine and fixed wing magnetic data. As a result, there is a "DC" shift of 47,847 nT between these two data sets (TMI-IGRF).

Shoreline Fault Zone Report, Appendix D Magnetics Page D-6 of 32

The magnetic profile modeling is simplified if a single azimuth profile is compatible with modeling faults with disparate strike (Hosgri, Shoreline, and Los Osos faults). To determine if this was feasible a test model was built in order to determine the effect of a change in line azimuth from 008' to 0580 (clockwise rotation relative to north). The trial magnetic model includes a dyke-like feature or magnetization along a fault plane, and a lateral contact between strongly magnetized and non-magnetic basement (Figure D- 16).

The difference between the 0080 azimuth magnetic anomaly (Figure D-16) and the 0580 azimuth model (Figure D- 17) is significant in terms of absolute amplitude but the relative responses are similar. Therefore, a single strike for the entire line profile was deemed acceptable for 2D magnetic modeling.

A new geological cross section was compiled by FWLA for 2D model profile 1. The cross section was geo-referenced and digitized using FGMS 2D modeling software. The digitized cross section elements were then transformed into polygons for the initial model. Magnetic depth to basement analyses were performed for the vicinity of the 2D model Line 1 only using FGMS MagprobeTM software (Li, 2003). FWLA confirmed that magnetic basement in this area may be ophiolite or melange, and perhaps not that of typical granitic or metamorphic composition. A total of 31 forward models were evaluated and successively modified to produce a model that provided a reasonable fit to the magnetic and gravity anomaly data along 2D model profile 1 that was consistent with available geologic and geophysical data, including first-order basement depth estimates in the offshore region (Figure D- 18). Depth-to-basement estimates in the onshore region may be biased by shallow source bodies and complex shallow structure, along the south side of the Pismo syncline, so the onshore depth-to-basement estimates were not used as constraints in this part of the section (Figure D-18).

In forward model 31, the narrow anomaly just offshore that coincides with the mapped position of the Shoreline fault is reproduced using a steeply-dipping shallow source body (Figure D-18). To determine if other source body configurations could fit the magnetic anomaly profile equally well as series of alternative forward models were constructed that focused on the region near the Shoreline fault shown in forward model 31 (Figure D-18). Forward model 33 evaluated a thin, vertical source body that extended nearly to basement (Figure D- 19). The calculated magnetic anomaly does not fit as well due to the shorter predicted wavelength than observed near the Shoreline fault (Figure D- 19).

Forward model 34 used a thicker, 200-m wide vertical source body that produced a much better fit near the Shoreline fault (Figure D-20). A north-dipping source body near the Shoreline fault in forward model 35 produces a much poorer fit to the magnetic anomaly data, particularly on the south side of the fault and the source body has to be shifted south of the Shoreline fault to match the position of the peak anomaly amplitude (Figure D-2 1).

The south-dipping source body in forward model 36 produces an excellent fit to the magnetic anonialy, but requires moving the top of the source body 200 m north of the Shoreline fault (Figure D-22). Thus, only a near-vertical Shoreline fault source body configuration that is consistent with both 2D modeling profile 1 (Figures D- 18 and D-20a and b), the RTP map (Figure D-10), the horizontal gradient map (Figure D-13), and the tilt-angle map (Figure D- 15). Significant north or south dips requiring displacement the top of the source body several hundred meters from the mapped trace of the Shoreline Shoreline Fault Zone Report, Appendix D Magnetics Page D-7 of 32

fault (Figures D-21 and AD-22). The 2D profile modeling shows that the source body along the Shoreline fault need not extend more than 200-300 m below the surface (Figure D- 18), but can extend to at least 4 km and fit the magnetic anomaly data equally well (Figure D-20a and b). Thus, the magnetic data indicate that the source body associated with the Shoreline fault is most likely to be nearly vertical, with a width on the order of 200 m, and extends from near-surface to a depth of several hundred meters to several km below the surface.

References Beyer, L.A. and McCulloch, D.S., 1988, Free-air gravity anomaly map of the offshore Santa Maria Basin, U.S. Geol. Surv. Open-FileRep 89-322 1 map, scale 1:125,000 Briggs, I.C., 1974, Machine contouring using minimum curvature: Geophysics, 39, 39-48.

Chapman, R.H., Beyer, L.A., and Youngs, L.G., 1989, Bouguer gravity and magnetic anomaly map of the south-central California Continental margin (Map 4c) , in Greene H.G. and Kennedy, M.P., eds., California Continental Margin Geologic Map Series: South-Central California continental margin- Area 4 of 7, California Division of Mines and Geology, Scale 1:250,000 Fugro Gravity and Magnetics Services, 2010, PG&E California Gravity and Magnetic Data Modeling, Interpretation Interim Report, 53 pp.

Langenheim, V.E., Jachens, R.C, and Moussaoui, K., 2009a, Aeromagnetic Survey Map of the Central California Coast Ranges, USGS Open File Report 2009-1044 http://pubs.usgs.izov/of/2009/1044/

Langenheim, V.E., Jachens, R.C., Graymer, R.W. and Wentworth, C.M., 2009b, Constraints on 3-Dimensional structure of the central California Coast Ranges from gravity and magnetic data (abs), Seism. Res. Lett., 80, 323 Li, X., 2003, On the use of different methods for estimating magnetic depth: The Leading Edge, v. 22, no. 11, p. 1090-1099 Li, X., 2008, Magnetic reduction-to-the-pole at low latitudes: observations and considerations, Practical considerations: The Leading Edge, v. , no. , 990-1002 McCulloch, D.S. and Chapman, R.H., 1977, Maps showing residual magnetic intensity along the California coast, Lat. 37' 30'N to 340 30'N, US Geological Survey Open File Rpt. 77-79, 14 aeromagnetic maps, scale 1:250,000 Miller, H. G., and V. Singh, 1994, Potential field tilt - a new concept for location of potential field sources: Journal of Applied Geophysics, v. 32, p. 213-217.

Shoreline Fault Zone Report, Appendix D Magnetics Page D-8 of 32

New Sense Geophysics, 2010, Point Buchon: Helicopter aeromagnetic geophysical survey for Pacific Gas and Electric Company, December 2 nd - December 5 th, 2009.

PG&E No. 2500257867, NSG No. HM191015, Toronto, Canada, 40 p.

Pacific Gas and Electric Company (PG&E), 1989a, Response to question Q43m-2, in Response to questions 43k, 431, 43m, 43n, 43o, and 43p, Gravity and magnetic contour maps, U.S. Nuclear Regulatory Commission Document No. 50-275 and No. 50-323, January 1989.'

Pacific Gas and Electric Company (PG&E), 1989b, Response to question Q43m-2, in Response to questions 43r, 43s, 43t, 43u, 43v, and 43w, Drill logs and core descriptions Honolulu Tidewater USL Heller Lease well, U.S. Nuclear Regulatory Commission Document No. 50-275 and No. 50-323, January 1989.

Sliter, R.W., Triezenberg, P.J., Hart, P.E., Watt, J.T., Johnson, S.Y., and Scheirer, D.S.,

2009, revised 2010. High-resolution seismic reflection and marine magnetic data along the Hosgri Fault Zone, Central California, USGS Open File Report 2009-1100, version 1.1 http://pubs.usgs.gov/of/2009/1100 /

Verduzco, B., J. D. Fairhead, C. M. Green, C. Mackenzie, 2004, New insights into magnetic derivatives for structural mapping: The Leading Edge, v. 23, no. 2, p. 116-119.

Watt, J.T., Johnson, S.Y., Langenheim, V.E., Scheirer, D.S., Rosenberg, L.I., Graymer, R.W., and Kvitek, R., 2009, Geologic mapping in the Central California coastal zone: integrating geology, geophysics, and geomorphology: Geological Society of America Abstracts with Programs, v. 41, n. 7, p. 283.

Shoreline Fault Zone Report, Appendix D Magnetics Page D-9 of 32

0 L.

Dt 0N Taft x

E r-A C3 NabonalFore CL Vo7 ......-... o-o't of CD aT.

E02008 C.,

aeromnagnetic survey tracklines; along

~the central California Coast Ranges

~(Langenheimn et al., 2009)

SHORELINE FAULT STUDY 11/22/2010

= ~Pacific Gas and Electric Company D-Shoreline Fault Zone Report, Appendix D Magn Page D-1 1 oft32

co 4) 04N . .

IDK CLJ 0

- Aeromagnetic Survey Map of the central California Coast Ranges 20 V_ LANGiEM1IA. FtQ .ACHENW. and K MCUSSA.MOP

'survey Aeromagnetic.. map. of th co 00 Vcentral California Coast Ranges CD(Langenheim et al., 2009)

Co SHORELINE FAULT STUDY 11/22/2010 0.

__ Pacific Gas and Electric Company Figure D-3 Shoreline Fault Zone Report, Appendix D Magnetics Page D-12o7 32

0 0 Notes: Blue lines represent ship tracks where high resolution marine seismic reflection and magnetic data were collected at 800 m spacing. Red lines are additional marine magnetics ship tracks for a net 400 m spacing (Watt et al., 2009).

Track line map of marine geophysical data collected in 2008 and 2009.

o N 51 co 2 2ý5

4) -16 3-25 o -32 N -38

,i- -44

-44

.58 a -53 o -57

-62 0-66 ~-72

= -80 U. -90

~-102 -128 Q Magnetic Anomalies nanotestas o 10,000 0 10,000 NWGS S4fW~a*zoneIO

• Magnetic Base Station

-12115' -120*50' q

CD c-SMagnetic anomaly map of the 2008-2009 marine survey area (Sliter et al., 2009)

SHORELINE FAULT STUDY 11122/2010 l-iPacific Gas and Electric Company Figure D-6 Shoreline Fault Zone Report, Appendix D Magnetics Page D-15 of 32

N o \62 (6 52 47 D 45 43 41

-N 40 0 338 0 37 35

\ . .,\34 R-a -"NI /

LL 31 29 /

Res dual nomalies N Re a mgMagnetic m nanote is 102000 (0treal0

.9215 120*50'

• LL 00 M

SO Rsdald magetric Cmanomlya

• SHORELINE FAULT STUDY 11/22/2010 oPacific Gas allow susrace Figure D-7 (U

Shoreline Fault Zone Report, Appendix D Magnetics Page D- 16 ot 3 2

Notes: Marine magnetic data were filtered to effectively place those data at the same height as the Merged aeromagnetic aeromagnetic data (- 305 m) and marine magnetics data (Langenheim et al., 2009)

SHORELINE FAULT STUDY 11/2212010 F Pacific Gas and Electric Company Figure D-8 ihoreline Fault zone Keport, Appendix 1) Magnetics Page D-L/ ot 32

-12 -12060' -120'4S 690000 695000 700000 705000 I '

I II

-9° Tota Magei Intensity 0

0 0

t==

C/)

=

LL. 0 Scae 1.100000 o N

0 Cd Cd,

~690000 ei6950 700 70500

... .0L . .

CoHelicopter total magnetic intensity

~re-gridded to 30 m spacing SHORELINE FAULT STUDY 11/22/2010 LL RIPacific Gas and Electric Company e D-9 Shoreline Fault Zone Report, Appendix D Magnetics Page 13-18 of 32

12V55" -12015a *12o045' 690000 695000 700000 705000 P.

690000 705000

-12(a5 -120'5Vf - 120'4*

Note: The approximate region of water depths of 40 m or more is southwest of the short-dashed white line Reduction to Pole Transformation of the Total Magnetic Intensity Map SHORELINE FAULT STUDY 11/30/2010 R Pacific Gas and Electric CompanyPag Figure D-1 0 Shoelie.Fult .oe..por..ApedixD....e.c ... t...

Shoreline Fault Zone Report, Appendix D Magnetics Page D- 19 of 32

1*=~MO

• iThor 1IUM depth 4)

(0 c.lJ Width for 40 m depth 4)

'0 cC)

CL 4)

• = Notes: The yellow double-arrow on the -40 m contour illustrates that the Shoreline fault anomaly width is about three times (smaller wider in the yellow regionarrow double of 40asmindicated).

water depthThethan thedashed white anomalycircle widthoutlines in 10 mthe of region water near the shore dominated by the

_dipole response to the DCPP.

C-c'T 0

MBathymetry in m (negative to denote depth below sea level) as white contours con the RTP field map SHORELINE FAULT STUDY 11/22/2010 CU 0.

4) Pacific Gas and Electric Company Figure D-11

.. re. n. . ul.. ne. e .A.e n .. .x . et.ag c - Pa ge..

I J-. . . . ..

Shoreline Fault Zone Report, Appen jx D Magnetics Page D-20 of 32

-1205w 700000

-120"55'

-12055 .120o50, 8900OO 695000 696000 700000

-A.-

li

( degrees)

TI* Angle ofRTP K -0

/ SO 08 CO 01 U,

0o Go -- Fi*.

bn~

00 Scale 1:100000

. . ..0 0

1000 0 1000 2000 3000 0,SSa41UyUM MVN 695000 700000 7 05000

-120*W' -1205W- 120'45

__j Tilt Angle Magnetic Anaomaly Map based on 2009 Helicopter RTP Anomaly Map SHORELINE FAULT STUDY 11/3012010 1Pacific Gas and Electric Company Figure D-12 Shoreline Fault Zone Report, Appendix D Magnetics Page D-21 of 32

-16900 -120'50 -120*45' 69OOO 69%000 700000 705000 1 T CJ 0

a, 7/

DCPP dipole 0

0 0

Horizontal Gradieht of RTP 0o C.

00

/ 20 Model Profile I 02 W

00 Shoreline Fault Double Peaks If'i Scale 1:100000 w Co 1000 0 1000 20OO 3000 (0 0

0 (mS Jr IO 0 M~ 0 690000 69*50 700000 . . 705000

-120W55 -120*50' -120*45' Notes: The approximate region of water depths of 40 m or more is southwest of the short-dashed white line Horizontal Gradient of Helicopter Reduction to Pole Anomaly Map SHORELINE FAULT STUDY 11130/2010 oe Sfoeie1al eot pedxVMgntc T Pacific Gas and Electric Company aeV2 Figure 13 D-13 shore ine Fault Zone Keport, Appendix J-)magnetics Fage V-22 of :32

-12055 -120 -12("45' 690000 695000 700000 705000 oo

• i ODCPP dipole 0 j0 2D Mod, Profie 1 0 7ii oto 0P Shoreline Fault Double Peaks

• -- In Horizont~al Gradient 0 >0 CL Scale 1:e100000 2 od Pi 6 0 1000 0 1000 2000 3M0 to CL r 0(MNNS) Cýoo 0690000 695000 700000 705000

-120*55' -1;20"0 -120"4W Notes: from 40 mshallower bathymetry water meters as the short-dashed white line separating deeper water to the southwest to thein northeast.

contour aprxmtc4 ahmer otu Grnd a x E 4 ri C ariyzonta 0Pacific Gas and Figure D-14 Shoreline Fault Zone Report, Appendix D Magnetics Page D-23 of 32

0 AL ýý 1 74 LEGEND Offshore faults from this study (2010) 21 2010 Helicopter Magnetic 1 Anomaly Survey

-21 3 lt angle of RTP 5 is shown in degrees Map scale: 1:60,000 Map projection: NAD 1983, UTM Zone 10 North 0 1 2 Miles Kiometers 0 1 2 Tilt-angle of RTP superimposed on terrain with offshore faults 1~ww

._4 0*

-o U--

1.

C, 7

N 0

4) 03 0.

co Test magneticComan model with line 6 azimuth of 0080 C.,

SHORELINE FAULT STUDY 11/2212010

-US RPacific Gas and Electic Company Figure D-16 Sfoein altLn-Kpr,-pedx )Mgntc Pg -i t3 Nhore me Fault Zone Keport, Appendix 1) magnetics Page 1)-25 of -32

04 JJ, 'N cL 4) a.

U-O 01

• q co 0

Test magnetic model with line azimuth of 0580 Cn SHORELINE FAULT STUDY 11/22/2010 0 Pacific Gas and Electric Company Figure D-1 7 Shoreline Fault Zone Report, Appendix 1) M Page D-26 ot.32

L:

0z 0

W Ca 0i '0 C)

Co geophysiAl cosrintsW 0

04 D' Notes: ~~~~~~~~Forward andgrvt(mdl)aoayft(edcvs)oobredaa(lucre)

Modelingofmagnetic(tp 0

C, Co Co 0.

oP IIW1Pcficrwasrad Modeting Cofmpany ti an Shoreline Fault Zone Report, Appendix D Magnetics Page D-27 of 32

Co 03 1ý-T Cu cc LL CD 07) 0X 0) 0.

04

~Magnetic sensitivity test with a thin, U.. ~vertical Shoreline fault magnetic body cD(model 33) cSHORELINE FAULT STUDY 11/22/2010 SPacific Gas and Electric Cmay Figure D-1 9 Shoreline Fault Zone Report, Appendix D Magnetics Page D-28 of 32

U21 Cnu ot~

tL 16rica Jhrln Lkal 34dl manei bodyo 4 M0IM.

mi U0&

a, DRA SN INN"mpany 0L 0I Magnetic sensitivity test with a 200-rn-thick vertical Shoreline fault magnetic body (model 34)

U, SHORELINE FAULT ZONE STUDY Pacific Gas and Electric CompanyTFigure D- 20a Shoreline Fault Zone Report, Appendix D Magnetics Page D-29 of 32

Magnetic sensitivity test with a 200-m-thick vertical Shoreline fault magnetic body (model 34)

SHORELINE FAULT STUDY Figure D - 20b I" Pacific Gas and Electric Company Shoreline Fault Zone Report, Appendix D Magnetics Page D-30 of 32

N 0

ca W

C"4 C>

L:

LL I

r*

0 CLJ 0

0 C>

ro Magnetic sensitivity test with a 200-rn-thick north-dipping Shoreline fault C1 magnetic body (model 35)

SHORELINE FAULT STUDY 11/22/2010 CL S)_Pacific U_1_ - Gas and Electric Company Figure D-21 Shoreline Fault Zone Report, Appendix 1) Magnetics Page D-31 ot32

N C14 CLA a.

CL C14 n,

q Co M

rMagnetic sensitivity test with a 200-m-thick south-dipping Shoreline fault magnetic body (model 36)

SHORELINE FAULT STUDY 11122/2010 PcfcGas and Electric Company Figure D-22 ir________________________7_Pa Shoreline Fault Zone Report, Appendix D Magnetics Page D-32 of 32