Table 2: The plate vector test above demonstrates that the NCEP model is a good match to independent estimates of strain that can potentially accumulate in the system. A more rigorous extension of this test is to sum the total seismic moment rate included in the model for the entire San Andreas fault system (Tables 3 and 4). The result is 6.8 x 10¹⁸ Nmyr^-1. For a simple test of this result, we postulate a single straight fault that runs the length of our modelled region as shown in Figure 12. This hypothetical structure extends 576 km from the north end of the 1966 Parkfield rupture (PKF) to the Mendocino triple junction (MTJ) and has the 12 km down-dip fault width typical of the northern Coast Range. Using the 39 mm/yr VLBI motion yields 8.1 x 10¹⁸ Nmyr^-1, but if one subtracts loss of seismic moment equivalent to the aseismic, 100-km-long creeping section of the San Andreas the result is6.8 x 10¹⁸ Nmyr^-1, identical to the sum of moment rate for the NCEP model.

Comparison of NCEP model to historical seismic moment rate

Because the historical period (~1850-present) is short (145 yr) compared to the length of a major regional seismic cycle (recurrence time of 1906 San Andreas fault event ~210 or ~250 yr), we cannot assume that the flux of seismic moment is uniform enough to test our model. Nevertheless, we have summed the historic moments for the San Andreas fault system as shown in Figures Figure 12 and Figure 13. Despite the incomplete representation of the entire seismic cycle in the historical record the regional total for seismic moment rate has been 6.8 x 10¹⁸ Nmyr^-1. This match of the net historic seismic moment rate to both the sum of moment rate for the entire NCEP model, and to the sum derived from the simplified plate boundary test above is either a remarkable coincidence, or it may suggest some spatial and temporal partitioning of slip release in the historic period. Most of the Hayward and Calaveras fault subsystems have accounted for much less moment rate in the historic period than the San Andreas-San Gregorio subsystem. Possibly these lesser faults release strain earlier in the seismic cycle in preparation for the great San Andreas earthquakes. The historic record permits this possibility and future paleoseismological investigations could be initiated to test this idea. Great Valley thrust earthquakes also match the historic seismicity rates well, perhaps because they are relatively smaller and regionally-speaking are more frequent than events on the larger faults. On the northeastern faults the historic rate of earthquakes has been low by a factor of 2 or more compared to our model. Most historical moment release in this region occurred between 1857-1887, demonstrating the highly episodic character of large earthquakes in the region.

Figure 13. Moment Rate of Model vs Historical Earthquakes. Only 50-70% of main seismic cycle has occurred in the historical period

CONCLUSIONS

We have summed the overall seismic moment rates of all potential sources of large earthquakes in the NCEP model for the San Andreas fault system in northern California. We have compared this sum to essentially independent data: global tectonic models, VLBI observations, regional crustal strain observations, and the historical earthquake record (Table 4 and preceding discussion). We find the close agreement of all of these diverse data to be a reassuring quality check for the needs of regional seismic hazard mapping.

Although we believe the model is internally consistent and fully adequate for the present purpose given our present state of knowledge, many useful avenues of research remain to be explored to improve future hazard maps and other forms of hazard mitigation such as forecasting future events. Of particular note is a continuing need for the extension of the geodetic monitoring network into the more remote regions of high hazard that are still poorly known such as the northern Coast Range and the Modoc Plateau. The distribution of creeping and locked behavior is virtually unknown for several fault segments in the northern Coast Range, thus, monitoring the region could have important impact on understanding both the local hazards and the San Andreas fault system as a whole. Many more paleoseismological studies are required along the major strike-slip faults in the urban areas to improve forecasting of future events. Paleoseismic work is also needed in the North Coast region to clarify system-wide issues of great seismic cycles and for a detailed understanding of each fault subsystem. Monitoring of microseismicity remains an essential tool both for delineating potential earthquake sources on poorly-mapped active faults and as a means of testing current ideas about the migration of stress changes in such complex tectonic regimes as the northern Coast Range.