PART II— DYNAMICS OF THE SOLID EARTH 



increase the likelihood of a shock 

 occurring in a neighboring region 

 over a time-scale of several years. 



Organization is an auto-correlation 

 problem — i.e., one must search for 

 predictive elements in the time-series 

 of shocks for a given region within 

 the series itself, without benefit of 

 comparison with other time-series. 

 Although a given earthquake catalog 

 does not appear to be wholly random, 

 the "signal-to-noise" ratio is small. 

 The differences from randomness are 

 small, and the problems of extracting 

 the organized part from the random 

 part has not yet been solved. 



The Limits of Prediction 



dictive capabilities for the second and 

 succeeding large shocks, after the 

 next one has occurred, will be much 

 better on all accounts, for we will 

 then know what we are looking for. 



In Stable Regions — Problems of 

 prediction are difficult enough in 

 regions of high activity such as the 

 circum-Pacific belt. They are almost 

 impossible in regions with little or no 

 history of seismicity. For example, 

 the region from the Rocky Mountains 

 to the Atlantic Coast is supposed to 

 be stable; yet two of the greatest 

 earthquakes in U.S. history occurred 

 east of the Rockies. Destructive 

 earthquakes of record occurred in 

 southeastern Missouri in 1811 (the 

 shock was felt over an area of two 



million square miles; it relocated the 

 Mississippi River) and near Charles- 

 ton, South Carolina, in 1886. 



Seismic-risk studies show the New 

 York area to have a hazard roughly 

 100 times smaller than southern Cali- 

 fornia. Does this mean that the larg- 

 est shocks on the southern California 

 scale would recur in the New York 

 area at an interval of 10,000 years? 

 Or are the largest possible shocks for 

 the New York area less than the larg- 

 est for southern California? The 1811 

 and 1886 experiences show that stable 

 regions are not immune. But we still 

 have no way of determining where in 

 stable United States a great earth- 

 quake is likely to occur — or when. 

 (See Figure II-7) 



In Active Areas — All three meth- 

 ods of prediction in active areas share 

 one major difficulty: even in Cali- 

 fornia, where most U.S. activity in 

 prediction research is concentrated, 

 the rate of occurrence of truly large 

 shocks is small. 



We have not had a great earth- 

 quake in California since careful seis- 

 mological records began to be kept. 

 The three great historical shocks — 

 San Francisco (1906), Lone Pine, or 

 Owens Valley (1872), and Fort Tejon, 

 near Los Angeles (1857) — all oc- 

 curred in earlier times. The historical 

 method postulates that the order of 

 small- and intermediate-sized shocks 

 can be used to predict when large 

 shocks will occur. However, seis- 

 mologists do not really know what 

 they are looking for, since a large 

 shock has not taken place in the 

 modern era of California seismology. 



The same criticism applies to the 

 other two methods. In the search for 

 premonitory effects and the measure- 

 ment of in situ stress, the presumption 

 is that the anomalous, or critical, states 

 will be obtained for the large shock 

 by studying these states for the small 

 or intermediate shocks. Whether this 

 is correct or not will be seen after 

 the next large shock. Indeed, our pre- 



Figure 11-7— SEISMIC RISK IN THE UNITED STATES 



J NO DAMAGE 



MINOR DAMAGE 

 INTENSITIES V AND VI 



MODERATE DAMAGE 

 INTENSITY OF VII 



MAJOR DAMAGE 



INTENSITY OF VIII AND HIGHER 



This figure delineates the areas where earthquakes have occurred and have caused 

 damage within the United States. The range is from areas of no damage in southern 

 Texas and Florida to areas of major damage such as the western coast of California. 

 Intensities are measured from to 8 in terms of the Richter scale. 



38 



