Comparison of Seismological Stations

Abstract

In the preparation of the 1940 tables there was originally much difficulty on account of widely differing standards of accuracy among stations, and this was mitigated by means of a ratio called the reliability (Jeffreys 1936). Stations were classified according to their P residuals after a solution had been made; the reliability is the ratio of the numbers in the range +4s to the whole. A station was used in further approximations if its reliability was 0.8 or more. The object at that stage was not to criticize bad stations but to select a set that could be said to be known to be good. Some that had appeared to be second-rate were found later to be first-rate when distances were recomputed with geocentric latitudes (Jeffreys 1937a). When the method of uniform reduction was applied the values o f 0 for the first-rate and secondrate stations were nearly equal; the difference was in p, which represents the trains of outlying observations. Many stations have since been greatly improved and the original reliabilities are mostly superseded. Some systematic differences appeared to exist even among the first-rate stations; the most important was that P beyond 90 and P K P tended to be read late on horizontal components. There was little sign of personal error or of relation with height above sea-level (Jeffreys 1936, 1937b). In further work systematic differences have mostly been ignored; at most they are of the same order as the standard errors of one observation and would mostly be randomized when earthquakes with different epicentres were combined. But they are now demanding attention. Allowance for height above sea level has usually been ignored. For vertical travel the direct effect would be about 0.23 s per kilometre. But there should be a further effect of.isostasy. If we treat the upper layers as of density 2.7 and the lower 3.3, the interface will be depressed by 4.5 km per km of height, and the time of vertical traved will be increased by about 0.45 s per km of height. For a continent we take a thickness of 30 km of material of density 2.7; for the ocean we may take roughly an upper layer of depth 15 km of density 3.0 instead of 30 km of density 2.7. With allowance for the weight of the water the elevation of the interface would be about 10 km. Then the difference in the times of vertical travel to sea level (on an oceanic island where the thickness would have to be increased by 5 km) would be about