Problems for probabilistic seismic hazard analysis

The use of b-values derived from the Gutenberg-Richter relationship as a phenomenological base for developing probabilistic seismic hazard analyses (PSHA) has been questioned for years. The relationship is still used because political demands require something for PSHA, one variable is easy to deal with, and no persuasive alternative has come forward. Using cumulative distribution probability plots, it can be shown that seismic magnitude-frequency data can be well described as one or more populations, each of which is normally distributed with respect to magnitude. This holds true for large earthquakes when sorted by mechanism, for earthquakes >400 lan deep, for the general USGS NEIC catalog, for the Harvard CMT catalog, for the CERI catalog of the New Madrid Zone, and for a Scandinavian catalog. In all instances, multiple normal populations provide a better fit to the data than does the Gutenberg-Richter relationship. Use of these multiple populations in PSHA emphasizes that the scientifically sound limits of magnitude projection are within the 4 sigma limit of the largest populations. Such graphs may make it easier to resist political requirement to extrapolate into scientifically unsound regions.