Criticism of some forecasts of the National Earthquake Prediction Evaluation Council

Abstract

The Working Group on California Earthquake Probabilities has assigned probabilities for rupture in the interval from 1988 to 2018 to various segments of the San Andreas fault on the basis of the lognormal distribution of recurrence times of characteristic earthquakes postulated by Nishenko and Buland (1987). I question the validity of those probabilities on the basis of three separate arguments: (1) The distributions of recurrence times of the four, best-observed, characteristic-earthquake sequences are each only marginally consistent with the Nishenko - Buland Iognormal distribution. (2) The range of possible 30-year conditional probabilities for many of the fault segments is so great due to uncertainty in the average recurrence time for that segment that the assigned probability is virtually meaningless. (3) The 1988 forecasts not subject to the foregoing objection are those in which there is a low probability of an earthquake in the near future (e.g., only a 5 per cent chance of rupture of the North Coast segment before the year 2049 and of the Carrizo segment before the year 2018). The same reasoning would assign only a 5 per cent chance of rupture before mid-1993 to the southern Santa Cruz Mountains segment, the segment that failed in October 1989. Finally, the forecast of the next Parkfield earthquake (95 per cent probability before 1993.0) by Bakun and Lindh (1985) depends upon an ad hoc explanation of the out-of-sequence 1934 earthquake. A less-contrived forecast would have assigned a conditional probability of about 60 ± 20 per cent to the 1985.0 to 1993.0 interval and 30 ± 15 per cent to the 1990.0 to 1993.0 interval.