Methodology for Evaluation of Characteristic Earthquake Models Using Paleoseismic Measurements of Fault Slip from Sites with Multiple Earthquakes

Monday, June 10, 2002 - 1:30pm - 2:20pm
Keller 3-180
Norm Abrahamson (Pacific Gas and Electric)
The main difficulty in developing earthquake recurrence models is that the historical observation period is short compared to the time scale of recurrence of large magnitude earthquakes. This leaves us two choices: trade space for time by combining observations from analogous regions around the world, or go back in time to increase the number of observations for a particular region (or fault). One method for going back in time is to use geological evidence of past earthquakes. Paleoseismic measurements are made by digging trenches across faults and looking for evidence of fault slip in past earthquakes. The measurements include the amount of slip and/or the date of the earthquake. In this study, we will only use data from the amplitude of the slip from past earthquakes.

To evaluate the hypothesis that faults have characteristic earthquake magnitudes, ideally, we would have multiple observations of large magnitude earthquakes on a specific fault. Using paleoseismic observations, we can look at sites (points along a specific fault) for which multiple past earthquakes can be observed. These observations have a small coefficients of variation (COV) of about 0.3 that implies highly characteristic behavior of the fault slip at a point. For comparison, an exponential distribution of earthquake magnitudes (with a b-value of 1.0) and a global model of slip vs magnitude (e.g. Wells and Coppersmith, 1994) leads to a COV for slip at a point of 0.9 to 1.0. Using a normal distribution of earthquake magnitudes on the fault (e.g. a characteristic earthquake model) the COV for slip at a point is 0.7 to 0.8.

The much smaller COV values from the paleoseismic data suggest that not only is the earthquake magnitude characteristic, but also the distribution of slip along the fault is highly characteristic.

There are several difficulties with the paleoseismic data that may tend to lead to underestimates of the COV. These include: (1) the number of observations at a point is still small (2 to 10); (2) the small slip events are not observed because they are below a detection threshold; (3) large slip events may mask earlier smaller slip events (even if they were above the detection threshold); (4) measurements are usually made at sites with the largest slips because they give the best measurements which may not be representative of the entire fault; (5) the slip is measured at a single site and smaller magnitude events may not rupture past the site, so they don't capture the range of magnitudes on the entire fault. Statistical methods for dealing with these features of the data sets will be discussed.