To develop a quantitative earthquake forecasting model, we experimentally estimated frictional parameters at a model plate interface during an earthquake generation cycle using realistic synthetic observational data. We used a 2-D fault model based on a rate- and state-dependent friction law to generate displacement data at observation points located above the fault plane during the latter half of an interseismic period, when fault slip is stable and slip rate is nearly constant. We estimated the likelihood of friction law parameters using sequential importance sampling (a particle filter). We accurately determined
A — B, which represents the rate dependence of steady-state frictional stress, for an aseismic slip area around a circular seismic slip area using typical noise levels and observation point spacings. However, the characteristic slip distance
dc we estimated for the aseismic slip area contained large estimation errors. We found that the likelihood distribution depended on the noise level of the data rather than on the interval between observation points. Moreover, we found two ranges of
A — B, one with high likelihood and the other with low likelihood, even when the noise level was too high or the data period was too short for detailed estimation. The likelihood data provided little information for determination of
dc, indicating that aseismic sliding during the interseismic period is much more sensitive to
A — B than to
dc. These results indicate that interseismic displacement data are useful for estimating friction parameters, especially
A — B values.
View full abstract