KERNEL ESTIMATION METHODS FOR SEISMIC HAZARD AREA SOURCE MODELING

In probabilistic seismic hazard analysis, the representation of seismi c sources by area zones isa standard means of data reduction. However, where the association between seismicity and geology is complex, as i t is in many tectonic regimes, the construction of zone geometry may b ecome contentiously subjective, and ambiguities may end up being resol ved through appeal to the nonscientific rule of conservatism or pragma tism. Although consideration of alternative zonations within a logic-t ree framework provides a channel for some of the uncertainty, it does not address the fundamental validity of the zonation procedure itself. In particular, neither the minimal assumption of uniform seismicity w ithin a zone nor the Euclidean geometry of a zone accord with the frac tal spatial distribution of seismicity, and the magnitude insensitivit y of zonation ignores the spatial extent and correlations of different -sized earthquakes. An alternative procedure for area source modeling avoids Euclidean zones and is based statistically on kernel estimation of the activity rate density inferred from the regional earthquake ca talog. The form of kernel is governed by the concepts of fractal geome try and self-organized criticality, with the bandwidth scaling accordi ng to magnitude. In contrast with zonal models for intraplate regions, the kernel estimation methodology makes provision for moderate earthq uakes to cluster spatially, while larger events may migrate over sizea ble distances.