Regression-based scenario earthquake selection for regional hazard-consistent risk assessments

Conventional probabilistic seismic hazard analysis (PSHA) is often repeated at many locations independently to develop uniform hazard maps. However, such maps are unsuitable for assessing risk to spatially distributed infrastructure because no single event will produce uniform hazard shaking intensities across a broad region. A robust but computationally expensive approach is to analyze spatially distributed infrastructure systems separately for every event considered in the seismic source characterization model used in the PSHA. This approach may not be practical when many scenario events are considered. An alternative is to select a manageable event subset that, in aggregate, approximately matches the hazard for single or multiple ground motion intensity measures across the spatially distributed system preserving contributions of different magnitudes and distances to the PSHA. We present a flexible and efficient regression-based method that meets these requirements using point-based PSHA results as inputs. The approach is illustrated with a case study of distributed infrastructure in southern California. We demonstrate the efficiency of the method by comparing it to a mixed-integer linear optimization method from the literature.