Risk assessment of building inventories exposed to large scale natural hazards.

摘要

Earthquakes, among the most devastating and unpredictable of natural hazards that affect civil infrastructure, have the potential for causing numerous casualties and significant economic losses over large areas. For example, the Mw 7.9 Sichuan, China Earthquake in 2008 caused at least 69,000 deaths, collapses of over 5 million buildings, and damage to more than 21 million buildings. While the Mw 8.8 Chile Earthquake in 2010 caused at least 500 deaths, at least 380,000 buildings were damaged or destroyed by the earthquake and tsunami that followed. The total estimated economic losses for the Sichuan and Chile earthquakes were ;Risk to distributed building inventories exposed to earthquake hazards is generally determined using one of two basic approaches: scenario earthquake risk assessment (SERA) and probabilistic seismic risk assessment (PSRA). The latter approach is required if the probable maximum loss (PML) decision metric, which often is used for underwriting purposes in the insurance industry, is to be estimated. The uncertainty of the SERA and PSRA, and in the estimate of the PML for building inventories, is dependent on correlation in building losses within the inventory, which is caused by spatial correlations in ground motion (demand), building response and damage. Previous studies that neglect these correlations underestimate the uncertainty in losses. Moreover, while correlation in demand has been considered in a few studies, response and damage correlations due to common building practices and occupancy characteristics have yet to be investigated.;This study proposes a new model for estimating correlations in response and damage for structural and nonstructural components within a building, and for structural and structural damage of building inventories. The total building inventory loss for a SERA, probabilities of loss exceedence (risk curves) for PSRA, and the PML decision metric are determined and the effect of correlations in demand and damage is included in the analyses. Since estimates of losses for a large building inventory requires numerous computational and numerical efforts, a sampling technique is developed that provides sufficiently accurate estimates. The sensitivity of the total structural loss and the combination of nonstructural and structural losses due to parameters that were assumed in modeling spatial correlation is investigated, and the role of epistemic uncertainty in the total structural losses and PML are examined. It is concluded that estimates of losses to building inventories made under the common assumption that the individual losses can be treated as statistically independent may underestimate the PML by a factor of range from 1.7 to 3.0, depending on which structural and nonstructural elements are included in the assessment.