Component correlations in structure-specific seismic loss estimation
This paper addresses correlations between multiple components in structure-specific seismic loss estimation. To date, the consideration of such correlations has been limited by methodological tractability, increased computational demand, and a paucity of data for their computation. The effect of component correlations, which arises in various forms, is however a significant factor affecting the results of structurespecific seismic loss estimation and therefore it is prudent that adequate consideration be given to their effect. This paper provides the details of a tractable and computationally efficient seismic loss estimation methodology in which correlations can be considered. Methods to determine the necessary correlations are discussed, particularly those that can be used in the absence of sufficient empirical data, for which values are suggested based on the judgement. The effects of various assumptions regarding correlations are illustrated via application to a case-study office structure. It is observed that certain correlation assumptions can lead to errors in excess of 50% in the lognormal standard deviation in the loss given intensity and loss hazard relationships, while full consideration of partial correlations is 50 times more computationally expensive than other assumptions.