The subject of this special issue is seismic reliability analysis, which considers (at least implicitly) the probabilistic aspects of seismic events and ground motion, as well as structural dynamic response and capacity. The 'probabilistic elements of seismic events' include the when, where, and how large of future earthquakes in the region of the site of the structure. 'Ground motion' deals with subjects such as the relative frequency content, phasing, and temporal variations of accelerograms over a range of general amplitude levels. 'Structural dynamic response and capacity' focuses on the more extreme nonlinear response range where damage and even collapse may occur. The seismic reliability problem consists of, in one form or another, representation of and integration over these multiple variabilities.
Many of the authors here assume that the reliability analysis problem can be decomposed into a single integral over the product of two functions. The first of these functions is the result of combining the distributions of seismicity and a scalar measure of the ground motion intensity, producing a 'seismic hazard curve.' The second function is the result of considering conditional distributions of response and capacity over a range of ground motion intensities, e.g. a fragility curve for the probability of collapse versus intensity level. Many variations and perturbations of this general outline are presented in the papers selected for this issue. Kafali and Grigoriu [1] for example, test the validity of the assumption implicit in this common decoupling, which is that the structural response is (conditionally) independent of event magnitude and distance given the level of the ground motion intensity. They consider the case, popular with authors herein, where the chosen measure is a spectral acceleration at a period near that of the first mode of the structure, S a (T 1 ). The limitations of this choice intensity measure are becoming better understood. Luco and Bazzurro [2] study the response bias that can be introduced by scaling records with respect to this intensity measure. Tothong and Luco [3] evaluate inelastic spectral acceleration as an intensity measure, finding that it avoids most of the potential difficulties of elastic spectral acceleration. Baker [4] has previously demonstrated the effectiveness of certain vector-valued intensity measures, and here he presents methods for estimating the conditional distributions of responses given a vector such as spectral acceleration and 'epsilon.' Zareian and Krawinkler [5] and Tothong and Luco [3] evaluate the effect of epsilon.
A number of the authors (e.g. Mackie and Stojadinovic [6], Montiel and Ruiz [7], Goulet et al.
[8] and Ellingwood et al. [9]) study the fragility curves and/or seismic failure probabilities of suites of structural models, seeking general conclusions that seismic reliability analysis may reveal. Such conclusions can guide future structural code changes to ensure more uniformity in the safety levels they induce. Several of this issue's papers go beyond structural performance to include damage and economic losses (Mackie and Stojadinovic [6] and Goulet et al. [8]) illustrating the role of seismic reliability analysis in the larger problem of performance-based earthquake engineering.
In an effort to reduce the computational effort, three papers focus on the application and assessment of nonlinear static procedures to study seismic reliability: Mori and Maruyama [10], Dolsek and Fajfar [11] and Koduru et al. [12]. In terms of integration methods used, direct