Component‐To‐Component Variability of Response Spectral Ordinates from Strike‐Slip Earthquakes

It is important to appropriately capture the uncertainty associated with ground motion predictions. This study evaluates the component‐to‐component variability of 5%—damped response spectral ordinates using 4331 horizontal ground motions recorded in strike‐slip earthquakes. This variability is quantified relative to three scalar measures of central tendency of ground motion intensity (namely, geometric mean, RotD50, and MaxRotD50). The influence of earthquake magnitude, source‐to‐site distance, and the spatial distribution of recording stations relative to the epicenter on the component‐to‐component variability around each scalar measure is also studied. When combined with within‐event and between‐event variabilities, component‐to‐component variability allows the quantification of the total uncertainty of the random horizontal component, which better represents demand on a structure. The definition used in the past to quantify component‐to‐component variability was found to be sensitive to the orientation of horizontal sensors, leading to under‐ or over‐estimation depending on the orientation of the as‐recorded components relative to the orientation of maximum spectral response at a given station. Equations for component‐to‐component variability that are independent of the orientation of the sensors are proposed to obtain the total variability of the randomly oriented horizontal component from either the geometric mean or the RotD50. Compared to the commonly used equation, the proposed equations yield larger variabilities in the medium‐ to long‐period ranges and have smaller event‐to‐event variability.