PGA Conditioning and Directional Response of Paired Earthquake Accelerograms

Paired acceleration time histories are essential inputs to nonlinear structural analysis, but a scalar peak ground acceleration (PGA) target does not specify how the horizontal components should be scaled. This study examines the influence of amplitude conventions on paired-record processing through template reconstruction, coordinate rotation and response-spectrum calculations. A 100 s template-reconstruction dataset yields component PGAs of 350.000 and 247.510 cm/s2, correlations with the adjusted reference of 0.988 and 0.994, and peak-normalized errors of 0.885% and 0.441%. Directional analyses use the manually processed Amatrice record pair from the Engineering Strong-Motion Database. Scaling the first rotated component to 350 cm/s2 produces median directional PGA values between 273 and 669 cm/s2. Applying a common factor to the pair using an RotD50 PGA target removes this dependence for rotations that preserve the tested angular grid while retaining the relative component amplitudes and temporal structure. The RotD100/RotD50 spectral ratio varies from 1.191 to 1.414 over periods of 0.05–5 s. A separate set of 240 deterministic duration-model evaluations characterizes sensitivity to magnitude, distance, site velocity and mechanism. These results demonstrate the importance of a consistent horizontal intensity definition and explicit processing metadata. Their scope is numerical verification and single-event record processing; predictive performance across earthquakes and sites requires independent evaluation.