Simplified Predictive Equations for Site‐Specific Adjustment of Ground Motion Spectral Shape and Duration Effects in Codified Performance‐Based Assessments

ABSTRACT A key challenge in seismic performance assessment is selecting ground motions that accurately represent site‐specific hazard characteristics. Although matching conditional distributions of spectral shape and duration ensures hazard consistency, it becomes computationally prohibitive for code‐based evaluations of large building inventories. Current codified approaches, such as FEMA P695 and Canada's performance‐based unified procedure, rely on incremental dynamic analysis (IDA) with generic records followed by site‐specific adjustment. However, existing adjustment equations are calibrated only for collapse and are based on regular ductile structures, limiting their applicability across noncollapse performance levels and diverse building archetypes where structural sensitivity to spectral shape and duration may differ. This study develops new predictive equations for spectral shape and duration adjustment factors across multiple performance levels, considering both component‐level and global performance criteria. A database of 45 code‐compliant reinforced concrete moment frames was analyzed using systematically selected ground motions. Regression analyses demonstrated that fundamental period and a normalized backbone area index are robust predictors of spectral shape and duration sensitivity, outperforming the period‐based ductility factor employed in current guidelines. Validation at two Canadian sites revealed that unadjusted IDA produced substantial bias, with the mean annual frequency of exceeding various performance levels overestimated by 50%–80%. Raw IDA underestimated median performance intensities by up to 22%, and FEMA P695 produced inconsistent adjustments, overestimating intensities in Vancouver by over 50% and underestimating them in Montreal by up to 16%. The proposed equations reduced these errors to below 11% across all performance levels and closely reproduced multiple stripe analysis results.