Kolli, Mohan KrishnaMohan KrishnaKolliBora, Sanjay SinghSanjay SinghBora2025-08-312025-08-312021-03-0110.1007/s10518-021-01052-w2-s2.0-85100955148http://repository.iitgn.ac.in/handle/IITG2025/25520The major challenge that remains with random vibration theory (RVT) based predictions of ground motion intensity measures (GMIMs) is the definition of the input ground motion duration. In literature it is reported that random vibration theory optimized duration (D<inf>rvto</inf>) (Bora et al. in Bull Seismol Soc Am 105(4):2192–2218, 2015, Earthquake Spectra 35(1):61–932019) can be a better measure of duration in situations when empirical models of FAS (Fourier Amplitude Spectrum) and duration are used in predicting response spectra but, such a measure of duration is often questioned for its physical significance. Moreover, no quantitative assessments are performed to analyze the performance of commonly used significant duration measures (D<inf>75</inf>, D<inf>95</inf> and 2D<inf>80</inf>) in comparison to D<inf>rvto</inf> in the RVT framework. In this study, we perform a comparative study to evaluate the performance of D<inf>75</inf>, D<inf>95</inf> and 2 D<inf>80</inf> in generating PGA (Peak Ground Acceleration) and response spectra using the observed FAS of ground motion. This study also investigates the physical significance of D<inf>rvto</inf>. Our main analysis is performed on the recorded acceleration traces compiled from the NGA (Next Generation of Attenuation)-West2 database. The efficacy of different measures of ground motion durations is performed using residuals analysis. The duration measure D<inf>75</inf> was found to be resulting in the least variation of residual spread in comparison to the other two duration measures. D<inf>95</inf> and 2D<inf>80</inf> were found to be longer measures of duration, resulting in smaller values of root mean square motion and, hence underprediction of ground motion. While D<inf>75</inf> was found to be performing better in the case of real data, we observed that in the case of stochastic simulations, 2D<inf>80</inf> performs better.Scaling of D<inf>rvto</inf> was found to be identical with that of D<inf>75</inf> and 2D<inf>80</inf> for real and simulated data, respectively.falseGround motion duration | NGA-West2 | Random vibration theory | Stochastic simulationsArticle157314561687-1707March 202116ArticleArticle13