The current national standards in China, such as the Seismic Ground Motion Parameters Zonation Map, the Code for Seismic Design of Buildings, etc, adopt design spectral accelerations primarily derived from foreign strong-motion records. Moreover, the number of records used in these statistical analyses was relatively limited. In recent years, however, strong-motion observation networks in China have developed rapidly, leading to a substantial increase in high-quality records. It is therefore necessary to reassess the applicability of the current design spectral accelerations using newly available domestic data. In this study, 788 strong-motion records from China, together with data from the NGA-West2 database and the KiK-net network, were used to statistically derive PGA-normalized spectral acceleration on site classes Ⅰ , Ⅱ , and Ⅲ. Recommended standardization parameters were also proposed.

Horizontal acceleration records with M_W≥5.0, source-to-site distance R≤200 km, and peak ground acceleration （PGA）≥30 cm/s² were selected to compute 5%-damped spectral accelerations, which were then normalized by PGA to characterize their shapes. The Jaccard similarity coefficients were calculated to quantify the similarity between the mean normalized spectra under different sample sizes of records, by which the minimum reliable sample size for each subgroup was determined to be 18.

A multivariate linear regression model was developed to examine the dependence of PGA-normalized spectral accelerations on earthquake magnitude, source distance, time-averaged shear-wave velocity of the upper 30 m, and overburden thickness. In addition, residuals between observed spectra and predictions from a simplified regression model, which included magnitude and v_S30, were analyzed. The results showed that earthquake magnitude and v_S30 were the dominant factors. In contrast, source distance and overburden thickness had relatively weaker effects. However, incorporating overburden thickness may further improve the characterization of site amplification effects.

Based on the classification of building sites in China, an empirical relationship between PGA-normalized spectral accelerations and earthquake magnitude was established using least-squares regression. Strong-motion records from China, Japan, the United States, and Europe were included. Comparative analyses revealed clear regional characteristics in PGA-normalized spectral accelerations. Compared with Europe and the USA, ground motions in western China and Japan exhibited relatively stronger short-period components and weaker long-period components. This finding indicates that China’s current design spectral accelerations, largely based on foreign ground-motion data, require revision.

For site classes Ⅰ and Ⅲ in the magnitude range of 7.5≤M≤8.0, the number of records was insufficient to meet the minimum reliable sample size. Based on the finding from the NGA-West2 database that normalized site amplification factors exhibit weak dependence on earthquake magnitude, normalized site amplification factors and then PGA-normalized spectral accelerations on site classes Ⅰ and Ⅲ versus site class Ⅱ in western China were derived. For site classes Ⅰ , Ⅱ , and Ⅲ at M6.0, M7.0 and M8.0, PGA-normalized spectral accelerations exhibit higher values at short periods and lower values at long periods relative to the design spectra from the Seismic Ground Motion Parameter Zonation Map of China. This discrepancy become more pronounced for smaller magnitudes and for sites with thinner and stiffer overburden layers.

The revised normalization parameters of design spectral accelerations on site classes Ⅰ , Ⅱ , and Ⅲ were proposed based on strong-motion records in western China, which were expected to improve seismic design codes. For engineering applications, it was strongly recommended to obtain matching PGA by attenuation relationships derived directly from strong-motion records in western China. It should be noted that records used in this study were from western China, and therefore the applicability of the results to eastern China required further verification. In addition, due to the limited number of records on site class Ⅳ and the lack of distinction between subclasses Ⅰ ₀ and Ⅰ ₁, further research was still needed.