Rapid and accurate prediction of peak ground acceleration (PGA) is an important basis for determining seismic damage through on-site earthquake early warning (EEW). The current on-site EEW uses the feature parameters of the first arrival P-wave to predict PGA, but the selection of these feature parameters is limited by human experience, which limits the accuracy and timeliness of predicting peak ground acceleration (PGA). Therefore, an end-to-end deep learning model is proposed for predicting PGA (DLPGA) based on convolutional neural networks (CNNs). In DLPGA, the vertical initial arrival 3–6 s seismic wave from a single station is used as input, and PGA is used as output. Features are automatically extracted through a multilayer CNN to achieve rapid PGA prediction. The DLPGA is trained, verified, and tested using Japanese seismic records. It is shown that compared to the widely used peak displacement (Pd) method, the correlation coefficient of DLPGA for predicting PGA has increased by 12–23%, the standard deviation of error has decreased by 22–25%, and the error mean has decreased by 6.92–19.66% with the initial 3–6 s seismic waves. In particular, the accuracy of DLPGA for predicting PGA with the initial 3 s seismic wave is better than that of Pd for predicting PGA with the initial 6 s seismic wave. In addition, using the generalization test of Chilean seismic records, it is found that DLPGA has better generalization ability than Pd, and the accuracy of distinguishing ground motion destructiveness is improved by 35–150%. These results confirm that DLPGA has significant accuracy and timeliness advantages over artificially defined feature parameters in predicting PGA, which can greatly improve the effect of on-site EEW in judging the destructiveness of ground motion.

A new characteristic parameter $Sdτ$ is proposed to improve the performance of magnitude estimation in earthquake early warning (EEW). $Sdτ$ is the product of summation of absolute displacement multiplied by the maximum predominant period (⁠ $τmaxP$⁠) for the first arriving seconds of a seismic wave. About 30,725 underground records at borehole stations for 3645 earthquakes with magnitude between 4.0 and 9.0 from the Japanese KiK‐net were used to compare the magnitude proxy performance based on the proposed $Sdτ$ with that based on either $τmaxP$ or peak displacement $Pd$⁠. The comparison results show that for a magnitude between 4.0 and 7.3, $Sdτ$ has a better correlation with magnitude and higher estimated accuracy than either $τmaxP$ or $Pd$⁠. Hypocentral distance is not required when using $Sdτ$⁠, but it can be used to further improve the accuracy of magnitude estimate. These results confirm that $Sdτ$ can significantly improve the accuracy and timeliness of continuous magnitude estimation in an EEW system.

Quantifying the source, propagation path, and site characteristics accurately is pivotal for evaluating regional earthquake hazards. The nonparametric generalized inversion technique of strong-motion records offers an effective approach for analyzing these three fundamental elements. However, this technique might overlook the spatial variation of regional attenuation and the inherent uncertainty of reference stations. To enhance the understanding of strong ground motion characteristics in the Sichuan area, this study employs an improved nonparametric inversion scheme to analyze 3338 sets of strong-motion data from 397 earthquakes (Ms 3–6.5) that occurred in the Sichuan region between 2007 and 2020. In terms of regional source characteristics, the setting of low-frequency cutoffs in strong-motion records may lead to the overestimation of corner frequency and stress drop for earthquakes with higher magnitudes. Conversely, for events with moment magnitudes less than 5, the regional stress drop levels tend to stabilize as the magnitude increases. Furthermore, stress drop is independent of depth and fault type. Regarding path attenuation, considering the influence of Moho reflection, the geometric attenuation is limited to R−1 within a hypocentral distance of 75 km. The nonparametric results from 75 to 300 km show that as the hypocentral distance increases, the geometric attenuation in this region first decreases and then stabilizes. Furthermore, the mean level of the inelastic attenuation factor Qs is 199.5f0.98. Pertaining to site effects, we observe that at the edge of the Sichuan basin, the amplification of sites in the long-period part is clearly characterized and tends to stabilize with increasing frequency. In addition, the site amplification factor of 33 logarithmically spaced frequency points between 0.5 and 20 Hz has a linear relationship with the soil layer parameter VS30 or VS20. The correlation coefficients of both show moderate correlation, among which VS30 is slightly better than VS20.

A new characteristic parameter _max or peak displacement P _d . The comparison results show that for magnitude between 4.0-7.3, S _dτ has better correlation with magnitude and higher estimate accuracy than either τ ^p _max or P _d . Hypocentral distance is not required when using S _dτ , but it can be used to further improve the accuracy of magnitude estimate. These results confirm that S _dτ can significantly improve the accuracy and timeliness of continuous magnitude estimation in an EEW system.