A Novel Method for Predicting Local Site Amplification Factors Using 1-D Convolutional Neural Networks

The analysis of site seismic amplification characteristics is one of the important tasks of seismic safety evaluation. Owing to the high computational cost and complex implementation of numerical simulations, significant differences exist in the prediction of seismic ground motion amplification in engineering problems. In this paper, a novel prediction method for the amplification characteristics of local sites was proposed, using a state-of-the-art convolutional neural network (CNN) combined with real-time seismic signals. The amplification factors were computed by the standard spectral ratio method according to the observed records of seven stations in the Lower Hutt Valley, New Zealand. Based on the geological exploration data from the seven stations and the geological hazard information of the Lower Hutt Valley, eight parameters related to the seismic information were presumed to influence the amplification characteristics of the local site. The CNN method was used to establish the relationship between the amplification factors of local sites and the eight parameters, and the training samples and testing samples were generated through the observed and geological data other than the estimated values. To analyze the CNN prediction ability for amplification factors on unrecorded domains, two CNN models were established for comparison. One CNN model used about 80% of the data from 44 seismic events of the seven stations for training and the remaining data for testing. The other CNN model used the data of six stations to train and the remaining station’s data to test the CNN. The results showed that the CNN method based on the observation data can provide a powerful tool for predicting the amplification factors of local sites both for recorded positions and for unrecorded positions, while the traditional standard spectral ratio method only predicts the amplification factors for recorded positions. The comparison of the two CNN models showed that both can effectively predict the amplification factors of local ground motion without records, and the accuracy and stability of predictions can meet the requirements. With increasing seismic records, the CNN method becomes practical and effective for prediction purposes in earthquake engineering.

Double-difference seismic attenuation tomography method and its application to The Geysers geothermal field, California

SUMMARY Knowledge of attenuation structure is important for understanding subsurface material properties. We have developed a double-difference seismic attenuation (DDQ) tomography method for high-resolution imaging of 3-D attenuation structure. Our method includes two main elements, the inversion of event-pair differential ${t^*}$ ($d{t^*}$) data and 3-D attenuation tomography with the $d{t^*}$ data. We developed a new spectral ratio method that jointly inverts spectral ratio data from pairs of events observed at a common set of stations to determine the $d{t^*}$ data. The spectral ratio method cancels out instrument and site response terms, resulting in more accurate $d{t^*}$ data compared to absolute ${t^*}$ from traditional methods using individual spectra. Synthetic tests show that the inversion of $d{t^*}$ data using our spectral ratio method is robust to the choice of source model and a moderate degree of noise. We modified an existing velocity tomography code so that it can invert $d{t^*}$ data for 3-D attenuation structure. We applied the new method to The Geyser geothermal field, California, which has vapour-dominated reservoirs and a long history of water injection. A new Qp model at The Geysers is determined using P-wave data of earthquakes in 2011, using our updated earthquake locations and Vp model. By taking advantage of more accurate $d{t^*}$ data and the cancellation of model uncertainties along the common paths outside of the source region, the DDQ tomography method achieves higher resolution, especially in the earthquake source regions, compared to the standard tomography method using ${t^*}$ data. This is validated by both the real and synthetic data tests. Our Qp and Vp models show consistent variations in a normal temperature reservoir that can be explained by variations in fracturing, permeability and fluid saturation and/or steam pressure. A prominent low-Qp and Vp zone associated with very active seismicity is imaged within a high temperature reservoir at depths below 2 km. This anomalous zone is likely partially saturated with injected fluids.

Waveform-based estimation of Q and scattering properties for zero-offset vertical seismic profile data

Estimating [Formula: see text] using downgoing waves in zero-offset vertical seismic profiles (VSPs) can be challenging when scattered waves from near-borehole heterogeneities interfere with direct arrivals. In any [Formula: see text] estimation method that assumes a downgoing plane wave, constructive and destructive wave-mode interference can cause errors in the estimate. For example, in the spectral-ratio method, such interference modulates the amplitude spectra introducing significant variations and even nonphysical negative [Formula: see text] (amplification) estimates. We have investigated this phenomenon using synthetic and field data sets from offshore Australia and developed a two-step waveform-based method to characterize scattering anomalies and improve [Formula: see text] estimates. Waveform information is key to deal with closely spaced band-limited seismic events. First, we solve an inverse problem to locate and characterize scatterers by minimizing the traveltime and waveform misfits. Then, using the estimated parameters, we model the scatterers’ contribution to the VSP data and remove it from the observed waveforms. The resulting spectra resemble those that would have been acquired in the absence of the scatterers and are much more suitable for the spectral-ratio method. By assuming a 1D medium and a simple scatterer shape (i.e., circular), we parameterize a scattering heterogeneity using five parameters (depth, distance, size, velocity, and density) and seek a solution using a grid search to handle the nonuniqueness of the VSP inversion. Instead, adaptive subtraction is required to fine-tune the modeled interference to better fit the observation. We successfully use this method to characterize and mitigate the strongest wave interference in the field data. The final [Formula: see text] estimates contain milder variations and much less nonphysical negative [Formula: see text]. Our results demonstrate that the proposed method, readily extendible to multiple scatterer cases, can locate discrete scatterers, remove the effects of their interference, and thus significantly improve the [Formula: see text] estimates from VSP data.