Frequency-dependent horizontal-to-vertical spectral ratios (HVSR) of Fourier amplitudes from three-component recordings can provide information on one or more site resonant frequencies and relative levels of amplification at those frequencies. Such information is potentially useful for predicting site amplification but is not present in site databases that have been developed over the last 15–20 years for the Next-Generation Attenuation (NGA) projects, which instead use the time-averaged shear-wave velocity (VS) in the upper 30 m of the site (VS30) as the primary site parameter and are supplemented with basin depth terms where available. As a consequence, HVSR parameters are also not used in NGA ground motion models. In order for HVSR-based parameters to be used in future versions of site databases, a publicly accessible repository of this information is needed. We adapt a relational database developed to archive and disseminate VS data to also include HVSR. The database provides relevant microtremor-based HVSR data (mHVSR) and supporting metadata. We consider the most relevant data to be the frequency-dependent mHVSR, where the horizontal is taken as the median component and also as a function of horizontal azimuth (referred to as polar plots). Relevant metadata includes site location information, details about the equipment used to make the measurements, and processing details related to windowing, anti-trigger routines, and filtering. We describe the database schema developed to organize and present this information. The relational database stores mHVSR data, but not site parameters derived from the data. Site parameters of potential interest for modeling purposes include: (1) a binary variable indicating whether an mHVSR plot contains a peak; (2) one or more peak frequencies; (3) peak amplitudes; and (4) peak widths. We present procedures for peak identification that we believe to be better suited to California conditions than the SESAME (2004) guidelines that are typically applied in current practice. These procedures are informed by analysts’ visual assessments and can largely reproduce peak selections developed by relatively “conservative” or “liberal” analysts (producing relatively few or many sites with peaks, respectively). These procedures use tree regression to identify peak-adjacent plateaus in HVSR, which in turn can be used to identify relative peak amplitudes and peak widths that are considered in the proposed peak identification criteria. The algorithm is coded in R and a Jupyter Notebook and performs the operation of peak identification, and for sites with peaks, peak fitting using a Gaussian function. These routines interact with the database via cloud computing, but are not directly part of the database. We investigate the consistency of mHVSRs derived from velocity seismometers and accelerometers, which shows a high rate of false negatives (missed peaks) from accelerometers, even when used with 24-bit data recorders. This might be due to the relatively high intrinsic noise and low sensitivity of accelerometers. We compare mHVSRs derived from co-located temporary iii instruments (as would be used in a site characterization study) vs. permanent instruments (which could be applied to characterize ground motion stations) to evaluate the effectiveness of the latter, enabling us to query microtremors from permanent stations to boost the HVSR database. We find about 60-80% consistency in this case, with no bias in the peak assessment from one type of noise measurement relative to the others. We compare mHVSR from velocity seismometers to those from earthquake recordings (eHVSRs). We find microtremors and earthquake recordings are consistent for 60-70% of sites, in the sense that both either do or do not have significant peaks, and when peaks are present, they occur at similar frequencies (i.e., differences between frequencies are < 20%). However, for sites where mHVSR identifies a peak, we find a false-positive rate of about 50%. Approximately one-third of those false-positives can be accounted for by the limited frequency range over which eHVSR results are well constrained (low frequencies are often missed). The false-negative rate from mHVSR is very low. Future adjustments of peak identification criteria (making them more conservative) could be undertaken if it were desirable to bring the false-positive and false-negative rates into closer alignment. These findings are important to consider when contemplating the development or use of site response models derived from HVSR-based parameters such as site frequency. In engineering applications, these parameters will almost always be derived from mHVSRs. However, for model development, it is tempting to use eHVSRs, because such information is most widely available for ground motion stations. Because mHVSRs and eHVSRs do not always match, it is important to derive models solely from mHVSRs to ensure consistency between parameters used in model development and forward applications. Moreover, because many (approximately 70%) of California sites do not have HVSR peaks, it is important for HVSR model development to consider indices on whether peaks are or are not present. Currently available models do not adequately account for these effects.
Horizontal-to-Vertical Spectral Ratios from California Sites: Open-Source Database and Data Interpretation to Establish Site Parameters
Zimmaro P.;
2021-01-01
Abstract
Frequency-dependent horizontal-to-vertical spectral ratios (HVSR) of Fourier amplitudes from three-component recordings can provide information on one or more site resonant frequencies and relative levels of amplification at those frequencies. Such information is potentially useful for predicting site amplification but is not present in site databases that have been developed over the last 15–20 years for the Next-Generation Attenuation (NGA) projects, which instead use the time-averaged shear-wave velocity (VS) in the upper 30 m of the site (VS30) as the primary site parameter and are supplemented with basin depth terms where available. As a consequence, HVSR parameters are also not used in NGA ground motion models. In order for HVSR-based parameters to be used in future versions of site databases, a publicly accessible repository of this information is needed. We adapt a relational database developed to archive and disseminate VS data to also include HVSR. The database provides relevant microtremor-based HVSR data (mHVSR) and supporting metadata. We consider the most relevant data to be the frequency-dependent mHVSR, where the horizontal is taken as the median component and also as a function of horizontal azimuth (referred to as polar plots). Relevant metadata includes site location information, details about the equipment used to make the measurements, and processing details related to windowing, anti-trigger routines, and filtering. We describe the database schema developed to organize and present this information. The relational database stores mHVSR data, but not site parameters derived from the data. Site parameters of potential interest for modeling purposes include: (1) a binary variable indicating whether an mHVSR plot contains a peak; (2) one or more peak frequencies; (3) peak amplitudes; and (4) peak widths. We present procedures for peak identification that we believe to be better suited to California conditions than the SESAME (2004) guidelines that are typically applied in current practice. These procedures are informed by analysts’ visual assessments and can largely reproduce peak selections developed by relatively “conservative” or “liberal” analysts (producing relatively few or many sites with peaks, respectively). These procedures use tree regression to identify peak-adjacent plateaus in HVSR, which in turn can be used to identify relative peak amplitudes and peak widths that are considered in the proposed peak identification criteria. The algorithm is coded in R and a Jupyter Notebook and performs the operation of peak identification, and for sites with peaks, peak fitting using a Gaussian function. These routines interact with the database via cloud computing, but are not directly part of the database. We investigate the consistency of mHVSRs derived from velocity seismometers and accelerometers, which shows a high rate of false negatives (missed peaks) from accelerometers, even when used with 24-bit data recorders. This might be due to the relatively high intrinsic noise and low sensitivity of accelerometers. We compare mHVSRs derived from co-located temporary iii instruments (as would be used in a site characterization study) vs. permanent instruments (which could be applied to characterize ground motion stations) to evaluate the effectiveness of the latter, enabling us to query microtremors from permanent stations to boost the HVSR database. We find about 60-80% consistency in this case, with no bias in the peak assessment from one type of noise measurement relative to the others. We compare mHVSR from velocity seismometers to those from earthquake recordings (eHVSRs). We find microtremors and earthquake recordings are consistent for 60-70% of sites, in the sense that both either do or do not have significant peaks, and when peaks are present, they occur at similar frequencies (i.e., differences between frequencies are < 20%). However, for sites where mHVSR identifies a peak, we find a false-positive rate of about 50%. Approximately one-third of those false-positives can be accounted for by the limited frequency range over which eHVSR results are well constrained (low frequencies are often missed). The false-negative rate from mHVSR is very low. Future adjustments of peak identification criteria (making them more conservative) could be undertaken if it were desirable to bring the false-positive and false-negative rates into closer alignment. These findings are important to consider when contemplating the development or use of site response models derived from HVSR-based parameters such as site frequency. In engineering applications, these parameters will almost always be derived from mHVSRs. However, for model development, it is tempting to use eHVSRs, because such information is most widely available for ground motion stations. Because mHVSRs and eHVSRs do not always match, it is important to derive models solely from mHVSRs to ensure consistency between parameters used in model development and forward applications. Moreover, because many (approximately 70%) of California sites do not have HVSR peaks, it is important for HVSR model development to consider indices on whether peaks are or are not present. Currently available models do not adequately account for these effects.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
