SCEC2024 Poster #032: A Preliminary Framework for Evaluating the Effectiveness of 1D Ground Response Analysis at Vertical Array Sites Utilizing the Spatial Variability of mHVSR

Group B, Poster #032, Seismology

A Preliminary Framework for Evaluating the Effectiveness of 1D Ground Response Analysis at Vertical Array Sites Utilizing the Spatial Variability of mHVSR

Francisco Javier G. Ornelas, Christopher A. de la Torre, & Jonathan P. Stewart

Poster Image:

Poster Presentation

2024 SCEC Annual Meeting, Poster #032, SCEC Contribution #13773 VIEW PDF

One-dimensional (1D) ground response analysis (GRA) is commonly used to evaluate site response but has limitations for situations in which the wave propagation phenomena at the site are not dominated by vertically propagating waves through nearly laterally continuous strata. We aim to establish an approach to assess, a priori, whether a given site is likely to be effectively modelled by GRA. Our hypothesis is that sites where 1D modeling is effective are likely to show relatively consistent horizontal-to-vertical spectral ratios (HVSRs) for different locations across the site. In contrast, sites where 1D modeling is less effective are likely to have relatively spatially variable HVSR ordinat...es.
We collected 300 microtremor recordings from 17 vertical downhole array sites across California (Afshari et al. 2019, Ornelas et al. 2023), which include sites in which 1D GRA have varying levels of effectiveness (Afshari and Stewart 2019, de la Torre et al. 2022). Using this data, we compare multiple microtremor HVSR (mHVSR) spectra at variable distances from a reference mHVSR measurement (nearly) co-located with the vertical array. The vertical array is used to assess site response empirically, which can be compared to GRA results. Among the 17 sites, 24% are poorly modelled by GRA, 52% are well-modelled, and 29% are intermediate cases.
We evaluate spatial variability in mHVSR curves using three correlation models: Longest Common Subsequence (LCSS), Pearson’s r Coefficient, and Mean Absolute Error (MAE). Among these, LCSS is preferred for its ability to adjust the model according to visually discernible trends, resulting in correlations that accurately reflect the spatial variability of mHVSR. Our general finding is that sites with high LCSS values (indicating spatial uniformity) are more likely to have their site response accurately modelled with GRA than sites with low LCSS. These analyses focused on the frequency intervals containing the low-frequency peaks in the transfer function and the corresponding peaks in HVSR. Peaks at higher frequencies are often less correlated, likely due to interference from cultural noise.
SHOW MORE