On 3 June 1932, an M s 8.2 shallow thrust subduction earthquake struck the Colima–Jalisco (CJ) region of Mexico at an epicentral distance of ∼250 km from Guadalajara, the second largest town in Mexico. The return period of these CJ earthquakes has been estimated from 77 to 126 years, which suggests the next is likely to occur between 2009 and 2058. As a step toward estimating the seismic hazard posed by potential M s 8.2+ events on Guadalajara, we present a study consisting of the following: (1) the analysis of the strong ground motions recorded at Guadalajara on 11 free‐field accelerographs for the CJ 9 October 1995 M w 8 earthquake; (2) the analysis of the site effects in Guadalajara observed for this earthquake; and (3) the estimation of the spatial distribution of peak ground acceleration (PGA) in Guadalajara for this event. We propose, validate, and apply a recurrent neural network (RNN) technique to the recorded PGA. Important site effects were identified (using the horizontal‐to‐vertical and H/Hrocksite techniques) on Guadalajara’s sandy soil layers with thicknesses h >20 m to the basaltic rock basement. The estimated PGAs in Guadalajara for the CJ 1995 earthquake varied from ∼2 to ∼27 cm/s2 for soil layers with h ≤5 m and h >50 m, respectively. We expect the hybrid technique to obtain broadband synthetics (Chavez et al. , 2011) and the proposed RNN methodology can be used to estimate Guadalajara’s seismic hazard for M w 8+ CJ scenario earthquakes. Online Material: Figures of Fourier amplitude spectra, H/V and H/H spectral ratios, and signal‐to‐noise spectral ratios.
Read full abstract