Determination of soil properties from earthquake data

Soil damping and site (system) dominant vibration frequency estimations were obtained by means of the Random Decrement Method (RDM) using numerically simulated time series of soil model responses upon random excitations and real earthquake records. Highly reliable estimations were obtained when the system response was dominated by a distinctive or preferential vibration mode. Different damping mechanisms did not play a significant role in the variability of damping estimations; however, the excitation type did. The damping estimations were highly dependent on how well the Randomdec signature was defined. The alternate methods to measure the decay of the Randomdec signature may produce large variability in the damping estimations. The most consistent and reliable estimations were obtained using the average of the decay every halfcycle of the Randomdec signature. Hurley’s method consistently underestimated the damping values by an average of 50%. The frequency estimations were highly consistent when the Randomdec signature is well defined. Horizontal-to-vertical (H/V) spectral ratios were used to characterize local sediment response, and 1-D wave propagation modeling was used to estimate soil properties and theoretical amplification factors of shallow marine sediment layers in an experimental site in the Gulf of Mexico (GOM). Relative to the vertical spectral amplitude, the horizontal spectral amplitudes increased by an order of magnitude at 0.35 Hz, and by at least two orders of magnitude at 1.9 Hz. A 50-mthick soil system parameterized as three solid layers resting over a half-space with a water layer at the top produced theoretical H/V spectral ratios largely consistent with the observed H/V spectral ratios. The modeling results were consistent with both earthquake and background noise records. The use of background noise offers the advantage of better defining the spectral characteristics of the signal when, during the averaging process, a large ensemble is used. Time-frequency signal processing techniques were used to characterize the time-frequency variability of earthquake transient signals and to quantify the frequency-temporal variability of the seismic phases that constitute the multicomponent earthquake records. The state-of-the-art time-frequency distributions when applied to earthquake records were studied, and the conceptual bases for using this technique for seismic wave parameters estimations were set. In order to overcome various shortcomings associated with the Short Time Fourier Transform (STFT), a class, known as Cohen's class, of time-frequency energy distributions has been developed in recent years. This class includes the WignerVille Distribution (WVD), Choi-Williams Distribution (CWD), Reduced Interference Distribution (RID), Radially Gaussian Kernel Distribution (RGKD), and the Adaptive Optimal Kernel Distribution (AOKD). Each one of these processes was applied to controlled signals numerically generated as well as to seismic records. These time-frequency distributions possess advantages and disadvantages, which were discussed in this dissertation. On the basis of this study, I conclude that the RGKD and the AOKD are the most suitable for the analysis of the time-frequency characteristics of the waves (multi-components) that constitute an earthquake record.