← Back to Sessions

A Robust Time-Domain Early-Arrival Seismic Waveform Inversion Method Using Data Uncertainties

Session: Critical Zone, Environmental and Cryospheric Seismology

Type: Oral

Date: 4/22/2021

Presentation Time: 06:00 PM Pacific

Description: 

Early-arrival waveform inversion (EWI) is a technique for high-resolution near-surface velocity estimation, by iteratively updating the velocity model to minimize the difference between observed and calculated diving wave and refraction waveforms. Traditional EWI uses least-squares error as a penalty function that is sensitive to outliers in the data. In addition, the iteration of EWI stops when there is no further decrease in data misfit or when the maximum number of iterations is reached. These stopping criteria can result in overfitting the noise in the data. To overcome the problem of sensitivity to outliers and overfitting, we develop a robust inversion method by estimating and incorporating data uncertainties into the penalty function of EWI. The data uncertainties are estimated based on waveform reciprocal errors. In the situation where full waveform reciprocity is not available in the data, we apply seismic trace interpolation to compensate for the missing reciprocities. Comparing with conventional EWI, the proposed method minimizes the weighted least-squares error and prevents the waveform inversion from closely fitting the data with low signal-to-noise ratio (SNR). The proposed method avoids overfitting by stopping the iterations when the normalized chi-square (χ²) waveform misfit of 1 is achieved. Numerical examples suggest the final models using EWI with data uncertainties is more accurate than the corresponding EWI model, at the same level of misfit. A practical example of applying the proposed method to a real 2D near-surface seismic refraction data is presented.

Presenting Author: Ao Cai

Student Presenter: Yes

Authors