Point-source Parameters from Noisy Waveforms: Error Estimate by Monte-Carlo Simulation

We introduce an approach to estimate the error in the determination of the point-source mechanism and corresponding source-time function by inverting waveform data contaminated by a noise. No a priori assumptions about the statistical characteristics of the noise are needed. The estimate of the confidence region of the retrieved source is constructed as a set of solutions of the inverse problem for individual realizations of noisy data. They are generated from a sample of observed noise by Monte-Carlo simulation: white noise is convolved with the observed noise sample yielding a particular realization of the noise, which is random but retains the spectral characteristics of the observed sample. This approach allows us to consider even the situations where the noise varies from station to station. However, the basic assumption is that the noise is random, thus the method cannot handle systematic errors, e.g., due to mislocation or incorrect model of the medium.Synthetic tests with noisy data simulating the configuration of the seismic network in Friuli, NE Italy, show that determination of the source-time function is rather uncertain even with low-noise contamination but that orientation of the mechanism is recovered very well. As an example, local waveforms of a magnitude 3 event from the 1988 Moggio Udinese swarm, Northern Italy, are processed.