Item

Abstract

In our study, we analyzed high-frequency DAS data from the FORGE EGS facility in Utah to develop signal-stacking-based detection algorithms for detecting microearthquakes triggered by hydraulic fracturing.In terms of data structure, it is found that the DAS data considered is constituted by a high-frequency coherent noise, while microearthquakes are characterized by irregularities of this noise structure in the time- and frequency-domain.We evaluate different preprocessing steps (e.g., different frequency-filters, trace-normalisations, coherent noise supression methods,...) with the goal of improving the SNR of stacked DAS-quantities.It turns out that, that unlike in conventional seismic signal-stacking, the components of the background noise do not cancel each other out during summation in the time-domain.Instead, the numerous frequencies of background noise interfere with each other, which causes respective large noise peaks in sections of particularly strong constructive interference, which is an additional limiting factor of the detection threshold.In addition, we used the Power-Spectral-Density (PSD) method, that analyzes the signals in the frequency-domain.We show, that the background noise interference do not occur during stacking of PSD-traces, since it neglects unwanted frequency-components in the calculations before summation.However, the PSD method responds to another DAS-data characteristic noise-burst-type, that is not resolved by the stacked quantities in the dime-domain. We present an detection algorithm, which superimposes the individual threshold exceedances of time- and frequency-domain DAS-stacks, to filter both observed noise sources from the results.This algorithm was tested on a subset of 3-hours continuous recordings, where it captured all microearthquakes with magnitudes greater than -1.4 with a reliable true/false ratio.