Seismic arrival enhancement through the use of noise whitening

Birnie, Claire; Chambers, Kit; Angus, Doug

doi:10.1016/j.pepi.2016.11.006

Cited by 9 publications

(4 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Whereas, stacking procedures typically require multiple nearby stations and are therefore not applicable for single station or sparse array applications, ruling out their use for a significant portion of hazard monitoring scenarios. Alternative approaches could be to include a noise suppression procedure prior to detection, however that would introduce an unwanted additional computational cost and due to the varying nature of noise signals it would be difficult to determine an automated noise suppression algorithm which is optimal [25,26].…”

Section: Discussionmentioning

confidence: 99%

Bidirectional recurrent neural networks for seismic event detection

Birnie¹,

Hansteen²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Real time, accurate passive seismic event detection is a critical safety measure across a range of monitoring applications from reservoir stability to carbon storage to volcanic tremor detection. The most common detection procedure remains the Short-Term-Average to Long-Term-Average (STA/LTA) trigger despite its common pitfalls of requiring a signal-to-noise ratio greater than one and being highly sensitive to the trigger parameters. Whilst numerous alternatives have been proposed, they often are tailored to a specific monitoring setting and therefore cannot be globally applied, or they are too computationally expensive therefore cannot be run real time. This work introduces a deep learning approach to event detection that is an alternative to the STA/LTA trigger. A bi-directional, long-short-term memory, neural network is trained solely on synthetic traces. Evaluated on synthetic and field data, the neural network approach significantly outperforms the STA/LTA trigger both on the number of correctly detected arrivals as well as on reducing the number of falsely detected events. Its real time applicability is proven with 600 traces processed in real time on a single processing unit.Recently, Machine Learning (ML) methodologies, in particular Deep Learning (DL), has seen a resurgence across all fields of seismology, and wider geoscience applications. The majority of applications have focused on the adaptation of computer vision approaches for aiding seismic interpretation. For example, salt body detection [8], fault detection [9], and horizon detection [10]. However, other studies have considered how ML methodologies can be utilised for seismic event detection. [11] proposed the combination of convolutional Neural Networks (NN) with k-means clustering for detection of seismic arrivals, whilst [12] presents one of the earliest studies that considered using NNs for event detection on broadband seismometers. Novelly, they combined three different back-propagation NNs that

show abstract

Section: Discussionmentioning

confidence: 99%

Bidirectional recurrent neural networks for seismic event detection

Birnie¹,

Hansteen²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Recently, the topic has gained more attention and a couple of specialized techniques for noise suppression in microseismic monitoring have been developed. The suggested methods include downweighting or removal of stations based on their noise level (Trojanowski 2019;Alexandrov et al 2020), coherent noise removal using multichannel convolution filters (Trojanowski 2019), curvlet-based noise filtering (Trojanowski 2019) and statistical noise whitening (Birnie et al 2017). These methods look promising but it is important to consider that filtering can potentially damage the signal.…”

Section: The Importance Of Pre-processingmentioning

confidence: 99%

“…Chambers et al 2010;Pesicek et al 2014;Zeng et al 2014). Recently, new noise suppression techniques have been suggested that are specially designed for microseismic monitoring (Birnie et al 2017;Trojanowski 2019;Alexandrov et al 2020).…”

Section: Introductionmentioning

confidence: 99%

Comparing the performance of stacking-based methods for microearthquake location: a case study from the Burträsk fault, northern Sweden

Beckel

Lund

Eggertsson

et al. 2021

Geophysical Journal International

View full text Add to dashboard Cite

Summary Traditional earthquake location relying on first arrival picking is challenging for microseismic events with low signal-to-noise ratio. Over the past years, alternative procedures have been explored based on the idea of migrating the energy of an earthquake back into its source position by stacking along theoretical traveltime curves. To avoid destructive interference of signals with opposite polarity, it is common to transform the input signals into positive timeseries. Stacking-based source location has been successfully applied at various scales, but existing studies differ considerably in the choice of characteristic function, the amount of pre-processing and the phases used in the analysis. We use a dataset of 62 natural microearthquakes recorded on a 2D seismic array of 145 vertical geophones across the glacially-triggered Burträsk fault to compare the performance of five commonly used characteristic functions: the noise filtered seismograms and the semblance, the envelope, the short term average/long term average ratio and the kurtosis gradient of the seismograms. We obtain the best results for a combined P- and S-wave location using a polarity-sensitive characteristic function, i.e., the filtered seismograms or the semblance. In contrast, the absolute functions often fail to align the signals properly, yielding biased location estimates. Moreover, we observe that the success of the procedure is very sensitive to noise suppression and signal shaping prior to stacking. Our study demonstrates the usefulness of including lower quality S-wave data to improve the location estimates. Furthermore, our results illustrate the benefits of retaining the phase information for location accuracy and noise suppression. To ensure optimal location results, we recommend carefully pre-processing the data and test different characteristic functions for each new dataset. In spite of the sub-optimal array geometry, we obtain good locations for most events within ∼30-40 km of the survey and the locations are consistent with an image of the fault trace from an earlier reflection seismic survey.

show abstract

“…Noise is particularly troublesome for microseismic monitoring where the signal often arises from a low magnitude event, and is therefore, hidden below the noise (Maxwell, 2014), which can result in errors and artefacts in detection and imaging procedures (Bardainne et al, 2009). As such, large efforts are made to reduce the noise from data collection (e.g., Maxwell, 2010;Auger et al, 2013;Schilke et al, 2014) to data processing (e.g., Eisner et al, 2008;Mousavi and Langston, 2016;Birnie et al, 2017), and to ensure monitoring algorithms are adequately tested under realistic noise conditions (Birnie et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Transfer learning for self-supervised, blind-spot seismic denoising

Birnie

Alkhalifah

2022

Front. Earth Sci.

Self Cite

View full text Add to dashboard Cite

Noise is ever present in seismic data and arises from numerous sources and is continually evolving, both spatially and temporally. The use of supervised deep learning procedures for denoising of seismic datasets often results in poor performance: this is due to the lack of noise-free field data to act as training targets and the large difference in characteristics between synthetic and field datasets. Self-supervised, blind-spot networks typically overcome these limitation by training directly on the raw, noisy data. However, such networks often rely on a random noise assumption, and their denoising capabilities quickly decrease in the presence of even minimally-correlated noise. Extending from blind-spots to blind-masks has been shown to efficiently suppress coherent noise along a specific direction, but it cannot adapt to the ever-changing properties of noise. To preempt the network’s ability to predict the signal and reduce its opportunity to learn the noise properties, we propose an initial, supervised training of the network on a frugally-generated synthetic dataset prior to fine-tuning in a self-supervised manner on the field dataset of interest. Considering the change in peak signal-to-noise ratio, as well as the volume of noise reduced and signal leakage observed, using a semi-synthetic example we illustrate the clear benefit in initialising the self-supervised network with the weights from a supervised base-training. This is further supported by a test on a field dataset where the fine-tuned network strikes the best balance between signal preservation and noise reduction. Finally, the use of the unrealistic, frugally-generated synthetic dataset for the supervised base-training includes a number of benefits: minimal prior geological knowledge is required, substantially reduced computational cost for the dataset generation, and a reduced requirement of re-training the network should recording conditions change, to name a few. Such benefits result in a robust denoising procedure suited for long term, passive seismic monitoring.

show abstract

Seismic arrival enhancement through the use of noise whitening

Cited by 9 publications

References 24 publications

Bidirectional recurrent neural networks for seismic event detection

Bidirectional recurrent neural networks for seismic event detection

Comparing the performance of stacking-based methods for microearthquake location: a case study from the Burträsk fault, northern Sweden

Transfer learning for self-supervised, blind-spot seismic denoising

Contact Info

Product

Resources

About