Random noise attenuation of sparker seismic oceanography data with machine learning

Abstract

Abstract. Seismic oceanography (SO) acquires water column reflections using controlled source seismology and provides high lateral resolution that enables the tracking of the thermohaline structure of the oceans. Most SO studies obtain data using air guns, which can produce acoustic energy below 100 Hz bandwidth, with vertical resolution of approximately 10 m or more. For higher-frequency bands, with vertical resolution ranging from several centimeters to several meters, a smaller, low-cost seismic exploration system may be used, such as a sparker source with central frequencies of 250 Hz or higher. However, the sparker source has a relatively low energy compared to air guns and consequently produces data with a lower signal-to-noise (S∕N) ratio. To attenuate the random noise and extract reliable signal from the low S∕N ratio of sparker SO data without distorting the true shape and amplitude of water column reflections, we applied machine learning. Specifically, we used a denoising convolutional neural network (DnCNN) that efficiently suppresses random noise in a natural image. One of the most important factors of machine learning is the generation of an appropriate training dataset. We generated two different training datasets using synthetic and field data. Models trained with the different training datasets were applied to the test data, and the denoised results were quantitatively compared. To demonstrate the technique, the trained models were applied to an SO sparker seismic dataset acquired in the Ulleung Basin, East Sea (Sea of Japan), and the denoised seismic sections were evaluated. The results show that machine learning can successfully attenuate the random noise in sparker water column seismic reflection data.

Mid-Ocean Ridge and Storm Enhanced Mixing in the Central South Atlantic Thermocline

“…In addition, we implement a denoising convolutional neural network (DnCNN) to suppress random noise after stacking. We use the recommended steps and parameters of 17 layers and a mini-batch size of 128 (Zhang et al, 2017;Jun et al, 2020). We train the DnCNN model for 40 epochs and the number of iterations within each epoch is 220.…”

Turbulent Heat Fluxes in a Mediterranean Eddy Quantified Using Seismic and Hydrographic Observations

“…However, those spectra could neither be fully identified as noise since their slopes were not as high as +2. Jun et al [55] showed that after applying the machine learning noise attenuated algorithm, the noise parts of the slope spectra were converted to IW and turbulence spectra. This result also supports our "spectra contamination" theory.…”

“…Thus, it is implied that, in general, seismic data could be a good substitute to estimate the spatial distribution of turbulence features, with the benefit of increased spatial resolution. Dissipation estimates from the turbulent spectral regimes were limited due to noise and thus could not be validated, although this ideally side-steps assumptions of fine-scale parameterizations: future work may apply the noise attenuation algorithm suggested by [55].…”

Machine-learning Based Noise Attenuation of Field Seismic Data using Noise Data Acquisition

“…탄성파 자료에 기록되는 잡음은 선박 소음, 해류의 흐름이나 파도, 장비 운용 등으로 인해 발생한 무작위 잡음 과 다중 반사파(multiple), 공기파(air wave), 표면파(surface wave) 등과 같은 일관성 잡음이 있다 (Ebadi, 2017;Hlebnikov et al, 2021). 탄성파 탐사를 수행하는 과정에서 잡음이 발 생하는 요인을 완전히 억제하는 것은 매우 어렵고 탐사환 경에 따라 상이하기 때문에 탄성파 자료마다 서로 다른 잡 음특성을 가지고 있다 (Kragh and Christie 2002;Nasser et al, 2016;Waage et al, 2019) (Li et al, 2018;Zhao et al, 2019;Liu et al, 2018;Jun et al, 2020)…”