Noise In Seismic Data Research Articles

Lithospheric S Wave Velocity Variations Beneath the Mackenzie Mountains and Northern Canadian Cordillera

AbstractThe Mackenzie Mountains (MMs) in the Yukon and Northwest Territories, Canada, are an enigmatic mountain range. They are currently uplifting (Leonard et al., 2008, https//doi.org/10.1029/2007JB005456), yet are about 700 km from the nearest plate boundary. Their arcuate shape is distinct and extends over 100 km eastward from the general trend of the Northern Canadian Cordillera. To better assess the cause and conditions of the current uplift, we processed ambient seismic noise data from a linear array of broadband seismographs crossing the mountains, along with other regional seismic stations, to estimate Rayleigh wave phase velocities between 6 and 40 s periods. From this, we estimated phase velocity dispersion and performed a tomographic inversion to estimate VS. Tomography reveals a low‐velocity structure that extends upward from the base of the ∼50–66 km thick lithosphere to the upper crust, and we hypothesize that inferred low density and low rigidity associated with the VS anomaly localizes the ongoing uplift and thrust‐dominated seismicity of the MMs. Additionally, we find relatively low crustal velocities that extend to the west of the MMs, suggesting that strain transfer from the Gulf of Alaska plate boundary plays a driving role as the crust translates to the northeast and buckles up against the craton consistent with the orogenic float hypothesis of Mazzotti and Hyndman (2002, https//doi.org/10.1130/0091-7613(2002)030〈0495:YCASTA〉2.0.CO;2). Finally, we observe lithospheric azimuthal anisotropy with an NW‐SE fast direction. This is nearly orthogonal to teleseismic shear wave splitting measurements in the central MMs, and suggests that asthenosphere flow and lithospheric strain are not aligned in this region.

Continental Reworking in the Eastern South China Block and Its Adjacent Areas Revealed by F‐J Multimodal Ambient Noise Tomography

AbstractThe eastern South China Block (SCB) has experienced complex and drastic reworking processes since the early Neoproterozoic, the mechanisms of which remain unclear. To better understand the continental reworking mechanisms in the eastern SCB, we imaged the shear‐wave velocity structure of the crust and upper mantle in the eastern SCB and its adjacent areas with the F‐J multimodal ambient noise tomography method. The seismic ambient noise data were acquired by 652 seismic stations in this area. The shear‐wave velocity model presents widespread mid‐crustal low‐velocity zones (MCLVZs) and upper mantle low‐velocity zones (UMLVZs), the distribution of which shows a good correlation with regional tectonics. The MCLVZs formed due to compositional and structural layering during the late Neoproterozoic to the early Phanerozoic. The disrupted MCLVZs near the Qinling‐Dabie orogen can be linked to metamorphism caused by a continental collision between the SCB and the North China Craton during the Triassic Indosinian event. The thin crustal and lithospheric thicknesses east of the North‐South Gravity Lineament may be the result of asthenospheric upwelling due to the subduction of the Pacific plate since the late Mesozoic. In the areas near the Shaoxing‐Jiangshan‐Pingxiang fault zone and the Tanlu fault zone, upheaval UMLVZs indicate that these ancient fault zones were reactivated by the subduction of the Pacific plate, providing channels for asthenospheric upwelling. Well‐preserved MCLVZs north of the Qinling‐Dabie orogen can be used to assess compressions and extensions induced by the subduction of the Pacific plate.

Application of six-component ambient seismic noise data for high-resolution imaging of lateral heterogeneities

SUMMARY We develop a novel approach for imaging subsurface lateral heterogeneities using six-component (6C) ambient seismic noise data, consisting of three translational components and three rotational components. We first derive the 6C cross-correlation functions (CCFs) from ambient seismic noise data for surface waves and then apply the modified frequency Bessel (MF-J) transform to the 6C CCFs for surface wave dispersion curves. The phase velocities of Rayleigh and Love waves can be directly calculated from the 6C CCFs of a single pair of stations. Traditional array-based surface wave methods derive the surface wave phase velocity based on the presumption of a horizontally layered model. When this assumption breaks down due to the presence of, for example, dipping layer or heterogeneities, the resulting phase velocity can be severely smeared. By contrast, the proposed approach is not limited by the layered model assumption because it relies on single-point measurements to calculate the localized dispersion relations of the formation right beneath a receiver. Our numerical modelling results demonstrate that this approach is applicable to heterogeneous models and can image small-scale subsurface anomalies with very high lateral resolution.

Seismic noise to public health signal: investigating the effects of pandemic guidance in Mexico

ABSTRACT Understanding public activities and developing thoughtful public health strategies are key goals in efforts to manage the COVID-19 pandemic. This paper explores how seismic noise data can be used as part of such efforts. We show that the fluctuation of seismic noise levels has the capacity to demonstrate aggregate human movement. When considered in relation to major public health efforts, these data can help us evaluate the effectiveness of public health communication strategies that seek to limit social activity. We show evidence that, broadly speaking, Mexican national efforts to encourage “lockdown” worked for a few months in areas around seismic stations, and broke down as time went on. Further, we suggest that changes in the levels of human activity detected in seismic noise can be read alongside social data that provide some clues as to why people respond or not to health recommendations. Our findings have implications for both efforts to understand the nature and effects of public trust in the Mexican state and also the practicalities of using seismic noise data in this manner. An interdisciplinary analysis allows us to address these data and their possible use in a way that takes seriously the opportunities and challenges that emerge in the context of contemporary biopolitics and emerging configurations of surveillance technologies. Analyzing anthropogenic seismic activity opens up new opportunities for ethical data collection and use.

Machine-learning-based porosity estimation from multifrequency poststack seismic data

Estimating porosity models via seismic data is challenging due to the low signal-to-noise ratio and insufficient resolution of the data. Although impedance inversion is often used in combination with well logs, to obtain subseismic scale porosity data, several problems must be addressed. Alternatively, we have proposed a machine learning-based workflow to convert seismic data into porosity models. A residual U-Net++ (ResUNet++)-based workflow is designed to take multiple poststack seismic volumes with different frequency bands as input and estimate a corresponding porosity model as output. This workflow is demonstrated in a 3D channelized reservoir to estimate the porosity model, and the R2 score of more than 0.9 is achieved for training and validation data. Moreover, a stress test is performed by adding noise to the seismic data to verify the expandability of applications, and the results find a robust estimation with 5% added noise. The additional two ResUNet++ are trained to only take the lowest or highest resolution seismic data as input to estimate the porosity model, but they exhibit underfitting and overfitting, respectively, supporting the importance of using multiscale seismic data for the porosity estimation problem. We mainly use experimental cases with simulated data. Therefore, scaling ResUNet++ for real data is needed in future research, such as considering coherent noise in seismic data, allowing uncertainty in petrophysical parameters, and expanding the size of ResUNet++ to the practical reservoir extent.

Crust and Uppermost Mantle Magma Plumbing System Beneath Changbaishan Intraplate Volcano, China/North Korea, Revealed by Ambient Noise Adjoint Tomography

AbstractChangbaishan volcano (CBV), straddling the border between China and North Korea, is one of the most intriguing and explosive volcanoes on Earth owing to its voluminous Holocene eruptions and unusual magma composition within a continental intraplate setting. Despite the enormous size and destructive past, its underlying magma plumbing system remains debated and poorly understood. We present a refined high‐resolution 3D crust and uppermost mantle S‐wave velocity (Vs) model of CBV by applying advanced 3D full‐waveform adjoint tomography on seismic ambient noise data recorded in both China and North Korea. The newly obtained seismic images detail a 3D architecture of magma plumbing systems beneath CBV, characterized by the presence of a prominent lower‐crustal mush zone with an estimated melt fraction of ∼1.5%–3.6%, which is sourced from upwelling asthenosphere through a vertical narrow conduit in the uppermost mantle region that contains an estimated melt fraction of ∼6%.

Ambient noise tomography for a high-resolution 3D S-wave velocity model of the Kinki Region, Southwestern Japan, using dense seismic array data

Research interest in the Kinki region, southwestern Japan, has been aroused by the frequent occurrence of microearthquake activity that do not always coincide with documented active fault locations. Previous studies in the Kinki region focused mainly on deep, large-scale structures and could not efficiently resolve fine-scale (~ 10 km) shallow crustal structures. Hence, characterization of the upper crustal structure of this region at an improved spatial resolution is required. From the cross-correlation of the vertical components of the ambient seismic noise data recorded by a densely distributed seismic array, we estimated Rayleigh wave phase velocities using a frequency domain method. Then, we applied a direct surface wave tomographic method for the measured phase velocity dispersion data to obtain a 3D S-wave velocity model of the Kinki region. The estimated velocity model reveals a NE–SW trending low-velocity structure coinciding with the Niigata–Kobe Tectonic Zone (NKTZ) and the active Biwako-seigan Fault Zone (BSFZ). Also, we identified fine-scale low-velocity structures coinciding with known active faults on the eastern side of the NKTZ, as well as sets of low-velocity structures across the Tanba region. Furthermore, sedimentary basins manifest as low-velocity zones extending to depths ranging from ~ 1.5 to 2 km, correlating with those reported in previous studies. Our results therefore contribute towards fundamental understanding of earthquake faulting as well as tectonic boundary and will be useful for hazard assessment and disaster mitigation.Graphical abstract

Correlations of Seismic Velocities and Elastic Moduli with Temperature in Superhot and Enhanced Geothermal Systems

This paper presents correlations derived by linear regression analysis of seismic velocities VP and VS and elastic moduli EP and ES with temperature in Los Humeros superhot (SHGS) and Acoculco enhanced (EGS) geothermal systems at depths down to 3 km below the surface and temperatures up to approximately 400 °C. In Los Humeros, the seismic velocity models were derived from the inversion of legacy active seismic survey data acquired in 1998, as well as from passive seismic monitoring and ambient seismic noise interferometry carried out during 2017–2019 by the GEMex consortium. In the Acoculco EGS, ambient seismic noise data were used. Steady-state formation temperatures were re-evaluated during and after the end of the GEMex project using measurements provided as a courtesy of the Federal Electricity Commission of Mexico (CFE). The density data needed for the calculation of elastic moduli were provided by the GEMex consortium, as derived from the inversion of regional and local gravity surveys. The analysis indicated that statistically significant correlations of seismic parameters to temperature exist in the vertical direction, namely exponential in Los Humeros superhot and logarithmic in Acoculco EGS, but no correlation was evident in the horizontal direction. This result suggests an indirect relationship among the considered variables due to interdependence on other parameters, such as pressure and vapor saturation. As the analysis was performed using only data obtained from sensing-at-surface methods, without direct geophysical calibration at depth, a distributed fiber-optic seismic and temperature sensing system at both surface and downhole is proposed for active-source and passive seismic monitoring, and seismic-while-drilling by the drill-bit source is considered for reverse vertical seismic profile (RVSP) recording whenever possible for future high-temperature geothermal applications.

Adaptive ground-roll attenuation using local nonlinear filtering

Source-generated coherent noise suppression is a long-standing issue in land seismic data processing. Traditional methods—such as f–k filtering, τ-p transform, and median filtering—often fail to remove coherent noise in irregular data, especially in the case that the noise highly overlaps with the signal. To address this issue, we developed a method based on local nonlinear filtering (LNF) to attenuate coherent noise in seismic data. In the proposed method, a novel similarity coefficient algorithm is presented to determine the optimal velocity of coherent noise. An initial model of coherent noise is constructed using a mixed median–mean filter and is then refined by the proposed weighting correction algorithm. After modeling the noise, coherent noise in seismic data can be then suppressed by adaptively subtracting the estimated noise from the raw seismic data. To verify the effectiveness of the proposed method, we applied it to regular and irregular synthetic data and a 3D field dataset containing strong ground roll. The results demonstrate that the new method successfully attenuates the ground roll and obtains better results in ground-roll attenuation and the preservation of useful signals compared with the three commonly used methods. Synthetic and real data examples indicate that the method is applicable to the attenuation of coherent noise in irregular data.

A first 3-D shear wave velocity model of the Ischia Island (Italy) by HVSR inversion

SUMMARY Following the Mw 3.9 earthquake that occurred in the Ischia island (Naples, southern Italy) on 21 August 2017, the local monitoring seismic network was significantly improved in terms of both number of stations and instrumentation performance. Due to the huge amount of collected seismic ambient noise data, in this paper we present a first 3-D shear wave velocity model of the island retrieved from the inversion of horizontal-to-vertical spectral ratio curves by fixing the shear wave velocities (Vs) and modifying the thicknesses to get the corresponding 1-D Vs models. We are confident about the robustness of the attained models since the inversion process provided a good convergence towards the best-fitting solutions. Then, a first 3-D velocity model was obtained by contouring all the 1-D models obtained for the selected seismic stations to highlight possible lateral variations of the layer thicknesses and to reconstruct the morphology of the deeper interface characterized by a high-impedance contrast. A good correspondence between the 3-D Vs model and the geological features of the island was observed, especially in the northern sector where most of the stations are installed. In particular, the top of the high-impedance contrast interface appears deeper in the northern coastal areas and shallower in the central sector. This result agrees with the structural settings of the island likely due to the resurgence of Mount Epomeo.

Site Effects Study in the Peshawar District using Seismic Noise

Surficial geology plays an important role during earthquakes in terms of its impact on the resulting ground shaking, referred to as site effects. Site effects influence the level of earthquake hazard and risk in a region. Peshawar, the provincial capital of Khyber Pakhtunkhwa, is one of the most seismically active cities in Pakistan. However, it lacks any such seismic characterization. In this work, a seismic ambient noise study was performed in Peshawar using observations from a total of 92 stations. The seismic noise data is recorded using a broadband seismometer for a period of about 30 minutes at every station point. The fundamental resonance frequency and a first-order amplification based on the seismic noise recording are estimated using the method of Nakamura. First-order estimates of the amplification range from 1.0 to 7.0, whereas the resonance frequency ranges between 0.2 to 2.5 Hz. The results suggest a complex geological structure with low seismic impedance and deep soil deposits in Peshawar. The depth of soil deposits is estimated to be more than 300 m, leading to a potential amplification of low-frequency seismic waves. These findings are crucial for tall structures with fundamental vibration periods greater than 0.40 sec. Doi: 10.28991/CEJ-2022-08-04-010 Full Text: PDF

Random noise attenuation via convolutional neural network in seismic datasets

With the explosive growth in seismic data acquisition and the successful application of convolutional neural networks to various image processing tasks within multidisciplinary fields, many researchers have begun to research convolutional neural networks based seismic interpretation techniques. Seismic random noise attenuation is a key step in seismic data processing. In seismic data interpretation, faults are an important geological structure that has great significance for accumulation and migration of oil and gas reservoirs. Random noise within seismic data will seriously affect the accuracy of subsequent data processing and interpretation. Therefore, it is crucial to eliminate random noise in seismic data. This paper aimed to improve the Signal-to-Noise Ratio of seismic data, and proposed an intelligent convolutional neural network noise reduction framework. In this paper, the median filtering, the mean filtering, and the proposed algorithm is used to denoise seismic fault data. Experimental results show that the method not only yields a higher Signal-to-Noise Ratio, but also preserves more useful fault information.

Surface-wave tomography for mineral exploration: a successful combination of passive and active data (Siilinjärvi phosphorus mine, Finland)

Abstract. Surface wave (SW) methods offer promising options for an effective and sustainable development of seismic exploration, but they still remain under-exploited in hard rock sites. We present a successful application of active and passive surface wave tomography for the characterization of the southern continuation of the Siilinjärvi phosphate deposit (Finland). A semi-automatic workflow for the extraction of the path-average dispersion curves (DCs) from ambient seismic noise data is proposed, including identification of time windows with strong coherent SW signal, azimuth analysis and two-station method for DC picking. DCs retrieved from passive data are compared with active SW tomography results recently obtained at the site. Passive data are found to carry information at longer wavelengths, thus extending the investigation depth. Active and passive DCs are consequently inverted together to retrieve a deep pseudo-3D shear-wave velocity model for the site, with improved resolution. The southern continuation of the mineralization, its contacts with the host rocks and different sets of cross-cutting diabase dikes are well imaged in the final velocity model. The seismic results are compared with the latest available geological models to both validate the proposed workflow and improve the interpretation of the geometry and extent of the mineralization. Important large-scale geological boundaries and structural discontinuities are recognized from the results, demonstrating the effectiveness and advantages of the methods for mineral exploration perspectives.

On the use of the microtremor HVSR for tracking velocity changes: a case study in Campo de Dalías basin (SE Spain)

SUMMARY The stability of the low-frequency peaks (<1 Hz) obtained in the passive seismic survey of Campo de Dalías basin (CDB) by applying the horizontal-to-vertical spectral ratio (HVSR) method was investigated. Three temporary seismic stations were installed in remote sites that enabled studying the stationarity of their characteristic microtremor HVSR (MHVSR) shapes. All stations began to operate in mid-2016 and recorded at least 1 yr of continuous seismic ambient noise data, having up to 2 yr in some. Each seismic station was counted with a monitored borehole in their vicinity, registering the groundwater level every 30 min. The MHVSR curves were calculated for time windows of 150 s and averaged hourly. Four parameters have been defined to characterize the shape of the MHVSR around the main peak and to compare them with several environmental variables. Correlations between MHVSR characteristics and the groundwater level are shown to be the most persistent. The robustness of MHVSR method for applications to seismic engineering was not found to be compromised since the observed variations were within the margins of acceptable deviations. The results of this study have widen the possibilities of the MHVSR method from being a reliable predictor for seismic resonance to also being an autonomous monitoring tool, especially sensitive to the S-wave modifications.

Crustal S-Wave Velocity Structure Beneath the Northwestern Bohemian Massif, Central Europe, Revealed by the Inversion of Multimodal Ambient Noise Dispersion Curves

The northwestern Bohemian Massif and adjacent areas are a tectonically active region associated with complex geodynamic activities, that manifest as Quaternary volcanism, earthquake swarms in the upper and middle crust, degassing of CO2, and crustal fluid migration. The intricate tectonic evolution and activities of this region reflect the complexity of the crustal structure therein. However, the crustal models derived from previous studies in this area offer different, even contradictory information regarding the existence of a mid-crustal low-velocity zone (LVZ). In this study, we apply the frequency-Bessel transform (F-J) method to extract the fundamental-mode and up to five higher-mode Rayleigh wave dispersion curves from ambient seismic noise data recorded in the study area and perform multimodal ambient noise dispersion curves inversion. The addition of higher-mode dispersion curves enhances the vertical resolution of the velocity structure inversion results. Our models support the view that the general S-wave velocity level of the crust is high within the study area. We detect two S-wave LVZs beneath the study area that are distributed mainly in the middle crust rather than the lower crust, and these LVZs are separated by a high-velocity zone. Considering the results of previous studies in the area, we infer that these S-wave LVZs may be the consequence of crustal fluids, plastic deformation and even partial melting of the felsic middle crust at relatively high crustal temperatures. Furthermore, these S-wave LVZs could be responsible for the origin and foci depth distribution of earthquake swarms. S-wave low-velocity anomalies are also observed in the uppermost mantle beneath the study area. These S-wave models based on the joint inversion of multimodal dispersion curves can provide new references for understanding the tectonic activity and geodynamic evolution of the northwestern Bohemian Massif and adjacent areas.

Planetary analogue study using microseismic analysis for near-surface lava tube detection and exploration

The paper presents new results from microseismic data analysis for the detection and delineation of near-surface lava tubes. Single-station, free-field seismic noise data were collected at the Corona Volcano area (Lanzarote, Canary Islands) as part of the PANGAEA-X 2017 European Space Agency (ESA) astronaut training campaign. The site was selected as it represents a suitable analogue for lunar maria and provided the opportunity to cross-validate and ground truth seismic and other geophysical studies. In this paper, we present our observations on the distribution of frequencies and amplitudes of standing waves generated by microtremors in the space between the ground surface and underground cavities by averaging amplitude spectra of microseismic records.This study shows that the frequency of the vertical component peaks is a relatively good indicator of presence, lateral extent and approximate dip of the cavities. However, only the vertical component peaks with a frequency either equal to or close to central to high amplitude (3.1–8.5) horizontal-to-vertical (H/V) peaks are related to lava tubes. If the P-wave velocities in the rocks overlying the lava tubes are known, then these frequencies can also be used to estimate the depth of such cavities.This study also identified some important pitfalls that may limit the use and accuracy of our approach. These include layering of rocks overlying the lava tubes, lateral changes in thickness, lateral variations of the P-wave velocity, and the presence of shallow void-rich zones.Overall, the results of this study confirmed that microseismic data are suitable for cavity detection. It is also, to our knowledge, one of the first studies dealing with microseismic data from near-surface lava tubes. With the recent increase in planetary exploration, this method can be easily adapted to support the geophysical exploration of volcanic terrains on the Moon and Mars.

A Method for Denoising Seismic Signals With a CNN Based on an Attention Mechanism

Suppressing random noise in seismic data is a significant problem in seismic data processing. Often, there is serious aliasing between the effective signal and random noise, affecting the identification of weak signals, and even resulting in great difficulties in the suppression of conventional seismic signals. We propose an improved attention-guided convolutional neural network (ADNet) to eliminate seismic interference noise. After a sufficient amount of training, the network removes noise by transferring seismic data features learned from a synthetic dataset to tests with complex field data. Our workflow consists of four parts. First, in the model, we improve the feature enhancement module (FEM) and attention module (AM), increase the convergence speed, and enhance the expressive ability. Second, we use 2-D synthetic data to verify the ability of the model to suppress noise in seismic records. Third, we use 2-D real seismic data to further verify the denoising effect of the improved ADNet. Fourth, we convert the 3-D simulated seismic data and field data into 2-D data for processing and reorganize the 2-D denoising results into 3-D data. By comparing the noise suppression outcomes of several classic denoising methods, simulations and actual experiments show that the improved ADNet effectively maintains the signal amplitude, reduces the network depth, and better suppresses seismic noise. Hence, we believe that our model can be widely applied in the field of seismic data processing.

NS2NS: Self-Learning for Seismic Image Denoising

Attenuation of incoherent noise is an effective way to improve signal-to-noise ratio (SNR) of seismic data. Recently, supervised deep learning based methods have been widely utilized for seismic image denoising, which often need plenty of noise-free data as training labels. However, noise-free seismic data are often unavailable in field applications. We propose an unsupervised learning method (NS2NS) to train a denoising network by using single noisy seismic data. The proposed model is based on two basic truths of seismic data: (1) High self-similarity of seismic data; (2) Spatially independence of incoherent noise in seismic data. To implement the proposed method, we first build a sampling workflow to generate paired noisy images based on single noisy seismic image. Moreover, we create similar noisy images that are similar but different with the original noisy image by using the proposed self-similar sampler. The original noisy images and generated similar noisy images are then fused by using a suggested Bernoulli sampler to create new paired noisy images. These new paired noisy images are used as the input and target of the denoising model, respectively. Next, an end-to-end convolutional neural network (CNN) is built for seismic image denoising, which aims to learn features of valid signals and suppress unpredictable random noise. Finally, we apply the proposed NS2NS method to both synthetic and field data. The results show that our proposed method can effectively suppress incoherent noise while preserving valid signals.

Applying Non-Local Means Filter on Seismic Exploration

The seismic reflection method is one of the most important methods in geophysical exploration. There are three stages in a seismic exploration survey: acquisition, processing, and interpretation. This paper focuses on a pre-processing tool, the Non-Local Means (NLM) filter algorithm, which is a powerful technique that can significantly suppress noise in seismic data. However, the domain of the NLM algorithm is the whole dataset and 3D seismic data being very large, often exceeding one terabyte (TB), it is impossible to store all the data in Random Access Memory (RAM). Furthermore, the NLM filter would require a considerably long runtime. These factors make a straightforward implementation of the NLM algorithm on real geophysical exploration data infeasible. This paper redesigned and implemented the NLM filter algorithm to fit the challenges of seismic exploration. The optimized implementation of the NLM filter is capable of processing production-size seismic data on modern clusters and is 87 times faster than the straightforward implementation of NLM.

DenoiseNet: Deep Generator and Discriminator Learning Network With Self-Attention Applied to Ocean Data

Well-organized seismic signals play a significant role in subsequent seismic data processing. The multi-scale learning of characteristic signals of complex structures by deep convolutional neural networks has obvious benefits in reducing random noise in seismic data. However, deep convolutional neural networks also have shortcomings. It cannot discover effective features in seismic data structures or recover highquality seismic signals just using convolution. Therefore, the paper presents a Generative Adversarial Network (GAN) architecture in conjunction with the U-Net network. To produce the mapping connection between clean seismic signals and noisy seismic data, the U-Net network is employed as the G network of GAN. Incorporating a self-attention mechanism to strengthen the correlation between seismic data, with the goal of improving the network’s reconstruction capacity on the continuity of seismic signals. The intelligent denoising of seismic data enabled by DsGAN enhances labor efficiency when compared to traditional approaches. When compared to the optimal state of current models such as DnCNN, DnGAN, the peak signal-to-noise ratio (PSNR) is enhanced by 1.52db of the DsGAN model, according to experimental data from simulated and actual seismic data. Not only that, the network also has the ability to learn complex unknown noise with strong generalization and robustness.