Source Wavelet Research Articles

Simulation of Spark Source Wavelet Under Multibubble Motion

Marine spark sources are widely used in high-resolution marine seismic surveys. The characteristic of a wavelet is a critical part in seismic exploration; thus, the formation and numerical simulation of spark source wavelets should be explored. In studies on spark source excitation, the acoustic field generated by the interaction between bubbles constitutes the near-field wavelet of a source. Therefore, this interaction should be revealed by studying complex multibubble motion laws. In this study, actual discharge conditions were combined to derive the multibubble equation of motion. Energy conservation, ideal gas equation, and environmental factors in the discharge of spark source wavelets were studied, and the simulation method of an ocean spark source wavelet was established. The accuracy of the simulation calculation method was verified through a comparison of indoor-measured signals using three electrodes and the spark source wavelet obtained in the field. Results revealed that the accuracy of the model is related to the number of electrodes. The fewer the number of electrodes used, the lower will be the model’s accuracy. This finding is attributed to the statistical hypothesis factor introduced to eliminate the coupling term of the interaction of the multibubble motion equation. This study presents a method for analyzing the wavelet characteristics of an indoor-simulated spark source wavelet.

Unified elimination of 1D acoustic multiple reflection

ABSTRACTMigration, velocity and amplitude analysis are the employed processing steps to find the desired subsurface information from seismic reflection data. The presence of free‐surface and internal multiples can mask the primary reflections for which many processing methods are built. The ability to separate primary from multiple reflections is desirable. Connecting Marchenko theory with classical one‐dimensional inversion methods allows to understand the process of multiple reflection elimination as a data‐filtering process. The filter is a fundamental wave field, defined as a pressure and particle velocity that satisfy the wave equation. The fundamental wave field does not depend on the presence or absence of free‐surface multiples in the data. The backbone of the filtering process is that the fundamental wave field is computed from the measured pressure and particle velocity without additional information. Two different multiples‐free datasets are obtained: either directly from the fundamental wave field or by applying the fundamental wave field to the data. In addition, the known schemes for Marchenko multiple elimination follow from the main equation. Numerical examples show that source and receiver ghosts, free‐surface and internal multiples can be removed simultaneously using a conjugate gradient scheme. The advantage of the main equation is that the source wavelet does not need to be known and no pre‐processing is required. The fact that the reflection coefficients can be obtained is an interesting feature that could lead to improved amplitude analysis and inversion than would be possible with other processing methods.

Variational Bayesian Learning for Decentralized Blind Deconvolution of Seismic Signals Over Sensor Networks

This work discusses a variational Bayesian learning approach towards decentralized blind deconvolution of seismic signals within a sensor network. Blind seismic deconvolution is cast into a probabilistic framework based on Sparse Bayesian learning developed for blind image deconvolution. The posterior distribution of the signals of interest is then approximated using a variational Bayesian method. Depending on a particular form of selected variational factors, the scheme is shown to generalize the state-of-the-art distributed seismic blind deconvolution algorithm. The algorithm operates by repeatedly alternating between two stages: (i) estimation of seismic source wavelet and (ii) reflectivity estimation. The wavelet is estimated in a distributed fashion using Alternating Direction Method of Multipliers. Based on this estimate, each sensor then locally obtains a sparse reflectivity estimate. Numerical evaluation with synthetic seismic data shows that the proposed method outperforms existing deconvolution algorithms in a high signal-to-noise ratio (SNR) region. In low SNR regime a higher sensitivity of sparse Bayesian learning regarding model mismatches in the estimated source wavelet is observed. Also, processing of seismic data generated with an acoustic wave equation showed that the proposed method is able recover the original reflectivities more accurately, with more distinct support of the reflectivities, as compared to existing methods despite present model mismatches. The algorithm is also applied to the estimation of real seismic data, and shows improved performance as compared to state-of-the-art estimation method.

Elastic prestack seismic inversion through discrete cosine transform reparameterization and convolutional neural networks

We have developed a prestack inversion algorithm that combines a discrete cosine transform (DCT) reparameterization of data and model spaces with a convolutional neural network (CNN). The CNN is trained to predict the mapping between the discrete cosine-transformed seismic data and the discrete cosine-transformed 2D elastic model. A convolutional forward modeling based on the full Zoeppritz equations constitutes the link between the elastic properties and the seismic data. The direct sequential cosimulation algorithm with joint probability distribution is used to generate the training and validation data sets under the assumption of a stationary nonparametric prior and a Gaussian variogram model for the elastic properties. The DCT is an orthogonal transformation that we used as an additional feature extraction technique that reduces the number of unknown parameters in the inversion and the dimensionality of the input and output of the network. The DCT reparameterization also acts as a regularization operator in the model space and allows for the preservation of the lateral and vertical continuity of the elastic properties in the recovered solution. We also implement a Monte Carlo simulation strategy that propagates onto the estimated elastic model the uncertainties related to noise contamination and network approximation. We focus on synthetic inversions on a realistic subsurface model that mimics a real gas-saturated reservoir hosted in a turbiditic sequence. We compare the outcomes of the implemented algorithm with those provided by a popular linear inversion approach, and we also evaluate the robustness of the CNN inversion to errors in the estimated source wavelet and to erroneous assumptions about the noise statistic. Our tests confirm the applicability of our proposed approach, opening the possibility of estimating the subsurface elastic parameters and the associated uncertainties in near real time while satisfactorily preserving the assumed spatial variability and the statistical properties of the elastic parameters.

Periods of refracted P-waves in coal seams and their applications in coal thickness estimations

The direct and accurate estimations of coal thicknesses are prerequisites for intelligent mining practices. One of the most effective methods for detecting the distributions of coal thicknesses in coal mining panels is the in-seam seismic (ISS) method. In the present study, after examining the formation processes and propagation characteristics of refracted P-waves in ISS data, it was concluded that the refracted P-waves in coal seams are mainly formed by the multiple transmission and reflection of the P-waves between the coal and rock interfaces of roof and floor at critical angles. This results in the refracted P-waves having strong periodicity, and these periods are proportional to the coal thicknesses. This study adopted numerical simulation models with different coal thicknesses, and the aforementioned periodicity characteristics were examined. It was found that the coal seam thicknesses could be calculated using the periods of the refracted P-waves. However, in thin- or medium-thick coal seams, it was found that multiple transmitted P-waves overlapped and the periods could not be read directly. Therefore, in order to solve this problem, this study composed source wavelets with the main frequency of the source signals and then composite synthetic P-waves by convoluting the source wavelets with the sequences of various coal thicknesses. The suitable estimated coal thickness corresponded to the minimum value of the errors between the synthetic and actual refracted P-waves. An experiment was conducted in the No. 42224 panel of the Chaigou Coal Mine in order to validate the proposed method. The experimental results revealed that the estimated coal thicknesses from the refracted P-waves were consistent with the actual geologic conditions in the coal mine. Due to the fact that the refracted P-waves arrive earlier than other waves in seismic records, the refracted P-waves could be easily identified and processed. Overall, the proposed method was found to be a simple application process for accurate coal thickness estimations.

An unsupervised deep-learning method for porosity estimation based on poststack seismic data

We propose to invert reservoir porosity from poststack seismic data using an innovative approach based on deep-learning methods. We develop an unsupervised approach to circumvent the requirement of large volumes of labeled data sets for a conventional learning process. We apply convolutional neural networks (CNN) on seismic data to predict the relative porosity that is to be added to a low-frequency prior component. We then apply a forward model to synthesize seismic data based on a source wavelet and an acoustic impedance converted from the network-determined porosity. The parameters in the CNN are iteratively updated to minimize the error between recorded and simulated seismic data. We test the capability of our deep-learning approach to estimate reservoir porosity using a synthetic rock-physics model with two different signal-to-noise ratios. We also apply the proposed method to a real case study of seismic data acquired for hydrocarbon exploration of clastic reservoirs in the Vienna Basin. Instead of randomly assigning neural parameters, we use pretrained weights and biases at a previous location as initialization values for the next location, to preserve the geologically lateral continuity of the layers’ physical properties. As shown by these analyses, the unsupervised CNN-based scheme provides more or equally accurate results than standard methods for porosity estimation from seismically inverted acoustic impedance, which makes it a promising tool in seismic reservoir characterization with less user intervention.

Q-interface imaging using accumulative attenuation estimation

A high-resolution Q model is beneficial for more accurate attenuation compensation and preferable for gas-related interpretation. Given an accurate velocity model, viscoacoustic/viscoelastic full-waveform inversion ( Q-FWI) could reconstruct a high-resolution Q model, but it requires significant computational cost due to the iterative process of solving viscoacoustic/viscoelastic wave equations. We have proposed an efficient high-resolution Q-interface imaging method through the following steps. First, we estimate the attenuated traveltime via inversion of the dynamic match filter between synthetic acoustic and observed viscoacoustic prestack records. Second, we derive virtual Q reflectivities via piecewise linear regression on the attenuated traveltime estimation. Finally, by convolving a source wavelet on the virtual Q reflectivities, we generate the virtual Q reflection gathers and migrate them through reverse time migration (RTM) to image the Q interfaces. The Q-interface information is essentially derived by comparing the accumulative attenuation effects estimated from near-offset primary reflections arriving at the same receiver successively in time, and the high resolution is assured by the piecewise linear regression based on prior knowledge of the Q-interface number along the depth. The key insight of our method is to use accumulative attenuation effects to derive immediate effects of Q interfaces (virtual Q reflections) in the prestack data domain, which are readily applicable for Q-interface imaging through simple acoustic RTM. Numerical examples demonstrate that our method produces unprecedented high-resolution images of Q interfaces along the vertical direction with satisfying positioning and interpretable polarity.

3D aquifer characterization of the Hermalle-sous-Argenteau test site using crosshole ground-penetrating radar amplitude analysis and full-waveform inversion

To improve the understanding of flow and transport processes in the critical zone, high-resolution and accurate estimation of the small-scale heterogeneity is essential. Preferential flow paths related to high-porosity layers and clay lenses in gravel aquifers greatly affect flow and transport processes in the subsurface, and their high electrical contrast to their surrounding matrix and limited extent can act as low-velocity electromagnetic waveguides. In the past decade, time-domain full-waveform inversion (FWI) of crosshole ground-penetrating radar (GPR) data has shown to provide 2D decimeter-scale resolution images of relative permittivity and electrical conductivity of the subsurface, which can be related to porosity and soil texture. Most studies using crosshole GPR FWI resolved high-porosity zones that were identified by an amplitude analysis approach. But clay lenses or zones with higher electrical conductivity that act as low-velocity waveguides are hard to distinguish in the measured data and amplitude analysis because of the absence of characteristic wave-propagation features. We have investigated a set of nine crosshole GPR data sets from a test site in Hermalle-sous-Argenteau near the Meuse River in Belgium to characterize the aquifer within a decimeter-scale resolution and to improve the understanding of a previously performed heat tracer experiment. Thereby, we extend the amplitude analysis to identify two different types of low-velocity waveguides either caused by an increased porosity or a higher electrical conductivity (and higher porosity). Combining the GPR amplitude analysis for low-velocity waveguide zones with the standard FWI results provided information on waveguide zones, which modified the starting models and further improved the FWI results. Moreover, an updated effective source wavelet is estimated based on the updated permittivity starting models. In comparison with the traditional FWI results, the updated FWI results present smaller gradient of the medium properties and smaller root-mean-squared error values in the final inversion results. The nine crosshole sections are used to generate a 3D image of the aquifer and allowed a detailed analysis of the porosity distribution along the different sections. Consistent structures of the permittivity and electrical conductivity show the robustness of the updated FWI results. The aquifer structures obtained by the FWI results agree with those results of the heat tracer experiment.

Time-varying wavelet estimation and deconvolution for nonstationary data based on a FWE function

• We present a nonstationary convolution model constituted of the reflectivity series and a time-varying wavelet matrix. • We propose a novel wavelet estimation method toextract time-varying wavelet at every instant for nonstationary seismic data . • We achieve blind deconvolution without assumptions about factor Q value and the source wavelet.

Correlative Full-Intensity Waveform Inversion

Full-waveform inversion (FWI) is considered an effective technique for building high-resolution velocity models by fitting observed seismology waveforms. When the observed waveforms lack low frequencies or when the starting model is dissimilar to the true model, FWI usually suffers from cycle-skipping problems. To mitigate this difficulty, we propose a new correlative objective function that matches the phase differences between the seismic-waveform intensities to provide a good starting model. The waveform intensity separates the frequency band of the original data into an ultralow-frequency part and a higher frequency part, even when the original data lacking in low-frequency information. As the low-frequency part of the intensity is less prone to the cycle-skipping problem, it can be used to build an initial model for FWI. Furthermore, the source wavelets, in practice, are estimated from the observed data, which may be inaccurate in amplitude, phase shift, and time delay. To mitigate the inaccurate-source problem, a Wiener filter is constructed to build a source-independent objective function. Applications to the Marmousi model and Chevron blind-test model demonstrate that the proposed method converges to an acceptable initial model for conventional FWI. It is highly efficient and insensitive to inaccurate source wavelets, including inaccurate amplitude, phase shift, and time delays.

Data processing and interpretation schemes for a deep-towed high-frequency seismic system for gas and hydrate exploration

The deep-towed Autonomous Cable Seismic (ACS) system is a high-resolution acoustic survey system designed for use in deep-water environments. This system uses a high-frequency acoustic source and a multichannel receiver cable. A common problem in the analysis of deep-towed ACS data is the unstable positioning of the source and receivers due to ocean currents and seafloor bathymetry. Since the data acquisition using high-frequency source with unstable source–receiver positions causes destructive interference on the final stack profile, correction of the unstable source-receiver is a crucial issue. In this study, we propose a method to solve the unstable source–receiver position problem and thus to construct an accurate final stack profile. We used deep-towed ACS data acquired in the Joetsu Basin in Niigata, Japan, where hydrocarbon features in the form of gas chimneys, gas hydrate, and free gas have been observed. Because sidelobes in the ACS source signature defocus the source wavelet and decrease the bandwidth frequency content, we designed a filter to focus the source signature. Our proposed approach considerably improved the quality of the final stack profile. Even though depth information was not available for all receivers, the velocity spectra in the velocity analysis were well focused. Furthermore, shaping the source wavelet considerably increased the bandwidth frequency of the source signature. We applied seismic attribute analysis to the post-stack profile to identify the distributions of free gas and hydrate. Our seismic attribute analyses for the high-frequency ACS data demonstrated that free gas accumulations are characterized by low reflection amplitude and an unstable frequency component, and that hydrate close to the seafloor can be identified by its high reflection amplitude.

Frequency extension and robust full-waveform inversion based on nth power operation

The conventional full-waveform inversion (FWI) often minimizes the objective function using some local optimization algorithms. As a result, when the initial model is not good enough, the inversion process will drop into a local minimum. The low-frequency components contained in seismic data are of vital importance for reducing the initial model dependence and mitigating the cycle-skipping phenomenon of FWI. In this research, a frequency extension method using the nth power operation is proposed, which compresses the seismic data in time domain and extends their frequency band. Based on this, we construct a new objective function using the nth power wavefield and derive the corresponding gradient formula. The new objective function shows better property to overcome local minimum than the conventional one. When conduct inversion, we can invert from high-order to low-order successively, which is a new multiscale strategy. Since seismic data is more sensitive to source wavelet errors after high-order operation, we make the method more robust by proposing a source-independent method to mitigate the effects of source wavelet inaccuracy. After that, we extend the proposed method to encoded multisource waveform inversion. The numerical examples on the Marmousi model demonstrate that the proposed method can effectively mitigate the cycle-skipping of FWI, and it also has good anti-noise property.

Matching pursuit-based sparse spectral analysis: Estimating frequency-dependent anomalies from nonstationary seismic data

Spectral decomposition has been widely used to detect frequency-dependent anomalies associated with hydrocarbons. By ignoring the time-variant feature of the frequency content of individual reflected wavelets, we have adopted a sparse time-frequency spectrum and developed a matching pursuit-based sparse spectral analysis (MP-SSA) method to estimate the sparse time-frequency representation of the seismic data. Further, we evaluate a generalized nonstationary convolution model concerning propagation attenuation and frequency-dependent reflectivity, and we mathematically evaluate the sparse time-frequency spectrum of the nonstationary seismic data as being equal to the product of the Fourier spectrum of the source wavelet, frequency-dependent reflection coefficient, and the cumulative attenuation during seismic wave propagation. Therefore, the reflectivity spectrum, which is a combination of the frequency-dependent reflectivity and the propagation attenuation, can be determined by dividing the sparse time-frequency spectrum of the seismic data by the Fourier spectrum of the source wavelet. Application of the matching pursuit-based decomposition methods to synthetic nonstationary convolutional data illustrates that the adopted MP-SSA spectrum shows a higher time resolution than the matching pursuit-based Wigner-Ville distribution and the matching pursuit-based instantaneous spectral analysis spectra. Notably, the MP-SSA method can avoid spectral smearing, which may introduce distortions to the frequency-dependent anomaly estimation. Application of the amplitude versus frequency analysis based on MP-SSA to field data illustrates the potential of using the sparse reflectivity spectral intercept and gradient to detect the hydrocarbon reservoirs.

The influence of source wavelet estimation error in acoustic time domain full waveform inversion

Abstract Almost all forms of full waveform inversion use a source wavelet estimate in the extrapolation of the down-going wavefield to produce the forward modelled data and also to create the gradient. This source wavelet can be modelled, extracted from the recorded data during the pre-processing, or sometimes a zero-phase band-limited synthetic wavelet can be used instead. Alternatively, the source information can be regarded as another unknown and the initial source estimate then updated within the inversion. The importance of reliable source wavelet information during the waveform inversion implementation has been widely implied or briefly mentioned by multiple authors, but little detailed study of the effects of poor wavelets has been presented in the literature. It is the purpose of this study to examine if shape errors in the source wavelet manifest themselves in a significantly damaging way in the velocity model obtained using conventional time-domain acoustic waveform inversion, and also to assess what effect such velocity change has on the subsequent migrated image positions. We conclude that for modest structural complexity, acoustic time-domain least-squares waveform inversion for refractions with a maximum frequency of 9 Hz, commencing from a non-cycle skipped model, is surprisingly robust with respect to source wavelet error.

Joint Minimization of the Mean and Information Entropy of the Matching Filter Distribution for a Robust Misfit Function in Full-Waveform Inversion

A full-waveform inversion (FWI) is a highly nonlinear inversion methodology. FWI tends to converge to a local minimum rather than a global one. We refer to this phenomenon as “cycle skipping” in FWI. A cost-effective solution for resolving this issue is to design a more convex misfit function for the optimization problem. A global comparison based on using a matching filter (MF) admits more robust misfit functions. In this case, we would compute an MF first by deconvolving the predicted data from the measured ones. When the velocity model is accurate, the predicted data resemble the measured ones, and the resulting MF would be an approximated Dirac delta function. If the velocity produces data that are different from the observed ones, a misfit function can be formulated by penalizing the energy away from the zero-lag time (the center). Here, we develop a general mechanism for an evolution of the MF to our objective in FWI. Specifically from the statistics point of view, rather than using a penalty, we propose a novel misfit by minimization of the mean and information entropy of the MF distribution. We show that the resulting misfit function can mitigate the “cycle skipping” as well as reduce the mean and variance of the resulting MF distribution. We use a modified Marmousi example to demonstrate the features of the proposed misfit. We also evaluate the robustness of the proposed method using inaccurate (rotation in phase) source wavelets and measured data with different levels of Gaussian random noise.

An Effective Approach for Motion Artifacts Suppression from EEG Signal

Electroencephalographic(EEG) is a vital signal to analysis the neurological diseases in human being. This EEG signal captured even in highly hospitalic and standard environment may currpted by some non-physiological signals which are termed as artifact in medical term. These artifacts may disturb the quality of signal. Thus, mitigation of these artifacts from EEG signal is an important step. In this work an improved filtering mechanism is proposed forsingle channel EEG signal motion artifacts eradication. The input single channel EEG signal isdecomposed into multi-channel signal. Moreover, this multichannel signal is applied to an cascaded approach of Blind Source Separation (BSS) and wavelet transform in order to eleiminate the artifacts as well as randomness available in the signal due to this artifats. The results are tested with the existing work in the EEG artifact removal which shows outperformance of the proposed method. Keyword : EEG, EEMD-ICA, CCA, DWT, EEMD-DWICA.

Processing of random roadway source signals based on a cross-correlation algorithm in the deconvolution domain

Advanced seismic detection technology utilising random roadway sources is advantageous method adapting to the development of dynamic intelligent technologies for the detection of hidden and disastrous geological structures. However, roadheaders generate complex signals, that are continuous and random with seismic wavelets that are wider and longer than conventional source wavelets, and these complex wavelets cannot be directly subjected to data processing. To address this problem, based on two basic techniques, namely, deconvolution and cross-correlation, a cross-correlation algorithm in the deconvolution domain is proposed to process continuous random roadheader signals, convert those signals into normal-impulse seismic data, and extract the arrivals of direct and reflected waves and other information. First, the feasibility of the algorithm is explained based on a derivation of theoretical formulas. Then, to verify the applicability of the algorithm to the processing of actual data, a random signal from a tamping source is processed; the continuous random signals are successfully converted into impulse signals, and the extracted in-phase direct waves are clear with good continuity. Moreover, the continuous random signal processing results are compared with an active source signal, and good consistency is found, verifying the effectiveness of the algorithm. Finally, an application analysis of the algorithm is carried out using an actual acquired roadheader source signal. After implementing the algorithm, the random roadheader signal is successfully converted into a normal-impulse seismic signal, and the direct wave arrivals are extracted. Based on the arrival times of the direct waves, the longitudinal wave velocity of the coal seam is calculated as 1918 m/s, which is consistent with the actual velocity. A comprehensive analysis of the test results confirms that the cross-correlation algorithm in the deconvolution domain is both feasible and effective and that the processed seismic signals can be used for subsequent processing and interpretation.

Source-independent efficient wavefield inversion

SUMMARYFull-waveform inversion (FWI) is an effective tool to retrieve a high-resolution subsurface velocity model. The source wavelet accuracy plays an important role in reaching that goal. So we often need to estimate the source function before or within the inversion process. Source estimation requires additional computational cost, and an inaccurate source estimation can hamper the convergence of FWI. We develop a source-independent waveform inversion utilizing a recently introduced wavefield reconstruction based method, which we refer to as efficient wavefield inversion (EWI). In EWI, we essentially reconstruct the wavefield by fitting it to the observed data as well as a wave equation based on iterative Born scattering. However, a wrong source wavelet will induce errors in the reconstructed wavefield, which may lead to a divergence of this optimization problem. We use a convolution-based source-independent misfit function to replace the conventional data fitting term in EWI to formulate a source-independent EWI (SIEWI) objective function. By convolving the observed data with a reference trace from the predicted data and convolving the predicted data with a reference trace from the observed data, the influence of the source wavelet on the optimization is mitigated. In SIEWI, this new formulation is able to mitigate the cycle-skipping issue and the source wavelet uncertainty simultaneously. We demonstrate those features on the Overthrust model and a modified Marmousi model. Application on a 2-D real data set also shows the effectiveness of the proposed method.

Q-factor estimation using bisection algorithm with power spectrum

The centroid frequency shift (CFS) method is a widely used [Formula: see text] estimation approach. However, the CFS approach assumes that the amplitude spectrum of a source wavelet has a particular shape, which can cause systematic error in [Formula: see text] estimation. Moreover, the amplitude spectrum at high and low frequencies is susceptible to random noise, which can reduce the robustness of [Formula: see text] estimation using the CFS method. To improve the accuracy and robustness of [Formula: see text] estimation, we have developed a [Formula: see text] extraction method using the bisection algorithm based on the centroid frequency shift of power spectrum (BPCFS). In the BPCFS approach, we first obtain the source and the received wavelet. Then, we calculate the centroid frequency of the attenuated wavelet and that of the received wavelet from the power spectrum. Based on the obtained centroid frequencies, we establish an equation containing only one variable — the [Formula: see text] factor. Introducing the Jeffrey divergence to measure the attenuation of the power spectrum, we prove that this equation has only one root when [Formula: see text] is greater than zero. The root of this equation, which is the desired [Formula: see text] factor, is obtained through the bisection algorithm; we do not make any assumption about the shape of the amplitude spectrum. The noise-free numerical tests indicate that the BPCFS gives more accurate results than the CFS, which demonstrates that the shape of the wavelet spectrum has little effect on the [Formula: see text] estimation accuracy for BPCFS. Gaussian random and blue noise tests also show that the stability of BPCFS is better than that of CFS. The frequency band selection for BPCFS is also more flexible when the amplitude spectrum of the seismic wavelet is band limited. The application to real vertical seismic profile data further demonstrate the effectiveness and feasibility of the new method.

Bayesian trans-dimensional full waveform inversion: synthetic and field data application

SUMMARYSeismic full waveform inversion (FWI) is a state-of-the-art technique for estimating subsurface physical models from recorded seismic waveform, but its application requires care because of high non-linearity and non-uniqueness. The final outcome of global convergence from conventional FWI using local gradient information relies on an informative starting model. Bayesian inference using Markov chain Monte Carlo (MCMC) sampling is able to remove such dependence, by a direct extensive search of the model space. We use a Bayesian trans-dimensional MCMC seismic FWI method with a parsimonious dipping layer parametrization, to invert for subsurface velocity models from pre-stack seismic shot gathers that contain mainly reflections. For the synthetic study, we use a simple four-layer model and a modified Marmousi model. A recently collected multichannel off-shore seismic reflection data set, from the Lord Howe Rise (LHR) in the east of Australia, is used for the field data test. The trans-dimensional FWI method is able to provide model ensembles for describing posterior distribution, when the dipping-layer model assumption satisfies the observed data. The model assumption requires narrow models, thus only near-offset data to be used. We use model stitching with lateral and depth constraints to create larger 2-D models from many adjacent overlapping submodel inversions. The inverted 2-D velocity model from the Bayesian inference can then be used as a starting model for the gradient-based FWI, from which we are able to obtain high-resolution subsurface velocity models, as demonstrated using the synthetic data. However, lacking far-offset data limits the constraints for the low-wavenumber part of the velocity model, making the inversion highly non-unique. We found it challenging to apply the dipping-layer based Bayesian FWI to the field data. The approximations in the source wavelet and forward modelling physics increase the multimodality of the posterior distribution; the sampled velocity models clearly show the trade-off between interface depth and velocity. Numerical examples using the synthetic and field data indicate that trans-dimensional FWI has the potential for inverting earth models from reflection waveform. However, a sparse model parametrization and far offset constraints are required, especially for field application.