Publications
71 results found
Debens HA, Knodel D, Umpleby A, et al., 2023, Enabling technologies for economical and efficient cloud-based FWI, Pages: 650-654, ISSN: 1052-3812
Seismic full-waveform inversion (FWI) is a computationally intensive but embarrassingly parallel procedure, where production workloads are typically distributed across large volumes of high-performance computing (HPC) resources. In the absence of available local HPC resources, public cloud HPC is a popular alternative due to the almost limitless scale and pay-as-you-go models offered by most providers, although these benefits often entail increased costs. In this work, we introduce and discuss several strategies for making cloud-based FWI more affordable without sacrificing stability or efficiency.
Warner M, Nangoo T, Umpleby A, et al., 2023, Automated salt model building: From compaction trend to final velocity model using waveform inversion, Leading Edge, Vol: 42, Pages: 196-206, ISSN: 1070-485X
Conventional seismic velocity model building in complicated salt-affected areas requires the explicit identification of salt boundaries in migrated images and typically involves testing of possible subsurface scenarios through multiple generations. The resulting velocity models are slow to generate and may contain interpreter-driven features that are difficult to verify. We show that it is possible to build a full final velocity model using advanced forms of full-waveform inversion applied directly to raw field data, starting from a model that contains only a simple 1D compaction trend. This approach rapidly generates the final velocity model and migrates processed reflection data at least as accurately as conventionally generated models. We demonstrate this methodology using an ocean-bottom-node data set acquired in deep water over Walker Ridge in the Gulf of Mexico. Our approach does not require exceptionally long offsets or the deployment of special low-frequency sources. We restrict the inversion so it does not use significant energy below 3 Hz or offsets longer than 14 km. We use three advanced forms of waveform inversion to recover the final model. The first is adaptive waveform inversion to proceed from models that begin far from the true model. The second is nonlinear reflection waveform inversion to recover subsalt velocity structure from reflections and their long-period multiples. The third is constrained waveform inversion to produce salt- and sediment-like velocity floods without explicitly identifying salt boundaries or velocities. In combination, these three algorithms successively improve the velocity model so it fully predicts the raw field data and accurately migrates primary reflections, though explicit migration forms no part of the workflow. Thus, model building via waveform inversion is able to proceed from field data to the final model in just a few weeks. It entirely avoids the many cycles of model rebuilding that may otherwise be required.
Mavropoulos C, Umpleby A, 2023, MPI-free FWI for Cloud Spot Markets: Faster and Cheaper Results
As the industry shifts to more computationally intensive data-driven applications, so does the need for more scalable and efficient processing power. Running such applications on the cloud is the obvious solution as the resources can scale per the requirements and stage of the project. We propose an Infrastructure as Code (IaC) environment: S-Cube Cloud (SCC) to launch and control large volumes of computational resources needed for new seismic processing applications. To effectively leverage the cloud, spot instances must be utilised, which are offered at a large discount but may be interrupted at any time. A key limitation we address is the absence of an efficient and fault-tolerant parallelisation scheme which is cloud-native as, without it, usage of discounted spot instances is unachievable. We propose RIPS(SCI) - Robust Inter Process Simple Socket Communication Interface - which allows for the utilisation of spot instances through its fault tolerance. Applied in real-world conditions, RIPS communicates between thousands of instances and handles spot instance interruptions. Furthermore, RIPS relieves major bottlenecks in the master process bypassing processing terabytes of data per iteration compared to MPI. Savings of 70%-80% are observed in processing workloads in customer workflows using spot instances enabled by RIPS.
Debens HA, Knodel D, Mancini F, et al., 2022, Semi-global multi-parameter FWI using public cloud HPC, Pages: 932-936, ISSN: 1052-3812
This paper explores the potential for hyperscale public cloud high-performance compute (HPC) to enable efficient deployment of a semi-global approach to multi-parameter full-waveform inversion (FWI) over large areas. We introduce several novel aspects to semi-global FWI that improve convergence and suppress crosstalk, while establishing that the algorithm's embarrassingly parallel nature is well suited for public cloud implementation. We describe how various public cloud services can be taken advantage of to reduce the cost of the inversion and provide a reference architecture for the deployment of semi-global FWI to the Amazon Web Services (AWS) platform. Finally, we apply semi-global FWI to raw data from a large-scale legacy surface seismic dataset acquired offshore Australia as part of a re-processing sequence undertaken recently. Our results demonstrate that semi-global FWI can be effectively parallelized across more than one million logical central processing units (CPUs) and is able to recover an anisotropic velocity model in a few hours and in an automated fashion.
Warner M, Armitage J, Umpleby A, et al., 2022, Full-elastic AVA extraction using acoustic FWI, Pages: 907-911, ISSN: 1052-3812
We demonstrate that accurate amplitude-vs-angle parameters can be extracted from raw unprocessed seismic data using purely acoustic full-waveform inversion. The resultant parameters incorporate the full elastic response of the observed data, and it is not necessary to use elastic FWI in order to determine AVA. This approach naturally corrects for a multitude of other amplitude effects including reflector geometry and transmission losses, and it deals correctly with near-critical and post-critical reflections. FWI-based workflows are consequently simpler than conventional workflows, and AVA parameters can normally be generated by FWI within days of raw field data first becoming available.
Debens HA, Knodel D, Mancini F, et al., 2022, ACCELERATED EXPLORATION VIA FWI, Pages: 2199-2203
Several recent studies have established that seismic full-waveform inversion (FWI) can be used to generate interpretable models of acoustic reflectivity from practically raw seismic data. Owing to their use of the full wavefield and an iterative least-squares approach to optimisation, these models, referred to as FWI images, offer an improvement in image quality over conventional approaches to depth migration, such as Kirchhoff pre-stack depth migration. Furthermore, the ability of FWI - when combined with an appropriate objective function - to begin from a basic initial model and unprocessed data means that these images can begin to be built shortly after acquisition. The effectively limitless scale of public cloud compute allows for these workloads to then be turned around quickly, while reasonable costs can be maintained by leveraging spare capacity markets. In an exploration setting, the availability of high-quality FWI images soon after acquisition can aid in improved and faster decision-making. In this abstract, we demonstrate our proposed workflow using a large subset of a modern surface-streamer dataset that was recently acquired for exploration purposes. 45 and 60 Hz FWI images were generated within weeks of the survey concluding and prior to a conventional fast-track image being delivered.
Warner M, Armitage J, Umpleby A, et al., 2022, AVO DETERMINATION USING ACOUSTIC FWI, Pages: 2054-2058
We demonstrate that purely acoustic FWI, run at true amplitude, can be used to extract accurate AVO parameters using offset-restricted subsets of raw unprocessed seismic data. Rather than producing conventional AVO parameters directly, acoustic FWI instead produces AVO anomalies showing the departure of the input data from the AVO displayed by a purely acoustic model. It is trivial to transform this into conventional AVO. AVO extraction via acoustic FWI applied to elastic synthetic data show that the AVO recovery using acoustic FWI is near perfect. When combined with full-bandwidth FWI, this approach removes any requirement for conventional data processing, model building, explicit migration or Kirchhoff-based AVO extraction. A final depth-migrated reflectivity volume, and accurate AVO parameters, can both be generated purely by acoustic FWI.
Nangoo T, Shah N, Lin T, et al., 2021, Accurate velocities and reduced cycle times from cloud-enabled full waveform inversion using XWI (AWI and RWI)
The central objective of advanced Full Waveform Inversion is to enable rapid turnaround of accurate velocities directly from raw seismic data. AWI with its convolutional filter-based residual represents a fundamental change in the way FWI is normally run, as an 'add on' to time-consuming velocity- model building performed on preprocessed data, where its role is to finesse a tomography starting model. Here we show the combined solution of AWI and RWI known collectively as XWI serves as a predictor for unseen drilling logs. The inversion is run on raw data from NW Australia (6 sailline validation test) demonstrating convergence to essentially the same result from two simple 1D starting models. It is able to predict deviations in a sonic log from a starting position over 1000 m/s away from the measured value. The cloud environment where XWI ingests traces directly from blob storage consists of a dynamic pool of interruptible compute instances. This allows for cost-effective frequency sweeps through the iterations and scalable hyperparameter scans. It also is configured for interfacing XWI with interactive processing software for seamless trace preparation and project start up.
Warner M, Nangoo T, Umpleby A, et al., 2021, Adaptive reflection waveform inversion: Faster, tighter, deeper, smarter, Pages: 582-586, ISSN: 1052-3812
We demonstrate that an appropriate combination of adaptive waveform inversion (AWI) (Guasch et al., 2019), kinematic reflection waveform inversion (RWI) (Warner et al., 2018), and quantum particle-swarm global optimization (qPSO) (Debens et al., 2015), is able to generate accurate well-resolved velocity models from unprocessed raw field data. We begin from simple one-dimensional starting models, we use minimal human intervention, and we recover velocity models, that are both kinematically accurate and highly resolved in space, to depths that lie well below the deepest penetration of refracted arrivals. Using this approach, refractions, reflections and multiples all contribute to the quality of the final velocity model. We demonstrate the efficacy of this approach using a realistic blind synthetic dataset in 2D, and using the corresponding reflection-dominated narrow-azimuth 3D field dataset that served, in part, as the motivation and archetype for the synthetic. For the field data, we demonstrate a close match to a blind well that penetrates below the refracted arrivals. This approach can build final velocity models in a small fraction of the time required for conventional depth velocity-model building. We show three types of inversion: (1) simple vanilla FWI which evolves towards local minima, leading to mis-convergence when starting far from the true answer, (2) AWI which increase the region of convexity that surrounds the global minimum, and (3) AWI-RWI which targets residual timing errors and that has explicit sensitivity to reflection moveout and so is able to recover deep macro-model velocity reflection updates.
Debens HA, Nangoo T, Mancini F, et al., 2019, Penetrating below the diving waves with RWI, AWI, and FWI: A NWS Australia case study
FWI has become a standard in velocity model building, however standalone FWI has not. To address this, FWI is brought into the model building sequence earlier by alternating RWI and AWI to recover the long-wavelength acoustic velocity model that is usually built by ray-based tomography. The corresponding long-wavelength anisotropy model is extracted using semi-global FWI. Least-squares FWI then has an adequate starting point to commence introducing the full range of length scales into the final model. The outcome is a high-resolution velocity model bypassing tomography, which penetrates over a kilometre deeper than the turning point of the deepest diving waves.
Debens HA, Nangoo T, Mancini F, et al., 2019, Penetrating below the diving waves with RWI, AWI, and FWI: A NWS Australia case study
© 81st EAGE Conference and Exhibition 2019. All rights reserved. FWI has become a standard in velocity model building, however standalone FWI has not. To address this, FWI is brought into the model building sequence earlier by alternating RWI and AWI to recover the long-wavelength acoustic velocity model that is usually built by ray-based tomography. The corresponding long-wavelength anisotropy model is extracted using semi-global FWI. Least-squares FWI then has an adequate starting point to commence introducing the full range of length scales into the final model. The outcome is a high-resolution velocity model bypassing tomography, which penetrates over a kilometre deeper than the turning point of the deepest diving waves.
Debens HA, Nangoo T, Mancini F, et al., 2019, Penetrating below the diving waves with RWI, AWI, and FWI: A NWS Australia case study
© 81st EAGE Conference and Exhibition 2019. All rights reserved. FWI has become a standard in velocity model building, however standalone FWI has not. To address this, FWI is brought into the model building sequence earlier by alternating RWI and AWI to recover the long-wavelength acoustic velocity model that is usually built by ray-based tomography. The corresponding long-wavelength anisotropy model is extracted using semi-global FWI. Least-squares FWI then has an adequate starting point to commence introducing the full range of length scales into the final model. The outcome is a high-resolution velocity model bypassing tomography, which penetrates over a kilometre deeper than the turning point of the deepest diving waves.
Warner M, Stekl I, Umpleby A, 2018, 3D wavefield tomography: Synthetic and field data examples, Pages: 3330-3334
Full wavefield tomography has become well established in two dimensions, but its extension into 3D for realistically sized problems is computationally daunting. In this paper, we present one of the first studies to apply 3D wavefield tomography to field data, and demonstrate that the method can solve useful exploration problems that that are not tractable by other methods.
Yao G, Shah N, Umpleby A, et al., 2018, High-resolution reflection FWI
We demonstrate reflection FWI on a less-than-ideal 3D narrow-azimuth towed-streamer dataset that contains little refracted energy and that is deficit in low frequencies. We begin from a very simple starting model built rapidly from stacking velocities. We fist use an FWI scheme that alternates between a migration-like and a tomography-like stage, showing that this can both recover the background velocity model and generate high vertical resolution. We follow this by using global inversion to build the long-wavelength anisotropy model. Finally, we use more-conventional reflection-based FWI to introduce the full range of wavelengths into the recovered velocity model, and show that this both migrates the reflection data and is structurally conformable with the reflections.
Yao G, Shah N, Umpleby A, et al., 2018, High-resolution reflection FWI
© 2018 Society of Petroleum Engineers. All rights reserved. We demonstrate reflection FWI on a less-than-ideal 3D narrow-azimuth towed-streamer dataset that contains little refracted energy and that is deficit in low frequencies. We begin from a very simple starting model built rapidly from stacking velocities. We fist use an FWI scheme that alternates between a migration-like and a tomography-like stage, showing that this can both recover the background velocity model and generate high vertical resolution. We follow this by using global inversion to build the long-wavelength anisotropy model. Finally, we use more-conventional reflection-based FWI to introduce the full range of wavelengths into the recovered velocity model, and show that this both migrates the reflection data and is structurally conformable with the reflections.
Yao G, Da Silva N, Warner M, et al., 2017, Improved FWI convergence using efficient receiver-side Spatial preconditioning employing ray theory
Spatial preconditioning can improve the convergence of full-waveform inversion (FWI) significantly. An accurate spatial preconditioning consists of the contribution of both the source and receivers. Prohibited by the unfordable computational cost of directly forming the receiver spatial preconditioning, source-only spatial preconditioning using the energy of the incident source wavefield is typically used to precondition the gradient of FWI. Although this is an efficient means of spatial preconditioning, the quality of the inversion result is still compromised. To improve the quality of spatial preconditioning, we here approximate the receiver spatial preconditioning using ray theory since ray tracing is much faster than numerically solving the two-way wave equation directly. In order to maintain the same time cycle as the inversion without receiver spatial preconditioning, we use an additional compute node to calculate the receiver spatial preconditioning in parallel with other compute nodes used for the usual gradient computation. The effectiveness is demonstrated by applying this technique to the Marmousi model.
Da Silva NV, Yao G, Warner M, et al., 2017, Global visco-acoustic full waveform inversion
Full Waveform Inversion aims to determine parameters of the subsurface by minimising the misfit between the simulated and recorded seismic data. The quality of such fit depends on many different aspects, as for example, the inversion algorithm and the accuracy of the constitutive laws. The latter is particularly important as if there are factors that are not taken into account in the seismic simulation then the inversion algorithm will compensate for their existence in the parameter(s) being estimated. One of such factors is attenuation. Here we introduce an approach that jointly estimates velocity and attenuation using a combination of Quantum Particle Swarm Optimisation with the conventional gradient descent method. This hybrid approach takes advantage of the fact that it is sufficient to estimate smooth models of Q and for this reason these can be represented with a sparse support, thus decreasing substantially the number of weights of the basis functions that have to be estimated and making the use of global algorithms practical. We demonstrate that the proposed method mitigates cross-talk between velocity and attenuation, while allowing the convergence towards accurate models of attenuation and velocity, thus being an effective method for velocity model building and consequently for seismic imaging.
Silverton A, Warner M, Morgan J, et al., 2015, Offset-variable density improves acoustic full-waveform inversion: a shallow marine case study, Geophysical Prospecting, Vol: 64, Pages: 1201-1214, ISSN: 0016-8025
We have previously applied three-dimensional acoustic, anisotropic, full-waveform inversion to a shallow-water, wide-angle, ocean-bottom-cable dataset to obtain a high-resolution velocity model. This velocity model produced: an improved match between synthetic and field data, better flattening of common-image gathers, a closer fit to well logs, and an improvement in the pre-stack depth-migrated image. Nevertheless, close examination reveals that there is a systematic mismatch between the observed and predicted datafrom this full-waveform inversion model, with the predicted data being consistently delayed in time. We demonstrate that this mismatch cannot be produced by systematic errors in the starting model, by errors in the assumed source wavelet, by incomplete convergence, or by the use of an insufficiently fine finite-difference mesh. Throughout these tests, the mismatch is remarkably robustwith the significant exception that we do not see an analogous mismatch when inverting synthetic acoustic data. We suspect therefore that the mismatch arises because of inadequacies in the physics that are used during inversion. For ocean-bottom-cabledata in shallow water at low frequency, apparent observed arrival times, in wide-angle turning-ray data, result from the characteristics of the detailed interference pattern between primary refractions, surface ghosts, and a large suite of wide-angle multiple reflected and/or multiple refracted arrivals. In these circumstances, the dynamics of individual arrivals can strongly influence the apparent arrival times of the resultant compound waveforms. In acoustic full-waveform inversion, we do not normally know the density of the seabed, and we do not properly account for finite shear velocity, finite attenuation, and fine-scale anisotropy variation, all of which can influence the relative amplitudes of different interfering arrivals, which in their turn influence the apparent kinematics. Here, wedemonstrate that the introduction of
Yao G, Debens HA, Umpleby A, et al., 2015, Adaptive Finite Difference for Seismic Wavefield Modelling, 77th EAGE Conference and Exhibition 2015
Yao G, Debens HA, Umpleby A, et al., 2015, Adaptive finite difference for seismic wavefield modelling, Pages: 650-654
We present an alternative scheme for calculating finite difference coefficients in seismic wavefield modelling. This novel technique seeks to minimise the difference between the accurate value of spatial derivatives and the value calculated by the finite difference operator over all propagation angles. Since the coefficients vary adaptively with different velocities and source wavelet bandwidths, the method maximises the accuracy of the finite difference operator. Numerical examples demonstrate that this method is superior to standard finite difference methods whilst comparable to Zhang's optimised finite difference method.
Debens HA, Warner MR, Umpleby A, 2015, Global anisotropic FWI, Pages: 4052-4056
Seismic anisotropy influences both the kinematics and dynamics of seismic waveforms. As such, if not taken into account properly during multi-parameter full-waveform seismic inversion (FWI), the anisotropy can manifest itself as significant error in the recovered P-wave velocity model. With this in mind we demonstrate a hybrid local-global inversion scheme to extract the low-wavenumber component of the anisotropy in combination with a high-resolution P-wave velocity model. This can then be followed by conventional local inversion for one or both parameters. Our results demonstrate that this technique suppresses the cross-talk between anisotropy and velocity, whilst producing more accurate final models for both parameters.
Debens HA, Warner M, Umpleby A, et al., 2015, Global anisotropic 3D FWI, Pages: 1193-1197, ISSN: 1052-3812
Seismic anisotropy influences both the kinematics and dynamics of seismic waveforms. If anisotropy is not adequately taken into account during full-waveform seismic inversion (FWI), then inadequacies in the anisotropy model are likely to manifest as significant error in the recovered P-wave velocity model. Conventionally, anisotropic FWI uses either a fixed anisotropy model derived from tomography or such, or it uses a local inversion scheme to recover the anisotropy as part of the FWI; both of these methods can be problematic. In this paper, we show that global rather than local FWI can be used to recover the long-wavelength anisotropy model, and that this can then be followed by more-conventional local FWI to recover the detailed model. We demonstrate this approach on a full 3D field dataset, and show that it avoids problems associated to cross-talk that can bedevil local inversion schemes. Although our method provides a global inversion of anisotropy, it is nonetheless affordable and practical for 3D field data.
Silverton A, Warner M, Umpleby A, et al., 2014, Non-physical water density as a proxy to improve data fit during acoustic FWI, Pages: 4135-4139
Major uplift in imaging is evident when migration is performed with a FWI velocity model for a North Sea hydrocarbon field. However, a small but significant, systematic mismatch in travel-time remains between the field data and synthetic data predicted using the final FWI model. However we perturb the model, the source, the number of iterations, the end result invariably returns to give the same final mismatch in which the predicted data are late. We know that both the synthetic and field data contain strong water-bottom multiples, and these affect the duration, bandwidth and amplitude decay of the coda. However, the finitedifference representation of the velocity model does not contain the seabed explicitly. We propose changing the assumed density of the water layer, which changes the seabed reflection amplitudes without affecting other aspects of the data, thereby properly modelling the seabed reflectivity. The wave-train is in reality an interference pattern between several arrivals, and as the relative strength of those arrivals changes, the interference pattern changes, thereby better fitting the travel-times. We find that decreasing the density of seawater improves the fit to the field data, and that we have to reduce the density by a greater factor as offset increases.
Warner M, Ratcliffe A, Nangoo T, et al., 2013, Anisotropic 3D full-waveform inversion, Geophysics, Vol: 78, Pages: R59-R80
We have developed and implemented a robust and practical scheme for anisotropic 3D acoustic full-waveform inversion (FWI). We demonstrate this scheme on a field data set, applying it to a 4C ocean-bottom survey over the Tommeliten Alpha field in the North Sea. This shallow-water data set provides good azimuthal coverage to offsets of 7 km, with reduced coverage to a maximum offset of about 11 km. The reservoir lies at the crest of a high-velocity antiformal chalk section, overlain by about 3000 m of clastics within which a low-velocity gas cloud produces a seismic obscured area. We inverted only the hydrophone data, and we retained free-surface multiples and ghosts within the field data. We invert in six narrow frequency bands, in the range 3 to 6.5 Hz. At each iteration, we selected only a subset of sources, using a different subset at each iteration; this strategy is more efficient than inverting all the data every iteration. Our starting velocity model was obtained using standard PSDM model building including anisotropic reflection tomography, and contained epsilon values as high as 20%. The final FWI velocity model shows a network of shallow high-velocity channels that match similar features in the reflection data. Deeper in the section, the FWI velocity model reveals a sharper and more-intense low-velocity region associated with the gas cloud in which low-velocity fingers match the location of gas-filled faults visible in the reflection data. The resulting velocity model provides a better match to well logs, and better flattens common-image gathers, than does the starting model. Reverse-time migration, using the FWI velocity model, provides significant uplift to the migrated image, simplifying the planform of the reservoir section at depth. The workflows, inversion strategy, and algorithms that we have used have broad application to invert a wide-range of analogous data sets.
Warner M, Nangoo T, Shah N, 2013, Full-waveform inversion of cycle-skipped seismic data by frequency down-shifting, Pages: 903-907, ISSN: 1052-3812
Full-waveform inversion can be severely compromised by problems of cycle skipping; this occurs when predicted and observed data differ by more than half a cycle, and it leads the inversion to recover a local rather than the global minimum model. Overcoming cycle skipping normally requires both a good starting model and low-frequency content in the field data. Here we present a scheme that uses a non-linear extrapolation to add missing low-frequencies into the field data. We demonstrate the scheme using a 3D OBC field dataset, and show that it can invert to recover the global minimum model even when the original un-extrapolated field dataset is significantly cycle skipped.
Nangoo T, Warner M, O'Brien GS, et al., 2013, The application of full waveform seismic inversion to a narrow-azimuth marine dataset, Pages: 4422-4426
We apply 3D anisotropic acoustic full-waveform inversion to a North Sea narrow-azimuth, marine-streamer dataset. We use a windowed strategy, with 3 stages, first focusing on mainly refracted arrivals with offsets up to (a) 1 km, (b) 2 km and then (c) 3 km with increasing iterations. We demonstrate that our recovered velocity model is realistic.
Buchan AG, Pain CC, Umpleby AP, et al., 2012, A sub-grid scale finite element agglomeration multigrid method with application to the Boltzmann transport equation, INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING, Vol: 92, Pages: 318-342, ISSN: 0029-5981
- Author Web Link
- Cite
- Citations: 7
Vieira Da Silva N, Morgan J, Warner M, et al., 2012, 3D constrained inversion of CSEM data with acoustic velocity using full waveform inversion, Pages: 605-609
The marine Controlled Source Electromagnetic Method (CSEM) has been used as exploration technology in the oil and gas industry. Inversion is the most widely used technique for interpretation of electromagnetic data. Nonetheless, resistivity images obtained from CSEM data generally have low resolution creating difficulties in the accurate characterization of geological reservoirs (containing hydrocarbons or for CO2 storage for example). Here is shown that by integrating a velocity field to guide the CSEM inversion leads to better focused resistivity images reducing ambiguities and source-receiver footprint. The proposed method integrates velocity models obtained through seismic full waveform inversion, in a constrained inversion algorithm for controlled source electromagnetic data. A synthetic example using a modified version of the Marmousi model is presented to validate the proposed methodology.
Shah N, Warner M, Nangoo T, et al., 2012, Quality assured full-waveform inversion: Ensuring starting model adequacy, Pages: 2562-2566
Successful full-waveform inversion (FWI) of 3D seismic data typically requires low-frequency content in the field data coupled with an accurate starting velocity model. In these circumstances, two fundamental questions always arise: (1) is the starting model sufficiently accurate given the data that are available, and (2) will the inversion iterate towards the global minimum, or will it instead become trapped locally leading to an erroneous final model? We present a robust and objective means to answer both these questions. The diagnostic feature that we use to achieve this is the spatial continuity of the phase difference between the predicted and observed field data, extracted at a single low frequency, after windowing in time around early arrivals. We show proof of principle on a simple 2D synthetic example, and demonstrate the application of quality assured full-waveform inversion (QA-FWI) to a full 3D field dataset that shows significant velocity anisotropy.
Liu F, Guasch L, Morton SA, et al., 2012, 3-D Time-domain Full Waveform Inversion of a Valhall OBC dataset, Pages: 2520-2524
By minimizing the difference between synthetic and field data sets, full waveform inversion (FWI) can produce high resolution and high fidelity earth model parameters that are not resolvable by commonly used ray-based tomography. In this paper, we share our experience on applying 3-D acoustic time-domain full waveform inversion to an OBC data set collected over Valhall field in North Sea.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.