A. Gisolf

Multi‐dimensional non‐linear full waveform inversion of gas cloud reflection data using a genetic algorithm and a blended acquisition approach

In this paper, we present a multi-dimensional non-linear full waveform inversion approach using an optimized Genetic Algorithm (GA), which includes a blended acquisition approach to invert the highly non-linear gas cloud reflection response for its effective medium parameters. This non-linear inversion process is vital to construct full waveform transmission operators (including the codas) from the gas cloud reflection response via an effective medium representation. However, extending this approach to a 2D non-linear full waveform inversion is not simple or straightforward, as multi-parameter inversion may become inefficient. Furthermore, to simulate the actual seismic wavefield in the subsurface we use the finite difference method as the forward modeling process and thus it becomes an expensive inversion procedure for the GA. We demonstrate that multi-dimensional non-linear full waveform inversion to obtain effective gas cloud medium parameters from the gas reflection response is viable using a GA. Several modifications to the conventional GA are suggested to overcome its limitations and therefore to gain faster convergence of the inversion process.

2D Full Waveform Transmission Deconvolution For True Amplitude Imaging Below Gas-clouds

Summary Conventional imaging processes for a situation with a gas cloud do not offer satisfactory solutions. Due to the complex wave propagation through the anomaly and the transmission imprint on the reflections from below these complexities, the image below the anomaly is usually not properly recovered. We aim at constructing full waveform transmission operators (including the codas) from the gas cloud reflection response via an effective medium representation. True amplitude imaging is achieved via multi-dimensional deconvolution for these full waveform transmission operators. In this paper, we present a feasibility study on 2D synthetic data, for which the operators are obtained by forward modeling through the gas cloud region. For the 1.5D case, it is demonstrated that a target-oriented full waveform inversion process to obtain effective gas cloud for medium parameters is viable.

Seismic imaging through gas clouds: a data-driven imaging strategy

Summary In Malaysian Basins, offshore Malaysia, there are many imaging problems related to the so-called ’gas clouds’ or ’gas chimneys’. Conventional imaging processes do not offer satisfactory solutions, because of the complex propagation effects that occur. The involved simplified Green’s functions that are used in these algorithms do not properly describe the kinematics as well as the dynamic behavior of the actual visco-elastic wave propagation. The proposed method in this paper aims at solutions beyond the traditional imaging approaches. The focusing of the wavefield is addressed via the so called Common Focus Point (CFP) method. In this way an optimum focusing of the energy below the anomaly is obtained. Furthermore, the estimated migration operators can be translated into a velocity-depth background model by tomography inversion. A synthetic dataset with a complex, low velocity anomaly was generated. The CFP imaging approach has contributed significantly toward the improvement in reflector continuity and amplitude restoration. Finally, initial results on a field dataset confirm the viability of the approach.

3D Focusing Operator Estimation Using Sparse Data

Estimation of focusing operators directly from 2D seismic data without the need of updating a velocity model, using the Common Focal Point (CFP)-technology, has proved to be a fruitful concept. In this paper we will discuss how the CFP-technology is extended to sparse 3D seismic data by incorporating a data-infill step in the focusing operator estimation procedure. This data-infill step is exploiting the large azimuthal redundancy that is available in seismic reflection data. The results of the focusing operator estimation procedure applied on a realistic synthetic sparse 3D dataset will be presented.

Data-Driven Meets Model-Driven - 3D-CFP Operator Updating

In this paper an iterative procedure will be presented that allows us to create 3D CFP-gathers and update 3D CFP-operators in case of a sparsly sampled acquisition geometry. The procedure is applied to a synthetic dataset and shows satisfactory results.

High-resolution Reservoir Rock Properties Via Joint Pre-stack Seismic Amplitude Inversion

Within the paradigm of inverse theory, the estimation of highresolution reservoir rock properties is cast as a linear-in-theparameters optimization problem. Through the formal separation of wave propagation and reflection effects, in addition to the inclusion of the seismic source wavelet’s ray-parameter dependence, all the primary information carried in the pre-critical seismic amplitudes can be exploited. Further improvement to the forward model is achieved through the simultaneous consideration of compressional and converted wave pre-stack seismic amplitudes. The established theory of regularized least-squares inversion is employed to overcome the ill conditioning of the problem. In addition, regularization is used to handle the noise corruption and band-limitation of seismic data. Numerical analysis shows this inverse problem to have an optimal solution that is an accurate broad-band reservoir rock properties estimate. When considered in the time-lapse situation, the inversion of each data vintage reveals which reservoir properties are crucial at each reservoir state while the inversion of the time-lapse seismic data reveals which reservoir properties are most affected by production.

INFILLING OF SPARSE 3D SEISMIC DATA FOR 3D FOCUSSING OPERATOR ESTIMATION

A scheme is presented for determination of 3D one-waytravel-time operators for Common Focus Point processing, from 3D data-sets without a full areal coverage of both sources and receivers. Examples are the parallel marine geometry and the cross-spread land geometry. Kinematically the operators are rigorously correct. When actually applied to imaging of the data the AVP information is fully preserved, but no azimuth dependency of the reflectivities and transmission losses are taken into account.