Rock physics modelling and inversion for saturation-pressure changes in time-lapse seismic studies

Time-lapse seismic data are generally used to monitor the changes in dynamic reservoir properties such as fluid saturation and pore or effective pressure. Changes in saturation and pressure due to hydrocarbon production usually cause changes in the seismic velocities and as a consequence changes in seismic amplitudes and travel times. This work proposes a new rock physics model to describe the relation between saturation-pressure changes and seismic changes and a probabilistic workflow to quantify the changes in saturation and pressure from time-lapse seismic changes. In the first part of this work, we propose a new quadratic approximation of the rock physics model. The novelty of the proposed formulation is that the coefficients of the model parameters (i.e. the saturation-pressure changes) are functions of the porosity, initial saturation and initial pressure. The improvements in the results of the forward model are shown through some illustrative examples. In the second part of the work, we present a Bayesian inversion approach for saturation-pressure 4D inversion in which we adopt the new formulation of the rock physics approximation. The inversion results are validated using synthetic pseudo-logs and a 3D reservoir model for CO₂ sequestration.     

Improved normalization of time‐lapse seismic data using normalized root mean square repeatability data to improve automatic production and seismic history matching in the Nelson field

Updating of reservoir models by history matching of 4D seismic data along with production data gives us a better understanding of changes to the reservoir, reduces risk in forecasting and leads to better management decisions. This process of seismic history matching requires an accurate representation of predicted and observed data so that they can be compared quantitatively when using automated inversion. Observed seismic data is often obtained as a relative measure of the reservoir state or its change, however. The data, usually attribute maps, need to be calibrated to be compared to predictions. In this paper we describe an alternative approach where we normalize the data by scaling to the model data in regions where predictions are good. To remove measurements of high uncertainty and make normalization more effective, we use a measure of repeatability of the monitor surveys to filter the observed time‐lapse data. We apply this approach to the Nelson field. We normalize the 4D signature based on deriving a least squares regression equation between the observed and synthetic data which consist of attributes representing measured acoustic impedances and predictions from the model. Two regression equations are derived as part of the analysis. For one, the whole 4D signature map of the reservoir is used while in the second, 4D seismic data is used from the vicinity of wells with a good production match. The repeatability of time‐lapse seismic data is assessed using the normalized root mean square of measurements outside of the reservoir. Where normalized root mean square is high, observations and predictions are ignored. Net: gross and permeability are modified to improve the match. The best results are obtained by using the normalized root mean square filtered maps of the 4D signature which better constrain normalization. The misfit of the first six years of history data is reduced by 55 per cent while the forecast of the following three years is reduced by 29 per cent. The well based normalization uses fewer data when repeatability is used as a filter and the result is poorer. The value of seismic data is demonstrated from production matching only where the history and forecast misfit reductions are 45% and 20% respectively while the seismic misfit increases by 5%. In the best case using seismic data, it dropped by 6%. We conclude that normalization with repeatability based filtering is a useful approach in the absence of full calibration and improves the reliability of seismic data.     

Post‐stack stratigraphic inversion workflow applied to carbon dioxide storage: application to the saline aquifer of Sleipner field‡

In the Norwegian North Sea, the Sleipner field produces gas with a high CO₂ content. For environmental reasons, since 1996, more than 11 Mt of this carbon dioxide (CO₂) have been injected in the Utsira Sand saline aquifer located above the hydrocarbon reservoir. A series of seven 3D seismic surveys were recorded to monitor the CO₂ plume evolution. With this case study, time‐lapse seismics have been shown to be successful in mapping the spread of CO₂ over the past decade and to ensure the integrity of the overburden. Stratigraphic inversion of seismic data is currently used in the petroleum industry for quantitative reservoir characterization and enhanced oil recovery. Now it may also be used to evaluate the expansion of a CO₂ plume in an underground reservoir. The aim of this study is to estimate the P‐wave impedances via a Bayesian model‐based stratigraphic inversion. We have focused our study on the 1994 vintage before CO₂ injection and the 2006 vintage carried out after a CO₂ injection of 8.4 Mt. In spite of some difficulties due to the lack of time‐lapse well log data on the interest area, the full application of our inversion workflow allowed us to obtain, for the first time to our knowledge, 3D impedance cubes including the Utsira Sand. These results can be used to better characterize the spreading of CO₂ in a reservoir. With the post‐stack inversion workflow applied to CO₂ storage, we point out the importance of the a priori model and the issue to obtain coherent results between sequential inversions of different seismic vintages. The stacking velocity workflow that yields the migration model and the a priori model, specific to each vintage, can induce a slight inconsistency in the results.     

Multiwell study of seismic attenuation at the CO2CRC Otway project geosequestration site: Comparison of amplitude decay,centroid frequency shift and 1D waveform inversion methods

At the CO2CRC Otway geosequestration site, the abundance of borehole seismic and logging data provides a unique opportunity to compare techniques of Q (measure of attenuation) estimation and validate their reliability. Specifically, we test conventional time-domain amplitude decay and spectral-domain centroid frequency shift methods versus the 1D waveform inversion constrained by well logs on a set of zero-offset vertical seismic profiles. The amplitude decay and centroid frequency shift methods of Q estimation assume that a seismic pulse propagates in a homogeneous medium and ignore the interference of the propagating wave with short-period multiples. The waveform inversion explicitly models multiple scattering and interference on a stack of thin layers using high-resolution data from sonic and density logs. This allows for stable Q estimation in small depth windows (in this study, 150 m), and separation of the frequency-dependent layer-induced scattering from intrinsic absorption. Besides, the inversion takes into account band-limited nature of seismic data, and thus, it is less dependent on the operating frequency bandwidth than on the other methods. However, all considered methods of Q estimation are unreliable in the intervals where subsurface significantly deviates from 1D geometry. At the Otway site, the attenuation estimates are distorted by sub-vertical faults close to the boreholes. Analysis of repeated vertical seismic profiles reveals that 15 kt injection of the CO₂-rich fluid into a thin saline aquifer at 1.5 km depth does not induce detectable absorption of P-waves at generated frequencies 5–150 Hz, most likely because the CO₂ plume in the monitoring well is thin, <15 m. At the Otway research site, strong attenuation Q ≈ 30–50 is observed only in shaly formations (Skull Creek Mudstone, Belfast Mudstone). Layer-induced scattering attenuation is negligible except for a few intervals, namely 500–650 m from the surface, and near the injection interval, at around 1400–1550 m, where Q_scat ≈ 50–65.     

An estimation method for effective stress changes in a reservoir from 4D seismics data

Advances in seismics acquisition and processing and the widespread use of 4D seismics have made available reliable production‐induced subsurface deformation data in the form of overburden time‐shifts. Inversion of these data is now beginning to be used as an aid to the monitoring of a reservoir's effective stress. Past solutions to this inversion problem have relied upon analytic calculations for an unrealistically simplified subsurface, which can lead to uncertainties. To enhance the accuracy of this approach, a method based on transfer functions is proposed in which the function itself is calibrated using numerically generated overburden strain deformation calculated for a small select group of reference sources. This technique proves to be a good compromise between the faster but more accurate history match of the overburden strain using a geomechanical simulator and the slower, less accurate analytic method. Synthetic tests using a coupled geomechanical and fluid flow simulator for the South Arne field confirm the efficacy of the method. Application to measured time‐shifts from observed 4D seismics indicates compartmentalization in the Tor reservoir, more heterogeneity than is currently considered in the simulation model and moderate connectivity with the overlying Ekofisk formation.     

Quantitative integration of 3D and 4D seismic impedance into reservoir simulation model updating in the Norne Field

The ultimate goal of reservoir simulation in reservoir surveillance technology is to estimate long-term production forecasting and to plan development and management of petroleum fields. However, maintaining reliable reservoir models which honour available static and dynamic data, involve inherent risks due to the uncertainties in space and time of the distribution of hydrocarbons inside reservoirs. Recent applications have shown that these uncertainties can be reduced by quantitative integration of seismic data into the reservoir modelling workflows to identify which areas and reservoir attributes of the model should be updated. This work aims using seismic data to reduce ambiguity in calibrating reservoir flow simulation model with an uncertain petro-elastic model, proposing a circular workflow of inverted seismic impedance (3D and 4D) and engineering studies, with emphasis on the interface between static and dynamic models. The main contribution is to develop an updating procedure for adjusting reservoir simulation response before using it in the production forecasting and enhance the interpretive capability of reservoir properties. Accordingly, the workflow evaluates consistency of reservoir simulation model and inverted seismic impedance, assisted by production history data, to close the loop between reservoir engineering and seismic domains. The methodology is evaluated in a complex, faulted, sandstone reservoir, the Norne benchmark field, where a significant reservoir behaviour understanding (about the static and dynamic reservoir properties) is obtained towards the quantitative integration of seismic impedance data. This leads to diagnosis of the reservoir flow simulation reliability and generation of an updated simulation model consistent with observed seismic and well production history data, as well as a calibrated petro-elastic model. Furthermore, as Norne Field is a benchmark case, this study can be considered to enrich the discussions over deterministic or probabilistic history matching studies.     

Integration of rock physics and seismic inversion for rock typing and flow unit analysis: A case study

Rock typing and flow unit detection are more challenging in clastic reservoirs with a uniform pore system. An integrated workflow based on well logs, inverted seismic data and rock physics models is proposed and developed to address such challenges. The proposed workflow supplies a plausible reservoir model for further investigation and adds extra information. Then, this workflow has been implemented in order to define different rock types and flow units in an oilfield in the Persian Gulf, where some of these difficulties have been observed. Here, rock physics models have the leading role in our proposed workflow by providing a diagnostic framework in which we successfully differentiate three rock types with variant characteristics on the given wells. Furthermore, permeability and porosity are calculated using the available rock physics models to define several flow units. Then, we extend our investigation to the entire reservoir by means of simultaneous inversion and rock physics models. The outcomes of the study suggest that in sediments with homogeneous pore size distribution, other reservoir properties such as shale content and cementation (which have distinct effects on the elastic domain) can be used to identify rock types and flow units. These reservoir properties have more physical insights for modelling purposes and can be distinguished on seismic cube using proper rock physics models. The results illustrate that the studied reservoir mainly consists of rock type B, which is unconsolidated sands and has the characteristics of a reservoir for subsequent fluid flow unit analysis. In this regard, rock type B has been divided into six fluid units in which the first detected flow unit is considered as the cleanest unit and has the highest reservoir process speed about 4800 to 5000 mD. Here, reservoir quality decreases from flow unit 1 to flow unit 6.     

An application of 4D seismic monitoring technique to modern coal mining

To study the impact of modern coal mining on the overlying formation, a full‐life‐cycle four‐dimensional seismic monitoring study has been carried out. Four seismic data campaigns have been performed using flexi‐bin geometry with square bins, with total duration of 171 days. The four seismic datasets have been processed with the same processing workflow and parameters; major problems such as statics correction, signal‐to‐noise ratio, resolution, and consistency processing are addressed taking into account the geological features of the research area. This guarantees that remaining four‐dimensional differences between the time‐lapse datasets show mostly geological factors due to the coal mining and effects such as surface subsidence. Our four‐dimensional seismic monitoring of modern coal mining shows that mined and unmined areas have significant zoning characteristics; coal mining has a direct impact on the overlying formation. The mining leads to obvious event subsidence, which reflects that overlying formations undergo subsidence during the mining process. The overlying formation appears as two zones called caving zone and fractured zone. We determine the fault dip of the overlying formation at one end of the working face to be 56°or so by calculation and conversion. We also see that, during the coal mining process, over time, the overlying formation has a self‐recovery capability, which gradually strengthens from the roof siltstone upward to the Aeolian sandstone near the surface. The stability of 20‐m coal pillars between working faces displays a strengthening trend and remains safe during the mining process due to both coal seam supporting and formation compaction effects.     

Review Paper: A semi‐empirical model of strain sensitivity for 4D seismic interpretation

We reformulate the original model of Hatchell and Bourne and Røste, Stovas and Landrø that couples fractional velocity change to subsurface strain via a fundamental constant R. The new model combines elastic compressibility of a dual‐porosity system for a sand–shale mixture with horizontal planes of inter‐granular weakness. The majority of observed R‐factor magnitudes from post‐stack 4D seismic data in both the reservoir and overburden can thus be explained. R is predicted to depend strongly on lithology and also initial strain state. The model is also extended to predict the observed angle‐dependence of time‐lapse time‐shifts from pre‐stack data. An expression for the gradient of time‐shift with incidence angle is obtained in terms of the background V_P/V_S, and also the ratio of tangential to normal compliances B_T/B_N representing loss or creation of inter‐granular coupling. If accurately estimated from data, this compliance ratio can be used as an additional parameter to assess the post‐production state of the overburden. It is concluded that whilst R remains the over‐arching parameter controlling the magnitude of time‐shifts measured from 4D seismic data, B_T/B_N is a subtler parameter that may also prove of future value.     

Analysis of time‐lapse travel‐time and amplitude changes to assess reservoir compartmentalization

Fluid depletion within a compacting reservoir can lead to significant stress and strain changes and potentially severe geomechanical issues, both inside and outside the reservoir. We extend previous research of time‐lapse seismic interpretation by incorporating synthetic near‐offset and full‐offset common‐midpoint reflection data using anisotropic ray tracing to investigate uncertainties in time‐lapse seismic observations. The time‐lapse seismic simulations use dynamic elasticity models built from hydro‐geomechanical simulation output and a stress‐dependent rock physics model. The reservoir model is a conceptual two‐fault graben reservoir, where we allow the fault fluid‐flow transmissibility to vary from high to low to simulate non‐compartmentalized and compartmentalized reservoirs, respectively. The results indicate time‐lapse seismic amplitude changes and travel‐time shifts can be used to qualitatively identify reservoir compartmentalization. Due to the high repeatability and good quality of the time‐lapse synthetic dataset, the estimated travel‐time shifts and amplitude changes for near‐offset data match the true model subsurface changes with minimal errors. A 1D velocity–strain relation was used to estimate the vertical velocity change for the reservoir bottom interface by applying zero‐offset time shifts from both the near‐offset and full‐offset measurements. For near‐offset data, the estimated P‐wave velocity changes were within 10% of the true value. However, for full‐offset data, time‐lapse attributes are quantitatively reliable using standard time‐lapse seismic methods when an updated velocity model is used rather than the baseline model.     

The added value of joint PP–PS inversion for reservoir characterization: A case study using Jubarte PRM seismic data (offshore Brazil)

The added value of the joint pre-stack inversion of PP (incident P-wave and reflected P-wave) and PS (incident P-wave and reflected S-wave) seismic data for the time-lapse application is shown. We focus on the application of this technique to the time-lapse (four-dimensional) multicomponent Jubarte field permanent reservoir monitoring seismic data. The joint inversion results are less sensitive to noise in the input data and show a better match with the rock physics models calibrated for the field. Further, joint inversion improves S-impedance estimates and provides a more robust quantitative interpretation, allowing enhanced differentiation between pore pressure and fluid saturation changes, which will be extremely useful for reservoir management. Small changes in reservoir properties are expected in the short time between the time-lapse seismic acquisitions used in the Jubarte project (only 1 year apart). The attempt to recover subtle fourth-dimensional effects via elastic inversion is recurrent in reservoir characterization projects, either due to the small sensitivity of the reservoirs to fluid and pressure changes or the short interval between the acquisitions. Therefore, looking for methodologies that minimize the uncertainty of fourth-dimensional inversion outputs is of fundamental importance. Here, we also show the differences between PP only and joint PP–PS inversion workflows and parameterizations that can be applied in other projects. We show the impact of using multicomponent data as input for elastic seismic inversions in the analysis of the time-lapse differences of the elastic properties. The larger investment in the acquisition and processing of multicomponent seismic data is shown to be justified by the improved results from the fourth-dimensional joint inversion.     

Sensitivity analysis and application of time‐lapse full‐waveform inversion: synthetic testing and field data example from the North Sea,Norway

Time‐lapse refraction can provide complementary seismic solutions for monitoring subtle subsurface changes that are challenging for conventional P‐wave reflection methods. The utilization of refraction time lapse has lagged behind in the past partly due to the lack of robust techniques that allow extracting easy‐to‐interpret reservoir information. However, with the recent emergence of the full‐waveform inversion technique as a more standard tool, we find it to be a promising platform for incorporating head waves and diving waves into the time‐lapse framework. Here we investigate the sensitivity of 2D acoustic, time‐domain, full‐waveform inversion for monitoring a shallow, weak velocity change (?30 m/s, or ?1.6%). The sensitivity tests are designed to address questions related to the feasibility and accuracy of full‐waveform inversion results for monitoring the field case of an underground gas blowout that occurred in the North Sea. The blowout caused the gas to migrate both vertically and horizontally into several shallow sand layers. Some of the shallow gas anomalies were not clearly detected by conventional 4D reflection methods (i.e., time shifts and amplitude difference) due to low 4D signal‐to‐noise ratio and weak velocity change. On the other hand, full‐waveform inversion sensitivity analysis showed that it is possible to detect the weak velocity change with the non‐optimal seismic input. Detectability was qualitative with variable degrees of accuracy depending on different inversion parameters. We inverted, the real 2D seismic data from the North Sea with a greater emphasis on refracted and diving waves’ energy (i.e., most of the reflected energy was removed for the shallow zone of interest after removing traces with offset less than 300 m). The full‐waveform inversion results provided more superior detectability compared with the conventional 4D stacked reflection difference method for a weak shallow gas anomaly (320 m deep).     

Quantitative estimation of water storage and residence time in the epikarst with time‐lapse refraction seismic

The hydrodynamic characterization of the epikarst, the shallow part of the unsaturated zone in karstic systems, has always been challenging for geophysical methods. This work investigates the feasibility of coupling time‐lapse refraction seismic data with petrophysical and hydrologic models for the quantitative determination of water storage and residence time at shallow depth in carbonate rocks. The Biot–Gassmann fluid substitution model describing the seismic velocity variations with water saturation at low frequencies needs to be modified for this lithology. I propose to include a saturation‐dependent rock‐frame weakening to take into account water–rock interactions. A Bayesian inversion workflow is presented to estimate the water content from seismic velocities measured at variable saturations. The procedure is tested first with already published laboratory measurements on core samples, and the results show that it is possible to estimate the water content and its uncertainty. The validated procedure is then applied to a time‐lapse seismic study to locate and quantify seasonal water storage at shallow depth along a seismic profile. The residence time of the water in the shallow layers is estimated by coupling the time‐lapse seismic measurements with rainfall chronicles, simple flow equations, and the petrophysical model. The daily water input computed from the chronicles is used to constraint the inversion of seismic velocities for the daily saturation state and the hydrodynamic parameters of the flow model. The workflow is applied to a real monitoring case, and the results show that the average residence time of the water in the epikarst is generally around three months, but it is only 18 days near an infiltration pathway. During the winter season, the residence times are three times shorter in response to the increase in the effective rainfall.     

Wavelet transform-based seismic facies classification and modelling: application to a geothermal target horizon in the NE German Basin

At the geothermal test site near Groß Schönebeck (NE German Basin), a new 3D seismic reflection survey was conducted to study geothermal target layers at around 4 km depth and 150°C. We present a workflow for seismic facies classification and modelling which is applied to a prospective sandstone horizon within the Rotliegend formation. Signal attributes are calculated along the horizon using the continuous Morlet wavelet transform. We use a short mother wavelet to allow for the temporal resolution of the relatively short reflection signals to be analysed. Time-frequency domain data patterns form the input of a neural network clustering using self-organizing maps. Neural model patterns are adopted during iterative learning to simulate the information inherent in the input data. After training we determine a gradient function across the self-organizing maps and apply an image processing technique called watershed segmentation. The result is a pattern clustering based on similarities in wavelet transform characteristics. Three different types of wavelet transform patterns were found for the sandstone horizon. We apply seismic waveform modelling to improve the understanding of the classification results. The modelling tests indicate that thickness variations have a much stronger influence on the wavelet transform response of the sandstone horizon compared with reasonable variations of seismic attenuation. In our interpretation, the assumed thickness variations could be a result of variable paleo-topography during deposition of predominantly fluvial sediments. A distinct seismic facies distribution is interpreted as a system of thicker paleo-channels deposited within a deepened landscape. The results provide constraints for the ongoing development of the geothermal test site.     

Bayesian inversion of time‐lapse seismic data for the estimation of static reservoir properties and dynamic property changes

Seismic conditioning of static reservoir model properties such as porosity and lithology has traditionally been faced as a solution of an inverse problem. Dynamic reservoir model properties have been constrained by time‐lapse seismic data. Here, we propose a methodology to jointly estimate rock properties (such as porosity) and dynamic property changes (such as pressure and saturation changes) from time‐lapse seismic data. The methodology is based on a full Bayesian approach to seismic inversion and can be divided into two steps. First we estimate the conditional probability of elastic properties and their relative changes; then we estimate the posterior probability of rock properties and dynamic property changes. We apply the proposed methodology to a synthetic reservoir study where we have created a synthetic seismic survey for a real dynamic reservoir model including pre‐production and production scenarios. The final result is a set of point‐wise probability distributions that allow us to predict the most probable reservoir models at each time step and to evaluate the associated uncertainty. Finally we also show an application to real field data from the Norwegian Sea, where we estimate changes in gas saturation and pressure from time‐lapse seismic amplitude differences. The inverted results show the hydrocarbon displacement at the times of two repeated seismic surveys.     

Target-level waveform inversion: a prospective application of the convolution-type representation for the acoustic wavefield

Nowadays, full-waveform inversion, based on fitting the measured surface data with modelled data, has become the preferred approach to recover detailed physical parameters from the subsurface. However, its application is computationally expensive for large inversion domains. Furthermore, when the subsurface has a complex geological setting, the inversion process requires an appropriate pre-conditioning scheme to retrieve the medium parameters for the desired target area in a reliable manner. One way of dealing with both aspects is by waveform inversion schemes in a target-oriented fashion. Therefore, we propose a prospective application of the convolution-type representation for the acoustic wavefield in the frequency–space domain formulated as a target-oriented waveform inversion method. Our approach aims at matching the observed and modelled upgoing wavefields at a target depth level in the subsurface, where the seismic wavefields, generated by sources distributed above this level, are available. The forward modelling is performed by combining the convolution-type representation for the acoustic wavefield with solving the two-way acoustic wave-equation in the frequency–space domain for the target area. We evaluate the effectiveness of our inversion method by comparing it with the full-domain full-waveform inversion process through some numerical examples using synthetic data from a horizontal well acquisition geometry, where the sources are located at the surface and the receivers are located along a horizontal well at the target level. Our proposed inversion method requires less computational effort and, for this particular acquisition, it has proven to provide more accurate estimates of the target zone below a complex overburden compared to both full-domain full-waveform inversion process and local full-waveform inversion after applying interferometry by multidimensional deconvolution to get local-impulse responses.     

Inversion of multicomponent seismic time shifts for reservoir pressure and length: a feasibility study

Pressure drops associated with reservoir production generate excess stress and strain that cause travel‐time shifts of reflected waves. Here, we invert time shifts of P‐, S‐, and PS‐waves measured between baseline and monitor surveys for pressure reduction and reservoir length. The inversion results can be used to estimate compaction‐induced stress and strain changes around the reservoir. We implement a hybrid inversion algorithm that incorporates elements of gradient, global/genetic, and nearest neighbour methods and permits exploration of the parameter space while simultaneously following local misfit gradients. Our synthetic examples indicate that optimal estimates of reservoir pressure from P‐wave data can be obtained using the reflections from the reservoir top. For S‐waves, time shifts from the top of the reservoir can be accurately inverted for pressure if the noise level is low. However, if noise contamination is significant, it is preferable to use S‐wave data (or combined shifts of all three modes) from reflectors beneath the reservoir. Joint wave type inversions demonstrate improvements over any single pure mode. Reservoir length can be estimated using the time shifts of any mode from the reservoir top or deeper reflectors. We also evaluate the differences between the actual strain field and those corresponding to the best‐case inversion results obtained using P‐ and S‐wave data. Another series of tests addresses the inversion of the time shifts for the pressure drops in two‐compartment reservoirs, as well as for the associated strain field. Numerical testing shows that a potentially serious source of error in the inversion is a distortion in the strain‐sensitivity coefficients, which govern the magnitude of stiffness changes. This feasibility study suggests which wave types and reflector locations may provide the most accurate estimates of reservoir parameters from compaction‐induced time shifts.     

Combining discrete cosine transform and convolutional neural networks to speed up the Hamiltonian Monte Carlo inversion of pre-stack seismic data

Markov chain Monte Carlo algorithms are commonly employed for accurate uncertainty appraisals in non-linear inverse problems. The downside of these algorithms is the considerable number of samples needed to achieve reliable posterior estimations, especially in high-dimensional model spaces. To overcome this issue, the Hamiltonian Monte Carlo algorithm has recently been introduced to solve geophysical inversions. Different from classical Markov chain Monte Carlo algorithms, this approach exploits the derivative information of the target posterior probability density to guide the sampling of the model space. However, its main downside is the computational cost for the derivative computation (i.e. the computation of the Jacobian matrix around each sampled model). Possible strategies to mitigate this issue are the reduction of the dimensionality of the model space and/or the use of efficient methods to compute the gradient of the target density. Here we focus the attention to the estimation of elastic properties (P-, S-wave velocities and density) from pre-stack data through a non-linear amplitude versus angle inversion in which the Hamiltonian Monte Carlo algorithm is used to sample the posterior probability. To decrease the computational cost of the inversion procedure, we employ the discrete cosine transform to reparametrize the model space, and we train a convolutional neural network to predict the Jacobian matrix around each sampled model. The training data set for the network is also parametrized in the discrete cosine transform space, thus allowing for a reduction of the number of parameters to be optimized during the learning phase. Once trained the network can be used to compute the Jacobian matrix associated with each sampled model in real time. The outcomes of the proposed approach are compared and validated with the predictions of Hamiltonian Monte Carlo inversions in which a quite computationally expensive, but accurate finite-difference scheme is used to compute the Jacobian matrix and with those obtained by replacing the Jacobian with a matrix operator derived from a linear approximation of the Zoeppritz equations. Synthetic and field inversion experiments demonstrate that the proposed approach dramatically reduces the cost of the Hamiltonian Monte Carlo inversion while preserving an accurate and efficient sampling of the posterior probability.     

遗传优化神经网络方法在桥梁震害预测中的应用

本文将遗传算法与神经网络相结合,从而建立了一种高效的、实用的桥梁震害预测方法。根据遗传算法具有局部寻优的特点,为避免BP神经网络陷入局部极小值,本文将二者结合起来形成GA-BP混合算法,以GA优化神经网络的初始权值和阈值,对网络进行训练。在大量收集梁式桥震害资料的基础上,将此算法引入桥梁的震害预测中,并与传统的单独BP神经网络相比较,结果表明该方法能够有效、准确地对桥梁结构进行震害预测。     

Monitoring and volumetric estimation of injected CO2 using 4D seismic,petrophysical data,core measurements and well logging: a case study at Ketzin,Germany

More than 50 000 tons of CO₂ have been injected at Ketzin into the Stuttgart Formation, a saline aquifer, at approximately 620 m depth, as of summer 2011. We present here results from the 1^st repeat 3D seismic survey that was performed at the site in autumn 2009, after about 22 000 tons of CO₂ had been injected. We show here that rather complex time‐lapse signatures of this CO₂ can be clearly observed within a radius of about 300 m from the injection well. The highly irregular amplitude response within this radius is attributed to the heterogeneity of the injection reservoir. Time delays to a reflection below the injection level are also observed. Petrophysical measurements on core samples and geophysical logging of CO₂ saturation levels allow an estimate of the total amount of CO₂ visible in the seismic data to be made. These estimates are somewhat lower than the actual amount of CO₂ injected at the time of the survey and they are dependent upon the choice of a number of parameters. In spite of some uncertainty, the close agreement between the amount injected and the amount observed is encouraging for quantitative monitoring of a CO₂ storage site using seismic methods.