Scott Singleton

Q Estimation using Gabor-Morlet Joint Time-Frequency Analysis Techniques

Summary Two new methods of Q estimation are presented for the first time. Both are based on Gabor-Morlet spectral decomposition. They differ substantially from traditional Q estimation methods which rely on the comparison of spectral characteristics of a shallow time window (approximating a seismic wave prior to encountering attenuation) with those of a deeper time window (after the wave has been attenuated). Instead, the fundamental principle in both new methods is a technique with which the spectral characteristics of each time interval are determined. We consider two data volumes. The first is the original data, while the second is the original data after it has compensated for attenuation effects by use of GaborMorlet spectral balancing. Both methods assume dispersion can be ignored, which of course is not strictly true because the Kramers-Kronig relations are then not satisfied. However, to a first approximation they give reasonable

Calibration of prestack simultaneous impedance inversion using rock physics

This paper represents the third installment in a sequence of reservoir characterization articles that follow a workflow whose ultimate goal is to understand and predict the rock properties of a reservoir away from well control. The first article (Singleton, 2008) described a method to detect anomalous seismic attenuation. This was important because seismic data at the well location have a steep falloff in amplitude caused by high attenuation in and above the reservoir interval. The second article (Singleton, 2009) dealt with the seismic gathers that were input into the inversion process. It is of the utmost importance that the quality of data input into a simultaneous prestack inversion be as high as possible. Errors in seismic gathers (low SNR, resolution loss from excessive NMO stretch, nonflat reflectors, offset amplitude problems, multiples, etc.) are broadcast straight into the inversion impedance volumes, so this is a critical step in geophysical reservoir characterization.

Novel Use of P- and S-Wave Seismic Attenuation for Deep Natural Gas Exploration and Development

The well selected for the application of our attenuation theory and extraction of attenuation attributes from seismic data is the Texaco well (API 177104132700) in Block 313 of Eugene Island in the Gulf of Mexico (Well 2700). The rock physics diagnostics indicates that the rock can be described by the uncemented (soft-sand) model. This model is used to predict the S-wave velocity that was missing in the original well data. The Pand S-wave inverse quality factors are computed according to our theoretical model. The ratio of these inverse quality factors (P-to-S) is small (on the order of one) in wet rock and large in the gas zone. The seismically-measured attenuation ratio may serve, therefore, as an indicator of hydrocarbons. The synthetic seismic traces computed using the well data and the ray-tracer with attenuation, specifically developed for this project, indicate that attenuation affects the seismic response and, therefore, can be extracted from real seismic data, including the P-to-P and P-to-S reflection amplitude. Rock Physics Diagnostics – Model for Velocity The gas saturation in the well was calculated from the resistivity curve while the clay content was estimated by linearly scaling the gamma-ray curve between its minimum and maximum values. It was assumed that the formation water has the bulk modulus 2.85 GPa and density 1.01 g/cc while the gas has the bulk modulus 0.14 GPa and density 0.26 g/cc. The total porosity was calculated from the bulk density by assuming that the density tool samples the virgin formation with gas saturation as calculated from resistivity. The measured impedance and P-wave velocity are compared to the curves due to the uncemented (soft-sand) model. The proximity of the data and model (Figure 1) indicates that this model is appropriate for the well under examination. This model was then use to predict the S-wave velocity (absent in the measured data) from the P-wave velocity. The in-situ impedance is plotted versus the total porosity and Poisson’s ratio (PR) in Figure 2 where the data are color-coded by gamma-ray and by water saturation. Similar cross-plots are shown in Figure 3 but for wet conditions where the elastic properties and density were calculated using the P-wave-only fluid substitution. The soft-sand model curves for water-saturated rock are superimposed upon the wet-condition data to further emphasize the relevance of this model. The curves are produced for varying porosity and each for fixed clay content. The latter variable changes from one to zero with step 0.2. These model curves fully encompass the well log data.

Evaluation of an inelastic (Q) synthetic seismic generator

Using a rigorous process for log correction and the appropriate rock physics models, a suitable log suite was prepared for usage in this experiment. Dvorkin’s Heterogeneous Q model was then applied to create a Q log. This input data was used to create several full waveform, normal incident Kennett synthetics, each showing a different wave effect or fluid case. These were then quantitatively analyzed.

Seismic Gas Hydrate Quantification using Cumulative Attributes (CATTs) at Milne Point, AK

Summary This study concludes that vast quantities of gas hydrate resources (>400 BCF) are trapped primarily by Eocene river sands of the Mikkelsen Tongue of Canning Fm. within a 12.5 mi 2 (33.375 km 2 ) area beneath Milne Point, on the North Slope of Alaska. This particular resource quantification was made possible by application of a new type of cumulative seismic attribute (CATT) that allows determination of accumulated gas hydrate volume in seismic data. Underlying this attribute is a rock physics effective-medium model that links hydrate saturation to seismic response. It is based on the“soft sand” model which includes methane gas hydrate as a solid constituent of the rock matrix.

The effects of seismic data conditioning on pre‐stack simultaneous impedance inversion

The demands that reservoir characterization place on seismic data far outweigh those of traditional structural interpretation. Because of this, gather conditioning is seen by many as a prerequisite to pre-stack inversion. This paper discusses three conditioning processes—signal/noise (S/N) improvement, stretch removal, and reflector alignment. It then seeks to document the improvements that these processes achieve in the gathers and in the inversion. Specifically, the gathers were measured for amplitude versus offset (AVO) fit using a two-term Shuey equation and found to be improved by 20%. A comparison of wavelets extracted from the angle stacks found amplitude and phase spectra to be much more stabilized, even out to the far angle stack. The far angle stack seismic/synthetic inversion residuals showed a 40% drop in amplitude and completely different frequency and reflector character. Finally, the acoustic impedance (AI) / shear impedance (SI) cross-plot showed a much more compact signature that allowed lithology and pay discrimination. Conversely, the raw data cross-plot contained noisy data that entered into the area of the polygon where the pay signature lay. Geobodies captured from improperly conditioned data are thus (1) inflated in size, and (2) have lower impedances. These errors, in turn, lead to incorrect rock property and reserve estimations.

Monte Carlo AVO Analysis For Lithofacies Classification

The objective of this work is to use AVO intercept and gradient, in conjunction with well-log petrophysics analysis, to discriminate and classify lithofacies in a shaly sand reservoir. Careful log and core analysis, and rock physics modeling was used to identify the important seismic litho-classes. Monte Carlo AVO simulations based on statistical rock physics were used to set up the classconditioned probability distributions (pdfs) of intercept and gradient. The effect of thin-layer anisotropy on the probability distributions of AVO intercept and gradient was considered by simulating various realizations of sand-shale thin layers. Monte Carlo simulations, by taking into account distributions of values instead of single average values, help to avoid the flaw of averages (Mukerji and Mavko, 2005). Monte Carlo simulations also give us confidence intervals and other measures of uncertainty. Computations using averages and average trends alone do not give any indication of the uncertainty due to the variability in the properties. The pdfs were then used to classify the seismic AVO intercept and gradient cubes to estimate the most-likely facies and obtain lithofacies probability cubes.