by Mr. Hamish Wilson, Ph.D. Candidate, School of Earth Sciences – University of Queensland
Thursday 23 May 2019, 11AM – 12PM, ARRC Auditorium
ABSTRACT: Moveout and wavefront parameter analysis, correction, and stacking are important steps in any seismic processing workflow. Moveout and wavefront parameters are utilised to approximate traveltime curves which best approximate the moveout of a reflector with offset and midpoint displacement. The arithmetic mean along each final approximated curve is then stored in a master trace at each time sample in a process known as correction and stacking. Global optimisation using the semblance operator as an objective function is utilised to determine the parameter set that best approximates the reflection surface moveout in the data at each time sample. Typically approximations for moveout in offset direction such as the Normal Moveout (NMO) approximation and approximations for moveout in both offset and midpoint displacement directions such as the Common Relfection Surface (CRS) and nonhyperbolic Common Reflection Surface (NCRS) approximations assume that the data in the subsurface is isotropic. This assumption is often too simplistic for real-world problems as real seismic data is generally anisotropic. Moveout approximations such as the generalized moveout approximation (GMA) have been utilised in recent times as a better approximation when it comes to determining nonhyperbholic perturbations associated with anisotropy and lateral heterogeneity than the NMO approximation. Although the GMA approximation has provided far superior accuracy and the ability to delineate anisotropy in 2D seismic data. There is a limitation in that it has no dependence on midpoint displacement that is accounted for in the CRS and NCRS approximations. Conversely the CRS approximations do not have any parameters associated with anisotropy.
Other limitations associated with parameter analysis and stacking are moveout stretch, and amplitude variations with offset (AVO). Stretching is the phenomena of lengthening the dominant wavelength of the reflection impulse with increasing offset and decreasing zero-offset time. This due to nonparallelism of the local traveltimes away from the onset of each reflection impulse which is a limitation of various moveout approximations. Class IIP AVO anomalies which are associated with a polarity reversal in the amplitudes with offset are known to augment the location of the true optima in the parameter space leading to incorrect parameter optimisation. Thus far I have outlined three issues that need to be remedied in the parameter analysis, correction, and stacking steps of the seismic workflow. A further step that may be an issue is the computational effort required to optimise large numbers of parameter sets at each zero-offset time.
In this talk I provide a new frame work that aims to remedy each of these issues. I firstly introduce a new nonhyperbolic moveout and CRS operator that amalgamates the GMA and NCRS approximations together to provide enhanced accuracy at the cost of more parameters in the optimisation process. This new approximation is called the GMA-NCRS approximation, it has the added advantage of being reducible to the GMA and NCRS approximations with various parameter substitutions, like the GMA approximation it can also embed different approximation types again with different parameter substitutions. I introduce two ways of finding the optimal parameter sets for the GMA-NCRS approximation at each layer rather than each sample . The first technique is partially user driven with users picking a point on each reflection that is utilised to constrain the potential zero-offset time locations for each other moveout parameter for the reflection of interest. The second technique uses multimodal optimisation via a sequential niching technique based on reflection filtering and removal from the the data. Both techniques can be utilised in conjunction with each other. These techniques aim to reduce the computational cost by only optimising each layer instead of sample whilst allowing user control on the amount of data-driven optimisation and user-driven optimisation.
To fix the issue of optima augmentation via AVO anomalies I introduce an AVO-friendly semblance operator that offers higher resolution for AVO-friendly moveout parameter analysis then the well known avo-friendly AB semblance operator. To remove stretch and interfering events I introduce a stretch free distribution for the proposed GMA-NCRS operator as a way to remove stretch from around the onset of a reflection surface and post stretch-free filtering to remove interfering events.
BIOGRAPHY: Hamish Wilson received a BSc (2011) in geological sciences from the University of Queensland, followed by honors majoring in exploration geophysics (2013), where he investigated the use of super-virtual interferometry for the enhancement of first-break picks for improved refraction statics. He is a PhD candidate in the School of Earth Sciences at the University of Queensland. Following his time as an undergraduate, he spent two years working full time as a geophysicist at Dayboro Geophysical, a seismic processing firm in Brisbane, Australia. In 2015, he started his PhD at the University of Queensland where he is investigating and developing techniques of moveout and velocity analysis to aid in the detection and interpretation of anisotropy. His research interests include methods of velocity analysis, semblance analysis, seismic processing methods, meta heuristic algorithms, machine learning, nonhyperbolic moveout analysis, anisotropy, common reflection surface stacking, moveout stretch removal, refraction statics, interferometry, and amplitude variation with offset methods.