Abstract
Volunteers are busy putting together the technical program for the SEG Annual Meeting in San Antonio. As members of the SEG Research Committee, Ray Abma and I are organizing the “Recent advances and the road ahead” special session. We invited several TLE special section editors to present hot topics. In addition, we invited a few authors of papers nominated by the editors of this column to present their cutting-edge research. I think it will be an interesting mix. The session will focus on seismic technology, and I wish we had more room for some of the great nonseismic innovative work. I am always impressed with the diversity and quality of the Geophysics papers nominated for this column, like the ones below.
The effect of the geometry of sample contaminants and the electrical response
The spectral induced polarization (SIP) method could be used to detect dense nonaqueous phase liquid contaminants (DNAPLs) in soil. To understand how NAPLs change the electrical response of sand, Johansson et al. use a new methodological approach in “Combining spectral induced polarization with X-ray tomography to investigate the importance of DNAPL geometry in sand samples.” High-resolution X-ray tomography and image analysis are used as visual and quantitative tools to aid the interpretation of SIP responses measured on the samples as shown in Figure 1. The results indicate that the geometrical distribution of DNAPLs in the pore space controls the SIP response. Current is transported along the surface of DNAPLs, and electrical charge-up effects are short circuited when the DNAPL is interconnected across several pores. X-ray tomography results also provide important insights into uncertainties related to the preparation of DNAPL-contaminated samples, since differences between the targeted and real DNAPL concentration are observed. Another interesting observation is made in SIP data from measurements repeated after X-ray tomography. The results show that conductivity increased significantly in the samples, which likely is an effect of X-ray interactions with sample materials. The combination of X-ray tomography and geophysical investigations of samples is a promising approach for understanding how geophysical signals arise in different materials.
Figure 1. (Figures 7c and 7d from Johansson et al.) 3D visualizations of X-ray tomography data showing the distribution of DNAPLs in two sand samples. All other sample constituents are transparent in the visualizations. The geometrical distributions of the DNAPLs were used for interpretation of measured SIP responses.
NMR measurements reveal permafrost thaw below arctic lakes
Parsekian et al. use surface nuclear magnetic resonance (NMR) to identify zones of thawed ground below lakes on Alaska's North Slope in their paper “Surface nuclear magnetic resonance observations of permafrost thaw below floating, bedfast, and transitional ice lakes.” NMR sounding is emerging as a powerful tool for groundwater investigations because of its direct sensitivity to unfrozen water. Surface NMR measurements are made using loops on the earth's surface rather than in boreholes. The loops for this study are 75 or 90 m circle loops located at the center of each lake of interest. The purpose of the investigation is to understand if annual lake ice conditions are related to the development of thawed permafrost below lakes. They find that bedfast lakes have no detectable unfrozen water below the lake, while floating ice lakes have high detectable water content from the lake and from sediments below the lake bottom. A key finding is that transitional ice lakes, which alternate between floating and bedfast ice conditions over multiyear timescales depending on winter ice growth and lake level conditions, have complex vertical unfrozen water content profiles attributed to sporadic periods of thaw. For the broader geophysics community, this paper may be interesting as an example of geophysical contributions to cryosphere research and for how geophysical field measurements play a role in understanding earth-system processes. Warming air temperatures could warm permafrost, resulting in mobilization of subsurface carbon gases into the atmosphere. This paper is open access, and the entire data set is freely available for download.
Time-lapse AVO inversion considering both solid and liquid phases of reservoirs
Zhi and Gu, in “Time-lapse amplitude variation with offset inversion using Bayesian theory in two-phase media,” present a joint rock-physics and seismic inversion model for time-lapse data used for monitoring and managing oil reservoirs. The conventional time-lapse amplitude variation with offset (AVO) inversion is based on Zoeppritz equations and assumes that the reservoir is a single-phase solid medium. The real-life reservoir is a two-phase medium, which consists of solid and fluid components. This means that Zoeppritz equations cannot describe the characteristics of the seismic reflection amplitudes in the reservoir in an accurate way. Zhi and Gu develop a method for time-lapse AVO inversion based on the theory of two-phase media. They use a reflection-coefficient equation in two-phase media, a rock-physics model, and a convolutional model to build a relationship between seismic records and reservoir parameters, which include porosity, clay content, saturation, and pressure. Assuming that the seismic data errors follow a zero-mean Gaussian distribution and that the reservoir parameters follow a four-variable Cauchy prior distribution, they use the Bayesian theory to construct the objective function for AVO inversion. In addition, they add a model constraint term to compensate for the lack of low-frequency information and improve the stability of inversion. Using the objective function of AVO inversion and the Gauss-Newton method, they derive the equation for time-lapse AVO inversion. This result can be used to estimate reservoir parameters and their changes accurately and robustly. Figure 2 illustrates some of the results for field data.
Figure 2. (Figures 18e and 18f from Zhi and Gu) Time-lapse AVO inversion results using field data showing (e) the change in water saturation and (f) the change in effective pressure. The red rectangle is the major region of reservoir variation.
Reliability of seismic wavefront attributes for complex environments
Xie and Gajewksi compare methods of obtaining 3D wavefront attributes in their paper, “Reliability of data-driven wavefront attributes in laterally heterogeneous media.” Wavefront attributes (slopes and curvatures) are often used in seismic processing methods such as prestack data enhancement, diffraction separation and imaging, and wavefront tomography. Such attributes are obtained by fitting an appropriate multiparameter stacking operator to the data. Because these operators describe the offset behavior of events using an analytical function, questions arise about their performance. How well do they perform in complex media, or more generally, how reliable is time processing in complex media? To quantify these questions, a comprehensive study is performed using the 3D SEG salt model using the common reflection surface operator for data fitting. Since the velocity model is known, wavefront attributes can be directly computed by kinematic and dynamic ray tracing (model-driven wavefront attributes) and compared with the wavefront attributes determined by data fitting from synthetic seismograms (data-driven wavefront attributes). Approximately 80% to 90% of the total picks show good match with a relative error of less than 10% when a semblance threshold of 0.1 is considered in the automatic picking process. Increasing the semblance threshold leads to fewer picks but a better match. The results confirm the validity of the determined wavefront attributes and lead to the conclusion that wavefront attributes can be reliably determined even in complex environments like the 3D SEG salt model.
Adaptive waveform inversion applied to field data
In “Adaptive waveform inversion: Practice,” Guasch et al. describe the theoretical background behind a previously published methodology aimed at overcoming cycle skipping in full-waveform inversion (FWI). They accompany it with a series of tests using field data examples and discussion of practical issues. The field tests are performed with a 3D ocean-bottom data set acquired over a pervasive gas cloud in the North Sea. They compare conventional FWI, which uses the standard misfit based on the subtraction of original and simulated data sets, and adaptive waveform inversion (AWI), which uses a misfit based on the ratio of two data sets in the frequency domain. The method can be implemented by designing Weiner filters that match one data set to the other and then modifying the earth model so that these filters trend toward unit-amplitude delta functions at zero time lag. When starting from 3 Hz and using similar good starting models, FWI and AWI generate similar results. When the starting model is less accurate and/or low frequencies are not present in the data, AWI can converge successfully when FWI cannot. They suggest that AWI can be used with raw unprocessed field data without a tomography step and then follow AWI with conventional FWI for improved resolution.
Efficient coding of large-scale seismic inversion algorithms
In “A large-scale framework for symbolic implementations of seismic inversion algorithms in Julia,” Witte et al. describe new developments in how to code complex geophysical algorithms in a concise way. Subsurface seismic imaging and parameter estimation are among the most computationally challenging problems in the scientific community. Codes for solving seismic inverse problems, such as FWI or least-squares reverse time migration (LS-RTM), need to be highly optimized, but at the same time, facilitate the implementation of complex optimization algorithms. Traditionally, production-level codes in the oil and gas industry were exclusively written in low-level languages, such as C or Fortran, with extensive amounts of manual performance optimizations, thus making code maintenance, debugging, and adoption of new algorithms prohibitively challenging. Witte et al. present a paradigm of software engineering for seismic inverse problems based on symbolic user interfaces and code generation with automated performance optimization. Inspired by recent deep learning frameworks, the Julia Devito inversion framework (JUDI; an open-source software package) combines high-level abstractions for expressing seismic inversion algorithms with a domain-specific language compiler called Devito for solving the underlying wave equations. Devito's generated code is compiled just in time and outperforms codes with manual performance optimizations. JUDI utilizes Julia's high-level parallelization, making the software easily adaptable to a variety of computing environments such as densely connected HPC clusters or the cloud. The numerical examples (Figure 3) demonstrate the ability to implement a variety of complex algorithms for FWI and LS-RTM in a few lines of Julia code and run it on large-scale 3D models. The paper concludes that abstractions and performance are not mutually exclusive, and use of symbolic user interfaces can facilitate the implementation of new and innovative seismic inversion algorithms.
Figure 3. (Figure 3 from Witte et al.) Demonstration of JUDI on a large-scale problem with a 1.2 TB data set and 9400 shots for a 3D overthrust model. Depth slice through (a) the model, (b) the initial model, and (c) the recovered model after 15 iterations. Some parts of the recovered model are cycle skipped, but overall the algorithm is able to converge.
Other papers nominated by the editors
Wu et al. — “Semiautomatic first-arrival picking of microseismic events by using the pixel-wise convolutional image segmentation method”
Liu et al. — “Detecting the propped fracture by injection of magnetic proppant during fracturing”
