Preconditioned high-resolution imaging
We examine the application of non-quadratic regularization methods to the solution of inverse problems that arise in seismic imaging. Non-quadratic regularization is implemented via operator pre-conditioning. In particular, we show that non-quadratic constraints can be used to mitigate imaging and sampling artifacts. Walk-away VSP data provides an excellent scenario to test these ideas. Limited aperture and sampling limitations are the main source of artifacts in walk -way VSP imaging. We device a non-quadratic regularization method that uses a priori information about the dip field of the image. The latter leads to an iterative algorithm that reduces imaging artifacts and enhance the resolution of seismic reflectors. We finally discuss potential application of preconditioned high-resolution imaging to global seismology.
Imaging Plumes and Slabs Using High-resolution Radon Transforms
The success of Radon-based methods in exploration seismology can be potentially duplicated in global seismic studies of crust and mantle reflectivity structures. Apart from the obvious scale difference between exploration (<20 km) and global (typically >100 km) problems, key objectives such as signal (or event) isolation and enhancement, noise reduction, and spatial/temporal interpolation are nearly independent of the applications. This study uses long-period, low-amplitude seismic arrivals to demonstrate the flexibility of joint Radon and frequency domain approaches. By exploiting the move-out or curvature of signal of interests we are able to accurately determine the depth and reflection amplitude of mantle reflectors beneath major hotspots and the western Pacific subduction zones. We observe a systematically depressed 410-km discontinuity relative to the global/ocean average beneath the majority of the hotspots, which we interpret as evidence of high temperatures above and/or within the upper mantle transition zone. The 660-km seismic discontinuity beneath hotspots is complex and its impedance contrast is smaller than expected from a spherically symmetric Earth model. In comparison, the transition zone phase boundaries are locally broadened along the Wadati-Benioff zones in the western Pacific subduction system. Superior depth anti-correlations and scattering amplitudes are identified beneath the Kermadec-Tonga subduction zone. Our high-resolution radon transform method also provides new evidence for a series of mantle reflectors at 200-300, 500-550, 800-850 and 900- 950 km depths beneath many of the hotspots and parts of the western Pacific. Amplitudes, ray angles and depths of these intermittent seismic reflectors are highly variable within each study area, as are their origins that likely involve both thermal and compositional variations within the top 1000 km of the mantle. The spatial variability of weak mantle reflectors is supported by results from Radon imaging of P wave records.
Adjoint Tomography: theory and applications
We show that the adjoint tomography approach provides a general framework to iteratively invert for the earth elastic and anelastic structural parameters as well as internal boundary undulations based upon an 3D initial model when 3D numerical simulations of wave propagation are used as the forward modeling technique. A key component of the adjoint tomography is to compute the Fréchet derivatives of a misfit function with respect to model parameters by interacting the forward wavefield s(x,t) and the adjoint wave field s†(x,t), generated by fictitious sources at the receivers that are related to the measurements. A direct application of the adjoint method involves computing the 'banana-doughnut' kernels at both global and regional scales. However, for actual tomography inversions, 'event kernels' are calculated instead to adapt to the setup of 3D numerical simulations, and innovative optimization schemes are applied to speed up the convergence of the iteration process. We also show potential applications of the adjoint tomography methods to improve our knowledge of the earth structure at both regional and global scales, including 3D tomography inversions in southern California, finite- frequency sensitivity kernels for seismic array analysis, and study of seismic structure at the CMB region.
Scattering - a probe to Earth's small scale structure
Much of the short-period teleseismic wavefield shows strong evidence for scattered waves in extended codas trailing the main arrivals predicted by ray theory. This energy mainly originates from high-frequency body waves interacting with fine-scale volumetric heterogeneities in the Earth. Studies of this energy revealed much of what we know about Earth's structure at scale lengths around 10 km throughout the Earth from crust to core. From these data we can gain important information about the mineral-physical and geochemical constitution of the Earth that is inaccessible to many other seismic imaging techniques. Previous studies used scattered energy related to PKP, PKiKP, and Pdiff to identify and map the small-scale structure of the mantle and core. We will present observations related to the core phases PKKP and P'P' to study fine-scale mantle heterogeneities. These phases are maximum travel-time phases with respect to perturbations at their reflection points. This allows observation of the scattered energy as precursors to the main phase avoiding common problems with traditional coda phases which arrive after the main pulse. The precursory arrival of the scattered energy allows the separation between deep Earth and crustal contributions to the scattered wavefield for certain source-receiver configurations. Using the information from these scattered phases we identify regions of the mantle that shows increased scattering potential likely linked to larger scale mantle structure identified in seismic tomography and geodynamical models.
Microseismic Monitoring Using Surface and Borehole Seismic Stations in an Oil Field, North Oman
Five shallow borehole seismic stations were installed to monitor microearthquake activities in a carbonate oil field in northern Oman since 1999. This shallow network of seismic station operated continuously until 2002 after which intermittent seismic recording took place due to lack of maintenance and failure of some stations. The objectives of the study are to determine the microseismic parameters in the oil field and to determine the spatial and temporal distribution of these events to evaluate possible triggering mechanism. Well over 400 microearthquakes per year were recorded in the first three years of operation and after that the level of seismic recording fell to less than 200 microearthquakes per year due to failure of some stations. In March 2008, temporary seismic experiment consisting of five near surface seismic stations were installed in the oil field to augment the shallow network station and to evaluate surface installment of seismic instrument to monitor microseismic activities. It has been recognized that microearthquakes data such as size, spatial, and temporal distribution provide information on the pressure waves initiated by either production of or injection of fluids into reservoirs. A total of 44 local microearthquake events were analyzed and located during the temporary seismic stations deployment using a non-linear location software that allows the use of variable accurate velocity model of the subsurface. The events location is confined to oil field reservoir boundary during the recording period and more events occurring at shallow depth. The correlation coefficient between gas production and number of events is the higher compared with the oil production or water injection. The focal plane solution for the largest event in the sequence indicates normal faulting with extensional stress consistent with the existing mapped normal faults in the oil field. Microseismic signal clearly detected by the collocated sensors of the near surface seismic experiment stations and augment the shallow borehole network signal for detection, analysis and locations.
Subsurface Images Retrieved From Ambient Noise Using Seismic Interferometry
In recent years, the process of generating seismic traces from the crosscorrelation of existing traces (Seismic Interferometry or SI) has gained rapidly in popularity among academia and industry. One application of SI is the retrieval of the Green's function form the crosscorrelation of ambient seismic noise. The complete Green's function is retrieved when the noise sources illuminate the recording stations from all directions. When the predominant recoded noise represents surface-wave arrivals, then the crosscorrelation will retrieve only the surface-wave part of the Green's function. To retrieve body-wave arrivals, like reflections, one needs to correlate noise arrivals that have propagated in the deeper subsurface. The retrieval of the reflection part of the Green's function has proven more challenging as the surface-wave noise normally drowns out the subtler body-wave noise. On the other hand, retrieval of reflections is very desirable, as they afford the construction of subsurface velocity models and subsurface reflection images with higher resolution than provided by surface-wave tomography. At the end of 2007, Shell carried out a passive seismic experiment recording approximately 11 hours of ambient noise in a desert in North Africa. The field geometry consisted of 8 parallel lines with 50 m station spacing and 500 m spacing between the lines. The recorded noise was dominated by strong surface waves, concentrated mainly below 6 Hz, which were caused by a traffic road that crossed the survey at its Northern section. We apply SI to the ambient noise with the aim to retrieve the reflection part of the Green's function. For this reason, we suppress the surface-wave energy before crosscorrelation. The result of the crosscorrelation is a so-called common-shot gather, i.e., a response from one virtual shot recorded by all receivers on a line. We use all the retrieved common-shot gathers in a processing scheme commonly used in the exploration seismics to extract velocity information of the subsurface and enhance reflection arrivals. The end result of the processing consists of stacked time-migrated subsurface sections showing the subsurface structures. Our results exhibit several laterally coherent events. We compare these sections to sections obtained from an active seismic reflection survey along the same lines. The comparison shows that several shallow marker events have been adequately reconstructed from the noise. Having confirmed that the coherent events are subsurface reflectors, we use the retrieved common-shot gathers in a more advanced processing step, so-called prestack depth migration. In this processing we use a constant-gradient subsurface velocity model, which is based on the velocities that we estimated from the retrieved common-shot gathers. The end result is an image of subsurface volume below the survey lines.