ABSTRACT No preview is available for this paper.

Summary A method of delay-time refraction analysis is proposed that provides robust estimates of both short and long period statics. The method derives an initial model of delay times and a spatially variant refractor slowness field from a slopeintercept analysis of first arrival times in the common midpoint domain. These initial model parameters are iteratively updated by inverting a combination of first arrival times and model constraints using a sparse conjugate gradient algorithm. Parameter constraints are applied by means of first or higher order regularization operators. Interpolation is used to provide a uniform grid that supports these operators. Stacks generated with statics derived from the algorithm will illustrate the effectiveness of the proposed method. Introduction Delay-time methods are conceptually simple, computationally efficient, and easily extended from two dimensional to crooked-line or three-dimensional acquisition geometries. Initial formulations of the delay-time (Gardner, 1939) and time-term (Scheidegger and Willmore, 1957; Willmore and Bancroft, 1960) methods express the total travel time of a refracted arrival between a source (S) and receiver (R) as an equation of three variables: ?? sr = ?? s + ?? r + x/v, (1) where ?? s is the source delay time, ?? r is the receiver delay time, X is the source-receiver offset, and V is a constant propagation velocity for the refractor. Given an estimate of the constant refractor velocity, a collection of first arrival times provides a system of equations that can be solved for source and receiver delay times in a least-squares sense. However, the refractor velocity is rarely a constant. Approximating it as such forces errors into the long wavelength components of the statics solution. To reduce the influence of dip on estimates of refractor depth made from slope and intercept calculations, MacPhail (1967) introduced a method of analyzing refraction arrival times in the CMP domain. Analogous to estimates of stacking velocity made from the analysis of reflection hyperbola in the CMP domain, the effects of dip on estimates of slope and intercept in the CMP domain are greatly reduced relative to the shot or receiver domains. Ruhl (1995) extended the CMP method to 3D. To accommodate spatially varying refractor geometries, Kirchheimer (1988) and Taner et al., (1988) each present solutions to the delay time equation that are dimensionally agnostic and model the refractor velocity field as an array of slowness cells whose values can vary spatially:

ABSTRACT Refraction seismic data has been conventionally processed for traveltime analysis for the purpose of resolving the near-surface velocity structures. In this study I present a refraction wavefield imaging approach that can migrate refractions from multiple refractors automatically. The approach involves preprocessing each reciprocal shot gathers and generates a number of new gathers by two loops of correlations. Migrating the resulted gathers with overburden velocities just like reflection migration can image refractors in depth. This approach does not require picking reciprocal times, and it can be applied for both shallow near-surface structure imaging with conventional recording geometry and deep subsurface imaging with long-offset recording geometry.

Summary A large 3D/3C seismic survey will be acquired in northeast British Columbia (NEBC) in 2012. In order to assist in the design of this survey, a high resolution multicomponent 2D line was acquired to give information about shear wave properties particularly for the near-surface. Shear (SH) and compressional (P) sources and 3-component geophones were used in this acquisition. Knowledge about the velocity-depth structure of the near-surface and feasibility of acquiring multicomponent data were the main drivers for this acquisition. Also, it is important to know the structure of the near-surface for the investigation and integration with deeper data. As a result, Vp/Vs analysis was carried out for both the shallow and deep formations. The analysis targeted both the near surface study and deeper horizon registration for PP and PS reflection data. From the shallow P-wave data, one refractor was detected and the presence of a glacial channel was confirmed at the east end of the line. The depth of this refractor ranges from 140 m to ~230 m in the channel. The average velocity for the first layer is 1950 m/s and for the second layer is 2800 m/s. From the S-wave data a different model was determined, with two refractors detected to the west end of the line and one refractor to the east of the line. The depth of the first refractor is ~70 m and the second ~140 m; to the east the refractor detected is at ~180 m. The S-velocity for the first layer is 350 m/s to 420 m/s, and 650 m/s to 800 m/s for the second layer to the west and 1400 m/s for the third layer. Finally static corrections for the reflection analysis were computed. For SH-wave data the static corrections range from -150 ms to -250 ms and for the P-wave data, values range from 10 ms to -15 ms. Vp/Vs analysis was performed for the near-surface structure with the values of velocities obtained from the shallow analysis and also through PP-PS registration for deeper structures. Good agreement was observed when comparing these results to Vp/Vs values from well log data. Values of Vp/Vs ranged from 5 in the near-surface to 2.2 in deeper formations.

Summary Characterization of the shallow structures was performed by using different approaches analysing both P- and Swave seismic data. In this way, we obtained seismic information with different resolution. The refraction tomography provided P and S velocity information about the first 80 m, while the conventional analysis gives a reasonable stacked velocity field until 300 m for both data. So, we estimated the Vp/Vs ratio and an empirical relationship between the two velocities. The conventional reflection seismic section and the multi-refractor image were obtained from P-data processing. In this way, we characterized the shallow strata (using tomographic velocity) and deep strata (using seismic images with different resolution. Moreover, the integration of different data and analysis gives an important contribution to the seismic interpretation. Introduction Shallow seismic refraction methods are usefully applied to many engineering, environmental and exploration targets. In addition, these methods are widely used to compensate for the weathered layer, which is one component of the static corrections, in the seismic reflection data processing. Recently, many authors suggested to use both reflected and refracted events to better image the subsurface (i.e., De Franco, 2005). Moreover, in the last years, the integration between compressional (P) and shear (S) wave analysis has been developed for shallow and deep targets. Shallow land seismic data are always affected by the presence of noise, such as environmental noise, ground roll and air wave. Often, it is very difficult to attenuate these noises, which can strongly attenuate the reflectors. So, the analysis of the refraction events can help to avoid this problem. Several techniques are available to analyze these events, and one of these is the multi-refractor imaging (De Franco, 2005). Multi-refractor imaging is a technique for constructing a single two-dimensional image of a number of refractors by stacking multiple convolved and crosscorrelated reversed shot records. The major advantage of the multi-refractor imaging method is the sum of all data in order to maximize the signal-to-noise ratio before the measurement of each traveltime. However, the signal-tonoise ratio can be further increased if only the traces that have arrivals from the same refractor are used, and if the correct reciprocal times or traces are employed. This method can produce a section similar to the familiar reflection section with substantially higher signal-to-noise ratio for equivalent interfaces. Integrating the two sections, it is possible to improve the interpretation of seismic data in the shallower part. The integration of the analysis of the P- (Vp) and S-wave (Vs) velocities allows information about the Poisson ratio, which is related to petrophysical characteristics of the subsoil. Moreover, the correlation between the Vp and Vs helps to detect fluid and the lithological changes at an interface. The Vs can be estimated by using two approaches. Sometimes, the seismic acquisition is performed using three-component geophones and a Pgenerator source; in this way, the P-converted wave is analyzed in order to obtain information about the Vs. This approach requires a complicate processing to obtain a stacked section.

ABSTRACT Karsting and other near-surface complexity often cause significant long and short wavelength statics problem in Permian Basin seismic data. Refraction statics using refracted first arrivals are often applied to the data in an attempt to model the near-surface adequately to resolve these statics problems. But even refraction statics, whether they be computed using refraction tomography or conventional delay-time methods, are often inadequate to resolve this complex issue. We present a dataset from the western Permian Basin where both tomographic and conventional delay-time first-break refraction approaches were deemed lacking overall, even though they partially attenuated the statics issues. Therefore, we took a somewhat radical approach and picked a refractor that for much of the survey was not a first break. This "refractor-driven" approach coupled with refractor model constraints resulted in a markedly superior result compared to the "first-break driven" solutions. Presentation Date: Monday, October 17, 2016 Start Time: 4:35:00 PM Location: 150 Presentation Type: ORAL

ABSTRACT. Measurements on the amplitude and pulse shape of seismic events are not usually made. In refraction work it is easy to do this and as magnetic recorders increase in dynamic range it may soon be possible with reflection records. This paper discusses the disturbing effects of variations in the source, the instruments, and the geophone plant. It is concluded that whereas they have no serious effect on the seismic wave form they cause a large amount of unwanted scatter in the amplitudes. Nevertheless, amplitudes give more information than do frequencies. Results are given for measurements on waves from refractors which are thick, thin, and anticlinal. For thick, flat-lying refractors the residual attenuation after allowing for the geometrical spread of energy is consistent with it being due mainly to absorption. For sedimentary refractors the residual attenuation is nearly 1.0 db/wave length for the predominant wave lengths whereas for a metamorphic and two igneous refractors it is about 0.1 db/wave length, or less. Practically no frequency variation was observed and this is not consistent with the residual attenuation depending upon frequency to any power higher than one. As the thickness gets less than about one third of a wave length the residual attenuation increases rapidly. This provides a useful method for determining whether or not the refracted arrivals come from a thin stringer. It should be possible eventually to determine the thickness of a layer by making amplitude and frequency measurements on the refracted arrival. When shooting over an anticline the refracted arrivals attenuate at an increased rate on the side of the crest closer to the shot-point and at a lesser rate on the opposite side. Guided waves are mentioned and it is suggested that they may be used to obtain the time correction for the low velocity layers when doing long distance refraction work. RESUME. On n'effectue généralement pas de mesures de l'amplitude et de la forme des impulsions des manifestations sismiques. Dans les travaux de réfraction, il est facile de le faire et, comme la gamme dynamique des enregistreurs magnétiques augmente, il devrait être bientôt possible d'enregistrer les réflexions. Cet exposé discute les effets perturbateurs des variations dans la source, des instruments et du placement du géophone. I1 conclut que, bien qu'ils n'aient pas d'effet sérieux sur la forme des ondes sismlques, ils provoquent un éparpillement excessif et indésirable dans les amplitudes. Néanmoins, les amplitudes donnent plus de renseignements que les fréquences. Le mémoire donne des résultats de mesurages sur des ondes partant de réfracteurs épais, minces et anticlinaux. Pour des réfracteurs épais et plats, I'attGnuation résiduelle, après avoir tenu compte de l'extension géométrique d'énergie est

Summary It is very difficult to obtain reasonable near-surface velocity models with rugged topography. Generally the conventional methods, such as a shallow refraction survey, up-hole survey and so on, do not work well themselves in areas with a rugged surface. The subsequent interpolation in building velocity models adds more to the unreliability of the so-built model. To solve this problem, we developed a new method for estimating near-surface velocity models with rugged topography using refraction travel-time based on the assumption of slow lateral near-surface velocity variation, and the assumption to the refractor is relaxed to two inclined flat ones. Since only the refraction travel times and the elevation differences between the corresponding sources and receivers are needed and the weathering layer velocity (LVL velocity) and interpolation are saved during obtaining near-surface velocities, more reliable nearsurface velocity models can be built with this method for areas with rugged topography. Application of this method to both synthetic data and real data has obtained satisfying results. Introduction At present, there are many different methods for estimating near-surface velocity models in seismic exploration. They fall into two main categories, the refraction-based methods and the tomography-based methods. The refraction methods include the intercept-time method, the plus-minus method, GRM, EGRM, decomposition and so on (Cox, 1999). With these methods, if the LVL velocity V0 is known, the velocities of the high-velocity layer can be directly calculated from refraction travel-times. Unfortunately, V0 often has to be obtained from shallow refraction or up-hole surveys and the interpolation in creating V0 and the subsequent velocity model building adds much error to the results. Therefore, near-surface velocity models built with these methods for areas with rugged topography often have very low reliability. For the tomography-based methods (Chen et al, 2002; Xu et al, 2002), many weak points encountered in the refraction-based methods are avoided, but two other crucial problems arise: efficiency and precision. Both of these are related to the initial model. The closer the initial model is to reality, the faster the convergence and the higher the precision of the inversion results, and vice versa. In practice, a constant velocity is often used as the initial model, and this means not only high computational cost, but also possible local convergence to an erroneous result. This is especially true for areas with rugged topography. Motivated by the above mentioned weak points of the existing methods, we propose a new method for creating near-surface velocity models for rugged areas. With the assumption of a slow lateral near-surface velocity variation, the method obtains near-surface velocities using only refraction travel times and the corresponding source and receiver elevations of a surface seismic survey. The refractions used could be from two different inclined refractors. Compared to the conventional refraction-based methods, this method avoids the requirement of V0 and interpolation, thus yielding better results with this method. For the tomography-based methods, a more realistic initial model can be obtained from this method. Therefore, more reliable near-surface velocity models can be built for rugged areas with this method.

ABSTRACT The surface conditions in Saudi Arabia demonstrate significant variability ranging from flat gravel plains to barchan sand dunes and desert washouts (wadis, ). Saudi Aramco has used a single layer velocity model from surface to datum for most of the prospective areas for both 2D and 3D seismic data. In 3D blocks where the model hasn't produced the required focusing or continuity, source gathers have been interpreted by evaluating the first breaks. This creates depth and velocity control input for a two layer model, from which static corrections are calculated. In this 3D block there are very large amplitude near-surface anomalies, the Plus/Minus technique has been applied to solve local areas. The implementation of the Plus/Minus technique utilizes a variable near-surface weathering velocity derived from the direct arrival velocity from source gathers. Multiple source pairs build up effective spreads, the method builds up a high fold of delay times at a location, and an average taken. The model is built by correcting from surface to base of weathering and from there to the datum. In areas where the Plus/Minus model refractor velocity and the two layer sub-weathering velocity converge the static models become equal.

ABSTRACT The theory and practice of refraction interferometry is presented. We show that the crosscorrelation between a nearby pair of refraction traces yields a head wave event kinematically equivalent to one generated by a source at position x along the refractor. This redatumed source location x is independent of the surface source location, so that all head waves arrive at the same time in a redatumed common geophone-pair gather (CPG). Thus, the traces in a redatumed CPG can be stacked together to yield an N-fold refraction trace, and combining different trace pairs yields N-fold refraction shot gathers. The benefits are that traveltime picking errors can be greatly reduced for noisy head wave arrivals such as far offset refraction events or secondary head-wave arrivals. We also show that head-wave events in a redatumed common geophone-pair gather will follow a flat trajectory in offset-time coordinates compared to the curved trajectory of a diving wave event. Thus, head waves can be distinguished from diving waves using redatumed CPGs, with the possibility of estimating the type of velocity gradient along an interface. This might be important for determining the lithology associated with important interfaces such as the Moho or an oil/gas bearing boundary.

ABSTRACT Various strategies for populating the shallow portion of the prestack depth migration (PSDM) velocity model are explored in a carefully-controlled synthetic environment which mimics a real-world Permian Basin unconventional play. Key findings from the synthetic experiments are corroborated by analogous observations on real data, suggesting that the experiments are capturing realistic effects. These synthetic experiments clearly demonstrate that gather flattening improves dramatically with application of the more sophisticated shallow model building approaches. In the case of the most primitive approaches (e.g., migration-from-flat-datum or migration from topography while populating the shallow model with a spatially homogenous "replacement" velocity), the migrated gathers exhibit significant residual moveout, and applying a tomographic velocity update to improve flattening leads to a significant error in event depth location (i.e., "depthing"), a finding that in turn suggests that downstream anisotropic parameter estimation will be compromised unless a more sophisticated shallow model building approach is employed. The concept of differential statics is introduced and is demonstrated to be a useful tool which can provide good gather flattening, accurate event depthing, and also improved lateral continuity of events in the common case where the near-surface velocity estimate from refraction statics analysis is not suitable for verbatim insertion into the shallow PSDM model. Presentation Date: Tuesday, September 17, 2019 Session Start Time: 1:50 PM Presentation Start Time: 3:55 PM Location: 303B Presentation Type: Oral

Abstract Despite the now-routine use of prestack depth migration (PSDM) for unconventionals, confusion abounds on the topic of how to best incorporate near-surface velocity estimates into the PSDM shallow-model-building process. The present work seeks to eliminate the confusion via a carefully-controlled synthetic experiment in which the (known) near-surface velocity distribution mimics typical Permian Basin shallow geology. In this experiment, various methods for near-surface model building are tested, ranging from simplistic to sophisticated, and PSDM results are compared against the ideal image. These tests clearly demonstrate that gather flattening improves dramatically with application of the more sophisticated shallow model building approaches. In the case of the most primitive approaches (e.g, migration-from-flat-datum or migration from topography where the shallow velocity cells are flooded with a spatially uniform “replacement” velocity), the migrated gathers exhibit significant residual moveout, and applying a tomographic velocity update to improve flattening leads to a significant error in event depth location (i.e, “depthing”). This depthing error suggests that downstream anisotropic parameter estimation will be compromised unless a more sophisticated shallow model building approach is employed. The concept of differential statics is introduced and is demonstrated to be a useful tool which can provide good gather flattening, accurate event depthing, and also improved lateral continuity of events in the common case where the near-surface velocity estimate from refraction statics analysis is not suitable for verbatim insertion into the shallow PSDM model. Key findings from the synthetic experiments are corroborated by analogous observations on real data, suggesting that the experiments are indeed capturing realistic effects. Introduction It is well known that near-surface heterogeneity can cause significant traveltime distortion of reflected signals, and, furthermore, that such distortion poses a major challenge in land seismic imaging. Addressing this challenge is particularly important in unconventional plays, where accurate depthing of subtle features is crucial for applications such as landing and steering optimization. Recently, some notable advances have been made, including the use of novel refraction statics techniques (Diggins et al., 2016), application of full-waveform inversion (e.g., Roy et al., 2017), and incorporation of gravity/EM data (Colombo et al., 2012), all of which seek to better estimate the near-surface velocity field. At the same time as these advances are unfolding, prestack depth migration is beginning to see widespread use in many North American unconventional shale plays (Rauch-Davies et al., 2018).

ABSTRACT The complexity in the near surface area can significantly affect land seismic data processing. If the near surface structures include a thin high-velocity layer above a weathering layer on top of the high-velocity bedrock, the first arrivals of a shot gather may exhibit a shingling pattern, where the first arrivals associated with the high-velocity layer are visible only within the near offset, and the bedrock refraction becomes a dominant arrival at the far offset. When this velocity reversal occurs, the effectiveness of conventional first-arrival traveltime tomography is limited. Nevertheless, the later refraction arrivals still contain substantial information about the weathering layer and the long-wavelength statics. In this study, we tested three different methods for statics and near surface velocity model estimation: the delay-time, generalized linear inversion (GLI), and combined traveltime tomography and refraction traveltime migration method using various synthetic models. Our goal is to assess the effectiveness of these methods in the areas with near surface velocity reversal. The test results suggest all three methods can recover reasonable statics solutions if the shallow velocity layers are simple and relatively flat. For more complex models with an undulated near surface high-velocity layer with variable thickness, better long-wavelength statics solutions are produced by the combined traveltime tomography and refraction traveltime migration method. In all tests, none of the derived near surface velocity models are entirely satisfactory. Nevertheless, the models derived from the combined traveltime tomography and refraction traveltime migration method are better and bear more resemblance to the true velocity model than the delay-time and GLI methods. Presentation Date: Monday, October 15, 2018 Start Time: 1:50:00 PM Location: Poster Station 3 Presentation Type: Poster

ABSTRACT Seismic interferometry is an effective approach in solving the data redatuming problems. Through interferometric redatuming, the source-receiver array becomes closer to the interesting target. However, to ensure the redatuming results accurate, we must utilize a large amount of data in general. In this study, we develop a unique high-resolution interferometric imaging method based on a new marine FreeCable acquisition geometry that allows applying interferometric transformation for each individual shot gather. This survey geometry sets a flat recording cable in deep water, and airgun shots right below water surface. Using direct waves from a source to correlate with reflection data from the same source, we virtually place the source to the same level of receivers in the water. In a numerical example, we demonstrate that FreeCable interferometric imaging result shows enhanced resolution. Presentation Date: Wednesday, September 27, 2017 Start Time: 10:35 AM Location: 351F Presentation Type: ORAL

Summary Seismic inversion products are routinely used for earth modeling and influence the drilling program. However, this process is not trivial as the deepest target intervals are at around 12,000ft and the drilling windows are very narrow, in the range of 20–40ft. We demonstrate that spending a large effort in the processing of the seismic data before inversion produces much better end results. In the Delaware Basin, large areas of modern 3D seismic data are available for seismic inversion. Based on wireline logs and core data, the Leonard Shale is known to correlate with density. By integrating inverted seismic and well data, we can predict PHIE (effective porosity) and TOC (total organic carbon) and map quality and distribution of the Leonard Shale in our development areas. The seismic processing included a special refraction statics solution that reduced the static error in this difficult area to a minimum. Noise attenuation was done in a cascading manner to avoid damaging the amplitudes. A 5D interpolation was performed to produce a regular sampled dataset and fill in acquisition holes. Vertical transverse isotropy (VTI) was corrected and horizontal transverse isotropy (HTI) was investigated but the effect was not large enough to influence the flatness of the gather events and therefore no HTI correction was done. After the full azimuthal pre-stack time migration, the CDP gathers went through additional noise cancellation and multiple attenuation processes before angle stacks were generated. The data were zero phased using log data from 16 wells. Simultaneous inversion was run by joint impedance and facies inversion. The software system is designed to extract facies and petrophysical properties from pre-stack seismic data. This methodology yields improved spatial resolution of facies and properties within a specific target. By cross-plotting well-derived acoustic impedance at seismic frequency with the inverted rock properties, we see an average correlation coefficient of 0.76. Wells with a full set of conditioned sonic and density logs demonstrate improved correlation coefficients over the average set. A previous attempt at inverting the seismic to rock properties failed as the original processing was not sufficient to obtain these high quality results.

As major oil and gas companies have been investing in shale oil and gas resources, even though has been part of the oil and gas industry for long time, shale oil and gas has gained its popularity back with increasing oil prices. Oil and gas industry has adapted to the low-cost operations and has started investing in and utilizing the shale oil sources significantly. In this perspective, this study investigates and outlines the latest advances, technologies, potential of shale oil and gas reservoirs as a significant source of energy in the current supply and demand dynamics of oil and gas resources. A comprehensive literature review focusing on the recent developments and findings in the shale oil and gas resources along with the availability and locations are outlined and discussed under the current dynamics of the oil and gas market and resources. Literature review includes a broad spectrum that spans from technical petroleum literature with very comprehensive research using SCOPUS database to other renowned resources including journals and other publications. All gathered information and data are summarized. Not only the facts and information are outlined for the individual type of energy resource but also the relationship between shale oil/gas and other unconventional resources are discussed from a perspective of their roles either as a competing or a complementary source in the industry. In this sense, this study goes beyond only providing raw data or facts about the energy resources but also a thorough publication that provides the oil and gas industry professional with a clear image of the past, present and the expected near future of the shale oil/gas as it stands with respect to other energy resources. Among the few existing studies that shed light on the current status of the oil and gas industry facing the rise of the shale oil are up-to-date and the existing studies within SPE domain focus on facts only lacking the interrelationship between heavy and light oil as a complementary and a competitor but harder-to-recover form of hydrocarbon energy within the era of rise of renewables and other unconventionals. This study closes the gap and serves as an up-to-date reference for industry professionals.

ABSTRACT The article presents results of permafrost scientific research in the Khatanga bay of the Laptev sea. Field works were organized in early spring 2017 by Rosneft Oil Company with participation of Lomonosov Moscow State University, Far Eastern Federal University and Arctic Research Centre. Research included geocryological drilling for subsea permafrost from the land fast ice, geophysical surveys and laboratory tests of thawed and frozen rock monoliths. In order to obtain information about subsea permafrost 13 core wells were drilled from the land fast ice in the Nordvik bay (sea depth ∼15 m) to a depth of 50 m with. Core data helped to define thermal characteristics of more than 300 permafrost monoliths. With the means of special equipment more than 1000 points of electromagnetic soundings were processed. Laboratory studies of the obtained samples combined with electromagnetic sounding information allowed to construct the permafrost horizon in the Nordvik bay and its immersion's behavior to a depth of 200 m. Estimation of the permafrost distribution to the depths of 500 m is presented in the paper as well as permafrost thickness dynamics simulation for 50 years. INTRODUCTION Rosneft Oil Company and Arctic Research Centre in cooperation with Far Eastern Federal University, Moscow State University and Tomsk Polytechnic University carried out a complex multidisciplinary geocryological research from the land fast ice in the Nordvik Bay, Laptev Sea (Fig. 1) in order to study geocryological conditions of the Khatanga Bay. The main aim was to obtain and summarize data on the geocryological conditions of the south-western part of the Laptev sea, which are necessary to assess the impact of adverse environmental parameters during industrial development of the area. The permafrost-geological structure and cryogenic processes, along with the presence of surface gas, are decisive geotechnical features of the shallow-water areas of the Arctic shelf, accounting and assessment of which are critically important at the first stages of development.

Near-surface characterization is an important part of seismic data processing, especially with land seismic data. The conventional approaches rely on refracted waves and estimate the compressional velocity models from the tomography of the first-break traveltimes ( Glushchenko et al., 2012 , Speziali et al., 2014 ). Despite its strong ability to image the subsurface, seismic tomography is a non-unique inverse problem ( Kanlı, 2009 , Mantovani et al. 2013 ). Because most inverse geophysical problems are non-unique, each problem must be studied to determine what type of non-uniqueness applies and, thus, determine what type of a-priori information is necessary to find a realistic solution ( Ivanov et al., 2005 ). There are several ways to incorporate the available a-priori information in the inverse problem; one of them is the definition of the initial model, which is the starting point of the inversion process. In this work, we present a data-driven approach that derives the initial velocity model for a refraction tomography workflow in an automated fashion, thus trying to minimize the amount of subjectivity that influences the starting model definition ( Osypov, 2001 ). We demonstrate the technique by mean of a synthetic, but realistic, 3D example.

ABSTRACT Near-surface characterization is an important part of seismic data processing, especially with land seismic data. Conventional approaches rely on refracted waves and using inversion techniques to estimate compression-wave velocity models from the first-break traveltimes (Glushchenko et al., 2012, Speziali et al., 2014). Despite its strong ability to image the subsurface, seismic tomography is a non-unique inverse problem (Kanli, 2009, Mantovani et al. 2013). Because most inverse geophysical problems are non-unique, each problem must be studied to determine what type of non-uniqueness applies and, thus, determine what type of a-priori information is necessary to find a realistic solution (Ivanov et al., 2005). There are several ways to incorporate the available a-priori information in the inverse problem; one of them is the definition of the initial model, which is the starting point of the inversion process. In this work, we present a data-driven approach that derives the initial velocity model for a refraction tomography workflow in an automated fashion, thus limiting to a minimum the amount of subjectivity that influences the starting model definition (Osypov, 2001). We demonstrate the technique by mean of a synthetic, but realistic, 3D example. Presentation Date: Thursday, October 18, 2018 Start Time: 8:30:00 AM Location: 208A (Anaheim Convention Center) Presentation Type: Oral