IN51C-1168 INVITED
The Noise/Data Conundrum in Gravity and Magnetic Surveys of Fluvial Sediments Near the Rio Grande, West Texas
Short wavelength, near-surface geologic variations play a frequently unquantifiable role in increasing noise or data errors in geophysical surveys targeting deeper structures. We have previously documented how large seasonal variations in the electrical properties of soils affect our ability to resolve longer wavelength, deeper features. In addition to examining time variations in geophysical properties, we have begun analysis of repeated, high density magnetic and gravity readings taken across a variety of fluvial soils within the Canutillo, Texas water well field located just west of the Rio Grande. Survey reduction techniques and operator experience represent controllable uncertainty sources. Short-wavelength variations in soil properties contribute irreducible noise to undersampled regional surveys. Initial data analysis suggests uncertainty estimates based on fractal models are applicable within the mapped surficial fluvial soil types of our study area, with channel deposits showing more variability than flood plain deposits.
IN51C-1169
Integrated 3D modelling, an effective way to improve geophysical data interpretation - the southwestern Barents Shelf as a case study.
We will demonstrate how seismic interpretation combined with density and magnetic modelling can help to better constrain the crustal structure below sedimentary basins and increase the accuracy of geophysical data interpretation. On the southwestern Barents Shelf, we applied 3D joint density and magnetic modelling to obtain maps of depth to the top basement and Moho and a basement characterisation in terms of density and magnetic properties of the crust and deep crust. We constrained our model using all available geological and geophysical data. The sedimentary succession is constrained by industrial depth-converted seismic horizons tied at wells. The top basement of our study area is complex and deeply buried under more than 15 km of sedimentary rocks in some areas. For an accurate assessment of the top basement and deep crustal structures we used seismic refraction models and a set of deep reflection profiles as well as 1D velocity laws extracted from the Barents50 model. Furthermore, a database compiling density, magnetic remanence and susceptibility measured on onshore samples from northern Norway was used to constraint the modelling values. The integration of all these data helps to avoid interpretation pitfalls and highlights discrepancies between published data and models. Overcoming these inconsistencies we propose a new 3D structural model. The 3D model contributes to understand the basement lithology distribution and the offshore prolongation of the Caledonian structures, well described onshore. The 3D model allows us also to discuss the tectonic evolution of the SW Barents Sea. A system with a unique Caledonides branch propagating toward the north and Caledonian nappes flowing asymmetrically in the West Barents Sea is confirmed. A unique Caledonian suture is proposed west of the Loppa High then propagating between the Svalbard and Franz Josef Land. The distribution of the Timanides structures possibly explains the limitation of the Caledonian nappes towards the NE. We suggest an asymmetric Proterozoic Timanian foreland sitting in the southwestern Barents Sea. We interpret the alignment of fault complexes trending N-S and N 50° as inherited from Caledonian weak zones expected in such system. The Billefjorden Fault Zone is consequently a complex Caledonian weak zone originating from mechanisms of terranes extrusions towards the north linked to an oblique collision of Laurentia and Baltica in that region. The combination of Timanian and Caledonian weak zones played an important role in the sedimentary basins evolution.
IN51C-1170 INVITED
Do end-member models reduce uncertainty in the structural interpretation of seismic data?
Remotely sensed structures are invariably open to multiple interpretations and seismic reflection data, the
principle source of imaging the upper crust, is no exception. In some cases, structural models can be
eliminated (evaluated as being highly implausible) as being geometrically (physically) unfeasible. Even
images of simple deformation structures such as extensional provinces with depth-dependent seismic velocity
structures can yield multiple interpretations of fault geometries. In thrust belts, which can have notoriously
complex seismic velocity structure, imaging can be less crisp. In these settings, existing analytical workflows
commonly rely on computer-facilitated section balancing and forward modelling of seismic interpretations to
validate structural restorations. These in turn build on a variety of alternative kinematic descriptions of the
relationships between distributed folding and localized thrusting that include fault-bend, fault-
propagation/trishear and detachment folding. Each predicts different amplification histories for folds and
distinct curvature evolutions that could in turn predict different second-order deformation histories within
prospective reservoirs or stratal juxtapositions along the main thrust zones. In industry, the different choices
of interpretation can have billion dollar impacts. Thus, the model variations can be assessed through multiple
scenario realizations to capture the range of interpretation uncertainty. A complementary approach is to
apply understanding of structural style from well-imaged areas into regions of more opaque seismic.
However, these approaches can be grossly misleading and the inferred levels of uncertainty substantially
over-optimistic. 3D seismic from Nigeria reveals structural style variations over 100s metres within prospects.
Outcrop analogue studies reveal substantial distributed strain, manifest as general layer parallel shortening,
generally unrepresented in kinematic models. Indeed all existing kinematic models employ end-member
descriptions of strain localization that do not address observable deviations such as folded thrust strands nor
honour theoretical research on multilayer deformation. Many observable forelimbs contain subsidiary folding
(consistent with multilayer deformation theory) that accommodate significant displacements and generate a
much greater range of stratal juxtapositions than simple models predict. Further, regional restorations of
gravity-driven thrust belts require significant volume loss (tectonic compaction). Strain and volume loss
represent substantial extra degrees of freedom that risk significant drilling surprises. Until these additional
deformations are available in restoration applications, the challenge is to generate the range of scenarios
using peer-review supplemented by a greater range of analogues. This requires community-based
approaches and are being pioneered by the Virtual Seismic Atlas.
http://www.seismicatlas.org
IN51C-1171 INVITED
Using tomograms for quantitative interpretation of processes and properties
Changes in subsurface hydrologic conditions that result from remediation or natural processes increasingly are monitored with near-surface geophysical imaging. In particular, time-lapse electrical resistivity tomography (ERT) and radar tomography (RT) are well suited to monitoring fluid-conductivity changes associated with injection of amendments for biostimulation, freshwater/saltwater dynamics in coastal aquifers, and aquifer-storage and recovery (ASR). Quantifying results from tomograms is complicated by limited resolution and estimation uncertainty that are inherent to geophysical inverse problems. Hydrologic and engineering parameters are difficult to estimate because survey geometry, measurement physics, experimental errors, regularization strategy, prior information, and parameterization also affect interpretation of geophysical measurements. Here, we review recent work to address the problem of "correlation loss" and to directly extract information about transport processes from geophysical time series. Both synthetic and field-experimental examples are presented. Case studies include applications of (1) time-lapse ERT and RT to monitor biostimulation injections at a former Department of Defense site in Brandywine, MD; (2) time-lapse ERT to monitor submarine ground-water discharge at Cape Cod, MA; and (3) time-lapse ERT to monitor an ASR experiment in Charleston, SC. In the first study, a spatially variable calibration between ERT-estimated bulk and sampled fluid conductivity was developed and applied to three-dimensional tomograms to infer the evolution of fluid conductivity after injections of amendments and pH adjustment. In the second study, the ability to resolve subsurface salinity contrasts is shown to vary strongly as a function of tide level. In the third study, electrical methods are used to monitor rate-limited mass transfer, and hydrologic parameters are estimated from the temporal moments of fluid and ERT-estimated conductivity; however, the applicability of this approach is shown to be limited by spatial resolution, temporal resolution, and the relative rates of advection and mass transfer.
IN51C-1172 INVITED
Characterization of Ground Motion in Regions of South Africa at Rock Sites
The objective of this evaluation is to determine the crustal amplification function of the seismic signal for different regions in South Africa. In order to model crustal amplification, models of the S-wave velocity and the shear quality factor Q are determined. All currently run seismic hazard assessment projects in South Africa are utilizing the attenuation equations mostly from central and eastern US. This is done on the assumption that both areas are intraplate regions characterised by low seismic activity. The quantification of similarities and differences between the South African and other crust models is vital for assessing the sensitivity of the hazard assessment. In South Africa most observations of shear wave velocity profiles are derived using a joint inversion method. The receiver function is jointly inverted with the Rayleigh wave dispersion. S-wave velocity models are obtained for 87 sites using teleseismic earthquakes, about 15 earthquakes are used to estimate S-wave velocity model under each seismic station. The shear wave quality factor Q is mostly based on surface wave attenuation measurements. Tectonically stable regions are usually characterized by high values of QLg (800-1200) and a weak frequency dependency. The region of low Q is the Cape Fold Belt at the southern tip of Africa. Q values obtained for craton Kaapvaal Craton are about 600. Exploratory data analysis was carried out by applying cluster analysis to determine groupings within the velocity profiles. Hierarchical clustering with the average linkage method was selected. The cluster results show quite clearly the grouping of velocity profiles of neighboring sites, i.e., there is a large degree of homogeneity of neighboring sites with respect to their velocity profiles. In the final step a synthetic signal is passed through the crustal structure to obtain the amplification function for different geological cluster-zones. Then, the same processing was applied to the generic crustal models for the CEUS in order to compare the amplification function in all regions. The results indicate that, in order to model crustal amplification in South Africa, a local model of the S-wave and Q- factor has to be used.
IN51C-1173 INVITED
Site Amplification Characteristics and Their Variations Derived from K-NET, KiK-Net, and JMA Strong Motion Records in Japan
To predict strong ground motions for future scenario earthquake in a broad-band frequency, we need to characterize both source spectra and site amplification. For short period we can use statistical method based on the observed records and the summation technique of observed waveforms or their synthetic substitutes. Thanks to the advent of the nationwide strong motion networks in Japan, we have now plenty of weak motion data, enough to construct statistical Greenfs functions. Since the site amplification effects on the sediment sites in an urbanized environment tend to be so large that we need to extract site amplification ratios in a statistical manner and construct amplification models for different site categories. In this study we perform first the analysis to separate the so-called source spectra, attenuation coefficient, and the site amplification factors from about 18,825 K-NET, KiK-Net, and JMA records observed at more than 1,700 stations in Japan. The separation method is the well-established one of Andrews (1980) and the resultant source spectra are modeled as omega-square spectra. As a reference site we use one rock station of KiK-Net in Yamaguchi Prefecture, from which we remove amplification due to the shallow weathered rock formation. Once we obtain site amplification factors, we try to reproduce them by using one-dimensional S- wave velocity structures below each site of K-NET and KiK-Net (in total, 1,300). We use Genetic Algorithm to invert the S-wave structures with fixed S-wave velocities in the shallow part. We succeed to reproduce site amplification factors at about one half of the sites very well. We then calculate average S-wave velocities for top 30 m, the so-called Vs-30. We also perform the same kind of separation analysis for three ground motion strength indexes, namely, PGA, PGV, and JMA Seismic Intensity. We found that the variations in spectral ratios between observed site amplification and averaged one is about 0.26 in a logarithmic scale, minimum (0.23) at 0.1 Hz and maximum (0.28) at 3 Hz. The variation at 3 Hz is largest probably because the fundamental peak frequency for soft sediments sites will be around 3 Hz so that small deviation in peak frequency for one event may result in relatively large deviation. Averaged spectral ratios for sites with different Vs-30 categories show peak values of about 10 irrespective of the Vs- 30 categories, although their frequencies tend to be higher as Vs-30 become larger. As for the strength indexes, the variation tend to be largest near the source and then monotonically decreasing to the minimum values at around 100 km away from the source. We also found that observed strength indexes tend to be higher in the near-field compared to the predicted values from the separated empirical models. The variations as a function of the strength index levels show quite stable values of around 0.2. However, the log-average ratios between the predicted and the observed (i.e., the prediction error) become minus if PGA becomes larger than 400 Gals or PGV larger than 30 cm/s, due to nonlinear response of soil deposits.
IN51C-1174 INVITED
Ambiguity diagram for gradational density model to reduce uncertainty in the interpretation of gravity data
Infinite models of geology at depth can be proposed in order to match their theoretical gravity anomaly. The problem so far does not offer a unique solution except in idealized cases. The uniqueness of solution or interpretation is not possible due to imperfection in the data and the basic ambiguity in case of potential fields which obey Laplace's equation. The first problem of imperfection in the data is the perennial one. The second difficulty, which is an inherent property, may also be reduced by placing restrictive but reasonable assumption on the admissible physical property, i.e. rigorous bounds on the density distribution. Similarly range of estimation of source parameters can be provided by construction of ambiguity diagram for gravity measurements, which provides an interval within which the unknown parameters are located with a given probability. Estimation of ambiguity of quantitative interpretation of gravity anomaly in the case of fault needs linearization but the gravity anomaly over a fault and gradational model contains logarithms, which can't be linearized. However it was shown that the anomaly curves are very similar to the curve of normal distribution function and therefore it was possible to replace the anomaly by the normal distribution function, which can be linearized to obtain the confidence limits of the parameters and to obtain range of ambiguity for a given fault or gradational model. In this paper ambiguity diagram for fault versus gradational model has been described in order to give a range of source parameters of a gradational density model. Synthetic examples for a fault and gradational density illustrate the use of ambiguity diagram and application is demonstrated through field gravity data.
IN51C-1175
What defines an Expert? Uncertainty in the interpretation of seismic data
Studies focusing on the elicitation of information from experts are concentrated primarily in economics and world markets, medical practice and expert witness testimonies. Expert elicitation theory has been applied in the natural sciences, most notably in the prediction of fluid flow in hydrological studies. In the geological sciences expert elicitation has been limited to theoretical analysis with studies focusing on the elicitation element, gaining expert opinion rather than necessarily understanding the basis behind the expert view. In these cases experts are defined in a traditional sense, based for example on: standing in the field, no. of years of experience, no. of peer reviewed publications, the experts position in a company hierarchy or academia. Here traditional indicators of expertise have been compared for significance on affective seismic interpretation. Polytomous regression analysis has been used to assess the relative significance of length and type of experience on the outcome of a seismic interpretation exercise. Following the initial analysis the techniques used by participants to interpret the seismic image were added as additional variables to the analysis. Specific technical skills and techniques were found to be more important for the affective geological interpretation of seismic data than the traditional indicators of expertise. The results of a seismic interpretation exercise, the techniques used to interpret the seismic and the participant's prior experience have been combined and analysed to answer the question - who is and what defines an expert?
IN51C-1176 INVITED
Quantifying Uncertainty in Velocity Models using Bayesian Methods
Quanitifying uncertainty in models derived from observed data is a major issue. Public and political understanding of uncertainty is poor and for industry inadequate assessment of risk costs money. In this talk we will examine the geological structure of the subsurface, however our principal exploration tool, controlled source seismology, gives its data in time. Inversion tools exist to map these data into a depth model but a full exploration of the uncertainty of the model is rarely done because robust strategies do not exist for large non-linear complex systems. There are two principal sources of uncertainty: the first comes from the input data which is noisy and bandlimited; the second, and more sinister, is from the model parameterisation and forward algorithms themselves, which approximate to the physics to make the problem tractable. To address these issues we propose a Bayesian approach. One philosophy is to estimate the uncertainty in a possible model derived using standard inversion tools. During the inversion stage we can use our geological prejudice to derive an acceptable model. Then we use a local random walk using the Metropolis- Hastings algorithm to explore the model space immediately around a possible solution. For models with a limited number of parameters we can use the forward modeling step from the inversion code. However as the number of parameters increase and/or the cost of the forward modeling step becomes significant, we need to use fast emulators to act as proxies so a sufficient number of iterations can be performed on which to base our statistical measures of uncertainty. In this presentation we show examples of uncertainty estimation using both pre- and post-critical seismic data. In particular, we will demonstrate uncertainty introduced by the approximation of the physics by using a tomographic inversion of bandlimited data and show that uncertainty increases as the central frequency of the data decreases. This is consistent with the infinite frequency approximation in the tomographic modeling step becoming increasing compromised.
IN51C-1177
Quantitative Integration of Multiple Geophysical Techniques for Reducing Uncertainty in Discrete Anomaly Detection
The objectives of this project are to (1) utilize quantitative integration of multiple geophysical techniques, (2) determine geophysical anomalies that may indicate locations of various archaeological structures, and (3) develop techniques of quantifying causes of uncertainty. Two sites are used to satisfy these objectives. The first, representing a site with unknown target features, is an archaeological site on the Tennessee River floodplain. The area is divided into 437 (20 x 20 m) plots with 0.5 m spacing where magnetic gradiometry profiles were collected in a zig-zag pattern, resulting in 350 km of line data. Once anomalies are identified in the magnetics data, potential excavation sites for archeological features are determined and other geophysical techniques are utilized to gain confidence in choosing which anomalies to excavate. Several grids are resurveyed using Ground Penetrating Radar (GPR) and EM-31 with a 0.25 m spacing in a grid pattern. A quantitative method of integrating data into one comprehensive set is developed, enhancing interpretation because each geophysical technique utilized within this study produced a unique response to noise and the targets. Spatial visualization software is used to interpolate irregularly spaced XYZ data into a regularly spaced grid and display the geophysical data in 3D representations. Once all data are exported from each individual instrument, grid files are created for quantitative merging of the data and to create grid-based maps including contour, image, shaded relief, and surface maps. Statistics were calculated from anomaly classification in the data and excavated features present. To study this methodology in a more controlled setting, a second site is used. This site is analogous to the first in that it is along the Tennessee River floodplain on the same bedrock units. However, this analog site contains known targets (previously buried and accurately located) including size, shape, and orientation. Four geophysical techniques are used to survey the area (EM-31 ground conductivity, EM-61 metal detector, Nogin (GPR) 250 MHz antennae, and Sensors and Software PulseEkko Pro (GPR) 100 MHz antennae). Data is integrated utilizing the same quantitative methodology used on the archaeological site. The study of results for the known targets are used to modify the integration methods to enhance uncertainty reduction, providing quality assurance in the manipulation of the data from the archeological site and resulting in an improved interpretation of the original geophysical data in the study.
IN51C-1178
Using an intelligent system to aid in tephra layer correlation of the tephra beds of the Mono-Inyo Craters, California
We are developing an intelligent system to correlate tephra layers by using the lithologic and geochemical characteristics of field samples, to aid geologists in interpreting eruption patterns in volcanic fields. Understanding the eruption history of a volcanic field from stratigraphic studies is important for forecasting future eruptive behavior and hazards. The intelligent system is used to define groups of tephra source vents and to correlate tephra layers based on a combination of geochemical data and lithostratigraphic characteristics. The tephra beds of the Mono-Inyo Craters, California, are used to test the ability of the intelligent system for tephra layer correlation. The data processing is performed by a suite of both unsupervised and supervised classifiers, built and combined within the framework of the Dempster-Shafer theory of evidence. We have developed algorithms to calculate isopleth maps of thickness, lithic and pumice size that are used in the processing of the lithostratigraphic data. This spatial information is important in the determination of eruption patterns and is used by an evidential nearest neighbor classifier to correlate tephra layers. Integrating a better isopleth approximation function and expert knowledge about stratigraphic order of the tephra layers into the classifier improves the lithostratigraphic correlation from 56% to 87% of layers correctly identified. Geochemical data for defining groups of tephra sources are processed by a suit of fuzzy k-means classifiers. Improved clustering results of geochemical data are achieved by the fusion of individual clustering results with an evidential combination method. The intelligent system aids correlation by showing matches and disparities between data patterns from different outcrops that may have been overlooked. The intelligent system produces a useful recognition result, while dealing with the uncertainty from sparse data and the imprecise description of layer characteristics.