Charles E. Glass

Location of subsurface targets in geophysical data using neural networks

Neural networks were used to estimate the offset, depth, and conductivity‐area product of a conductive target given an electromagnetic ellipticity image of the target. Five different neural network paradigms and five different representations of the ellipticity image were compared. The networks were trained with synthetic images of the target and tested on field data and more synthetic data. The extrapolation capabilities of the networks were also tested with synthetic data lying outside the spatial limits of the training set. The data representations consisted of the whole image, the subsampled image, the peak and adjacent troughs, the peak, and components from a two‐dimensional (2-D) fast Fourier transform. The paradigms tested were standard back propagation, directed random search, functional link, extended delta bar delta, and the hybrid combination of self‐organizing map and back propagation. For input patterns with less than 100 elements, the directed random search and functional link networks gave ...

Neural network pattern recognition of subsurface EM images

Abstract Neural networks are computer simulations of the brains neural functions; as such they perform well on the same types of problems on which humans perform well, namely pattern recognition. Neural networks have shown the capability to learn human speech, read handwritten signatures and recpgnize human faces. Applied to geophysical data, neural networks offer the ability to estimate model parameters in near realtime. A backpropagation neural network was trained to estimate the spatial location (offset and depth) of a target given an image of the electromegnetic ellipticity. Three components of the magnetic field were measured from which the ellipticity was calculated. Theoretical ellipticity images were used for training the neural network; field data were used to test it. The input data representation was important in obtaining results with 10% error or less from the neural network; generally, smaller input vectors yielded more accurate results. Five different representations were examined: the whole image, the subsampled image, trough-peak-trough, peak amplitude and frequency domain. The frequency-domain representation estimated the target locations in the field data with the least error, 0.4% for the offset and 1.5% for the depth. The network was examined for its ability to generalize, to extrapolate beyond the spatial limits of the training set and to ignore discrepancies between synthetic and field data. The generalization from synthetic training data to synthetic test data had errors near 5% for most offset estimates and near 2% for most depth estimates. We considered extrapolation errors satisfactory (10%) up to 1.5 model spacings beyond the limits of the training set.

Cokriging—a computer program

Abstract Cokriging is a process wherein several variables can be jointly estimated on the basis of intervariable and spatial structure information. Presented herein is the program COKRIG, for punctual cokriging, a program in a simple form to demonstrate the utility of cokriging. Equation solution follows a modification of a method developed for the solution of large-scale linear systems. Several example problems show that, at least for earthquake data, the inclusion of intervariable information results in a more accurate BLUE (best linear unbiased estimator).

Digitally processed topographic data for regional tectonic evaluations

Siting and design of major structures such as nuclear power plants, waste repositories, dams, and so forth require a thorough evaluation of earthquake hazard, including an assessment of the tectonic regime in the site region. All available data sources such as geologic maps, geophysical maps, and synoptic remote sensing images should be fully utilized for regional tectonic evaluations. Nearly all lineaments mapped on synoptic-scale images, however, are expressions of topographic features. It seems desirable, therefore, to include analysis of digital elevation data in regional phases of tectonic evaluations. These data have higher spatial resolution than Landsat and are less expensive than both Landsat and SLAR. Digital elevation data are inherently geometrically correct and, although of lower resolution than SLAR, are more versatile in accentuating geologic structure with less cultural and vegetation interference. Digital elevation data in the form of synthetic relief images and stereo pairs provide ideal data bases for guiding aerial reconnaissance and acquisition of large-scale images for detailed evaluations.

Recent Activity of the Wasatch Fault, Northwestern Utah, U.S.A.

Abstract The Wasatch fault trends north—south for a total length of 370 km, between Gunnison, Utah and Malad City, Idaho. The fault is in the eastern Basin-and-Range Province, adjacent to the Middle Rocky Mountains and the Colorado Plateau. The most recent fault displacements indicate normal slip, with the western block faulted down relative to the Wasatch Mountains. The Wasatch fault is within the Intermountain Seismic Belt of the western U.S.A.; however, it has been associated with only minor seismicity within historic time (140 yrs). There is no conclusive evidence of significant historic surface faulting along the Wasatch, although there is a concentrated 50–100-m-wide zone of damaged streets, curbs, houses, and buildings where the fault trends through Salt Lake City. This damage may be interpreted as evidence of tectonic creep; however, other possible interpretations include fault-controlled damage related to slope instability along the steep scarp or fault-controlled damage related to groundwater. Precise geodetic leveling across the Salt Lake City segments suggests that tectonic creep may be occurring, although the time span (7 yrs) has not been long enough to establish a consistent trend. Geomorphic evidence and tree-ring dating suggest that the most recent events of appreciable surface faulting are no older than a few hundred years. Fresh scarps in geologically young glacial and post-glacial deposits are 13–20 m high. Surface manifestations of active faulting along the Wasatch include multiple-fault scarplets in geologically young materials, graben on the downthrown block, backward tilting of the downthrown block near the fault, triangular facets along the fault scarp, lines of springs, and slickensided fault surfaces. In addition to the geologic evidence of active faulting, there are numerous landslides and large debris flows that have been faulted subsequent to their deposition, as linear scarps and other geomorphic features indicative of surface faulting cross them. Our current geologic research on the activity of the Wasatch fault is directed toward determining the recent history of surface displacements and the relationship of the fault with large-magnitude earthquakes, which is based upon the faulted sequence of Pleistocene Lake Bonneville beds and glacial deposits along the Wasatch Front.

Treatment of earthquake ground motion using regionalized variables

Earthquake ground motion is a spatial phenomenon associated with local random behavior and global attenuation away from a source. By this description, ground motion is a regionalized variable. Analyses of directional variograms for earthquake ground motion reveal ansiotropic attenuation. For some earthquakes, a spatial structure exists in all directions and an average variogram can be developed. For other earthquakes, either a spatial structure can be identified only for restricted directions, or a spatial structure cannot be found to exist in any direction. A valid application of regionalized variables theory to the spatial analysis of earthquake ground motion must therefore be predicated on an initial, exhaustive analysis of estimated variograms. To estimate these data, an application of ordinary kriging is preferable to other more complex regionalized variables techniques to simplify the spatial analysis of earthquake ground motion.

An application of disjunctive kriging for earthquake ground motion estimation

For earthquake ground motion studies, the actual ground motion distribution should be reproduced as accurately as possible. For optimal estimation of ground motion, kriging has been shown to provide accurate estimates. Although kriging is accurate for this application, some estimates it provides are underestimates. This has dire consequences for subsequent design for earthquake resistance. Kriging does not provide enough information to allow an analysis of each estimate for underestimation. For such an application, disjunctive kriging is better applied. This advanced technique quantifies the probability that an estimate equals or exceeds particular levels of ground motion. Furthermore, disjunctive kriging can provide improved estimation accuracy when applied for local estimation of ground motion.

Recognizing EM ellipticity patterns with neural networks

The interpretation of surface electromagnetic ellipticity measurements is treated as a pattern recognition problemvis a vis an inverse problem. A paralleldistributedprocessing or neural network approach is used for the pattern recognition. A feed forward, backpropagation network is trained to recognize the spatial location of an anomaly given ellipticity patterns from 20 theoreticalmodels. Thenetwork has demonstrated an ability to form abstractions and generalizations when presented with ellipticity patterns that are not part of the training set. the potential for one to match patternseffectively. Tosearch manually through numerous patterns for a visual match, however, is a laborious task: it is also a task for which the Von Neuman architecture of modern computer systems is poorly equipped.

A focus on the September 1985 Michoacan earthquakes

SummaryThe severity of damage to Mexico City as a result of the 19 September 1985 Michoacan earthquake was unusual given the citys distance (350 km) from the zone of seismic energy release. To explain the damage many authors have suggested that unusual source or transmission path characteristics contributed to enhanced ground motion in Mexico City. The purpose of this paper is to present a summary of results obtained from data recorded during the earthquake related to possible anomalous source characteristics.It is concluded that although the Michoacan earthquake was a large earthquake indeed, in terms of energy output, spectral content, geometry and source mechanics it was not remarkable or anomalous relative to other subduction zone earthquakes in Mexico or elsewhere. In fact the future may well see a larger earthquake generated along the Guerrero seismic gap which is significantly closer to Mexico City.

The Thermal Infrared

Thermal infrared (TIR) wavelengths compose the electromagnetic spectrum from approximately 2.5 μm to 14 μm. Atmospheric absorption dominates the TIR with the exception of two windows, one from approximately 2.5 μm to 5 μm, and another from 8 μm to 14 μm. Scanning systems rather than photographic products must be employed in the TIR, so TIR images can not be used for stereographic interpretation. Thermal scans are expensive, and contractors willing to conduct them are few. So make sure that there are good reasons to expect that thermal contrasts will give you an advantage in identifying hazards on the ground.