Thirty-nine construction blasts were monitored by mobile and fixed seismic arrays to develop a crustal velocity model in the Bath County, Virginia locale (June-August, 1982). The results indicate an upper-layer with P and S velocities of 5.45 km/sec and 3.07 km/sec respectively, and a thickness 3 km. Data from the 2 most distant observing stations indicate a second layer with a P velocity of 6.04 km/sec. Based on other studies in Virginia, lower crustal layers with P and S velocities 6.05 km/sec and 3.52 km/sec (11.7 km thickness) and 6.53 km/sec and 3.84 km/sec (36.0 km thickness) and upper mantle velocities of 8.18 km/sec and 4.79 km/sec are assumed. These layers make up the discrete velocity layer model. The observed travel-time data cannot be distinguished from theoretical travel-times calculated from either of two other models utilizing linear increases in velocity with depth. The first of these models uses a linear increase in P velocity from 5.45 km/sec at the surface to 6.53 km/sec at 50.7 km depth. The surface VP/VS ratio of 1. 74 is assumed for this entire layer. Beneath this layer is the mantle with P and S velocities 8.18 km/sec and 4.79 km/sec. The second of these models uses a linear increase in P velocity from 5.45 km/sec at the surface to 6.05 km/sec at 14.7 km depth. The surface VP/VS ratio is again assumed for this layer. Beneath this layer is the 36.0 km thick 6.53 km/sec layer underlain by the mantle. A test of the locational capabilities of the Bath County Network utilizing construction and quarry blasts was carried out for the 3 different velocity models. All three models gave virtually identical locations. The tests indicate an average of less than 1 km epicentral and depth errors inside the network. On the edge of the network, accuracy degrades to 3 km epicentral error with poor depth control. The Bath County area is seismically quiescent. Portable seismographs recorded over 3,000 hours of low noise seismic data in June, July and August of 1982 and failed to detect any local earthquake activity. Network monitoring by a permanent 4 station microearthquake network from November 1978 to November 1982 resulted in 11 recorded events. Three of these events were too small to be located. The other 8 events were located using all 3 velocity models. The linear increase in velocity over mantle model was eliminated from further consideration due to poor performance in event location. The other two models gave virtually identical locations. For these two models, the events form an apparent east-west trend. Reliable focal wave polarities and SV/P amplitude ratios, mechanisms, using both P indicate one nodal plane striking east-west and dipping to the south. The other nodal plane, which defines the mode of faulting, is poorly constrained. / Master of Science

Estimates of the apparent quality factors Q_Lg (attenuation determined from the spatial decay of Lg waves) and Q_c; (attenuation determined from the temporal decay of seismic coda waves) are made for the crust of Virginia and its environs. The results are presented in the form Q_(Lg,C)(f) = Q₀f^N, where Q₀ = Q_(Lg,C)(1 Hz) and N represents the frequency dependence. The study area is located in the Appalachian region of Virginia and eastern Tennessee, containing three areas of regionally high seismicity: the central Virginia, Giles County, and eastern Tennessee seismic zones. The attenuation of the Lg phase was studied using vertical component digital recordings from Virginia Tech Seismological Observatory network stations. The seismic sources were ten regional surface mine explosions and six regional earthquakes. It was determined that Q_(Lg,C) can be represented by Q₀ = 186, σ_logQ₀ = 0.05, and N = 1.1 ± 0.1 for the frequency band 1-4 Hz. A site effect corrected estimate of Q_(Lg,C) was also determined for the study area. This was accomplished using a spectral ratio method in which station site effects and instrument responses are canceled out. For the frequency band 1-10 Hz the site independent apparent quality factor can be represented by Q₀=155, σ_logQ₀ = 0.1, and N=1.2±0.2. Station site factors were estimated using a mean residual technique. The decay of seismic coda waves across the Giles County, Virginia seismic network was studied to estimate Q_c for western Virginia. A relatively new spectral method was used. The seismic sources were four local earthquakes. For the frequency band 1-10 Hz, the results can be represented by Q₀= 111, σ_logQ₀ = 0.07, and N = 1.3 ± 0.07 . These values agree with a limited number of results obtained using a bandpass, time domain method which showed Q₀ = 132, N = 1.3. The results obtained for the Virginia area differ significantly in the 1 - 3 Hz range from those reported in most previous studies of the eastern United States. Previous studies have generally shown 800 ≤ Q₀ ≤ 1000 and 0.3≤ N≤0.5, but many of those results are for much larger regions and determined using different analysis techniques. Several reasons that could account for the different results include 1) estimates of attenuation may be affected by incorrect geometrical spreading models, 2) the size of the study area may affect the estimates, and 3) estimates of Q_(Lg,C) made for broad regions may be biased by zones of differing tectonic activity. Of these factors, only the effects of changing geometrical spreading coefficients and scattering models (related to study area) can be quantified. Neither of these affect the results by more than a factor of two. The high frequency dependence values (N≃1.1) are probably influenced by the lack of definition of higher frequency (≃10 Hz) data at the path distances studied. Future studies should employ more extensive data sets covering a larger geographic area. At greater distances, the attenuation of higher frequency waves may be more easily observable. The large frequency dependence values are probably indicative of an area where scattering dominates over anelastic attenuation. The folded and thrusted Appalachian provinces may, indeed, be such a region of high scattering. Such a mechanism may also help to explain southeastern United States meizoseismal areas that are small relative to the total felt areas. Large frequency dependence results for Q_Lg and Q_c are relevant with respect to seismic hazard. We do not believe the results are overly biased by station site effects or varying source effects and if they hold for magnitudes greater than those studied here (m < 4.2) , they indicate a greater potential for damage by higher frequency waves to engineering structures in Virginia and its environs than previously assumed. / Master of Science

A technique has been developed to estimate the directions of principal stresses from focal mechanism solutions, under the assumption that the stress is homogeneous throughout the seismic zone. That method is called the Multiple Solution per Earthquake Technique (MSET), and utilizes each member of multiple focal mechanism solution set as a possible solution. The MSET is useful when applied to small data sets, and differs from existing techniques in that (1) the use of multiple focal mechanisms for individual earthquakes allows for a range in the possible orientation of the fault geometry, while preserving fit to the original polarity and amplitude ratio data, and (2) the differences between the observed and theoretical fault slip is used as a weighting scheme for the results of the tensor estimation. Other methods, which rotate the single observed focal mechanism solution from its original configuration to estimate misfit, do not take into consideration the fit of that final solution to the original input data but assume that a minimization of errors between the theoretical stress model and the focal mechanism solution indicates a reasonable fit. For large data sets that assumption is likely to be met. The MSET was applied to a set of 32 earthquakes to estimate the principal stress orientations for seismically active zones in the Southeastern United States, using focal mechanism solution sets derived by the program FOCMEC. Eight events were studied for the Giles County Seismic Zone, 13 for the Central Virginia Seismic Zone, and 11 for the Eastern Tennessee Seismic Zone. After testing against approximately 25000 theoretical solutions, an average of sixteen focal mechanism solutions fit the input polarity and (SV/P)₂ amplitude ratio data for each earthquake. The P-axes of the multiple focal mechanism solutions were averaged to determine a provisional single P-axis direction for each earthquake. P-axes for the Giles County and Eastern Tennessee Seismic Zones trended generally NE-SW, while those of the Central Virginia Zone varied with depth, with the P-axes of events above approximately 8 km trending NE-SW, and those below trending NW-SE. Application of the MSET resulted in consistent principal stress orientations for the Giles County and Eastern Tennessee Zones, with the horizontal component of the maximum compressive stress direction (σ₁) trending about N40°E and N50°E, respectively. Results for the Central Virginia Zone also suggested differently oriented stress regimes above and below a depth of 8 km. The direction of the σ₁ axis above that boundary was N70°E, while below it was east-west, with a shallow plunge to the west. While those results were not as pronounced as suggested initially by the P-axis data alone, the hypothesis of two stress tensors produced better MSET results than for a single stress tensor for the combined data set. The technique developed for this study produces comparable results to other methods when applied to identical data sets. Estimation of error is based on subjective criteria, and includes the lit of the original seismic polarity and amplitude ratio data with the focal mechanism solutions. The error associated with each step in the process (e.g. distribution and reliability of the polarity and ratio data, calculation of focal mechanism solutions and estimation for the stress field) is very difficult to parameterize, and thus, no formal statistical analyses were undertaken. After the estimate of the homogeneous stress field was made for each zone, a single best focal mechanism solution for each earthquake could be objectively chosen by constraining the slip associated with each mechanism to be aligned with the resolved stress derived from the principal stress directions. In that manner, focal mechanism solutions could be identified which fit the sparse input polarity and amplitude ratio data, but which were not compatible with the calculated stresses. Also, in that same procedure, the fault plane was chosen from the set of two nodal planes for each focal mechanism solution by examination of the theoretical slip on each of those planes. The faults within the Giles County Seismic Zone matched the direction found in previous seismic reflection surveys, with an average strike of N25°E. In the Eastern Tennessee Seismic Zone, faulting also occurred on planes oriented predominantly NE-SW. For the five shallow Central Virginia events, faults trended NW-SE, while for the deeper events there was no consistent trend. A comparison was made between the P-axis and σ₁ axis derived for each earthquake. Although in 81% of the cases σ₁ was within 35° of at least one P-axis of the focal mechanism solution set, no further empirical relationship was found. The MSET has proven itself useful in two ways when applied to sparse data sets. First of all, the primary seismic data (polarities and amplitude ratios) are not overlooked when deriving the orientation of a stress tensor associated with local faulting. Secondly, the MSET is an objective method for defining the best fitting solution among a family of focal mechanism solutions by requiring compatibility with the regional stresses. In the future, after integration with a program such as FOCMEC, regional stress tensors may be derived by the MSET which incorporate reasonable statistical parameters based on the fit of that primary data. / Ph. D.