A back‐propagation neural network is successfully applied to pick first arrivals (first breaks) in a background of noise. Network output is a decision whether each half‐cycle on the trace is a first or not. 3D plots of the input attributes allow evaluation of the attributes for use in a neural network. Clustering and separation of first break from non‐break data on the plots indicate that a neural network solution is possible, and therefore the attributes are suitable as network input. Application of the trained network to actual seismic data (Vibroseis and Poulter sources) demonstrates successful automated first‐break selection for the following four attributes used as neural network input: (1) peak amplitude of a half‐cycle; (2) amplitude difference between the peak value of the half‐cycle and the previous (or following) half‐cycle; (3) rms amplitude ratio for a data window (0.3 s) before and after the half‐cycle; (4) rms amplitude ratio for a data window (0.06 s) on adjacent traces. The contribution of the attributes based on adjacent traces (4) was considered significant and future work will emphasize this aspect.

We have developed an improved Levenburg‐Marquart technique to rapidly invert Bouguer gravity data for a 3-D density distribution as a source of the observed field. This technique is designed to replace tedious forward modeling with an automatic solver that determines density models constrained by geologic information supplied by the user. Where such information is not available, objective models are generated. The technique estimates the density distribution within the source volume using a least‐squares inverse solution that is obtained iteratively by singular value decomposition using orthogonal decomposition of matrices with sequential Householder transformations. The source volume is subdivided into a series of right rectangular prisms of specified size but of unknown density. This discretization allows the construction of a system of linear equations relating the observed gravity field to the unknown density distribution. Convergence of the solution to the system is tightly controlled by a damping parameter which may be varied at each iteration. The associated algorithm generates statistical measures of solution quality not available with most forward methods. Along with the ability to handle large data sets within reasonable time constraints, the advantages of this approach are: (1) the ease with which pre‐existing geological information can be included to constrain the solution, (2) its minimization of subjective user input, (3) the avoidance of difficulties encountered during wavenumber domain transformations, and (4) the objective nature of the solution. Application to a gravity data set from Hamilton County, Indiana, has yielded a geologically reasonable result that agrees with published models derived from interpretation of gravity, magnetic, seismic, and drilling data.

No abstract

No abstract

We operated an eleven-station network of digital instruments in the Wabash Valley region from November 1995 through June 1996 in order to investigate seismic activity in the Wabash Valley seismic zone. One station of the network was a ten-element, three-component, high-frequency, phased array. The array was primarily responsible for lowering the detection threshold by approximately 1.5 magnitude units below that achieved previously, to magnitudes of 1.2 to 1.5. We observe a significant excess of events in the region from that expected by extrapolation of the historical and earlier instrumental catalogs. We show that the excess is related to a cluster of earthquakes near New Harmony, Indiana. We argue that their shallow depth, similarity ofwaveform characteristics, and proximity to producing oil wells and underground coal mines suggest that these small-magnitude events may be artificially induced. We find that discarding the events from this cluster leads to seismicity rates more consistent with previous data.