Safety in Mines Research Advisory Committee Final Project Report Improved seismic locations and location techniques by A Cichowicz, SM Spottiswoode, LM Linzer, D Drent, PS Heyns, and MF Handley. Research Agency: Project Number: Date: University of Pretoria SIM020304 2 March 2005 Executive summary The location of a seismic event is critically important because not only does a good location provide a guide for the rescue effort, but also all other parameters (e.g. moment, and energy) are dependent on it. Location errors scatter the seismic locations, causing them to plot as diffuse clouds, which obscure any seismic patterns that might be emerging. This impacts negatively on seismic hazard assessment, which is often based on spatial-temporal seismicity trends, hence location error has a direct effect on mining safety. There are two broad groups of event location methodologies, namely absolute location methods, and relative location methods. An absolute location is defined in this report as a once-off location of a single seismic event using some or all the seismograms recorded by the seismic system for the event. A relative location uses some or all of the seismograms from two or more neighbouring seismic events to locate them relative to each other. The rationale for relative locations originates from the observations of Omori, a Japanese seismologist, who in 1905 noted that horizontal pendulum seismometers produced similar motions for neighbouring earthquakes. Relative locations tend to be more accurate than absolute locations because of the increased utilisation of data from each neighbouring event in the group. At present both earthquake and mining seismologists use absolute event locations as standard. Absolute location errors in the mining industry range from 30 m to about 100 m, depending on the sensitivity of the system. With errors of this order it is usually possible to pinpoint which working place is closest to the event, and to draw reasonably credible event location – damage distribution correlations. It is also possible to positively identify the most seismically active areas on a mine, and to assess the effectiveness of measures to reduce the seismic hazard. However, these benefits can only be approximately determined at best, which places mine leadership in the invidious position of having to implement short-term measures to minimise employee exposure to the seismic hazard. Short-term measures consist of deciding whether to evacuate, therefore potentially losing face shifts, to improve stope support temporarily to minimise the probability of a rockburst, or a seismically induced fall of ground, or to continue work as before, thereby exposing employees to a perceived seismic hazard. Since seismic data remains insufficiently reliable for such decisionmaking, the mining industry has rightly steered clear of any implementation of such short-term measures, excepting in the most exceptional circumstances. One effective way to improve absolute location accuracy is to install more geophones or accelerometers, that is, to increase the sensitivity of the system so that more data are included in the location. Since the seismic system becomes more sensitive, it records many more events than before, which places a larger load on seismic processing staff. Another rationale for increasing system sensitivity is to improve the reliability of the data on which to base implementation of short-term measures to minimise exposure to the seismic hazard. In a nutshell, increasing system sensitivity improves the potential to reliably predict seismic events, which would have a major impact on mine safety. This project aims to explore the potential of an alternative route: get the best out of the seismic system already in place at the lowest possible cost, while at the same time improving safety through more reliable seismic hazard assessment. 2 This option proposes to leave the seismic system unchanged, and to use relative location techniques to improve location accuracy. All this requires is modifications and additions to existing seismic location software, and will be free from all the other problems associated with more data, and higher processing loads. The relative location techniques can be performed automatically, so they should not add to processing loads on seismic staff. However, biases can creep in, so methods to check this are necessary. The most effective way to ensure the quality of relative locations is to locate the events in relation to a master event of known location. In mining, catalogues of master events could be built up from signals recorded from the daily blast. This may not always be possible, so the mine seismologist could turn to one of the many methods used by earthquake seismologists, who seldom have artificially produced master events of known location to work with. Improved seismic locations using relative location techniques have been demonstrated in numerous case studies in the literature (see main report for more details). In many instances the authors of these case studies have indicated that merely relocating events using relative relocation techniques will produce only marginally improved results unless other aspects of the seismic data such as signal to noise ratio, P- and S-pick inconsistencies, seismic signal quality, velocity models, and arrival time accuracy are addressed simultaneously. This often requires painstaking work on the part of the operator, which is not practical for day-to-day seismic monitoring in mines. This study aims to address all these problems and constraints simultaneously, in a way that would be practically implementable in the mining environment. In order for this to be possible the following objectives for a new location algorithm must be met: 1. The relative location algorithm must be able to handle large amounts of data, of the order of 10000 events per day; 2. It must be able to extract constrained locations from the seismic data obtained from existing seismic systems using one or more of the relative location techniques, with location errors of the order of one magnitude smaller than those currently obtained with absolute location methods; 3. The algorithm must provide clear seismic clustering patterns (i.e. clear images of seismically active features in the rockmass) in the located data, together with a viable means of demonstrating that there are no systematic biases or errors in the images provided; 4. The algorithm must be sufficiently automated in order that it requires minimum operator intervention, thereby minimising location and quality control workloads, while at the same time providing more time for analysis and interpretation; 5. It must provide significantly improved seismic data for management and research purposes; 6. It must be able to relocate seismic events in existing databases retrospectively, thereby creating better quality databases. In a proposal for this research written in 2001, the following outputs were identified for this project: 1. The primary output is a greatly improved seismic location algorithm compared with those currently available - the error of location determination should be one order of magnitude smaller than that possible with current algorithms, all other variables such as the seismic system, rock mass structure and rock mass properties being constant; 3 2. An automatic location algorithm that accurately picks P- and S-onsets in both good signal:noise seismograms and noisy seismograms with an error level of 1 in 104 picks; 3. The secondary outputs include more accurate seismic parameter quantification e.g. energy and scalar moment; reliable automatic P- and Swave onset determination, or a means of constraining these picks so that they provide reliable locations; 4. A significant reduction in the manpower needed to locate the events, manage the data, and perform quality checks. The proposal went on to outline the principal research thrusts that would be necessary to achieve the above goals and outputs: 1. Investigate weaknesses and strengths of the waveform similarity technique and other location improvement techniques to define the applicability of each as a general location algorithm; 2. Identify and quantify sources of noise in seismograms. Use the University of Pretoria servo-hydraulic equipment to carry out this task with geophone and accelerometer clusters; 3. Determine options for identifying seismic signals in noisy seismograms; alternatively determine the signal to noise ratio relationship with P- and Sonset determination accuracy, and define an acceptable signal: noise ratio criterion. 4. Develop waveform similarity software for inclusion in current seismic location algorithms 5. Review and refine automatic means of P- and S-onset identification in all seismic signals (including noisy ones); alternatively develop an automatic means to separate signals with acceptable signal: noise ratio for P- and Sonset determination; 6. Test improved seismic location techniques, determining errors of location and reliability of automatic location. Refine seismic location algorithms during this test phase; 7. Estimate the error of seismic parameter determination (confine this to location error, scalar moment, and energy, considering site effects, seismic system variables, mining factors, and geotechnical conditions), backing this up with laboratory experiments if need be; 8. Comprehensive report on the project, covering all aspects of the research, the location algorithms and their implementation, together with limitations and capabilities; 9. Workshops on the software, its capabilities and limitations, estimated error of determination of location, scalar moment, moment tensor, and energy, and the relationships between these errors and site factors, seismic system factors, mining factors, and geotechnical conditions. The second point in the above list addresses noise, noise sources, and their effects on the seismograms because this aspect cannot be excluded from a holistic approach to the problem, as has been suggested in the case studies reviewed. All the above-mentioned work should be confined to mining-induced seismic events with local magnitudes of 0.0 and larger, and aimed at typical mine-wide seismic systems with 10 to 30 sensors covering an area of about 25 sq.km. If the results achieve the stated goals of an order of magnitude reduction in location error and a reliable automatic location algorithm, then the door to real-time seismic monitoring and analysis will have been opened. 4 The project began on 1 April 2002, continuing for two years to 31 March 2004. The project proposal appears in Appendix IV. The cost of the project totalled R3.2-million with the bulk going to the two outside sub-contractors, namely the CSIR Division of Mining Technology, and Integrated Seismic Systems International (ISSI), who were both to develop and integrate relative location algorithms into existing seismic location software for use on the mines. The department of mechanical engineering at the University of Pretoria also carried out some research on geophone noise, and data distortion because of site effects and geophone misalignment. The full reports from the three sub-contractors appear in Appendices I-III. A summary of the project results follows. The work by CSIR Miningtek has resulted in significant improvements in location methodologies. In particular, the hybrid location method, an offshoot of the double difference location method, was developed and shown to provide significant improvements in location accuracy. Additional work by the CSIR Miningtek group included showing the potential benefits of using development blasts to improve locations and advancing the robust use of waveform similarity in automatic locations. An example of how locations can be improved appears in Figure E1. More details of analyses and ways of evaluating the error of improved locations appear in Appendix I. All the software used to carry out this work has been implemented in the PRISM suite of programs. Figure E1: Relocated event clusters (dark gray) are far more concentrated around development blast locations than the absolute clusters (light gray) in a deep level mine shaft pillar. ISSI, through a series of case studies, has used several relocation techniques to improve seismic locations. The techniques include group locations using a single master event, group locations using multiple master events, and double difference methods for relative relocation of large event clusters. All methods show improvement in location accuracy, but the double difference method produces the best results, similar to those produced by Miningtek. ISSI also had the brief to continue with development of an automatic location algorithm, with results shown in Figure E2. The automatically located results are very good, and would be amenable to relocation using relative location methods, together with P- and S-pick inconsistency minimization methods. At present, there is no reliable automatic Swave picker because so many seismigrams have S-onsets in the P-wave coda, thereby making it alost impossible to obtain a reliable pick. The results shown rely on P-wave picks only, where reliability is extremely good, with fewer than one in 10000 5 P-picks in error. This means that routine automatic location without operator intervention and quality checking is very close. Figure E2: Comparison of absolute locations obtained manually (top) and automatically (bottom). Details of all the analyses appear in Appendix II. All the software required to perform the research has been included and implemented in the ISSI suite of seismic software. The report also contains tutorials aimed at the mine seismologist on how to use the relative relocation techniques. The techniques can be used in real-time (i.e. 6 locating events as they happen), or off-line (reprocessing the daily seismic data to improve locations during quiet periods, say at 5am every morning). Appendix III contains the report from the University of Pretoria, which was to determine the effect of installation conditions on data recorded by geophones in South African mines. The two major aspects considered were geophone alignment (horizontal and vertical) and the grouting conditions. Misalignments of a few degrees have been shown to have significant effects on geophone output, thereby introducing systematic errors into seismic locations. A method to accurately determine geophone alignment at geophone sites underground has been developed. This can be used to correct geophone outputs, thus removing systematic bias from location data. This project has been successful in that it has achieved all the objectives originally set for it. Like all projects there were management problems that arose from unexpected deviations from the original work plan. These did not occur because of poor management, but because of incorrectly estimated workloads for each enabling output. For example enabling output 4: Noise in seismograms and its removal proved to require far less effort than anticipated, while it became evident that, once the project had started, a literature survey was critical to the future direction of the project. No provision for a literature survey had been made in the proposal. The case studies and error determinations went off as planned, as did the implementation of the relative location algorithms in the seismic location software. The laboratory work carried out at the University of Pretoria quickly grew to much more than had been anticipated by the proposal, and is still ongoing as an MSc project at the University. The proposal envisioned only laboratory work, but the project had to be expanded to the underground environment in order to develop a reliable method of determining geophone alignments in-situ. The prime objective, to reduce location error by an order of magnitude (i.e. 10-fold), was considered too ambitious by all the sub-contractors, even though researchers in earthquake seismology had claimed 10-fold error reductions using the double difference method, and a 5-fold reduction in error using the waveform similarity method. These results, obtained from smaller datasets than mining datasets, required meticulous and time-consuming work by the authors, which is impractical in the mining environment because of the manpower cost and the large number of events involved. What this project has achieved is a 2- to 5-fold reduction in location errors when compared with absolute locations, depending on circumstances such as the validity of velocity models, and the accuracy of master event locations. This achievement has been balanced by another, which is to obtain improved results with minimum operator input and no extra cost, a far more difficult task than had been anticipated in the proposal. As experience with the software grows, and mine-wide velocity models improve with implementation of the software, it is anticipated that the 10-fold reductions in location error will become routine within five years. Finally, the automatic location algorithm, combined with an automated relative location algorithm and an automatic P-and S-pick inconsistency algorithm, should be able to produce seismic data of a consistently good quality. This will meet another goal, namely to reduce operator workloads, freeing up time for data analysis and interpretation, at no extra cost. This will have significant effects on seismic hazard management, safety, and future research projects, while at the same time making viable seismic prediction a real possibility. Further research should be undertaken in this direction. 7 Table of contents 1 2 3 4 5 6 7 Introduction ........................................................................................................ 11 Literature survey ................................................................................................ 14 2.1 Improved location algorithms ...................................................................... 15 2.1.1 Waveform similarity ............................................................................. 15 2.1.2 Joint hypocentral location (JHD).......................................................... 17 2.1.3 Double difference relocation method ................................................... 18 2.1.4 Other relocation methods .................................................................... 18 2.1.5 Location uncertainty using relative location methods .......................... 19 2.1.6 Brief comparison of mining induced seismicity and earthquakes ........ 19 2.2 Improvement of seismic data ...................................................................... 20 2.2.1 Site effects on data distortion .............................................................. 21 2.2.2 Managing noisy seismograms and automatic picks ............................ 21 2.3 Guidelines for research strategy from literature survey .............................. 23 Research work ................................................................................................... 25 Discussion of project conformance with stated goals ........................................ 25 Recommendations for future work ..................................................................... 26 References......................................................................................................... 27 Bibliography ....................................................................................................... 30 APPENDIX I…………………………………………………………………………..51 APPENDIX II……………………………………………………………………….. ..91 APPENDIX III………………………………………………………………………. 134 APPENDIX IV……………………………………………………………………….162 8 List of figures Figure E1: Relocated event clusters (dark gray) are far more concentrated around development blast locations than the absolute clusters (light gray) in a deep level mine shaft pillar.................................................................................................... 5 Figure E2: Comparison of absolute locations obtained manually (top) and automatically (bottom).......................................................................................... 6 Figure 2.2.2.1: Binomial probability of mislocated event vs. no. of stations used in location............................................................................................................... 22 9 List of tables Table 1: Enabling Outputs and Work Split ............................................................... 24 10 1 Introduction Current seismic event location algorithms are based on time residuals and velocity models (Eccles and Ryder, 1984; Gibowicz and Kijko, 1994; and Mendecki and Sciocatti, 1997). Such locations are referred to as absolute locations, in that they are obtained from some or all the seismic signals from a single seismic event recorded by the seismic system. Absolute locations for the typical mine-wide seismic monitoring system have errors of the order of 30 to 100 metres, depending on the distribution of the seismic sensors, their distance apart from each other, and factors such as the mining geometry and the geotechnical environment. Absolute location accuracy has not materially improved over the last ten years, and there is little potential for a significant reduction in location errors. With errors of the order of 30-100 m it is usually possible to pinpoint which working place is closest to the event, and to draw reasonably credible event location – damage distribution correlations. It is also possible to positively identify the most seismically active areas on a mine, and to assess the effectiveness of measures to reduce the seismic hazard. However, these benefits can only be approximately determined at best, and it is impossible to pinpoint the greatest seismic hazards, for example a potentially seismically active fault or dyke ahead of a face. One effective way to improve absolute location accuracy is to install more geophones or accelerometers, that is, to increase the sensitivity of the system so that more data are included in the location. Since the seismic system becomes more sensitive, it records many more events than before, which places a larger load on seismic processing staff. Another rationale for increasing system sensitivity is to improve the potential to predict seismic events. This project aims to explore the potential of an alternative route: get the best out of the seismic system already in place at the lowest possible cost, while at the same time improving safety through more reliable seismic hazard assessment. This option proposes to leave the seismic system unchanged, and to use relative location techniques to improve location accuracy. All this requires is modifications and additions to existing seismic location software, and will be free from all the other problems associated with more data, and higher processing loads. The relative location techniques can be performed automatically, so they should not add to processing loads on seismic staff. A relative location uses seismograms from two or more seismic events, to locate them relative to each other. Relative locations are generally more accurate than absolute locations because more data (i.e. more seismograms) are used to locate them. However, biases can creep in, so methods to check this are necessary. The most effective way to ensure the quality of relative locations is to locate the events in relation to a master event of known location. In mining, catalogues of master events could be built up from signals recorded from the daily blast. This may not always be possible, so the mine seismologist could turn to one of the many methods used by earthquake seismologists, who seldom have artificially produced master events of known location to work with. Uncertainties in event locations (or hypocentre scattering) can be grouped into two categories: random errors and systematic errors. Random errors are generally caused by errors in the arrival times of picked wave phases. Such errors include picking inconsistencies, the misidentification of seismic phases, variations in signalto-noise ratios, and human subjectivity. Systematic errors are caused by the variation in the rock mass structure between the seismic source and the receiver. 11 Such variations could be due to the local geology, or the presence of stopes and associated fractured rock that result in ray path focusing and defocusing of the elastic wave between source and receiver. All these errors combine to a greater or lesser degree in the location error associated with the absolute location of each seismic event. Thus, the quest for more data in order to improve the seismic service is probably incorrect beyond a certain threshold quantity. It appears that if a seismic system is sensitive down to magnitude zero, it provides a manageable amount of data while at the same time allowing sufficiently effective seismic interpretation for adequate seismic hazard management. Van Aswegen (2004) has demonstrated improved seismic hazard analysis and management by increasing seismic system sensitivity locally, for example in an area on the mine or around a geological feature that is known to be seismically hazardous. An alternative approach is to improve the quality of the data already being recorded and processed without adding any sensors to the existing systems, in other words, without increasing costs. This approach is not in any way at odds with SIMRAC Project GAP 601, which advocates increased sensitivity for improved data analysis and interpretation. Instead, this project actually complements GAP 601, because the results obtained will lead to improved seismic locations obtained from seismic systems with enhanced sensitivity. This project aims to improve seismic locations by an order of magnitude; that is, to reduce the error of location by an order of magnitude, without requiring any changes or additions to existing mine-wide seismic systems. Note that the principal output of this project is equally applicable in systems with improved sensitivity. The only techniques available to improve location accuracy without changes to the seismic system are relative location techniques, which use seismic information from two or more seismic events to constrain the location of the event of interest. The waveform similarity technique is one relative location method that has already demonstrated its ability to achieve this goal. There are other relative location techniques, such as the Joint Hypocentre Determination or JHD method, the Doubledifference Method, and relative relocations based on the known positions of master events. Improved seismic locations using one of the above-named relative location techniques have been demonstrated in numerous case studies in the literature (see Literature Review below). In many instances the authors of these case studies have indicated that merely relocating events using relative relocation techniques will produce only marginally improved results unless other aspects of the seismic data such as signal to noise ratio, P- and S-pick inconsistencies, seismic signal quality, velocity models, and arrival time accuracy are addressed simultaneously. This often requires painstaking work on the part of the operator, which is not practical for dayto-day seismic monitoring in mines. This study aims to address all these problems and constraints simultaneously, in a way that would be practically implementable in the mining environment. In order for this to be possible the following objectives for a new location algorithm must be met: 1. The location algorithm must be able to handle large amounts of data, of the order of 10000 events per day; 2. It must be able to extract constrained locations from the seismic data using one or more of the relative location techniques, with location errors of the 12 order of one magnitude smaller than those currently obtained with absolute location methods and existing mine-wide seismic systems; 3. The algorithm must provide clear seismic clustering patterns (i.e. clear images of seismically active features in the rockmass) in the located data, together with a viable means of demonstrating that there are no systematic biases or errors in the images provided; 4. The algorithm must be sufficiently automated in order that it requires minimum operator intervention, thereby minimising location and quality control workloads, while at the same time providing more time for analysis and interpretation; 5. It must provide significantly improved seismic data for management and research purposes. In a proposal for this research written in 2001, the following outputs were identified for this project: 1. The primary output is a greatly improved automatic seismic location algorithm compared with those currently available; 2. The secondary outputs include more accurate seismic parameter quantification e.g. energy and scalar moment; reliable automatic P- and Swave onset determination, or a means of constraining these picks so that they provide reliable locations; a significant reduction in the manpower needed to locate the events, manage the data, and perform quality checks. The proposal went on to outline the principal research thrusts that would be necessary to achieve the above goals and outputs: 1. Investigate weaknesses and strengths of waveform similarity technique and other location improvement techniques and define the applicability of each as a general location algorithm; 2. Identify and quantify sources of noise in seismograms. Use the University of Pretoria servo-hydraulic equipment to carry out this task with geophone and accelerometer clusters; 3. Determine options for identifying seismic signals in noisy seismograms; alternatively determine the signal to noise ratio relationship with P- and Sonset determination accuracy, and define an acceptable signal: noise ratio criterion. 4. Develop waveform similarity software for inclusion in current seismic location algorithms 5. Review and refine automatic means of P- and S-onset identification in all seismic signals (including noisy ones); alternatively develop an automatic means to separate signals with acceptable signal: noise ratio for P- and Sonset determination; 6. Test improved seismic location techniques, determining errors of location and reliability of automatic location. Refine seismic location algorithms during this test phase; 7. Estimate the error of seismic parameter determination (confine this to location error, scalar moment, and energy, considering site effects, seismic system variables, mining factors, and geotechnical conditions), backing this up with laboratory experiments if need be; 8. Comprehensive report on the project, covering all aspects of the research, the location algorithms and their implementation, together with limitations and capabilities; 9. Workshops on the software, its capabilities and limitations, estimated error of determination of location, scalar moment, moment tensor, and energy, and 13 the relationships between these errors and site factors, seismic system factors, mining factors, and geotechnical conditions. The second point in the above list addresses noise, noise sources, and their effects on the seismograms because this aspect cannot be excluded from a holistic approach to the problem, as has been suggested in the case studies reviewed. All the above-mentioned work should be confined to mining-induced seismic events with local magnitudes of 0.0 and larger, and aimed at typical mine-wide seismic systems with 10 to 30 sensors covering an area of about 25 sq.km. If the results achieve the stated goals of an order of magnitude reduction in location error and a reliable automatic location algorithm, then the door to real-time seismic monitoring and analysis will have been opened. 2 Literature survey The location of a seismic event is the most important seismic parameter because all other parameters (e.g. moment, and energy) are dependent on it. In addition, the interpretation of seismogenic features is critically dependent on the accuracy of the estimated hypocentre positions, while hypocentral scattering due to location error affects other parametric data, for example, spatial-temporal seismicity trends, and rockmass structure imaging. Thus seismic hypocentre location and relocation is as old as seismology itself, which began to grow as a science with the development of the first seismometer by L Palmieri in 1855, and the development of horizontal pendulum seismometers by John Milne in Japan in 1880 (Encylcopædia Britannica, 2003). Some twenty-five years later, Omori (1905) noted that horizontal pendulum seismometers produced similar motions for neighbouring events. Perhaps the greatest boost to the development of seismology came with the San Francisco earthquake of 18 April 1906, which was followed by the formation of the Seismological Society of America on 20 November 1906 (Byerly, 1964). Even at these early stages, the precise location of a seismic source was of primary importance, and ways were sought to reduce errors. This did not meet with immediate success until telemetry (for remote data transmission and seismic station management) and clock synchronisation (for seismic wave arrival times at different stations) had improved considerably. Relocation techniques to improve locations from existing datasets therefore remained relatively undeveloped until the 1960s, when researchers began to devise and publish techniques to improve hypocentre location errors. The imperative for developing these techniques came not only from earthquake seismology, but also from the need for accurate location and characterisation of nuclear detonations during and after the Cold War. For example Thurber and Engdahl (2000), in a review of advances in seismic location techniques, state that global concerns about nuclear armament testing and proliferation demand improvements in global and regional seismic event location capability. They state that improved locations can come from: • • • • Improved data quality and data usage; Improved velocity models; Better station and/or source-region travel-time corrections; Better location algorithms. The above points can be divided into two main groups, namely improved data quality, and improved location algorithms. The literature survey that follows will be split up into these two categories, since they both form major thrusts in the current research. 14 2.1 Improved location algorithms There are many ways to optimise the location of an event, amongst which the Nelder and Mead Simplex Optimisation method (Nelder and Mead, 1965) has been applied by seismologists. This method is a non-linear numerical algorithm that is more efficient than the derivative-based Gauss-Newton location algorithms, especially in cases where only a few P-waves are detected. Other methods described in the literature to reduce hypocentral scattering are based on time residuals and a velocity model. Earthquake relocation techniques have recently been markedly improved by using a number of different techniques, several of which are highlighted in this report. There are numerous case studies in which reprocessed seismic data using various techniques has resulted in dramatic improvements in imaging seismically active structures. A few examples of the most commonly used relocation techniques will be covered to provide an indication of the degree of success of these methods, their advantages and disadvantages, and their applicability to mining induced seismicity. Coverage is confined to more recent results only, because there are hundreds of papers on event relocation in the literature, and the latter relocation case studies are probably more dramatic than earlier studies. A bibliography containing a sample of over 300 papers written on relocation and related subjects appears after the references. This bibliography is by no means complete, thus it serves to show how extensively these techniques have been reported in the literature. The second purpose of the bibliography is to provide a reliable guide to the serious student of seismic hypocentre relocation techniques. 2.1.1 Waveform similarity Waveform similarity relocation techniques are based on Omori (1905), who observed that events with neighbouring locations possess similar waveforms. The physical reasons for this are that mechanisms of neighbouring events are likely to be similar, and the hypocentre–geophone ray paths for the two events should pass through similar rock. Based on this it is possible to improve P- and S-wave arrival time measurements by making them more consistent in the dataset. The accuracy of the relative locations can be further enhanced if the events are clustered spatially and have similar waveforms. The waveform similarity is exploited to pick precise arrival times (e.g. Got et al. 1994; Nadeau et al., 1995; Spottiswoode and Milev, 1998; Rubin et al., 1999). The maps of the enhanced locations determined using these techniques show details not evident in absolute locations computed using conventional methods. Slunga et al. (1995) discuss three case studies of relocated event clusters in Iceland which not only fit fault plane solutions obtained from the data, but also fit fault plane positions determined by dip and strike measurements at surface. Nadeau et al. (1995) used waveform similarity techniques on microseismic data from the San Andreas Fault, which showed clustering on the fault surface. This data was also used to assess loading of the fault. Got et al. (1994) used multiplet (multiple similar waveforms) relative location of seismicity below Mount Kilauea, Hawaii, to resolve subsurface structural features responsible for the seismicity. There are several more papers in the bibliography on seismicity associated with volcanicity at Mount Kilauea and its implications for subsurface structure and tectonism. Phillips et al. (1997) used the similar waveform technique to relocate some 11000 seismic events in a geothermal reservoir at Fenton Hill, New Mexico. Nearly all the relocated events fell into spatially distinct, planar clusters of 80-150 events, which revealed the activation of joints in the rockmass. Many of the clusters showed sharp 15 linear edges, which indicate the boundaries of slipping joints truncated by aseismic joints. The authors conclude that they were able to determine a detailed joint structure from the relocated seismic data. Menke (1999) discusses waveform similarity (waveform cross-correlation) and differential traveltimes to constrain earthquake locations, for application in nuclear nonproliferation. He claims (for worldwide networks) that differential traveltimes can be measured to millisecond precision, allowing relative location precise to a few metres. The author briefly presents two case studies, firstly a synthetic dataset containing 64 events with 99% improvement in misfit, and secondly, using the principal of reciprocity, to locate eleven seismic stations in a linear array using seismograms measured at the stations from a single earthquake. The solution in the second case reduces the data misfit by 95%, and recovers the linearity of the station locations and their relative order. Although the paper does not contain much detail of the computation-intensiveness of the procedure, the author notes that about 100000 iterations-requiring a few minutes on a computer workstation-are necessary to obtain a solution. The program is available in the public domain, and should be tested in the mining environment. Aster and Rowe (2000) assert that inconsistent P- and S-wave arrival estimates in large seismic datasets will obscure true hypocentre distribution in space, thereby masking important seismogenic structures such as faults or abyssal magma movements in volcanoes. Available algorithms have resulted in dramatic improvements in discerning seismogenic structure, but they still require substantial analyst intervention, therefore limiting their applicability to small datasets. They summarise progress toward a fully automated technique based on waveform similarity that removes P- and S-wave pick inconsistencies in large datasets. In a follow-up case study using the above techniques on large datasets, Rowe et al. (2001) were able to reduce the median hypocentre discrepancy distance from 31m to 7m at the Soultz Geothermal Reservoir. They conclude in this study that large-scale automatic reprocessing will not provide the flexibility and adaptability of human analysis for some time. As far as is known to date, this algorithm is not yet implementable automatically, still requiring considerable operator intervention to ensure reliable seismic data. Although this approach could be applicable in the mining situation because it promises to reduce operator load, it is considered not yet suitable for reprocessing mine events, because many mine arrays detect in excess of 1000 seismic events daily. In the only known publication of the use of waveform similarity in a mining situation, Young et al. (2001) extended the work of Riviere-Barbier and Grant (1993) to analyse clusters of similar waveforms from mining-site blasts in New Mexico and Arizona. The startling aspect of this work is that the authors were able to relocate the clusters on the respective mines, which lay up to 300 km beyond the boundaries of the 15station NMTSN (New Mexico Tech Seismic Network). Their method is complicated and computer-intensive, and is probably not suitable for daily use on a seismically active mine. Waveform similarity techniques clearly produce startling improvements to seismic datasets, to the point that diffuse clouds of absolute locations resolve into images of seismically active structures. Although the results are good, waveform similarity techniques rely on extensive operator intervention and judgment. The authors of the above papers report that it is painstaking and slow work, and is not suited to large seismic catalogues. The waveform cross-correlation methods are also computer- 16 intensive, and not always reliable. It therefore appears that this technique will not be easily automated, and it is thus not appropriate for relocating large mining-induced seismicity catalogues. The waveform similarity method is probably the most extensively used relocation technique, becoming increasingly popular in earthquake studies. For example, the most recent issue of the Bulletin of the Seismological Society of America (Volume 92 - issue 5, 2002) has two out of 33 papers on the topic, namely that of Rubin (2002), and Got et al. (2002). A common thread to all the papers to date on the topic is that doublets or multiplets with almost identical waveforms are fairly common, but not the rule. The degree of similarity decreases with inter-event distance, but in a controlled manner that can be used with advantage. When processing for large mine data sets, each event of large data sets should be compared with all nearby events. This approach may not be foolproof, because hypocentral scattering may bring relatively distant, non-similar events together. Conversely, widely scattered similar events may never be compared. 2.1.2 Joint hypocentral location (JHD) The effects of errors in the knowledge of the earth structure can be minimised by using relative location methods (e.g. Got et al., 1994). These methods apply if the hypocentral separation between two earthquakes is small compared to the eventsensor distance and the scale length of the velocity heterogeneity. Under these conditions, the ray-paths between the source region and receiver can be considered common, and the difference in travel times for two events observed at one station can be attributed to the spatial offset between the events with high accuracy. JHD, or Joint Hypocentre Determination (Dewey 1971, Pujol 2000) is a relatively simple but effective way to account for lateral velocity variations not included in onedimensional velocity models used to locate seismic events. Experience shows that JHD always improves relative locations, but may introduce error to absolute locations, for example a systematic shift in true locations in some circumstances. It is however, an effective semi-quantitative method for three-dimensional velocity assessment, and therefore may have application in developing evolving velocity models for the mining situation. Jones and Stewart (1997) presented a statistical approach to reducing hypocentral ‘fuzziness’ caused by pick inconsistencies. This method is also referred to as ‘collapsing’ (see also Fehler et al., 2000 below), and is based on the assumption that the pick errors form a Gaussian distribution. According to Rowe et al. (2002) synthetic tests as well as comparisons to results from manual repicking of recorded data sets have demonstrated considerable success. These researchers suggest that collapsing, in combination with techniques for adjusting the velocity model and/or near-surface static corrections, may be a good approach to improve delineation of seismogenic features. The advantage of collapsing is that waveform data is not required. However, the collapsing technique can be prone to artifacts. Fehler et al. (2000) propose a method known as ‘JHD-collapsing’. This method is based on the same basic assumption of Jones & Stewart (1997) that there is a greater clustering on actual earthquake locations than there is in locations determined using conventional techniques. The difference is that the method is implemented as part of the Joint Hypocentre Determination (JHD) process so it operates on raw travel-time data rather than on derived hypocentres. These researchers found that the JHD-collapsing method resulted in the hypocentres being significantly more clustered than those found using JHD alone. The authors present one case study based on that by Phillips et al. (1997), where the researchers had painstakingly re-picked the relative arrival times of events in a catalogue. JHD- 17 collapsing produced comparable results to the manually re-picked results from Phillips et al. (1997) with considerable savings in time and manpower. 2.1.3 Double difference relocation method Waldhauser and Ellsworth (2000) have developed a ‘double-difference’ location algorithm, which is an offshoot of the JHD-collapsing method described above. Details of the method appear in Appendix II. It is effective in sharpening the seismogenic source volume image of large numbers of earthquakes. The location method incorporates absolute travel-time measures (obtained from catalogue data) and/or cross-correlations of P- and S-wave differential travel-times (derived from cross-spectral methods). The method is based on the premise that if the interhypocentre distance is much less than the source-sensor distance, then the seismic waves must have travelled along similar paths through essentially the same rock, and the resulting travel-time residuals must therefore be an accurate measurement of the offset between the two hypocentres (see also Got et al., 1994). The residuals between the observed and theoretical travel-time differences are known as ‘doubledifferences’ and are minimised for pairs of earthquakes at each station while linking together all observed event-station pairs. These researchers applied the double-differences algorithm in two separate case studies to relocate earthquakes on the northern Hayward Fault, California. The double difference relocations improved location uncertainty by an order of magnitude, condensing the two clusters on the fault surface into sharply defined features. The relocated on-fault seismicity revealed a nearly planar and vertical fault zone extending beneath the surface trace of the Hayward Fault. In an earlier study, Waldhauser et al. (1999) revealed a focused picture of events on the Hayward Fault aligning in depth along linear, horizontal streaks. Waldhauser (2001) later reported on the computer program HypoDD, which was written to perform the doubledifference relocations. The software is available in the public domain. This method is automated, fast, can be applied to dense clusters of seismic events (in fact it works best in this situation), and can be applied to large event catalogues, giving it strong potential for use in the mining situation. 2.1.4 Other relocation methods Myers and Schultz (2000) use a master event location technique to calibrate travel times along certain paths. Because events (nuclear detonations) with known space and time coordinates are very restricted, a technique of imperfect calibration has been developed using well-located earthquakes. This work does not appear to have application in the mining industry unless calibration blasts with known space and time coordinates are carried out on seismically active mines. Sambridge and Kennett (2001) applied a recently developed direct search method for inversion known as the Neighbourhood Algorithm (NA) to the location problem. The algorithm uses randomised sampling of a 4-dimensional hypocentral parameter space, to search for solutions with an acceptable fit to data. At each stage, the hypocentral parameter space is partitioned into a series of convex polygons called Voronoi cells. Each cell surrounds a previously generated hypocentre for which the fit to the data has been calculated. New hypocentres are generated randomly in the neighbourhood of those hypocentres with smaller data misfit as the algorithm proceeds. In this way, all previous hypocentres guide the search and the more promising regions of parameter space are preferentially sampled, until a solution is found with acceptable data fit. 18 The procedure is repeated on an event-by-event basis, thus enabling the procedure to be fully automatic. The authors present two case studies of which the first is the Marinas event relocation, which quickly converges to the optimal position. The second case study involves a nuclear detonation at the Semipalatinsk Test Site in east Kazakhstan, for which the Soviets had supplied the authors the precise coordinates (Bocharov et al., 1989). The result showed good convergence with best fits in a block 5 km in a North-South direction by 3 km in an East West direction. This result was based on sparse data. The authors claim that the method provides good quality locations with far fewer travel-time evaluations than any previous non-linear location procedure. This procedure shows promise for the mining situation because it is automatic, it provides good locations both inside and outside the seismic array, and the procedure requires only sparse information to obtain a location. As far as is known, this procedure has never been applied in a mining situation. Rabinowitz (2000) bases his Flexible Tolerance Method (FTM) on that of Nelder and Mead (1965), considering it a natural extension, and an efficient location optimization algorithm. This scheme has applications in automatic location algorithms, where there is a real need to incorporate nonlinear restraints to minimise the possibility of an incorrect location. The constraints that can be applied are the time-order of Parrivals, back azimuths to some stations, and source-to-station distances from S-P separation information. All these are effectively incorporated in the FTM scheme, improving its efficiency, and down-weighting possible outlying observations. This scheme is especially effective where data is few, and where the hypocentre lies outside the array, which will often cause large location errors with standard location algorithms. 2.1.5 Location uncertainty using relative location methods Jones and Stewart (1997) say that there is always greater clustering at actual earthquake locations than there is in absolute locations obtained by conventional techniques. This means that diffuse clouds of seismicity, which do not lend themselves to much further analysis and interpretation, are brought into sharper focus using an improved location technique. It would seem that most of the authors above are satisfied with results purely on the basis that imaging of seismogenic structures has improved by the denser clustering of events around them. Some of the results are compelling, e.g. spatial correlation of imaged structures with surface fault traces in Iceland (Slunga et al., 1995), and the Northern Hayward Fault, California (Waldhauser and Ellsworth, 2000). Very little error analysis has been undertaken in the recent literature, certainly with respect to the relocation methods discussed above. Rodi and Toksös (2001) have developed a general theoretical and computational framework for characterising the uncertainty in seismic locations, especially in situations where the seismic data is sparse. The authors show the uncertainty distributions surrounding event locations using sparse data in three separate case studies, concluding that elliptical uncertainty regions derived from a linear Gaussian analysis do not adequately characterise the uncertainty. This work is to be followed by developing a methodology, which is likely to be based on seismic catalogues obtained from local and regional networks in Turkey. 2.1.6 Brief comparison of mining induced seismicity and earthquakes Seismic data from the mining environment shows three fundamental differences with earthquake data. The first difference is that mining-induced seismicity is dynamic. By the term dynamic we mean that different rockmass structures will be excited (or 19 induced to release seismic energy) at different times as the mining advances into previously unmined ground. This should place constraints on the choice of clusters, since newer events in the cluster may not have taken place on the same structure as earlier events. The implications of this are not known, although it is certain that waveform similarity techniques will probably be less applicable than in the natural environment, where seismicity is essentially static (i.e. confined to one or several seismogenic structures, and for all intents and purposes remaining approximately constant in our lifetime). The second significant difference is that mining data are generally not sparse, since data from eight or more geophones is often available from seismic events with ML ≥ 0.0 on the typical mine-wide seismic system. The third difference is that mininginduced seismic events are small (approximate effective range -4.0 ≤ ML ≤ 5.0) in comparison with earthquakes (approximate effective range -4.0 ≤ ML ≤ 8.6), yet the hypocentre size in mines in relation to the source-sensor distance is often considerably larger than the case for earthquakes. These fundamental differences, and the lack of objective error analysis, require that objective relocation uncertainty estimates should be addressed in the mining situation. Although it has been implied in this project, no direct provision has been made for uncertainty analysis, or for the objective evaluation of bias-free relocated datasets. 2.2 Improvement of seismic data Good locations are born of good seismic data. Good data is only possible if all of the following criteria have been met: • • • • • • • • • • • Vertical geophones are vertical; Horizontal geophones are horizontal, and correctly aligned in the desired directions; The sensors are well coupled with the rockmass (i.e. solidly grouted into the borehole); The sensors are sited in solid rock; Mining induced stress changes at the sensor site are expected to remain low so that there is no stress-induced damage to the surrounding rockmass or the borehole grout; There are no sources of either mechanical or electrical noise near the sensor installation, e.g. fans, or electrical interference; The sensor cluster, and the borehole grout-sensor interface do not have any undesirable natural frequencies that introduce noise to the seismograms at typical seismic frequencies; The sensor cluster is properly sealed against the ingress of water or any other substance that could result in corrosion or other damage to the installation; The sensors are good quality, and correctly specified for the environment and seismic signals they will record; The sensors are connected to data collection and data transfer systems which will reproduce the measured signals with a high degree of fidelity; The installation as a whole is adequately protected from any damage that may occur as a result of normal mining operations, or from vandalism. This work is aimed at the first five points, since most of these have been inadequately addressed in the past, while solutions have already been found and implemented for the latter six points. 20 2.2.1 Site effects on data distortion Geophone calibration studies are primarily aimed at calibrating the geophone sensor before installation (Faber and Laroo, 2001; Kim and Lee, 2000). The manufacturer typically supplies these data to the user along with the sensors purchased. No postinstallation calibrations are described excepting for geophone misalignment corrections in the horizontal plane (see below). Feroci et al. (2000) discuss grouting conditions for survey purposes, while the effect of grouting conditions in the mines has not been covered in the available literature. Most of the research found by the literature study was aimed at using geophones in order to obtain a more detailed image of the subsurface geological structure of certain areas (Feroci et al., 2000; Elkibbi and Rial, 2003). Very little research has been done on the vertical alignment of geophones, or the effect that vertical misalignments may have on the data. As far as is known, no work has been done on how to quantify the magnitude or direction of misalignment in the vertical direction with the use of the electrical response of the geophone itself. Geophone alignment in the horizontal plane has already been explored by many researchers and mining experts (e.g. Bland and Stewart, 1996; Zou and Wu, 2001), while a lot of work has been done on the calibration of geophone installations in the horizontal plane (Luna and Jadi, 2002; Elkibbi et al., 2003). Most of this work utilises events with known time and magnitude in order to calibrate the geophone data. As a result of the fairly extensive research in this area, most seismic analysis programs contain software with the capability to rectify misaligned geophones in the horizontal plane. 2.2.2 Managing noisy seismograms and automatic picks Noise in seismograms has the effect of reducing the quality of the P- and S-picks for location purposes, while at the same time introducing errors to the scalar energy and moment estimations. There are two types of noise: regular, cyclic, or predictable noise; and white or random noise, which is irregular and unpredictable. Most noise in nature is a combination of the two. Regular noise can be easily analysed for frequency content and then subtracted from seismograms where it is present. All seismic location software available includes this facility. White noise is far more difficult, if not impossible to eliminate. Imaging of subsurface structures by active seismic survey uses the signal stacking technique to eliminate random noise when weak reflections fall below the noise threshold (e.g. Goldstein, 1998). The premise behind stacking is that if all the returning signals are similar while the noise is generally random, one can “stack” the signals i.e. add them together resulting in the weak reflections superimposing upon each other, while the random noise will tend to cancel itself out, resulting in a readable reflection signal. This of course will require several repeated soundings, which is not possible in the case of seismicity, since each event only happens once. There may be potential for signal stacking highly similar waveforms, although this is extremely unlikely because determining waveform similarity is impossible for signals swamped in noise; it would have to be done on a trial and error basis with an expected very low success rate. Researchers have scarcely ventured into this area with perhaps Niewiadomski (2002) being the sole researcher who has developed a neural network algorithm to remove noise from seismograms. Numerous authors refer to the subject in passing, but merely as an issue to be addressed when it poses a problem. For example, Zhang et al. (2003) describe an automatic P-onset picker that is capable of dealing with noisy seismograms. The authors claim it is able to pick P-onsets with 95% 21 reliability. If the probability of a successful pick is distributed binomially, the probability of a mislocated event versus the number of stations used in the location is shown in Figure 2.2.2.1 below. Probability of Mislocated Event 0.5 0.4 0.3 0.2 0.1 0.0 4 5 6 7 8 9 10 11 12 Number of Geophones used in Location Figure 2.2.2.1: Binomial probability of mislocated event vs. no. of stations used in location The result shows that the automatic picking algorithm has a 19% probability of mislocating an event using 4 stations, increasing to 46% for 12 stations using P-Picks only. Better constraints to the location using S-picks are possible, but far more difficult. This is especially true when the S-onset lies in the P-coda; Cichowicz (2002) claims that there is still no solution in sight. This is still far from satisfactory, and would require extensive operator intervention to manage the quality of the database. In order to locate 99.99% of all recorded events reliably, the probability of an unsuccessful P-pick will have to be of the order of 0.00001 (i.e. one in one hundred thousand), assuming a 12-station location. Noise reduction and elimination in seismograms does not hold much promise at the moment for the following reasons: • • • There are several successful techniques to eliminate P-and S-pick inconsistencies in seismic datasets, although these still require extensive operator intervention; Automatic location algorithms would improve dramatically if P- and S-pick inconsistencies could be eliminated from large datasets using an automatic algorithm; Relative relocation methods such as the Double Difference Method need only time residuals to obtain reliable locations, which could be made even more reliable using calibration blasts (master events) and improved velocity models (which are a natural result of relative relocation methods). Noisy seismograms can be dealt with in several ways, either by routine maintenance of the seismic system, or by software methods: • Replace the geophone cluster, or relocate it elsewhere if there is a nearby source of mechanical or electrical noise; 22 • • • • Investigate possible sources of noise or interference in data transmission lines/components and rectify or replace where necessary; Discard the data if the problem is intractable, and consider replacing the geophone/data link; Subtract the noise from the seismogram using one of the many frequency analysers available, and already implemented in seismic location software; Routinely apply Niewiadomski’s (2002) neural network algorithm to all seismic data before it is processed for location – this is only possible if the neural network is adequately “trained” to remove reasonably predictable noise, and impractical if the neural network has to be “retrained” for each incoming event. Experience with seismic location suggests that the signal: noise ratio must be at least 1.25:1.00 for the first swing of the P-wave for P-onset picking, and at least 1.75:1.00 for S-onset picking. There is no simple criterion determinable for this problem, because these figures will vary according to the type of noise, whether the picks are manual or automatic, and the location accuracy desired. It is concluded that this area does not require further investigation, because the rewards are far outweighed by the costs, and results are far more readily obtainable from relative location methods. 2.3 Guidelines for research strategy from literature survey Although results have been spectacular, the literature survey indicates that the oldest and most popular technique, waveform similarity, may not be an ideal approach for the mining environment because it requires extensive operator intervention, intensive computational input, while it still suffers from waveform cross-correlation pitfalls, and has been confined to relatively small datasets. Finally, it has not yet been automated, and due to the complexity of the procedure, it still requires the flexibility and innovation of human intervention, thereby making it generally unsuitable for the mining environment at present. Menke’s (1999) procedure may have the most applicability, although it must be borne in mind that his procedure relates to a very special class of earthquakes, namely nuclear detonations. JHD-collapsing (Fehler et al., 2000), the Double Difference Method (DDM) (Waldhauser and Ellsworth, 2000), the Flexible Tolerance Method (FTM) (Rabinowitz, 2000), and the Neighbourhood Algorithm (NA) (Sambridge and Kennett, 2001) all promise to be applicable in the mining environment because they have been automated, are capable of dealing with large datasets, and will require only minimum operator intervention. Of these JHD-collapsing, and the DDM offer the best promise because they have both produced good results in fairly extensive testing. The remaining two approaches should not be tested yet unless there is clear evidence that the first two methods will not be applicable in the mining environment. In the long term they should be investigated because they may have applicability with as-yet unforeseen advantages. Since so little work has been done on vertical misalignment of geophones, this should be investigated in the laboratory, both for its effects on the vertical as well as the horizontal geophones. These are bias determinations, and should not, as far as is known, result in random noise. Grouting effects should also be investigated in order to determine their influence on noise. Geophone site effects should also be tested, using in-situ tests underground. This should result in a viable method to 23 determine geophone misalignment underground, and then to correct for it in the seismograms. Table 1: Enabling Outputs and Work Split Enabling Output Number 1 2 3 4 5 6 7 8 9 Description Subcontractor Investigate weaknesses and strengths of waveform similarity technique and other location improvement techniques and define the applicability of each as a general location algorithm Identify and quantify sources of noise in seismograms. Use the University of Pretoria servo-hydraulic equipment to carry out this task with geophone and accelerometer clusters Determine options for identifying seismic signals in noisy seismograms; alternatively determine the signal to noise ratio relationship with P- and S-onset determination accuracy, and define an acceptable signal: noise ratio criterion Develop waveform similarity software for inclusion in current seismic location algorithms Review and refine automatic means of P- and S-onset identification in all seismic signals (including noisy ones); alternatively develop an automatic means to separate signals with acceptable signal: noise ratio for P- and S-onset determination Test improved seismic location techniques, determining errors of location and reliability of automatic location. Refine seismic location algorithms during this test phase Estimate the error of seismic parameter determination (confine this to location error, scalar moment, and energy, considering site effects, seismic system variables, mining factors, and geotechnical conditions), backing this up with laboratory experiments if need be Comprehensive report on the project, covering all aspects of the research, the location algorithms and their implementation, together with limitations and capabilities Workshops on the software, its capabilities and limitations, estimated error of determination of location, scalar moment, moment tensor, and energy, and the relationships between these errors and site factors, seismic system factors, mining factors, and geotechnical conditions Miningtek BE@UP MF Handley Miningtek ISSI Miningtek, ISSI Miningtek, ISSI All All Noise removal from seismograms should not be pursued any further because this is an intractable problem that has only been tackled in passing by most authors in pursuit of other goals, while only a very few have ever tackled the problem as a primary goal. It is far more prudent to adopt effective seismic system maintenance policies, and where P-and S-picks are inconsistent because of noise or other system errors, to improve them using relocation software. 24 3 Research work In a meeting held at the ISSI offices in Stellenbosch in May 2002, it was mutually agreed that the task of developing methods to remove noise from seismograms was too difficult (if not impossible) and therefore this should be reviewed only in a literature survey, which had not been planned for the project. The literature survey was undertaken because it became apparent that there was little sense in duplicating research already described in the literature, and that there were many case studies and other fundamental work that would provide clear guidelines for the project. The research work was split between the subcontractors as shown in Table 1. Late in 2002 it became apparent that the laboratory work scheduled for geophone distortion testing would not be enough, again indicated by a literature review and initial laboratory trials (see Appendix III). This led to the decision to incorporate the work in an MSc, which would go into far more detail than, was provided for in the proposal. New planning meant taking the study underground, where actual in-situ conditions at geophone installations could be tested and measured. This required devising a technique to achieve this, using a modal hammer and instrumented pin. Details of the research and research results for all subcontractors are included in Appendices I-III. 4 Discussion of project conformance with stated goals This project has been successful in that it has achieved all the objectives originally set for it. Like all projects there were management problems that arose from unexpected deviations from the original work plan. These did not occur because of poor management, but because of incorrectly estimated workloads for each enabling output. For example enabling output 4: noise in seismograms and its removal proved to require far less effort than anticipated, while it became evident that, once the project had started, a literature survey was critical to the future direction of the project. No provision for a literature survey had been made in the proposal. The case studies and error determinations went off as planned, as did the implementation of the relative location algorithms in the seismic location software. The laboratory work carried out at the University of Pretoria quickly grew to much more than had been anticipated by the proposal, and is still ongoing as an MSc project at the University. The proposal envisioned only laboratory work, but the project had to be expanded to the underground environment in order to develop a reliable method of determining geophone alignments in-situ. The prime objective, to reduce location error by an order of magnitude (i.e. 10-fold), was considered too ambitious by all the sub-contractors, even though researchers in earthquake seismology had claimed 10-fold error reductions using the double difference method, and a 5-fold reduction in error using the waveform similarity method. These results, obtained from smaller datasets than mining datasets, required meticulous and time-consuming work by the authors, which is impractical in the mining environment because of the manpower cost and the large number of events involved. What this project has achieved is a 2- to 5-fold reduction in location errors when compared with absolute locations, depending on circumstances such as the validity of velocity models, and the accuracy of master event locations. Source parameter estimation errors are reduced proportionately using the standard 25 seismogram-source parameter relationships in seismology. Errors should be further reduced when geophone misalignments and site effects have been quantified and corrected. The achievement of 2- to 5-fold location error reduction has been balanced by another, which is to obtain improved results with minimum operator input and no extra cost, a far more difficult task than had been anticipated in the proposal. As experience with the software grows, and mine-wide velocity models improve with implementation of the software, it is anticipated that the 10-fold reductions in location error will become routine within five years. Finally, the automatic location algorithm, combined with an automated relative location algorithm should be able to produce seismic data of a consistently good quality. This will meet another goal, namely to reduce operator workloads, freeing up time for data analysis and interpretation, at no extra cost. This will have significant effects on safety, management of seismicity, and future research projects, while at the same time making viable seismic prediction a real possibility. 5 Recommendations for future work Since this project is not considered complete, there are several recommendations for future work, listed below. 1. Annually audit the level of industry implementation of the outputs from this project; 2. Automate the P- and S-pick inconsistency algorithm so that it can be easily and quickly applied to large datasets; 3. Couple the automated P- and S-pick inconsistency algorithm with the automatic location algorithm to eliminate inconsistencies in all automatic locations, thereby improving the data; 4. Build and maintain refined velocity models of all rockmass volumes currently monitored in the mining industry; 5. Develop and automate a reliable, unbiased, measure of relocation quality; 6. Continue with practical site characteristic determination methods, including geophone/rockmass coupling, and accurate geophone orientation, together with automatic compensation methods for seismograms when orientation errors etc., have been determined. In the case of the relocation quality measure, development is almost non-existent both in earthquake and mining seismology. During this project two methods were developed, and a third proposed, namely: 1. Comparing relocated blast data with the actual blast locations (this could be improved by introducing a system trigger so that not only blast location but time are available as master event data) 2. Cichowic’s relocation stability criterion (see Appendix II) 3. Spottiswoode and Linzer’s seismic cluster dimension criterion (does the cluster form and position conform with known geological or mining features?). New work should investigate the reliability of the three methods outlined above, and then to develop towards a reliable relocation quality measure. Once these goals have been met, other research such as the relationship between geological features and mining induced seismicity, and seismic prediction could be more effectively researched and developed. 26 6 References Aster R and Rowe C (2000): Automatic phase pick refinement and similar event association in large seismic datasets. In: Thurber CH and Rabinowitz N (Eds.): Advances in Seismic Event Location. Dordrecht: Kluwer Academic Publishers, pp. 231-263. Bland HC, and Stewart RR (1996): Geophone orientation, location and polarity checking for 3-C seismic surveys. Crewes Consortium for Research in Wave Exploration Seismology: Crewes Research Report Vol. 8, pp.3-1 to 3-5, University of Calgary, Canada. Bocharov VS, Zelentsev A, and Mikhailov VN (1989): Characteristics of 96 underground nuclear explosions at the Semipalatinsk Test Site (in Russian). Atomnaya Energiya 67 (3). Byerly P (1964): History of the Seismological Society of America. Bull. Seismol. Soc. Am. 54, pp.1723-1741. Cichowicz A (2002): Personal Communication. ISSI, Stellenbosch, May 2002. Eccles CD, and Ryder JA (1984): Seismic location algorithms: a comparative evaluation. In: Gay NC and Wainwright EH (Eds.): Rockbursts and Seismicity in Mines, SAIMM Symposium Series No. 6, Johannesburg, 1984. Elkibbi M, and Rial JA (2003): The Geysers geothermal field I: Result for shear wave splitting analysis in a fractured reservoir. Department of Geological Sciences, University of North Carolina. Encylcopædia Britannica (2003): Ultimate Reference Suite on CD. Faber CAM, and Laroo HA (2001): The effect of geophone specifications on vector fidelity. 63rd Meeting: Eur. Assn. Geosci. Eng., Session P176. Fehler MC, Phillips WS, Jones RH, House LS, Aster RC, and Rowe CA (2000): A method for improving relative earthquake locations. Bull. Seism. Soc. Am. Short Notes 90, pp.775-780. Feroci M, Orlando L., Balia R, Bosman C, Cardarelli E, and Deidda G. (2000): Some considerations on shallow seismic reflections surveys. J. Appl. Geophys. 45 No.2, pp.127-139. Goldstein P (1998): Signal Stacking Subprocess Manual – Seismic Analysis Code (SAC). Lawrence Livermore National Laboratory. Got J-L, Fréchet J, and Klein FW (1994): Deep fault plane geometry inferred from multiplet relative relocation beneath the south flank of Kilauea. J. Geophys. Res. 99 No. B8, pp.15375-15386. Got J-L, Okubo P, Machenbaum R, and Tanigawa W (2002): A real-time procedure for progressive multiplet relative relocation at the Hawaiian Volcano Observatory. Bull. Seism. Soc. Am. 92 No. 5, pp.2019–2026. 27 Jones RH, and Stewart RC (1997): A method for determining significant structures in a cloud of earthquakes. J. Geophys. Res. 102, pp.8245-8254. Kim DS, and Lee JS (2000): Propagation and attenuation characteristics of various ground vibrations. Soil Dynamics and Earthquake Engineering 19, pp.115-126. Luna R, and Jadi H (2002): Determination of dynamic soil properties using geophysical methods. Proceedings of the First International Conference on the Application of Geophysical and NDT Technologies to Transportation Facilities and Infrastructure – Geophysics 2000, Federal Highway Administration, St Louis, MO, Paper No. 3-1. Menke W (1999): Using waveform similarity to constrain earthquake locations. Bull. Seismol. Soc. Am. 89, pp.1143-1146. Myers SC, and Schultz, CA (2000): The Use of Imperfect Calibration for Seismic Location. US Department of Energy, Office of Nonproliferation Research and Engineering; Office of Defense Nuclear Nonproliferation, National Nuclear Security Administration, Contract Number W-7405_ENG-48. Nadeau R, Foxall W, and McEvilly T (1995): Clustering and periodic recurrence of microearthquakes on the San Andreas Fault at Parkfield, California. Science 267, pp.503-507. Nelder JA and Mead R (1965): A simplex method for function minimization. Computer Journal Vol. 7, pp. 308-313. Niewadomski J (2002): Detection of seismic signals on noisy seismograms. Acta Geophysica Polonica 50 No.2, pp.159-166. Omori F (1905): Horizontal pendulum observations of earthquakes in Tokyo: Similarity of the seismic motions originating at neighboring centers. PubI. Earthquake Invest. Comm. Foreign Lang. 21, pp.9-102. Phillips WS, House LS, and Fehler MC (1997): Detailed joint structure in a geothermal reservoir from studies of induced microearthquake clusters. J. Geophys. Res. 102, pp.11745-11763. Phillips WS (2000): Precise microearthquake locations and fluid flow in the geothermal reservoir at Soultz-sous-Forêts, France. Bull. Seism. Soc. Am. 90, pp.212-218. Pujol J (2000): Joint event location – the JHD technique and applications to data from local seismic networks. In: Thurber CH and Rabinowitz N (Eds.): Advances in Seismic Event Location. Dordrecht: Kluwer Academic Publishers, pp. 163-204. Rabinowitz N (2000): Hypocenter location using a constrained nonlinear simplex minimization method. In: Thurber CH and Rabinowitz N (Eds.): Advances in Seismic Event Location. Dordrecht: Kluwer Academic Publishers, pp. 23-49. Riviere-Barbier F, and Grant LT (1993): Identification and location of closely spaced mining events. Bull. Seism. Soc. Am. 83, pp.1527-1546. 28 Rodi W, and Toksös MN (2001): Uncertainty analysis in seismic event location. 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, pp.324-332. Rowe CA, Aster RC, Phillips WS, Jones RH, Borchers B, and Fehler MC (2001): Using automated, high-precision repicking to improve delineations of microseismic structures at the Soultz Geothermal Reservoir. Pure Appl. Geophys. 159, pp.563– 596. Rowe CA, Aster RC, Borchers B, and Young CJ (2002): An automatic, adaptive algorithm for refining phase picks in large seismic data sets. Bull. Seism. Soc. Am. 92, No.5. Rubin AM, Gillard D, and Got J-L (1999): Microseismic lineations along creeping faults. Nature, 400, pp.635-641. Rubin AM (2002): Using repeating earthquakes to correct high-precision earthquake catalogs for time-dependent station delays. Bull. Seismol. Soc. Am. 92 No. 5, pp.1647-1659. Sambridge MS, and Kennett BLN (2001): Seismic event location: nonlinear inversion using a neighbourhood algorithm. Pure Appl. Geophys. 158, pp.241-257. Slunga R, Rögnvaldsson ST, and Bödvarsson (1995): Absolute and relative locations of similar events with application to microearthquakes in southern Iceland. Geophys. J. Int. 123, pp. 409-419. Spottiswoode SM, and Milev A (1998): The use of waveform similarity to define planes of mining-induced seismic events, Tectonophysics 289, pp.51–60. Thurber CH and Engdahl ER (2000): Advances in global seismic event location. In: Thurber CH and Rabinowitz N (Eds.): Advances in Seismic Event Location. Dordrecht: Kluwer Academic Publishers, pp. 3-22. van Aswegen G (2004): Private communication. AngloGold West Wits Region, 2 August 2004. Waldhauser F, Ellsworth W, and Cole A (1999): Slip-parallel seismic lineations on the northern Hayward Fault, California. Geophys. Res. Lett. 26, pp.3525-3528. Waldhauser F and Ellsworth WL (2000): A double-difference earthquake location algorithm: Method and application to the northern Hayward fault. Bull. Seismol. Soc. Am. 90, pp.1353-1368. Waldhauser F (2001): hypoDD - - A program to Compute Double-Difference Hypocenter Locations. US Geol. Surv. Open File Report 01-113. Young CJ, Merchant BJ, and Aster RC (2001): Comparison of cluster analysis methods for identifying regional seismic events. 23rd Seismic Research Review: Worldwide Monitoring of Nuclear Explosions, pp.229-238. Zhang H, Thurber C, and Rowe C (2003): Arrival detection and picking with multiscale wavelet analysis and single component recordings. Bull. Seism. Soc. Am. 93, pp.1904-1912. 29 Zou DH, Wu YK, (2001) Investigation of blasting-induced fracture in rock mass using reversed vertical seismic profiling. J. Appl. Geophys. 48 No. 3, pp.153-162. 7 Bibliography Aki K, and Richards P (1980): Quantitative Seismology: Theory and Methods. W.H. Freeman, New York. Aktar M, Ergin M, Özalaybey S, Tapirdamaz S, Yörük A, and Bicmen F (2000): A lower-crustal event in the Northeastern Mediterranean: The 1998 Adana Earthquake (Mw=6.2) and its aftershocks. Geophys. Res. Letters 27, pp.2361-2364. Albright JN, and Pearson CF (1982): Acoustic emissions as a tool for hydraulic fracture location: Experience at the Fenton Hill Hot Dry Rock Site. J. Soc. Pet. Eng., 22, 523-530. Amorese D, Lagarde JL, and Laville E (1999): A point pattern analysis of the distribution of earthquakes in Normandy (France). Bull. Seismol. Soc. Am. 99, pp.742-749. Anderson KR (1981): Epicentral location using arrival time order. Bull. Seism. Soc. Am. 71, 541-546. Anderson KR (I982): Robust earthquake location using M-estimates. Phys. Earth Plan. Int. 30, pp.119-130. Ando M (1979): The Hawaii earthquake of November 29, 1975: Low dip angle faulting due to forceful injection of magma. J. Geophys. Res. 84, pp.7616-7626. Andrews M, Mooney W, and Meyer R (1985): The relocation of microearthquakes in the northern Mississippi embayment. J. Geophys. Res. 90, pp.10223-10236. Antolik M, Ekström G, and Dziewonski A (2001): Global event location with full and sparse datasets using three-dimensional models of mantle P wave velocity. J. Geophys. Res. 158, pp. 291-317. Antolik M, Nadeau R, Aster RC, and McEvilly T (1996): Differential analysis of coda Q using similar microearthquakes in seismic gaps. Part 2: Application to seismograms recorded by the Parkfield High Resolution Seismic Network, Bull. Seis. Soc. Am. 86, pp.890-910. Árnadòttir T, and Segall P (1994): The 1989 Loma Prieta earthquake imaged from inversion of geodetic data. J. Geophys. Res. 99, pp.21835-21855. Árnadòttir T, and Segall P (1996): Reply to comment. J. Geoplys. Res. 101, pp.20137-20140. Árnadòttir T, Segall P, and Delaney P (1991): A fault model for the 1989 Kilauea south flank earthquake from leveling and seismic data. Geophys. Res. Lett. 18, pp.2217-2220. Aster R and Meyer R (1988): Three-dimensional velocity structure and hypocenter distribution in the Campi Flegrei caldera, Italy. Tectonophysics 149, pp.195-218. 30 Aster RC and Scott J (1993): Comprehensive characterization of waveform similarity in microearthquake data sets. Bull. Seis. Soc. Am. 83, pp.1307-1314. Aster RC, Shearer P, and Berger J (1990): Quantitative measurements of shearwave polarizations at the Anza seismic network, southern California - implications for shear-wave splitting and earthquake prediction. J. Geophys. Res. 95, pp.1244912474. Aster RC, Slad G, Henton J, and Antolik M (1996): Differential analysis of coda Q using similar microearthquakes in seismic gaps Part I: Techniques and application to seismograms recorded in the Anza seismic gap, southern California. Bull. Seis. Soc. Am. 86, pp.868-889. Balch B, Hanse H, Sanford A, and Lin K (1997): A new map of the geographic extent of the Socorro midcrustal magma body. Bull. Seismol. Soc. Amer. 87, pp.174-182. Bame D, and Fehler MC (1986): Observations of long-period earthquakes accompanying hydraulic fracturing. Geophys. Res. Let. 13, pp.149-152. Baria R, Baumgartner J, Gerard A, Jung R, and Garnish J (1999): European HDR research programme at Soultz-sous-Forêts (France) 1987-1996. Geothermics 28, pp.655-669. Barton CA, Zoback MD, and Moos D (1994): Fluid flow along potentially active faults in crystalline rock. Geology 23, pp.683-686. Beroza G, Cole A, and Ellsworth W (1995): Stability of coda wave attenuation during the Loma Prieta, California earthquake sequence. J. Geophys. Res. 100, pp.39773987. Bianchi R, Casacchia R, Coradini A, Duncan A, Guest J, Kahle A, Lauciano P, Pieri D, and Poscolieri (1990): Remote sensing of Italian volcanoes. EOS, Trans. Am. Geophys. Union 71, pp.1789-I791. Bijwaard H and Spakman W (1999): Fast kinematic ray tracing of first- and laterarriving global seismic phases. Geophys. J. Int. 139, pp.359-369. Bijwaard H, Spakman W, and Engdahl ER (1998): Closing the gap between regional and global travel time tomography. J. Geophys. Res. 103, pp.30055-30078. Billings S (1994): Simulated annealing for earthquake location. Geophys. J. Int. 118, pp.680-692. Billings S, Kennett BLN, and Sambridge MS (1994): Hypocentre location: Generic algorithms incorporating problem-specific information. Geophys. J. Int. 118, pp.6937O6. Billings SD, Sambridge MS, and Kennett BLN (1994): Errors in hypocenter location: Picking, model and magnitude dependence. Bull. Seism. Soc. Am., 84, pp.19781990. Billington S and Isacks B (1975): Identification of fault planes associated with deep earthquakes. Geophys. Res. Lett. 2, pp.63-66. 31 Bjarnason I, and Einarsson P (1991): Source mechanism of the 1987 Vatnafjöll earthquake in South Iceland. J. Geophys. Res. 96, pp.4313-4323. Bjarnason I, Menke W, Flóvenz ÓG, and Caress D (1993): Tomographic image of the plate boundary in Southwestern Iceland. J. Geophys. Res. 98, pp.6607-6622. Block L (1991): Joint Hypocenter-Velocity Inversion of Local Earthquake Arrival Time Data in Two Geothermal Regions. Ph.D. Dissertation, Massachusetts Institute of Technology, 448pp. Block LV, Pheng CH, Fehler MC, and Phillips WS (1994): Seismic imaging using microearthquakes induced by hydraulic fracturing. Geophysics 59, pp.102-112. Boeve R, Spakman W, Bijwaard H, and Engdahl ER (1999): 3-D earthquake location with grid search methods: Application to nuclear explosions (abstract). International Workshop on Tomographic Imaging of 3-D Velocity Structures and Accurate Earthquake Location (PAPHOS99), Platres, Cyprus, 5-9 July 1999. Bolt BA (1960): The revision of earthquake epicentres, focal depths, and origin times using a high-speed computer. Geophys. J. Roy Astr. Soc. 3, pp.433-440. Bratt SR, and Bache TC (1988): Locating events with a sparse network of regional arrays. Bull. Seismol. Soc. Am. 78, pp.780-798. Brown SR (1987): Fluid flow through rock joints: The effect of surface roughness. J. Geophys. Res. 92, pp.1337-1347. Bryan CJ (1992): A possible triggering mechanism for large Hawaiian earthquakes derived from analysis of the June 1989 Kilauea south flank sequence. Bull. Seismol. Soc. Am. 82, pp.2368-2390. Bryan CJ, and Johnson CE (1991): Block tectonics of the island of Hawaii from a focal mechanism analysis of basal slip. Bull. Seismol. Soc. Am. 81, pp.491-507. Buland R (1976): The mechanics of locating earthquakes. Bull. Seismol. Soc. Am. 66, pp.173-187. Buland R (I986): Uniform reduction error analysis. Bull. Seismol. Soc. Am. 76, pp.217-230. Buland R and Chapman (1983): The computation of Seismic travel times. Bull. Seismol. Soc. Am. 73, pp.1271-1302. CardwelI R, and Isacks B (1976): Investigation of the 1966 earthquake series in northern China using the method joint epicenter determination. Bull. Seismol. Soc. Am. 66, pp.1965-1982. Carr, D, Young C, Harris J, Aster RC, and Zhang X (1999): Cluster analysis for CTBT seismic event monitoring. Seism. Res. Lett. 70, pp.227-228. Chen WP, and Nabelek J (1990): Source parameters of the June 26, 1989 Hawaiian earthquake. EOS Trans. AGU 71, 562. 32 Chiu JM, Johnston A, and Yang YT (1992): Imaging the active faults of the central New Madrid seismic zone using PANDA array data. Seism. Res. Lett. 63, pp.375393. Christoffersson AF, Husebye ES, and Ingate SF (1988): Wavefieid decomposition using ML probabilities in modelling single site 3-component records. Geophys. J. Int. 93, pp.197-213. Claerbout F, and Muir F (1973): Robust modeling with erratic data. Geophysics 38, pp.826-844. Cleary J, and Hales AL (1966): An analysis of the travel times of P-waves to North American stations, in the distance range 32º to 100º. BulI. Seismol. Soc Am. 56, pp.467-489. Console R, and Di Giovambattista R (1987): Local earthquake relative location by digital records. Phys. Earth Planet. Inter. 47, pp.43-49. Cook RD and Weisberg S (1982): Residuals and Influence in Regression. Chapman and Hall, New York, 230 pp. Cornet FM, and Julien P (1989): Stress determination from hydraulic test data and focal mechanisms of induced seismicity. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 30, pp.789-795. Cornet FM and Jones K (1994): Field evidence on the orientation of forced water flow with respect to the regional principal stress directions. Proc. 1st North Amer. Rock Mech. Symp. Austin, Texas, pp.61-69. Cornet FM and Yin J (1995): Analysis of induced seismicity for stress field determination and pore pressure mapping. Pure Appl. Geophys. 145, pp.677-700. Cornet FM, Shapiro SA, Audigrance P, and Royer JJ (2000): Large-scale in situ permeability tensor of rocks from induced microseismicity: discussion and reply. Geophys. J. Int. 140, pp.465-473. Corrado G, Guerra I, Lo Bascio A, Luongo G, and Rampoldi R (1976-1977): Inflation and microearthquake activity of Phlegraean Fields, Italy. Bull. Volcanology 40-3, pp.169-188. Crosson RS, and Endo BT (1982): Focal mechanisms and locations of earthquakes in the vicinity of the 1975 Kalapana earthquake aftershock zone 1970-1979: Implications for tectonics of the south flank of Kilauea volcano, island of Hawaii. Tectonics 1, pp.495-542. Crosson RS, and Endo ET (1981): Focal mechanisms of earthquakes related to the 29 November 1975 Kalapana, Hawaii, earthquake: The effect of structure models. Bull. Seismol. Soc. Am. 71, pp.713-729. Davis J (1986): Statistics and Data Analysis in Geology. J. Wiley and Sons, New York. De Natale G, and Pingue F (1992): Seismological and geodetic data at Campi Flegrei (southern Italy): constraints on volcanological models. In: P Gasparini, R Scarpa. and K. Aki (Eds.) Volcanic Seismology, Springer-Verlag, pp.484-502. 33 Deichmann N and Garcia-Fernandez M (1992): Rupture geometry from highprecision relative hypocentre locations of microearthquake clusters. Geophys. J. Int. 110, pp.501-517. Delaney PT, Wyss M, Lipman PW, and Okamura AT (1992): Comment on “Precursors to the Kalapana M = 7.2 earthquake” by Max Wyss, F.W. Klein, and AC Johnston. J. Geophys Res. 97, pp.4839-4841. Delaunay BN (1934): Sur la sphere vide. Bull. Acad. Science USSR VII: Class. Sci. Math. pp.793-800. Dewey J (1972): Seismicity and tectonics of western VenezueIa. Bull. Seismol. Soc. Am. 62, pp.1711-1751. Dewey JW (1971): Seismicity Studies with the Method of Joint Hypocenter Determination. Ph.D. Thesis, University of California, Berkeley. Dietz L, and Ellsworth W (1990): The October 17, 1989, Loma Prieta, California, earthquake and its aftershocks: Geometry of the sequence from high-resolution locations. Geophys. Res. Lett. 17, pp.1417-1420. Dietz L, and Ellsworth W (1997): Aftershocks of the Loma Prieta earthquake and their tectonic implications. In: P Reasenberg (Ed.), The Loma Prieta, California, Earthquake of October 17, 1989 - Aftershocks and Postseismic Effects. U.S. Geol. Surv. Prof. Pap. 1550-D, D5-D47. Dodge D (1996): Microearthquake Studies Using Cross-correlation Derived Hypocenters. Ph.D. thesis, Stanford University, 146pp. Dodge DA, Beroza G, and Ellsworth W (1996): Detailed observations of California foreshock sequences: implications for earthquake initiation processes. J. Geophys. Res. 101, 22371-22392. Dodge DA, Beroza GC, and Ellsworth WL (1995): Foreshock sequence of the 1992 Landers, California earthquake and its implications for earthquake nucleation. J. Geophys. Res. 100, pp.9865-9880. Douglas A (1967): Joint epicentre determination, Nature 215, pp.47-48. Dreesen DS, and Nicholson RW (1985): Well completion and operations for MHF of Fenton Hill well EE-2. Trans. Geotherm. Resour. Counc. Part 2 9, pp.105-110. Dvorak JJ, Klein FW, and Swanson DW (1994): Relaxation of the south flank after the 7.2 magnitude Kalapana earthquake, Kilauea volcano, Hawaii. Bull. Seism. Soc. Am. 84, pp. 133-141. Dziewonski AM, and Anderson DL (1981): Preliminary reference Earth model. Phys. Earth Planet. Int. 25, pp.297-356. Dziewonski AM, and Anderson DL (1983): Travel times and station corrections for P waves at teleseismic distances. J. Geophys. Res. 88, pp.3295-3314. Eberhart-Phillips D and Stuart W (1992): Material heterogeneity simplifies the picture: Loma Prieta. Bull. Seism. Soc. Am. 82, pp.1964-1968. 34 Einarsson P (1991): Earthquakes and present-day tectonism in Iceland. Tectonophysics, 189, pp.261-279. Einarsson P (1994): Holocene surface fractures in the South Iceland seismic zone. In: Earthquake prediction research in the South Iceland test area: state-of-the-art, Thorkelsson B (ed.), Icelandic Meteorological Office, Reykjavik. Einarsson P, and Eiriksson J (1982): Earthquake fractures in the districts Land and Rangárvellir in the South Iceland Seismic Zone. JökuIl, 32, pp.113-120. Eissler HK, and Kanamori H (1987): A single-force model for the 1975 Kalapana, Hawaii, earthquake. J. Geophys. Res. 92, pp.4827-4836. Engdahl ER (1998): Development of an archive of seismic ground truth events globally in support of monitoring under the CTBT. Proc. 20th Annual Seismic Research Symposium on Monitoring a Comprehensive Test Ban Treaty (CTBT). Santa Fe, NM Sept. I998, pp.11-I8. Engdahl ER and Bergman EA (1992): The International Seismological Observing Period in Africa. Tectonophysics, 209, pp.1-16. Engdahl ER and Gunst RH (1966): Use of a high speed computer for the preliminary determination of earthquake hypocenters. BuIl. Seism. Soc. Am. 56, pp.325-336. Engdahl ER, and Billington S (1986): Focal depth determination of central Aleutian earthquakes. Bull. Seism. Soc. Am. 76, pp.77-93. Engdahl, ER, van der HiIst R, and Buland R (1998): Global teleseismic earthquake relocation with improved travel times and procedures for depth determination. Bull. Seism. Soc. Am. 88, pp.722-743. Everitt BS (1993): Cluster analysis. Edward Arnold, London. Evernden JF (1969): Precision of epicenters obtained by a small number of world wide stations. Bull. Seism. Soc. Am. 59, pp.1365-1398. Evernden JF (1969): Precision of epicenters obtained by small numbers of worldwide stations. Bull. Seism. Soc. Am. 59, pp.1365-1398. Everndon JF (1969): Identification of earthquakes and explosions by use of teleseismic data. J. Geophys Res. 74, pp.3828-3856. Fehler MC (1989): Stress control of seismicity patterns observed during hydraulic fracturing experiments at the Fenton Hill Hot Dry Rock Geothermal Energy Site, New Mexico. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 26, pp.211-219. Fehler MC (1990): Identifying the plane of slip for a fault plane solution from clustering of locations of nearby earthquakes. Geophys. Res. Letters 17, pp.969972. Fehler MC, and Johnson P (1989): Determination of fault planes at Coalinga, California by analysis of patterns in aftershock locations. J. Geophys. Res. 94, 74967507. 35 Fehler MC, and Phillips WS (1991): Simultaneous inversion for Q and source parameters of microearthquakes accompanying hydraulic fracturing in granitic rock. Bull. Seismol. Soc. Am. 81, pp.553-575. Fehler MC, House LS, and Kaieda H (1987): Determining planes along which earthquakes occur: Method and application to earthquakes accompanying hydraulic fracturing. J. Geophys. Res. 92, pp.9407-9414. Fehler MC, House LS, Phillips WS, and Potter R (1998): A method to allow temporal variation of velocity in travel-time tomography using micro-earthquakes induced during hydraulic fracturing. Tectonophysics 289, pp.189-202. Fiske RS, and Koyanagi RY (1968): The December 1965 eruption of Kilauea Volcano, Hawaii. U.S. Geol. Surv. Prof. Pap. 607, 21p. Fletcher CR (1981): Practical Methods of Optimization, Vol.2. John WiIey & Sons, New York 224 pp. Flinn EA (1960): Local earthquake location with an electronic computer. Bull. Seism. Soc. Am. 50, pp.467-470. Flinn EA (1965): Confidence regions and error determinations for seismic event location. Rev. Geophys. 3, pp.157-185. Forsythe G, Malcom M, and Moler C (1977): Computer Methods for Mathematical Computations. Prentice-Hall. Franke PR, and Nunz GJ (1985): Recent developments in the hot dry rock geothermal energy program. Trans. Geotherm. Resour. Counc. Part 2. 9, pp.95-98. Frankel A (1982): Precursors to a magnitude 4.8 earthquake in the Virgin Islands: Spatial clustering of small earthquakes, anomalous focal mechanisms, and earthquake doublets. Bull. Seismol. Soc. Am. 72, pp.1277-1294. Fréchet J (1985): Sismogenèse et Doublets Sismiques. Thèse d'Etat, Université Scientifique et Médicale de Grenoble, 206 pp. Freedman HW (1967): A statistical discussion of P residuals from explosions, II. Bull. Seismol. Soc. Am. 57, pp.545-561. Frémont M-J and Malone SD (1987): High precision relative locations of earthquakes at Mount St. Helens, Washington. J. Geophys. Res. 92, pp.10223-10236. Frohlich C (1979): An efficient method for joint hypocenter determination for large groups of earthquakes. Computers and Geosci. 5, pp.387-389. Gao F (1999): High Resolution Upper Crustal P and S Velocity Structures in the Upper Mississippi Embayment. M.S. thesis, The University of Memphis. Gao F, Chiu JM, Schweig E, Pujol J, and Street R (1999): Three-dimensional upper crustal Vp and Vs structures in the central New Madrid seismic zone. Eastern Section Seism. Soc. Am. 75th Annual Meeting, Memphis. 36 Gaucher E, Cornet FH, and Bernard P (1998): Induced seismicity analysis for structure identification and stress field determination. Proceedings of the SPE-ISPM Rock Mechanics in Petroleum Engineering Conference 1, pp.545-554. Geiger L (1910): Herdbestimming bei erdbeben aus den Antkunftszeiten. K. Ges. Wiss. Gött. 4, pp.331-349. Geiger L (1912): Probability method for the determination of earthquake epicenters from the arrival time only. Bull. St. Louis University 8, pp.60-71. Geller RJ, and Mueller CS (1980): Four similar earthquakes in central California. Geophys. Res. Lett. 7, pp.821-824. Genter A, and Traineau H (1996): Analysis of macroscopic fractures in granite in the HDR geothermal well EPS-1, Soultz-sous-Forêts, France. J. Volc. Geotherm. Res. 72, pp.121-141. Gibowicz SJ, and Kijko A (1994): An Introduction to Mining Seismology. Academic Press, San Diego, 399p. Gillard D, Rubin AM, and Okubo P (1996): Highly concentrated seismicity caused by deformation of Kilauea's deep magma system. Nature, 384, pp.343-346. Goldstein P, Dodge D, Firpo M, and Ruppert S (1998): What's new in SAC2OOO? Enhanced processing and database access. Seis. Res. Lett. 69, pp.202-205. Gomberg JS, Shedlock KM, and Roecker SW (1990): The effect of S-wave arrival times on the accuracy of hypocenter estimation. Bull. Seismol. Soc. Am. 80, pp.1605-1628. Gower JC (1988): Classification, geometry, and data analysis. In: Classification and Related Methods of Data Analysis, HH Bock, ed., Elsevier, North-Holland. Gudmundsson A (1992): Formation and growth of normal faults at the divergent plate boundary in Iceland. Terra Nova 4, pp.464-471. Gudmundsson A, and Brynjólfsson S (1993): Overlapping rift-zone segments and the evolution of the South Iceland Seismic Zone. Geophys. Res. Lett. 20, pp.1903-1906. Haase J, Shearer P, and Aster RC (1995): Constraints on temporal variations in velocity near Anza, California, from analysis of similar event pairs. Bull. Seis. Soc. Am. 85, pp.194-206. Hackman MC, King GCP, and Bilham R (1990): The mechanics of the South Iceland Seismic Zone. J. Geophys. Res. 95, pp.17339-17351. Harris M, and Young C (1997): MatSeis: a seismic GUI and toolbox for MATLAB. Seism. Res. Lett. 68, pp.267-269. Herrin E and Taggart J (1968): Regional variations in P travel times. Bull. Seism. Soc. Am. 58, pp.1325-1337. Herrin E, Tucker W, Taggart J, Gordon DW, and Lobdell JL (1968): Estimation of surface focus P travel times. Bull. Seism. Soc. Am. 58, pp.1273-1291. 37 Herrmann R, Park S, and Wang C (1981): The Denver earthquakes of 1967-1968. Bull. Seism. Soc. Am. 71, pp.731-745. Himmelblau DM (1972): Applied Nonlinear Programming. McGraw-Hill Book Company, 498 pp. House LS (1987): Locating microearthquakes induced by hydraulic injections in crystalline rock. Geophys. Res. Lett. 14, pp.919-921. House LS, and Jensen B (1987): Focal mechanisms of microearthquakes induced by hydraulic injection in crystalline rock (abstract). EOS Trans. AGU, 68, p.1346. House LS, and Phillips WS (1991): Imaging of a fractured rock volume using the codas of seismograms from induced microearthquakes (abstract). EOS Trans. AGU, 72(44), Fall Meet. Suppl., 306. Hurukawa N, and Imoto M (1992): Subducting oceanic crusts of the Philippine Sea and Pacific Plates and weak-zone-normal compression in the Kanto district, Japan. Geophys. J. Int. 109, pp.639-652. Husebye ES and Dainty AM (1996): Monitoring a Comprehensive Test Ban Treaty. Kluwer Academic Publishers, Dordrecht, 864 pp. Ishida M, and Kanamori H (1978): The foreshock activity of the 1971 San Fernando earthquake, California. Bull. Seismol. Soc. Am. 68, pp.1265-1279. Ito A (1985): High resolution relative hypocenters of similar earthquakes by crossspectral analysis method. J. Phys. Earth 33, pp.279-294. Ito A (1990): Earthquake swarm activity revealed from high-resolution relative hypocenters-clustering of microearthquakes. Tectonophysics 175, pp.47-66. Jeffreys H (1932): An alternative to the rejection of observations. Proc. R. Soc. Lond. 137A, pp.78-87. Jeffreys H (1939): Theory of Probability. Oxford University Press, London. Jeffreys H (1973): On travel times in seismology. In: Collected Papers of Sir Harold Jeffreys on Geophysics and Other Sciences Vol.2, H. Jeffrey (Editor), Gordon and Breach Sc. PubI., London, pp.36-120. Jeffreys H, and Bullen KE (1940): Seismological Tables. British Association for the Advancement of Science, London. Jenkins GM, and Watts DG (1968): Spectral Analysis and Its Applications. HoldenDay, Oakland, Calif. Jóhannesson H, and Sæmundsson K (1989): Geological map of Iceland, 1:500000. Bedrock geology, 1st edn., Icelandic Museum of Natural History and Iceland Geodetic Survey, Reykjavik. Jóhannesson H, Jakobsson SP, and Sæmundsson K (1990): Geological map of Iceland, sheet 6, South Iceland. 3rd edn., Icelandic Museum of Natural History and Iceland Geodetic Survey, Reykjavik. 38 Johnson PA (1989): Cluster Analysis of Eastern Washington Seismicity. M.S. Thesis, Univ. of Washington, Seattle. Johnson PA, and McEvilly TV (1994): Fluids and faulting: Suggestive evidence and crucial tests at Parkfield. US. Geol. Surv. Proc. Workshop LXIII Open File Rep. pp.94-228. Johnson PA, and McEvilly TV (1995): Parkfield seismicity: Fluid driven? J. Geophys. Res. 100, pp.12937-12950. Johnson PA, and Zollweg JE (1987a): The 1986 Darrington, Washington, earthquake swarm: Evidence for tightly clustered asperities? (abstract), EOS Trans. AGU 68, 46. Johnson PA, and Zollweg JE (1987b): High-precision relative locations for a northwestern Washington earthquake swarm. Seismol. Res. Lett. 58, p.28. Johnston A, and Nava S (1990): Seismic-hazard assessment in the central United States. In: E Krinitzsky and D Slemmons (Eds.), Neotectonics in Earthquake Evaluation. Geol. Soc. Am. Rev. Eng. Geol. 8, pp. 47-58. Johnston A, and Schweig E (1996): The enigma of the New Madrid earthquakes of 1811 – 1812. Ann. Rev. Earth Planet Sci. 24, pp.339-384. Jones RH, Beauce A, Jupe A, Fabriol H, and Dyer BC (1995): Imaging induced microseismicity during the 1993 injection tests at Soultz-sous-Forets, France. Proceedings of the World Geothermal Congress, Florence, pp.2665-2669. Jordan TH, and Sverdrup KA (1981): Teleseismic location techniques and their application to earthquake clusters in the South-central Pacific. Bull. Seism. Soc. Am. 71, pp.1105-1130. Jung (1991): Hydraulic fracturing and hydraulic testing in the granitic section of Borehole GPK1, Soultz-sous-Forets, France. Geotherm. Sci. Technology, 3, pp.149198. Jupe AJ, Jones RH, Dyer B, and Wilson S (1998): Monitoring and management of fractured reservoirs using induced microearthquake activity. Soc. Petroleum. Engineers contribution SPE 47315. In: International Collaborative Project on Mapping of Engineered Geothermal Reservoirs, MTC Project Report. Jupe AJ, Jones RH, Willis-Richards J, Dyer B, Nicholls J, and Jacques P (1994): Report on HDR Phase 4 - Soultz Experimental Programme 1993/1994. CSM Associates Ltd, 1R02/12. Keck RG, and Withers RJ (1994): A field demonstration of hydraulic fracturing for solids waste injection with real-time passive seismic monitoring. Pap. SPE 28495, Soc. of Pet. Eng., Richardson, Tex. Kennett BLN (1992): Locating oceanic earthquakes - The influence of regional models and location criteria. Geophys. J. Int. 108, pp.848-854. Kennett BLN (1996): Event location and source characterisation. In: Monitoring a Comprehensive Test-Ban-Treaty. Husebye FS, and Dainty A, Eds., Kluwer, Dordrecht, pp.501-520. 39 Kennett BLN, and Engdahl ER (1991): Traveltimes for global earthquake location and phase identification. Geophys. J. Int. 105, pp.429 465. Kennett BLN, and Engdahl FR (1991): Travel Times for global earthquake location and Phase Identification. Geophys. J. Int. 105, pp.429-465. Kennett BLN, and Sambridge MS (1992): Earthquake location: Genetic algorithms for Teleseisms. Phys. Earth. Planet. Int. 75, pp.103-110. Kennett BLN, Engdahl ER, and Buland R (1995): Constraints on seismic velocities in the Earth from travel times. Geophys. J. Int. 122, pp.l08-124. Kissling E (1988): Geotomography with local earthquake data. Rev. Geophys. 26, pp.659-698. Kissling E, Ellsworth WL, Eberhart-Phillips D, and Kradolfer U (1994): Initial reference models in local earthquake tomography. J. Geophys. Res. 99, pp.1963519646. Klein FW (1978): Hypocenter Location Program HYPOlNVESE. US Geological Survey Open-File Report 78-694, 113 pp. Klein FW (1981): A linear gradient crustal model for South Hawaii. Bull. Seismol. Soc. Am. 71, pp.1503-1510. Klein FW (1989): User's Guide to HYPOINVERSE, a program for VAX computers to solve for earthquake locations and magnitude. US Geological Survey Open-File Report 89-314. Klein FW, Koyanagi RY, Nakata JS, and Tanigawa WR (1987): The seismicity of Kilauea's magma system. Volcanism in Hawaii, 2, U.S. Geol. Surv. Prof Pap. 1350, pp.1019-1186. Kovach RL, and Hill DP (1987): The 1975 Kalapana earthquake revisited. Paper presented at Hawaii Symposium on How Volcanoes Work, Hawaiian Volcano Observatory, Hilo, Hawaii. Kvaerna T, and Doornbos DJ (1986): An integrated approach to slowness analysis with arrays and three-component stations. Semiannual Technical Summary, 1 October 1985-31 March 1986, NORSAR Sci. Report No.2-85/86, KjeIIer, Norway. Lance GN and Williams WT (1967): A general theory of classificatory sorting strategies: 1. Hierarchical systems. Comp. J. 9, pp.373-380. Lawson CL, and Hanson RJ (1974): Solving Least Squares Problems. Prentice-Hall, New Jersey, 340p. Lee WHK, and Lahr JC (1975): HYPO71 (revised): A computer program for determining hypocenter, magnitude, and first motion pattern of local earthquakes. US Geological Survey Open-File Report 75-311, 114 pp. Lee WHK, and Stewart SW (1981): Principles and Application of Microearthquake Arrays. Academic Press, New York, 293 pp. 40 Lees JM (1998): Multiplet analysis at Coso geothermal. Bull. Seis. Soc. Am. 88, pp.1127-1143. Leinert BR, Berg E, and Frazer LN (1986): HYPOCENTER: an earthquake location method using centered, scaled, and adaptively damped least squares. Bull. Seism. Soc. Am. 76, 771-783. Lilwall R and Douglas A (1970): Estimation of P-wave travel times using the joint epicentre method. Geophys. J. R. Astr. Soc. 19, pp.165-181. Lipman PW, Lockwood JP, Okamura RT, Swanson DA, and Yamashita KM (1985): Ground deformation associated with the 1975 magnitude-7.2 earthquake and resulting changes in activity of Kilauea volcano, Hawaii. U.S. Geol. Surv. Prof. Pap. 1276, 45 pp. Lirer L, Luongo G, and Scandone (1987): On the volcanological evolution of Campi Flegrei. EOS, Trans. Am. Geophys. Union 68, pp.226-234. Liu Z (1997): Earthquake Modeling and Active Faulting in the New Madrid Seismic Zone. Ph.D. Dissertation, Saint Louis University. Liu Z, Hermann R, and Xie JK (1996): Microearthquake source parameters in the central New Madrid seismic zone from waveform modeling. Seism Res. Let. 67 No.2, p.45. Ludwig JA and Reynolds JF (1985): Statistical Ecology. J. Wiley and Sons, New York. Ludwig JA and Reynolds JF (1988): Statistical Ecology: A Primer on Methods and Computing. J. Wiley & Sons, New York. Mc Laren J, and Frohlich C (1985): Model calculations of regional network locations for earthquakes in subduction zones. Bull. Seis. Soc. Am. 75, pp.397-413. Mendecki AJ, and Sciocatti M (1997): Location of seismic events. In: Mendecki AJ (ed.) Seismic Monitoring in Mines, Chapman and Hall, London, pp.87-106. Menke W (1984): Geophysical Data Analysis: Discrete Inverse Theory. Academic Press, Orlando, Fl. Menke W, Lerner-Lam A, and Mithal R (1991): Spatial coherence of 5-25 Hz seismic wavefields at a hard rock site. Structural Safety 10, pp.163-179. Menke W, Lerner-Lam A, Dubendorff B, and Pacheco J (1990): Polarization and coherence of 5-30 Hz seismic wave fields at a hard rock site and their relevance to velocity heterogeneities on the crust. Bull. Seism. Soc. Am. 80, pp.430-449. Michael AJ (1988): Effects of three-dimensional velocity structure on the seismicity of the 1984 Morgan Hill, California, aftershock sequence. Bull. Seism. Soc. Am. 78, pp.1199-1221. Michelini A, and Bolt B (1986): Application of the principal parameters method to the 1983 Coalinga, California aftershock sequence. Bull. Seismol. Soc. Am. 76, pp.409420. 41 Minster J-BH, Williams NP, Masters TC, Gilbert JF, and Haase JS (1995): Application of evolutionary programming to earthquake hypocenter determination. In: Proc. 4th Annual. Conf. of Evol. Prog. eds. McDonnell, JR, Reynolds, RC, and Fogel, DB. MIT Press, Cambridge 1995, pp.3-17. Molnar P (1979): Earthquake recurrence intervals and plate tectonics. Bull. Seismol. Soc. Am. 69, pp.115-133. Morelli A, and Dziewonski AM (1993): Body wave traveltimes and a spherically symmetric P- and S-wave velocity model. Geophys. J. Int. 112, pp.l78-194. Moriya H, Nagano K, and Niitsuma H (1994): Precise source location of AE doublets by spectral matrix analysis of triaxial hodogram. Geophysics, 59, pp.36-45. Moriya H, Nakazato K, and Niitsuma H (2001): Detailed fracture system of Soultzsous-Forets HDR field evaluated by using microseismic doublet/multipIet. Pure Appl. Geophys., submitted. Mosteller F and Tukey JW (1979): Data Analysis and Regression. Addison-Wesley, Reading, Mass. Motoya Y, and Abe K (1985): Waveform similarity among foreshocks and aftershocks of the October 18, 1981, Eniwa, Hokkaido, earthquake. Earthquake Predict. Res. 3, pp.627-636. Mueller K, Champion J, Guccione M, and Kelson K (1999): Fault slip rates in the modern New Madrid seismic zone. Science 286, pp.1135-1138. Myers SC, and Schultz CA (2000): Improving sparse network seismic locations with Bayesian Kriging and teleseismically constrained calibration events. Bull. Seism. Soc. Am. 90, pp.199-211. Nadeau RM, Antolik M, Johnson PA, Foxall W, and McEvilly TV (1994): Seismological studies at Parkfield III: Microearthquake clusters in the study of faultzone dynamics. Bull. Seismol. Soc. Am. 84, pp.247-263. Nanjo K, Hagahama H, and Satomura M (1998): Rates of aftershock decay and the fractal structure of active fault systems. Tectonophysics 287, pp.173-186. Nelder JA, and Mead R (1965): A simplex method for function minimization. Computer J. 7, pp.308-313. Niitsuma et al. (1999): Current status of seismic and borehole measurements for HDR/HWR development. Geothermics 128, pp.475-490. Noble B, and Daniel J (1977): Applied Linear Algebra. Prentice-Hall. Nuttli O (1973): The Mississippi Valley Earthquakes of 1811 and 1812: Intensities, ground motion, and magnitude. Bull. Seis. Soc. Am. 63, pp.227-248. Nuttli O, Stauder W, and Kisslinger C (1969): Travel time tables for earthquakes in the Central United States. Earthquake Notes 40 No.4, pp.19-28. Okabe A, Boots B, and Sugihara K (1992): Spatial Tessellations: Concepts and Applications of Voronoi Diagrams. John Wiley & Sons, New York. 42 Olsson DM, and Nelson LS (1975): The Nelder-Mead Simplex procedure for function minimization. Techtonometrics 17, pp.45-51. Paige CC and Saunders MA (1982): LSQR: Sparse linear equations and least squares problems, ACM Transactions on Mathematical Software 8/2, pp.195-209. Park J, Lindberg CR, and Vernon III FL (1987): Multitaper spectral analysis of highfrequency seismograms. J. Geophys. Res. 92, pp.12675-12684. Parker RL, and McNutt MK (1980): Statistics for the one-norm misfit measure. J. Geophys. Res. 85, pp.4429-4430. Pavlis G, and Hokanson N (1985): Separated earthquake location. J. Geophys. Res. 90, pp.12777-12789. Pavlis GL (1986): Appraising earthquake hypocenter location errors: a complete, practical approach for single-event locations. Bull Seism. Soc. Am. 76, pp.16991717. Pavlis GL (1992): Appraising relative earthquake location errors. Bull. Seimol. Soc. Am. 82, pp.836-859. Pavlis GL, and Booker JR (1983): Progressive Multiple Event Location (PMEL). Bull. Seism. Soc. Am. 73, pp.1753-1777. Pearson C (1981): The relationship between microseismicity and high pore pressures during hydraulic stimulation experiments in low permeability granitic rocks. J. Geophys. Res. 86, pp.7855-7864. Pechmann JC and Thorbjarnardóttir BS (1990): Waveform analysis of two preshockmainshock-aftershock sequences in Utah. Bull. Seism. Soc. Am. 80, pp.519-550. Pechmann JC, and Kanamori H (1982): Waveforms and spectra of preshocks and aftershocks of the 1979 Imperial Valley earthquake. Bull. Seismol. Soc. Am. 72, pp.237-258. Phillips WS, House LS, and Fehler MC (1996): Vp/Vs and the structure of microearthquake clusters. Seismol Res. Lett. 63, p.56. Phillips WS, Rutledge JT, Fairbanks TD, Gardner TU, Miller ME, and Schuessler BK (1996): Reservoir fracture mapping using microearthquakes: Austin chalk, Giddings field, TX, and 76 Field, Clinton Co., KY, Pap. SPE 36651, Soc. of Pet. Eng., Richardson, Tex. Phillips WS, Rutledge JT, House LS, and Fehler MC (2001): Induced microearthquake patterns in hydrocarbon and geothermal reservoirs: A review. Pure Appl. Geophys. 159, pp.345-369. Polak E, (1971): Computational Methods in Optimization. Academic Press, New York. Poupinet G, Ellsworth WL, and Fréchet J (1984): Monitoring velocity varations in the crust using earthquake doublets: An application to the Calaveras Fault, California. J. Geophys. Res. 89, pp.5719-5731. 43 Press WH, Flannery BP, Teukolsky SA, and Vetterling WT (1989): Numerical Recipes in C. Cambridge University Press. Prugger AF, and Gendzwill DJ (1988): Microearthquake location: a nonlinear approach that makes use of simplex stepping procedures. Bull. Seism. Soc. Am. 78, pp.799-815. Pujol J (1992): Joint hypocentral location in media with lateral velocity variations and interpretation of the station corrections. Phys. Earth Planet. Interiors 75, pp.7-24. Pujol J (1995): Application of the JHD technique to the Loma Prieta, California, mainshock-aftershock sequence and implications for earthquake location. BulI. Seism. Soc. Am. 85, pp.129-150. Pujol J (1996a): Comment on: “The 1989 Loma Prieta earthquake imaged from inversion of geodetic data" by Àrnadòttir A and Segall P. J. Geophys. Res. 101, pp.20133-20136. Pujol J (I996b): An integrated 3D velocity inversion-joint hypocentral determination relocation analysis of events in the Northridge area. Bull. Seism. Soc. Am. 86, S138S155. Pujol J (2000): Comment on “Laboratory measurements of ultrasonic wave velocities in rocks from the Campi Flegrei volcanic system and their relation to other field data” by Zamora M, Sartoris G, and Chelini W. J. Geophys. Res. 105, pp.21537-21542. Pujol J, and Aster RC (1990): Joint hypocentral determination and the detection of low-velocity anomalies. An example from the Phlegraean Fields earthquakes. Bull. Seism. Soc. Am. 80, pp.129-139. Pujol J, Herrmann R, Chiu SC, and Chiu JM (1998): Constrained relocation of New Madrid seismic zone earthquakes. Seism. Res. Lett. 69, pp.56-68. Pujol J, Johnston A, Chiu JM, and Yang YT (1997): Refinement of thrust faulting models for the central New Madrid seismic zone. Engineering Geology 46, pp.281298. Pujol, J. (1988): Comments on the joint determination of hypocenters and station corrections. Bull. Seism. Soc. Am. 78, pp.1179-1189. Rabinowitz N (1986): On identifying influential arrivals by means of density information matrix. Bull. Seism. Soc. Am. 76, pp.1817-I821. Rabinowitz N (1988): Microearthquake location by means of nonlinear simplex procedure. Bull. Seism. Soc. Am. 78, pp.380-384. Rabinowitz N, and Kulhanek O (1988): Application of a nonlinear algorithm to teleseismic locations using P-wave readings from the Swedish seismographic network. Phys. Earth. Plan. Int. 50, pp.111-115. Rabinowitz N, and Steinberg DM (1988): Effect of single arrival times on the hypocentral local depth determination. Geophys. J. 94, pp.567-570. 44 Ratchkovsky N, Pujol J, and Biswas N (1997): Relocation of earthquakes in tile Cook Inlet area, south-central Alaska, using the joint hypocenter determination method. BuIl. Seism. Soc. Am. 87, pp.620-636. Richards-Dinger KB, and Shearer PM (2000): Earthquake locations in southern California obtained using source-specific station terms. J. Geophys. Res. 105, pp.10939-10960. Ringdal F, and Kvaerna T (1989): A multi-channel processing approach to real-time network detection, phase association, and threshold monitoring. Bull. Seismol. Soc. Am. 79, pp.1927-1940. Roberts RG, Christoffersson A, and Cassidi F (I989): Real time events detection, phase identification, and source location estimation using single station component seismic data and a small PC. Geophys. J. Int. 97, pp.471-480. Roecker SW (1981): Seismicity and Tectonics of the Pamir-Hindu Kush Region of Central Asia. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, 294 pp. Roff A, Phillips WS, and Brown DW (1996): Joint structures determined by clustering of microearthquakes using waveform amplitude ratios. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 33, pp.627-639. Rögnvaldsson ST, and Slunga R (1993): Routine fault plane solutions for local networks: A test with synthetic data. Bull. Seism. Soc. Am. 83, pp.1232-1247. Rögnvaldsson ST, and Slunga R (1994): Single and joint fault plane solutions for microearthquakes in South Iceland. Tectonophysics, 237 pp.73-86. Röhm AHE, Trampert J, Paulssen H, and Snieder RK (1999): Bias in reported seismic arrival times deduced from the lSC Bulletin. Geophys. J. Int. 137, pp.163174. Rowe CA (2000): Correlation-Based Phase Pick Correction and Similar Earthquake Family Identification in Large Seismic Waveform Catalogues. Ph.D. Dissertation, New Mexico Tech. Rowe CA, and Aster RC (1999): Application of automatic, adaptive filtering and eigenspectral techniques to large digital waveform catalogs for improved phase pick consistency and uncertainty estimates. EOS 80, F660. Rowe CA, Aster RC, Fehler MC, Phillips WS, and Kyle RR (1998): Hypocenter relocation in tectonic and volcanic seismic data sets using automatic phase correlation: Application of cross-correlation, signal envelope, and principal components methods to improve pick consistency. EOS 79, F588. Rubin AM, Gillard D, and Got J-L (1998): A reinterpretation of seismicity associated with the January 1983 dike intrusion at Kilauea Volcano, Hawaii. J. Geophys. Res. 103, pp.10003-10015. Russ D (1979): Late Holocene faulting and earthquake recurrence in the Reelfoot Lake area, northwestern Tennessee. Geol. Soc. Am. Bull. Part 1 90, pp.1013-1018. 45 Russ D (1982) Style and significance of surface deformation in the vicinity of New Madrid, Missouri. In: F McKeown, and L Pakiser (Eds.), Investigations of the New Madrid, Missouri, Earthquake Region. US Geol. Survey Prof. Paper 1236, pp.95114. Ruud BO, Huesebye ES, Ingate SF, and Christoffersson A (1988): Event location at any distance using seismic data from a single, three-component station. Bull. Seism. Soc. Am. 78, pp.308-325. Sambridge MS (1998): Exploring multidimensional landscapes without a map. Inverse Problems 14, pp.427-440. Sambridge MS (1999a): Geophysical inversion with a neighbourhood algorithm - I. Searching a parameter space. Geophys. J. Int. 138, pp.479-494. Sambridge MS (1999b): Geophysical inversion with a neighbourhood algorithm - II. Appraising the ensemble. Geophys. J. Int. 138, pp.727-746. Sambridge MS, and Gallagher K (1993): Earthquake hypocenter location using genetic algorithms. Bull. Seismol. Soc. Am. 83, pp.1467-1491. Sambridge MS, and Kennett BLN (1986): A novel method of hypocentre location. Geophys. J. R. Astr. Soc. 87, pp.313-331. Schaff DP, Bokelmann GHR, Beroza GC, Waldhauser F, and Ellsworth WL (2001): High resolution image of Calaveras Fault seismicity. J. Geophys. Res. 107, pp.663668. Scherbaum F and Wendler J (1986): Cross spectral analysis of Swabian Jura (SW Germany) three-component microearthquake recordings. J. Geophys. 60, pp.157166. Scherbaum F, and Wendler J (1986): Cross spectral analysis of Swabian Jura (SW Germany) three-component microearthquake recordings. J. Geophys. 60, pp.157166. Schultz CA, Myers SC, Hipp J, and Young C (1998): Nonstationary Bayesian Kriging: A predictive technique to generate corrections for detection, location, and discrimination. Bull. Seismol. Soc. Am. 83, pp.1275-1288. Schweig E, and Ellis M (1994): Reconciling short recurrence intervals with minor deformation in the New Madrid seismic zone. Science 264, pp.1308-1311. Scotti O, and Cornet FH (1994): In-situ evidence for fluid-induced aseismic slip events along fault zones. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 31, pp.347-358. Shearer PM (1997): Improving local earthquake locations using the L1 norm and waveform cross correlation: Application to the Whittier Narrows, California, aftershock sequence. J. Geophys. Res. 102, pp.8269-8283. Shearer PM (1998): Evidence from a cluster of small earthquakes for a fault at 18 km depth beneath Oak Ridge, Southern California. Bull. Seism. Soc. Am. 88, pp.1327l336. 46 Sherilt R, and Geldart L (1995): Exploration Seismology. Second Edition, Cambridge University Press, 575 pp. Slunga R (1981): Earthquake source mechanism determination by use of body-wave amplitudes - an application to Swedish earthquakes. Bull. Seism. Soc. Am. 71, pp.25-35. Slunga R, Norrmann P, and Glans A-C (1984): Baltic shield seismicity, the results of a regional network. Geophys. Res. Lett. 11, pp.1247-1250. Slunga R, Rognvaldsson S, and Bodvarsson R (1995): Absolute and relative locations of similar events with application to microearthquakes in southern Iceland. Geophys. J. Int. 123, pp.409-419. Smith E (I982): An efficient algorithm for routine joint hypocentre determination. Phys. Earth Planet Int. 30, pp.135-144. Smith GP, and Ekstrom G (I996): Improving teleseismic earthquake locations using a three-dimensional Earth model. Bull. Seismol. Soc. Am. 86, pp.l23-132. Sneath PHA and Sokal RR (1973): Numerical taxonomy. Freeman, San Francisco. Snoke JA (1987): A stable determination of (Brune) stress drops. Bull. Seism. Soc. Am. 74, pp.530-538. Spudich P, and Bostwick T (1987): Studies of the seismic coda using an earthquake cluster as a deeply buried seismograph array. J. Geophys. Res. 92, pp.l0526-I0546. Starzec P, Fehler MC, Baria R, and Niitsuma H (2000): Spatial correlation of seismic slip at the HDR-Soultz-geothermal site: qualitative approach. Bull. Seism. Soc. Am. 90, pp.1528-1534. Stauder W, and Ryall A (1967): Spatial distribution and source mechanism of microearthquakes in Central Nevada. Bull. Seismol. Soc. Am. 57, pp.1317-1345. Stauder W, Kramer M, Fischer G, Schaefer S, and Morrissey S (1976): Seismic characteristics of southeast Missouri as indicated by a regional telemetered microearthquake array. Bull. Seism. Soc. Am. 66, pp.1953-1964. Stefánsson R and Halldórsson P (1988): Strain release and strain build-up in the South Iceland Seismic Zone. Tectonophysics 155, pp.267-276. Stefánsson R, Bödvarsson R, Slunga R, Einarsson, P, Jakobsdóttir J, Bungum H, Gregersen S, Havskov J, Hjelme J, and Korhonen H (1993): The SIL project, background and perspectives for earthquake prediction in the South Iceland Seismic Zone. Bull. Seism. Soc. Am. 83, pp.696-716. Stefánsson R, Bungum H, Bödvarsson R, Hjelme J, Husebye E, Johansen H, Korhonen H, and Slunga R (1986): Seismiskt datasamlingssysrem för Sodra Islands Iågland (A seismic data acquisition system for South Iceland Lowland; in English with Icelandic and Swedish summary). Icelandic Meteorological Office report, Reykjavik. Swanson DA, Duffleld WA, and Fiske RS (1976): Displacement of the south flank of Kilauea volcano: The result of forceful intrusion of magma into the rift zones. U.S. Geol. Surv. Prof Pap. 963, 1-39. 47 Sweeney J, (1998): Criteria for Selecting Accurate Event Locations from NEIC and ISC Bulletins. UCPL-JC-130655. Sykes LR (1970): Earthquake swarms and sea-floor spreading. J. Geophys. Res. 75, pp.6598-6611. Thomson JD (1982): Spectrum estimation and harmonic analysis. Proceedings of the IEEE 70, pp.1055-1096. Thurber C, Haslinger F, and Trabant C (1999): Testing event location capability with ground truth events in Kazakhstan. Proc. 21st Seismic Research Symposium, 283293. Thurber C, Roecker S, McEvilly T, Haslinger F, Steinberg D, and Ellsworth W (2000): Earth-scope at Parkfield, California: setting the stage for drilling. EOS 81, S318. Thurber CH (1983): Earthquake locations and three-dimensional crustal structure in the Coyote Lake area, central California. J. Geophys. Res. 88, pp.8226-8236. Thurber CH, and Gripp AE (1988): Flexure and seismicity beneath the south flank of Kilauea volcano and tectonic implications. J. Geophys. Res. 93, pp.4213-4248. Tong C and Kennett BLN (1995): Towards the identification of later seismic phases. Geophys. J. Int. 123, pp.948-958. Tong C and Kennett BLN (1996): Automatic seismic event recognition and later phase identification for broad-band seismograms. Bull. Seismol. Soc. Am. 86, pp.1896-1909. US Geological Survey Staff (1990): The Loma Prieta, California, earthquake: an anticipated event. Science 247, pp.286-293. Van Arsdale R, Kelson R, and Lumsden C (1995): Northern extension of the Tennessee Reelfoot scarp into Kentucky and Missouri. Seism. Res. Lett. 66, pp.5762. Van Decar JC, and Crosson (1990): Determination of teleseismic relative phase arrival times using multi-channel cross-correlation and least-squares. Bull. Seism. Soc. Am. 80, pp.150-159. van der HiIst RD, and Engdahl ER (1992): Step-wise relocation of ISC earthquake hypocenters for linearized tomographic imaging of slab structure. Phys. Earth Planet. Inter. 75, pp.39-54. van der Hilst RD, Widiyantoro S, and Engdahl ER (I997): Evidence for deep mantle circulation from global tomography. Nature 386, pp.578-584. Vernon FL, Pavlis GL, Owens TJ, McNamara DE, and Anderson PN (1998): Nearsurface scattering effects observed with a high-frequency phased array at Pinyon Flats, California. Bull. Seism. Soc. Am. 88, pp.1548-1560. Vidale JE, Ellsworth WL, Cole A, and Marone C (1994): Variations in rupture process with recurrence interval in a repeated small earthquake. Nature 368, pp.624-626. 48 Viret M, Bollinger G, Snoke J, and Dewey J (1984): Joint hypocenter relocation studies with sparse data sets - A case history: Virginia earthquakes. Bull. Seism. Soc. Am. 74, pp.2297-2311. Voronoi MG (1908): Novelles applications des paramètres continus à la théorie des formes quadratiques. J. reine angew. Math. 134, pp.198-287. Warpinski NR, Wright TB, Uhl JE, Engler BP, Drozda PM, Peterson RE, and Branagan PT (1996): Microseismic monitoring of the B-sand hydraulic fracture experiment at the DOE/GRI Multi-Site Project. Pap. SPE 36450, Soc. of Pet. Eng., Richardson, Tex. Wiemer S, and Wyss M (1997): Mapping the frequency-magnitude distribution in asperities: An improved technique to calculate recurrence times? J. Geophys. Res. 102, pp.15115-15128. Wigner E (1960): The unreasonable effectiveness of mathematics in the natural sciences. Comm. Pure Appl. Math. 13, pp.1-14. Williams WT, Lance GN, Dale MB, and Clifford HT (1971): Controversy concerning the criteria for taxomonometric strategies. J. Comp. 14 pp.162-165. Withers M, Aster R, and Young C (1999): An automated local and regional seismic event detection and location system using waveform correlation. Bull. Seismol. Soc. Am. 89, pp.657-669. Withers M, Aster R, Young C, and Chael F (1996): High-frequency analysis of seismic background noise and signal-to-noise ratio near Datil, New Mexico. Bull. Seism. Soc. Am. 86, pp.1507-1515. Wittle P (1971): Optimization under Constraints. Wiley, London. Wylie BN, Boyack KW, Davidson GS, and Johnson DK (2000): Visualization of information spaces with Vxlnsight. Sandia Labs Report #SAND2000-3100. Wyss M, and Kovach RL (1988): Comment on "A single-force model for the 1975 Kalapana, Hawaii, earthquake" by Holly K Eissler and Hiroo Kanamori, J. Geophys. Res. 93, pp.8078-8082. Wyss M, Klein FW, and Johnston AC (1981): Precursors to the Kalapana M=7.2 earthquake. J. Geophys. Res. 86, 3881-3900. Zamora M, Sartoris G, and Chelini W (1994): Laboratory measurements of ultrasonic wave velocities in rocks from the Camp Flegrei volcanic system and their relation to other field data. J. Geophys. Res. 99, pp.13553-13561. Zoback, ML, and Zoback MD (1980): State of stress in the conterminous United States. J. Geophys. Res. 85, 6113-6156. Zucca JJ, and Hill DP (1980): Crustal structure of the southeast flank of Kilauea volcano, Hawaii, from seismic refraction measurements. Bull. Seismol. Soc. Am., 70, 1149-1159. 49 Zucca JJ, Hill DP, and Kovach RL (1982): Crustal structure of Mauna Loa volcano, Hawaii, from seismic refraction and gravity data. Bull. Seismol. Soc. Am., 72, pp.1535-1550. 50
© Copyright 2024 Paperzz