Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A teleseismic study of the Northern Cordilleran upper mantle beneath the SNORCLE transect Frederiksen, Andrew William 1996

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1997-0062_1.pdf [ 11.82MB ]
JSON: 831-1.0052525.json
JSON-LD: 831-1.0052525-ld.json
RDF/XML (Pretty): 831-1.0052525-rdf.xml
RDF/JSON: 831-1.0052525-rdf.json
Turtle: 831-1.0052525-turtle.txt
N-Triples: 831-1.0052525-rdf-ntriples.txt
Original Record: 831-1.0052525-source.json
Full Text

Full Text

A Teleseismic Study of the Northern Cordilleran Upper Mantle Beneath the SNORCLE Transect by Andrew William Frederiksen B . S c , McGi l l University, 1994 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF T H E REQUIREMENTS FOR T H E DEGREE OF M A S T E R OF SCIENCE in T H E FACULTY OF G R A D U A T E STUDIES D E P A R T M E N T OF E A R T H AND O C E A N SCIENCES We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH COLUMBIA December 1996 ©Andrew William Frederiksen, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract The study area of the SNORCLE Lithoprobe transect comprises the northern Canadian Cordillera and the northwestern Canadian shield. An array of five portable broadband seis-mographs has been deployed along the trend of the transect, to complement five permanent stations of the Canadian National Seismic Network and seven Alaskan short-period instruments. The objective of the experiment is to examine the physical state of the upper mantle along the transect. P-wave travel-time residuals up to 2 seconds have been measured, and analyzed using a non-linear tomographic technique, thereby recovering velocity structure between 100 and 600 km depth for the western portion of the transect. Two significant P-wave mantle velocity anomalies have been located. The first, a relatively shallow high-velocity feature located at the western edge of the model, has been interpreted as being the edge of the Pacific slab from the southern Alaska subduction zone. The second is a large, tabular low-velocity anomaly centered at 60°N by 136°W, elongate northwest-southwest, dipping southeast, and reaching a depth of 450-500 km. This low-velocity anomaly is judged to reflect a thermal anomaly of the order of 100°C, with a possible compositional component. Multiple interpretations of the low-velocity feature are considered, the two main hypotheses being a mantle plume or a flow feature related to the proximity of the subducting slab and the opening of the northern Cordilleran slab window. The latter hypothesis is favored, due to the absence of other evidence for a plume in this region. In addition, the upper part of the low-velocity anomaly may reflect the influence of strain heating at lithospheric levels, related to the convergence of the Pacific and North American plates and the uplift of the St. Elias Mountains. ii C o n t e n t s Abstract ii Table of Contents iii List of Tables v List of Figures vii Acknowledgements viii 1 Introduction 1 1.1 Area of interest 1 1.2 History of plate motions 2 1.3 Geophysical data 7 1.4 Post-accretionary history of magmatism in the northern Cordillera 10 1.5 Terrane transport: controversies 11 1.6 Summary and motivation 13 2 The experiment 14 2.1 Rationale 14 2.2 Instrumentation 15 iii CONTENTS ' iv 2.3 Deployment and events 16 2.4 Data quality and processing 19 3 Travel-time analysis 24 3.1 Rationale 24 3.2 Determination of travel times 26 3.3 The inverse problem 29 3.4 Linear inversion technique 32 4 Results 36 4.1 Residuals 36 4.2 Choice of inversion parameters 39 4.3 Resolution 44 4.4 Observed features 52 5 Discussion of results 59 5.1 The nature of the observed anomalies 59 5.2 Plume models for the observed low-velocity feature 63 5.3 Plate-related models for the observed low-velocity feature 65 5.4 Summary 70 Bibliography 73 A List of earthquakes used in this experiment 78 List o f Tables 2.1 Stations used in this experiment List of Figures 1.1 Terranes of the Canadian cordillera 3 1.2 Detail: terranes and faults of the southern Yukon 4 1.3 Reconstruction of Pacific plate motions 5 1.4 Results of western-hemisphere S-wave tomography from Grand (1994) 8 1.5 Potential-field data for the northern Cordillera 9 2.1 Map of seismometer array 18 2.2 Event locations for travel-time inversion 19 2.3 Data quality for two events 21 2.4 Sample event section 23 3.1 Rationale for travel-time analysis . 25 3.2 Cross-correlation and arrival-time picking 28 3.3 Grid knots used for travel-time inversion 31 4.1 Some contour maps of travel-time perturbations 37 4.2 Contour maps of perturbations after removal of station averages 38 4.3 Sample polar plots of travel-time perturbations gathered by station 40 4.4 The effect of varying the degree of regularization for a linear inversion 42 4.5 Tradeoff curves for two linear iterations 43 vi LIST OF FIGURES vii 4.6 Effect of non-linearity 44 4.7 Synthetic model from large-anomaly resolution test 46 4.8 Recovered model from large-anomaly resolution test 47 4.9 Depth sections from the checkerboard resolution test 48 4.10 Depth sections from the checkerboard resolution test (cont.) 49 4.11 E-W vertical sections from the checkerboard resolution test 50 4.12 N-S vertical sections from the checkerboard resolution test 51 4.13 Horizontal cross-sections through final model 53 4.14 Horizontal cross-sections through final model (cont.) 54 4.15 SW-NE cross-sections through final model 55 4.16 NW-SE cross-sections through final model 56 4.17 Event statics for final model 56 5.1 Position of the Alaskan slab 60 5.2 Geometry of a slab window 67 5.3 Cordilleran slab windows 68 5.4 Conceptual model of flow due to the opening of a slab window 70 Acknowledgements Many people participated in the project this thesis describes. In particular, I would like to acknowledge the participation of Michael Bostock, who designed and supervised this project; Andy Langlois, Scott Dodd, and the operators of the portable stations; John Cassidy and Roger Hansen, for providing permanent-station data from both sides of the border; John VanDecar, for his excellent software and advice; Carl-Georg Bank for advice, moral support, and Fortran-to-English translation; the members of the TJBC seismology group, John Amor, and Gerry Grieve, for advice and technical support; and the members of my thesis committee, Bruce Buffett, Derek Thorkelson, and Bob Ellis. Arthur Calderwood, Ben Edwards, Don Francis, and Shi Lang provided helpful geological insights. The use of the first person plural throughout the thesis is intended to reflect the collaborative nature of this project. viii Chapter 1 Introduction 1.1 Area of interest The Yukon Territory lies at the northern end of the Cordilleran fold and thrust belt, where the main belts of the Cordillera converge and turn a corner into Alaska. It consists of a collage of terranes of Mesozoic to Paleozoic (and possibly Precambrian) ages, divided by faults, and intruded by plutons, and locally overlain by Cenozoic volcanic and sedimentary strata. To the east of the terrane assemblage are the folded and thrusted sedimentary rocks of the ancient North American margin, which lie along the North American craton. This area lies adjacent to the Pacific active margin of the North American Plate, opposite a corner in the Pacific Plate where strike-slip motion along the Queen Charlotte-Fairweather transform fault becomes subduction along the Aleutian Trench. The region at the corner is tectonically complicated; at its apex lies the Yakutat Terrane, a block which was carried north along the Queen Charlotte fault and is colliding with the North American plate at the trench. A strike-slip transition fault between this terrane and the Pacific Plate may be taking over, as the strike-slip motion vaults westward to avoid the stopped terrane (von Huene, 1989). A map of the principal terranes inland is shown in figure 1.1. They consist of a mix of per-1 CHAPTER 1. INTRODUCTION 2 icratonic terranes (containing sedimentary material related to the North American craton) and accreted terranes (brought in from elsewhere, such as from island arcs). The entire Cordillera has been overprinted by arc magmatism from mid-Cretaceous to early Tertiary time (Monger and Nokleberg, 1996). The most interesting recent structural features of this area are the Tintina and Denali fault zones (figure 1.2). Both are major dextral strike-slip fault zones striking northwest-southeast. The Denali Fault follows the boundary between the Coast Belt and the Insular Belt; the Tintina Fault, which lies about 300 km east of the Denali, lines up with the Rocky Mountain Trench farther south. The estimated displacements for these faults are approximately 400 km for the Denali, and 450 km or more for the Tintina (Lowe et al., 1994). The Tintina Fault displays little seismicity, and is not thought to be currently active. The Denali Fault Zone displays considerable seismic activity, and is currently slipping. However, it is the Tintina Fault that marks a more major divide in the geophysical data (see section 1.3). The Teslin Fault is a poorly understood fault between the Denali and Tintina faults (Lowe et al., 1994). 1.2 History of plate motions Several very complete reconstructions of the motions of oceanic plates in the northeastern Pacific have been performed, retracing their motions back to the mid-Cretaceous (Atwater, 1989, Engebretson et al., 1985, Stock and Molnar, 1988). Although there is appreciable variation in the timing and plate velocities determined by different methods, the overall history of Pacific plate motions off the coasts of what are now Alaska and northern British Columbia is well understood as far back as 120 Ma. Figure 1.3 summarizes one of these reconstructions. In the mid-Cretaceous, the Farallon Plate dominated the eastern Pacific. Its motion was obliquely convergent with the North American Plate, but the relative velocity between the two CHAPTER 1. INTRODUCTION 3 . NAC North American craton NAM North American margin craton fragments in collage ja KilbucMdono; MOMonashee Pericratonic terranes . proximal terranes AA Arctic Alaska; CACass ian D L DiSnger; MY Mystic; NX Nixon Fork; P C Porcupine: • WS Wickcrsham; YO York plstal terranes C O Cofctfoot K O Kootenay. RB Ruby; SD Seward; YT Yukon-Tanana Accreted terranes A A A ^ SS / v /vv-j A A A A ^ ^. / \ /N. H *<v<v <<<<< V V \S V . v v v NJ ^ V V V I ' v v v v l I I II I I I I I M I I I I t I I I I I I I I I f I I I I I I I I arc terranes inner O F Olds Ferry; O N Quesnellia; ST Stikinia Wermediafe.-' CO CadwaOader; C K ChiUiwack River; XL tzee; MT Methow; WA Wallowa outer: A X Alexander; KY KoyukulqNY. Nyac; P E Peninsular; TG Togiak; WR Wrangellia accretJonarv complex terranes (typically chert-rich) A G Angayucham; BA Baker; B R Bridge River, C C Cache Creek; CQ Chugach; (part); GD Goodnews; SM Slide Mountain ; (typically clastic-rich) CGChugach;(part);GSGrindstone; HO Hoh; O C Olympic Core ; P R Paciric R i m ; PW Prince William; SZ Siletzia; YA Yakutat; oceanic plates: JFJuan de Fuca; PAC Pacific Figure 1.1: Map of the terranes of the Canadian cordillera, from Monger and Nokleberg (1996). CHAPTER 1. INTRODUCTION 4 HI Y a k u t a t t=< C h u g a c h W?i A l e x a n d e r (H W r a n g e l l i a B2J K l u a n e S c h i s t a r i d I n t r u s i v e R o c k s o f t h e C o a s t P l u t o n i c C o m p l e x E53 C a c h e - C r e e k Y u k o n — T a n a n a lag S l i d e M o u n t a i n 123 C a s s i a r I I M i o g e o c l i n e W W i n d y - M c K i n l e y , N i s l i n g I-. A I S t i k i n a Figure 1.2: Detail of the terranes and major faults of the southern Yukon. From Lowe et al. (1994). was quite small; the strike-slip component may have changed from left-lateral to right-lateral between 140 Ma and 110 Ma (Engebretson et al., 1985). At about 83 Ma, a spreading center formed in the Farallon Plate, dividing it into southern and northern pieces. The northern portion is known as the Kula Plate; the southern portion retained the Farallon name. There is some controversy over the location of the divide between the two plates between California and Mexico, but by all accounts the Kula plate was the one present along the coast adjacent to our area of interest (Atwater, 1989). The Kula plate was quite remarkable for its velocity; in the hot-spot reference frame, it was moving northward at approximately 120 mm/yr (Atwater, 1989). Its motion relative to the North American plate was obliquely convergent, but dominated by the right-lateral strike-slip component. This relative motion became increasingly convergent into the Paleocene. Due to its high velocity, the Kula Plate was quite short-lived (approximately 83 Ma to 43 CHAPTER!. INTRODUCTION 5 PRESENT (0-5) 65 Ma (65-71) 37Ma (37-43) i . • 80Ma(74-85) 56Ma(56-6I) 110 Ma (100-115) Figure 1.3: Paleogeographic map of the plates of the western Pacific, Cretaceous to present. The arrows indicate plate velocities relative to the hot-spot reference frame, their lengths cor-responding to 10 Ma of motion. From Atwater (1989). CHAPTER 1. INTRODUCTION 6 Ma); as it subducted beneath North America and eastern Siberia, the Pacific Plate gradually moved eastward. The Kula Plate was gone by the mid-Eocene, most of it having subducted and the remainder having merged with the Pacific plate when the Pacific-Kula ridge shut down. A section of the Farallon remained west of Mexico and South America, eventually becoming the present-day Cocos and Nazca plates, while some small sections remained to the north, including the present-day Juan de Fuca and Explorer plates (Atwater, 1989; Engebretson et al., 1985). The issue of how these different plates interacted with North America is far from resolved. The uncertainties stem from the fact that the relative motions between North America and the various oceanic plates that interacted with it were generally highly oblique; therefore, there could have been a mix of subduction and strike-slip faulting at different points along the boundary, much as there is today. This leads in to a debate about the origin and transport of the accreted terranes that make up western North America - to what extent are they volcanic arcs rafted in by subduction, as opposed to blocks carried up the coast by strike-slip faulting (Irving et al., 1995)? As for the Yukon - how much strike-slip motion has there been, and can it be reconciled with motion farther south? This history of plate motions also has important direct implications for the evolution of the upper mantle beneath the northern Cordillera. The subduction of the Kula Plate, in particular, is likely to have exerted considerable influence on the upper mantle, by both disrupting any existing structures and driving mantle circulation. The strike-slip motion between the Pacific and North American plates may also have driven shearing flow in the upper mantle. Finally, the ongoing subduction of the East Pacific Rise and separation of the Explorer and Pacific slabs may have complicated flow patterns beneath the northern Cordillera. This will be discussed in more detail in section 5.3. CHAPTER 1. INTRODUCTION 7 1.3 Geophysical data Although no previous regional seismic tomography experiments have taken place in the north-ern Cordillera, the region has been included in global and near-global tomographic studies. Grand (1994) performed a tomographic S-wave study of the mantle beneath the Americas and surrounding oceans. Although our area of interest cannot be well resolved by such a large-scale experiment, it does appear as a region of low to moderate upper-mantle velocity, the velocity being particularly low in the upper 100 km. Overall, the area exhibits higher upper-mantle velocities than more southern regions of the Cordillera, and lower velocities than Alaska (see figure 1.4). A local-earthquake tomography experiment for southern Alaska is described in Zhao et al. (1995). An image of the crust and uppermost mantle was obtained to a maximum depth of 190 km. The main feature located was, not surprisingly, the subducting Pacific slab below southern Alaska and the Aleutian islands (its presence was included as a priori information). In addition, the authors located low-velocity anomalies associated with active volcanism at the surface, which dip to landward and descend to just below 100 km. The location and orientation of the slab at its eastern edge is somewhat uncertain; there is a tremendous drop-off in Benioff zone seismicity east of 145°W, associated with a bend in the descending slab, and so the location of the slab east of 145°W is poorly constrained (Stephens et al., 1984; Page et al., 1989). Much regional geophysical work on this area is planned as part of the SNORCLE Lithoprobe transect; at the time of writing, however, only gravity and aeromagnetic data are available. Lowe et al. (1994) examined aeromagnetic and gravity data (figure 1.5). The authors located a gravity high and magnetic low associated with the northern Yukon-Tanana terrane (interpreted as a region of young extension in Lowe and Cassidy (1995)). In addition, they remarked on a considerable change in gravitational and magnetic character across the Tintina Fault, suggesting that it is a crustal-scale feature and not a shallow tear fault. Figure 1.4: Results of S-wave mantle tomography from Grand (1994) for four different depths. Open circles represent low velocities; filled circles represent areas of high mantle S-wave velocity. The area of interest of this study is enclosed in a rectangle. CHAPTER 1. INTRODUCTION 9 Figure 1.5: Bouger gravity anomaly (top) and total-field magnetic (bottom) data for the north-ern Cordillera, from Lowe et al (1994). The major faults are indicated. CHAPTER 1. INTRODUCTION 10 In Lowe and Cassidy (1995), receiver-function analysis of teleseismic earthquakes at two permanent broad-band stations (Dawson Lake and Whitehorse, also used in this study) was combined with Bouger gravity data to determine the variation in crustal thickness between the Tintina and Denali faults. The authors conclude that there is an abrupt transition between thicker crust (w 39 km) south of 63°N and thinner crust ( « 35 km) farther north. This difference was attributed to young extension in the northern region, associated with stress transfer between the Denali and Tintina fault zones. 1.4 Post-accretionary history of magmatism in the northern Cordillera As the southern Yukon lies near an active margin and is composed of accreted terranes of various ages, it has been much affected by arc volcanism. Most of the area was included in the mid-Cretaceous to early Tertiary magmatic arc (Monger and Nokleberg, 1996) generated by the subduction of the Kula plate. This arc, and its associated plutonism, volcanism, and metamorphism, culminated in the mid-Eocene (circa 43 Ma) with the final consumption of the Kula plate (Erdmer and Mortensen, 1993). Much of the arc-related volcanic rocks from this period have been eroded away, and the plutons have been exposed. It is possible that some of the rocks deposited during the lifetime of this arc may not be arc-related. The Late Cretaceous Carmacks Group (around 70 Ma), which once covered much of the southwest Yukon, includes flood basalts and has been interpreted as being plume-related (Johnston et al., 1996); this interpretation is examined further in section 1.5. After the mid-Eocene, the change to strike-slip motion on the Queen Charlotte-Fairweather fault ended arc magmatism through most of this region, the exception being the Tertiary Wrangell Volcanic Belt northwest of the Saint Elias Mountains, which is of mixed arc-intraplate CHAPTER 1. INTRODUCTION 11 character (Souther and Yorath, 1991; Skulski et al, 1991). A l l other volcanism in this region since then has been comparatively minor (Francis, pers. comm.; Souther and Yorath, 1991). Tertiary and Quaternary volcanism in the northern Cordillera exclusive of the Wrangell Volcanic Belt has been limited to small-volume eruptions of alkaline to tholeiitic magma (such as those of the Stikine Volcanic Belt); such eruptions have occurred at points all along the Canadian Cordillera, and may be related to minor intervals of extension. One example is the Atlin area volcanics in northwest British Columbia, on the southern boundary of our study area, described in Edwards et al. (1996). 1.5 Terrane transport: controversies The tectonic history of the terranes of the northern Cordillera is a matter of some debate. At issue is the importance of strike-slip motion in terrane transport, and the degree to which the terranes of the Cordillera were rigidly attached after accretion. The principal difHculty in answering this question is that estimates of geological displacements along faults have generally yielded far less strike-slip motion than paleomagnetic measurements require; as well, estimates of fault slip from the Yukon do not match those from farther south (Monger and Nokleberg, 1996; Johmston et al, 1996; Irving et al., 1995). Paleolatitudes determined magnetically from Cretaceous igneous rocks generally indicate over 1000 km of motion since Cretaceous time. In Umhoefer et al (1989), paleomagnetic data from rocks in B . C . and Alaska are used to argue that a large block of Alaska, the Yukon, British Columbia and northern Washington (referred to as "Baja British Columbia") lay at the paleolatitude of Mexico around 90 M a (requiring about 2500 km of northward displacement since in the Coast Belt), and that this is consistent with reconstructions of oceanic plate motions for the region, assuming that this block was carried northward by the Kula plate. Irving et al. (1995) described paleomagnetic measurements for the Spences Bridge Group, CHAPTER 1. INTRODUCTION 12 a mid-Cretaceous volcanic group lying in the Intermontane Belt in south-central B.C., which they placed at a paleolatitude of about 51°, requiring about 1100 km ( ± 600 km) of northward displacement. This is not incompatible with the Baja B.C. model, assuming that the "Baja B.C." block did not remain rigid, and therefore that post-Cretaceous displacement increases westward. The Spences Bridge Group results can be reconciled with northern Cordilleran geology, as it is not difficult to accommodate this much displacement on the existing faults; the Denali and Tintina faults together could accommodate this much. In the southern Cordillera, however, it is difficult to accommodate this much motion along the known faults; the authors therefore proposed a major unrecognized dextral strike-slip fault located somewhere in the Omineca Belt, active during the Late Cretaceous or Paleocene. Paleomagnetic measurements have also been performed on the rocks of the Carmacks Group (briefly described in section 1.4); these rocks, located between the Denali and Tintina faults, were found to have been displaced by 1900 ± 700 km, placing them at the present-day latitude of Oregon at 70 Ma. This would place them in the vicinity of the Yellowstone hotspot at the time of their deposition; given geochemical evidence suggesting a possible plume origin for these rocks, the authors propose that they represent the Late Cretaceous expression of the Yellowstone hotspot (Johnston et al., 1996). In order to reconcile these paleomagnetic results with Southern Cordilleran geology, two possibilities present themselves: either there is a large-scale source of error in the paleomag-netic data (such as unrecognized tilting, although tilting was corrected for in the paleomagnetic results), or there has been considerable unrecognized strike-slip motion in the southern Cana-dian Cordillera. This study is unlikely to shed much direct light on this issue; however, major transitions in the crust may reflect major transitions in the underlying mantle. As well, the subduction of the Kula plate is likely to have exerted a considerable influence on the upper mantle in this area. CHAPTER 1. INTRODUCTION 13 1.6 Summary and motivation The northern Cordillera has experienced continuous tectonic activity since the Cretaceous: oblique subduction up to the mid-Eocene, and strike-slip deformation afterward. The latter is expressed by the Denali and Tintina fault zones (particularly the Denali after the Eocene). The total amount of post-Cretaceous strike-slip motion is in dispute, as paleomagnetic data require over 1000 km of motion, an amount which is difficult to reconcile with what is known about strike-slip motion in the southern Canadian Cordillera. The history of magmatism in this area reflects this plate-tectonic history: there was extensive arc magmatism until the mid-Eocene, after which magmatism has been limited to relatively small-volume alkaline volcanos which may be extension-related. Some of the Cretaceous rocks (the Carmacks group) may be plume-related. The structure and history of surface crustal rocks are likely to reflect the structure and evolution of the upper mantle beneath the northern Cordillera. We may expect some form of transition in the upper mantle between the more stable eastern portion of the Cordillera and the western portion where the most extensive deformation occurred, as well as between the Cordillera and the stable North American craton. As well, the structure of the upper mantle may provide clues as to the origin of recent magmatism in the region. The plate tectonic history of this area places limits on the age of upper mantle structures; in particular, it is unlikely that any pre-Eocene structures persist in the sub-northern Cordilleran mantle, due to the disruptive effect of the subduction of the Kula plate. With these factors in mind, the objective of this thesis is to examine large-scale upper mantle structure beneath the northern Cordillera and so provide constraints on the current state and past history of the region. Chapter 2 T h e experiment 2.1 Rationale Obtaining a detailed seismic image of the upper mantle beneath a given region is difficult to accomplish by conventional means. The main problem for controlled-source work is the diffi-culty in transmitting sufficient energy from a near-surface, anthropogenic source to appreciable distances below the Moho. Some recent large-scale refraction lines have been designed specif-ically to recover upper-mantle information, such as the 1995 Deep Probe experiment (Gorman et al., 1996). Peaceful nuclear explosions have been used for this purpose in the Soviet Union (Ryberg et al., 1996). In any case, nothing short of a nuclear explosion will produce seismic energy comparable to that which is generated on a daily basis by earthquakes. Therefore, pas-sive seismic techniques are an obvious means to obtain information about the upper mantle to considerable depth. Although the study area is seismically active, northern Cordilleran regional seismicity occurs entirely at shallow depths, and so fails to effectively illuminate the mantle, as the direct S and P arrivals propagate along paths which are predominantly crustal. This study is therefore based on distant earthquakes, more specifically events occurring in the teleseismic range. Teleseismic 14 CHAPTER 2. THE EXPERIMENT 15 earthquakes are those that occur sufficiently distant from the receiver that the main S and P phases bottom below the upper mantle discontinuities (i.e. below 660 km depth), and sufficiently close that the phases bottom above the core-mantle boundary, thus avoiding triplications that can complicate extraction of accurate travel times. This translates to an epicentral distance range of approximately 30° to 110°. A n advantage of using teleseismic earthquakes is that much of the ray path outside of the area of interest is common for all stations, and may therefore be accounted for by a simple static correction; this is explained in more detail in chapter 3. 2.2 Instrumentation The Geological Survey of Canada operates a sizable network of permanent seismic stations, known as the Canadian National Seismic Network. Five of these stations were used for this experiment: stations D A W Y (Dawson City, Yukon), D L B C (Dease Lake, B.C. ) , H Y T (Haines Junction, Yukon), W H Y (Whitehorse, Yukon), and Y K W 3 (one of four broad-band stations included in the Yellowknife Array, Yellowknife, N .W.T. ) A l l of these are broad-band, three-component stations, with the exception of station H Y T , which is a vertical short-period instru-ment. In addition, through the assistance of Roger Hansen at the University of Alaska (Fairbanks), seismograms from seven Alaskan short-period permanent stations were incorporated into the data set. These stations provide coverage at the western edge of the study area. To complement these stations and better cover the region of the S N O R C L E transect, five portable seismic stations were installed in the Yukon and eastern Northwest Territories: sta-tions F L S N (Fort Liard, N .W.T. ) , F P S N (Fort Providence, N . W . T . ) , FSSN (Fort Simpson, N .W.T . ) , L K S N (Lutsel Ke, N .W.T. ) , and W L S N (Watson Lake, Yukon). The locations of these instruments were constrained by the practical difficulties involved in locating year-round sites with dependable electric power in such a remote area. The instruments used were Guralp CHAPTER 2. THE EXPERIMENT 16 CMG-3T 3-component broadband seismometers, recorded onto Interay NARS CSD 20 data loggers. These data loggers record onto foitr-rniUimeter digital audio tape (DAT) cartridges that are easily changed in the field. See table 2.1 for detailed information about the stations used in this study. 2.3 Deployment and events Figure 2.1 shows the locations of the stations deployed for this experiment. The general ori-entation is along a line trending approximately east-west, with broader off-line coverage in the Cordillera. The portable stations were active in this deployment for the one-year period from July, 1995, to July, 1996; in addition, a few earlier archived events from the permanent sta-tions were used to improve the event coverage from the east and north. The portable stations were deployed at sites with available AC power; local station managers monitored the sites and mailed the tapes to Vancouver on a monthly basis. The deployment of the instruments and servicing of the sites were performed by Andy Langlois, a technician for the Geological Survey of Canada based in Yellowknife. The event set used for travel-time inversion is plotted in figure 2.2; in addition, a large number of events from Alaska (particularly the Aleutian Islands) were recorded, and may be used in future work, although they occurred too close to the array to be used as teleseismic events. As the figure shows, azimuthal coverage is good. The majority of events occur along the western Americas-western Pacific "ring of fire", which is unfortunately largely in line with the array, and so does not provide broad azimuthal coverage. Fortunately, there are a large number of southern Pacific events (such as from Fiji and Tonga) which fall just inside the edge of the teleseismic range and neatly fill in much of the southwestern quadrant. A scattering of Eurasian (Alpine-Himalayan) events provide some coverage to the north. The main gaps are the range of back azimuths from 40° to 110°, in which there are only a small number of CHAPTER 2. THE EXPERIMENT 17 Table 2.1: Stations used in this experiment. Station Type Lat. (°N) Lon. (°W) El. (m) Samp. (Hz) Components BCP Alaska short-period 59.953 139.635 396 120 Vertical CYK Alaska short-period 60.083 142.485 10 120 Vertical DAWY CNSN broad-band 64.053 139.432 346 40 3-comp. DLBC CNSN broad-band 58.437 130.027 978 40 3-comp. FLSN Portable broad-band 60.402 123.820 700 20 3-comp. FPSN Portable broad-band 61.050 117.438 119 20 3-comp. FSSN Portable broad-band 61.774 121.299 161 20 3-comp. HYT CNSN short-period 60.825 137.504 1416 100 Vertical LKSN Portable broad-band 62.400 110.730 500 20 3-comp. PNL Alaska short-period 59.668 139.397 585 120 Vertical SSP Alaska short-period 60.179 142.841 460 120 Vertical WAX Alaska short-period 60.448 142.851 991 120 Vertical WHY CNSN broad-band 60.660 134.881 1292 40 3-comp. WLSN Portable broad-band 60.115 128.751 675 20 3-comp. YAH Alaska short-period 60.359 141.745 2135 120 Vertical YKU Alaska short-period 59.554 139.725 40 120 Vertical YKW3 CNSN broad-band 62.562 114.605 198 20 3-comp. CHAPTER 2. THE EXPERIMENT 18 220° 230° 240° 250° 220° 230° 240° 250° Figure 2.1: A map of the array of seismometers used in this experiment. Permanent stations are shown in black; temporary stations are shown in gray. 19 Mid-Atlantic Ridge earthquakes, and the range from 180° to 220°. Figure 2.2: A map of the event locations used for the P-wave travel-time inversion, using a pro-jection equidistant from the approximate center of the array. The circles mark 30° increments in distance. 2.4 Data quality and processing Treatment of the data from the temporary stations presented a number of problems, due to both dimculties with the sites and problems with the instrumentation. The Interay data loggers depend on GPS antennae for accurate timing; unfortunately, the GPS systems were not locked CHAPTER 2. THE EXPERIMENT 20 to the satellites for a considerable proportion of the recording time, some of the timing outages lasting over a month. Due to the difficulties involved in performing regular maintenance on stations located in such a remote area, the problem was not detected until well into the recording period. In addition, the internal timing system of the data loggers proved to be somewhat unpredictable during periods of no GPS lock, occasionally exhibiting large clock shifts with no apparent cause. Fortunately, header information included in the field tapes provided complete information on the state of the GPS system and the behavior of the internal clock. We therefore were able to reconstruct reasonably accurate times over the shorter outages by linearly interpolating the starting and ending times. As such interpolation would be quite inaccurate over long periods of GPS failure, travel-time data from the longer outages were not used. The quality of waveform data from the temporary stations was variable. As the portable system requires AC electric power (a battery being present to cover short outages), stations had to be placed in settlements close to sources of anthropogenic noise. Some of this noise was reduced by band-pass filtering. WLSN proved to be the highest-quality station, rivalling the permanent stations in data quality. FSSN and FLSN were generally good, although intermittent anthropogenic noise was a problem. FPSN proved problematic due to frequent periods of noise, while LKSN operated for a shorter period than the others and so provided relatively few good events. Little processing was required for the permanent stations, as the CNSN and Alaskan sites were quite carefully chosen, and had the benefit of permanent vaults and reliable timing, as well as the considerable skills of the site maintainers. The quality of data for these stations was generally very high, with some variation from station to station (figure 2.3). In order to evaluate general data quality, event sections were plotted, with seismograms arranged by distance from the earthquake's epicenter; an example is shown in 2.4. For large CHAPTER 2. THE EXPERIMENT A 21 Time after earthquake (sec.) • i i — ' — i — • — i — ' — i — ' — i — • — i — B OAWY -jJ\|siV^ WHY" —1 1 1 . 1 . 1 p r -T . , . 1 1 1 • i 1 j — - r^ | j J | |W HYT * T ' 1 • 1 • 1 • 1 1 1 1 1 1 1 -- r \jvmf^»J^4^ L K S N v-l — l 1 r l ( • 1 1 1 . 1——• 1 1 | — i 1 1 1 1 1 1 1 1 1 650 700 750 800 850 900 Time after earthquake (sec.) Figure 2.3: Waveform data for two events recorded by the array used in this study. Event A was a deep subduction event (Columbia, Aug. 19, 1995; 126 km depth); note the sharp P-wave arrival. Event B (Mid-Atlantic Ridge, June 2, 1996) was shallower (10 km) and displays a more highly oscillatory character. Note the difference between the broad-band and short-period responses. Before plotting, this data was band-pass filtered using corner frequencies at 0.05 Hz and 5 Hz. CHAPTER 2. THE EXPERIMENT 22 events, multiple arrivals were recorded, including surface waves; this broad spectrum, along with the presence of three-component instruments, allows for the analysis of the entire three-dimensional waveform, although we concentrated on travel-time analysis for this study. For P-wave travel-time purposes, the longer-period waves were removed with a band-pass filter (usually with corner frequencies at 0.5 Hz and 5 Hz), leaving sharply-defined P-wave arrivals. CHAPTER 2. THE EXPERIMENT Figure 2.4: A sample event section (for the event in figure 2.3 B). Chapter 3 Travel-time analysis 3.1 Rationale Figure 3.1 is a cartoon illustrating the general configuration of a teleseismic tomography experi-ment. Two important considerations serve to simplify the analysis: the array is small compared to the distance to the source, and the first arrivals of P and S energy are always simple direct ray paths that bottom in the lower mantle. As a consequence, the rays for a given event follow almost identical paths for most of their length, diverging significantly only in the upper mantle on the receiver side. In addition, due to the large source-receiver separations and the relatively low velocities present in the crust, the ray paths are nearly vertical during their passage through the crust. Suppose, then, that we represent the travel time from earthquake i to station j by the equation tij = ejj -f rriij + rij, where ejj is the component due to travel from the source to the point in the mid-mantle where the rays diverge, m -^ is the component due to travel in the upper mantle beneath the array, and ry is the component due to crustal travel below the stations. As the paths from the earthquake focus to the mantle beneath the array are nearly identical for a given event, we can assume that, to a good approximation, the travel-time components for the 24 CHAPTER 3. TRAVEL-TIME ANALYSIS 25 Figure 3.1: The ray paths for teleseismic events are nearly identical for all stations, except when close to the array. In the crust below the stations, the rays are almost vertical, and so nearly identical for all events. This is the justification for the use of static corrections to absorb effects from outside the model. CHAPTER 3. TRAVEL-TIME ANALYSIS 26 event are also identical, and so we can rewrite ejj as ej, a component depending only on the source. In a similar vein, the vertical nature of the crustal ray paths requires all the crustal ray paths beneath a given station to be nearly identical; therefore, to a good approximation, we can replace by rj, which depends only on the station. We have thus reduced our problem to one involving only tomography in the upper man-tle beneath the array (as represented by the rriij component) and simple event and receiver corrections (e^  and TV,-, which are analogous to the static corrections used in the processing of seismic lines). If the static corrections can be correctly determined, the majority of remaining travel-time perturbations must be due to structure in the upper mantle beneath the array. 3.2 Determination of travel times Accurate determination of the arrived time of an earthquake at a receiver is a more difficult problem them it would seem at first glance. The principal difficulty lies in ensuring that, between multiple seismograms, the equivalent point is consistently picked. This is eased in the teleseismic case by the consistency of the waveform between stations: as the teleseismic range contains no triplications, large phase shifts are unlikely. However, teleseismic events are generally low-frequency (averaging around 1 Hz for the P-wave), and so picking the same point on each seismogramis likely to give imprecise travel times, as well as being difficult to do when the signal-to-noise ratio is low. It is therefore preferable to pick arrivals based on some measure of the correspondence of the first few cycles of the event on each seismogram. For the type of travel-time analysis we will perform, only relative travel times are required; that is, it is not necessary to pick the precise first arrival of energy on every seismogram, as long as the same point is chosen on each. The point chosen should, however, be early in the wave train, in order to avoid complications due to crustal multiples. CHAPTER 3. TRAVEL-TIME ANALYSIS 27 Consider two seismograms for the same event recorded at different stations, as shown in figure 3.2. The shift that best aligns the two seismograms is that which maximizes the cross-correlation between the two. This is superior to visual picking of maxima in that it is not highly dependent on high-frequency noise, and can easily be automated. However, as there are local minima at every alignment of peaks and troughs (particularly in narrow-bandwidth data, such as that from a short-period instrument), there is a considerable risk of cycle-skipping. Thus, using cross-correlation to improve the accuracy of visually-chosen picks was judged to be the appropriate technique to use (VanDecar and Crosson, 1990). Cross-correlating more than two seismograms imposes an additional complication: the dif-ficulty of reconciling the best shifts for different pairs of traces. In general, these maximum cross-correlation shifts will not be entirely consistent, and so some sort of best-fit approxima-tion must be used. For this project, John VanDecar's mccc program was employed (VanDecar and Crosson, 1990). It calculates the maximum cross-correlation shifts in a window around the hand-chosen picks for every possible pair of seismograms. The result is the system of equations Tj — Ti — Ary, where TJ is the absolute time for the ith trace and Ary is the best shift com-puted from the cross-correlation of traces i and j. Adding the constraint that YA=I T » = 0 (to ensure that the system is completely determined), we have an overdetermined linear system of equations, small enough to be easily solved in a least-squares sense. The program performs this calculation, and then writes the resulting picks to an output file (VanDecar and Crosson, 1990; VanDecar, 1991). The basic picking procedure employed is as follows: first, the traces are band-pass filtered, and predicted picks (calculated from the IASP91 model) are written to the data files. The hand picking is then performed (using Seismic Analysis Code [SAC] [Tapley and Tull, 1992]) and low signal-to-noise ratio traces are rejected. The multi-channel cross-correlation is then performed, and the resulting improved picks are plotted to check for cycle-skipping. The corner CHAPTER 3. TRAVEL-TIME ANALYSIS 28 ~ i i i i I i i r seconds 1 1 1 1 1 1 1 1 1 1 1 1 i i • i i i i i i i i | i i i i Cross-correlation / of A and B / v \ A A Optimum shift: =-2.5 sec. i i i t 1 i i i i 1 i i i i i i i i i i i i i i i p i i — -15 15 Shift of A with respect to B (seconds) Figure 3.2: Seismograms A and B consist of the same event recorded at different stations. Their cross-correlation is plotted as a function of the offset applied to seismogram A. The shift that provides the largest positive cross-correlation between the two is clearly that which best aligns them; note, however, that there are local maxima corresponding to misalignments of multiples of the dominant wavelength. CHAPTER 3. TRAVEL-TIME ANALYSIS 29 frequencies employed vary from event to event; generally, the 0.5 Hz to 5.0 Hz band, covered by both short-period and broad-band instruments, is used, but some events require a wider band. Large earthquakes, unless occurring at great depth, generally appear highly oscillatory in the usual high-frequency band, causing the first arrival to be difficult to recognize; these events are therefore generally picked in a lower frequency band (0.05 Hz to 3.0 Hz). A consequence of this is that data from the short-period instruments can seldom be used for these large events. 3.3 The inverse problem The forward problem of determining seismic travel times is formally expressed as where Uj is the travel time from source i to receiver j, is the ray path from source i to receiver j, and v(r) is the seismic velocity function of the Earth (P- or S-wave as the case This equation is almost that of a linear problem; however, non-linearity enters through the ray path Tij, which is dependent on the slowness function s(r). Therefore, the inversion method used must be non-linear. As a first step, we must decide how the model is to be described. For reasons explained in section 3.1, we need only consider that part of the upper mantle located immediately below the array, over a depth range between 100 and 900 kilometers. Within this region, we need a finite parameterization of the slowness function; the parameterization to use is determined by the nature of the data set. As all of the rays used are direct arrivals, with no reflections may be). If we introduce the slowness s(r), defined by s(r) — the equation becomes the linear-looking CHAPTER 3. TRAVEL-TIME ANALYSIS 30 or refractions within the area of interest, we have no means of detecting sharp interfaces; in any event, including sharp changes in velocity in a tomographic inversion causes the inverse problem to be highly non-linear and unstable. Thus, it is best to represent the slowness as a smooth function; this will capture large-scale variations in the region of interest, although it will smooth out discontinuities (such as the phase transitions at 410 and 660 km depth). In addition, the model is best described in terms of perturbations around a standard Earth model; the IASP91 standard model (Kennett and Engdahl, 1991) was selected for this purpose. The parameterization chosen, then, is that of a tensioned spline over a mesh of grid nodes (VanDecar, 1991). This provides a smooth, simple function given fixed values at the nodes. Figure 3.3 is a plot of the grid used for this experiment. The area covered by the model runs from 54°N to 68°N, and from 147°W to 103°W, extending vertically to 1400 km depth. This, as the figure demonstrates, means that the model extends in all directions considerably beyond the area in which we would expect to obtain good resolution. The presence of these extra nodes allows the model perturbations to be damped down to zero at the edges, as well as allowing anomalies to be placed outside the region of best expected resolution should that best fit the data (VanDecar, 1991). The nodes are spaced every half-degree in latitude, and every degree in longitude, the outermost nodes being spaced farther apart. In addition to these true model parameters, we will invert for three additional parameter sets: static corrections for the stations and the events, and focus mislocations for the events. The reasons for the use of the first two are explained in section 3.1. The earthquake mislocations allow for errors in the foci and have the effect of removing travel-time trends across the array. The station locations are assumed to be accurate, and in the case of the portable stations, were measured using the data loggers' in-board GPS units. The inverse problem then becomes the following: given the relative travel times of the events to the receivers, determine the seismic velocities of the grid nodes of a taut-spline model, along CHAPTER 3. TRAVEL-TIME ANALYSIS 31 Figure 3.3: A plot of the grid used for the travel-time inversion. The slowness function is fixed at the grid knots, where the grid lines intersect. The region outlined in yellow is that in which we expect to be able to resolve structure. CHAPTER 3. TRAVEL-TIME ANALYSIS with static event and receiver corrections and a set of event mislocation vectors. 32 3.4 Linear inversion technique A first attempt at solving a non-linear inverse problem may be made by finding an approxim-ation of the non-linear problem as a linear problem; this is also a necessary step for most true non-linear inversion techniques. Recall from the previous section that our forward problem can be stated as Suppose we know the predicted travel times and ray paths for a standard radial Earth model (e.g. IASP91 (Kennett and Engdahl, 1991), used in this study). If we represent the ray paths and slowness function for this model as T°j and a°(r), respectively, and express the true model as a perturbation around this starting model (i.e., s(r) = s°(r) + f5s(r) and T y = I y + STij), then we can rewrite the forward problem as *a= L (A') + 6<*))*r • According to Fermat's principle, the true ray path is a stationary point of the travel time function. Therefore, assuming that the model perturbation Ss(r) is small, we can neglect the influence of the changed ray path on the travel time and make the first-order linear approxim-ation «« = /„ (s°(r) + Ss(r))d7. Expressing the travel times in terms of perturbations about the travel times predicted from the starting model (i.e. iy = *?• + tfiy), the forward problem may be rewritten as Stij = f Ss(r)dl. CHAPTER 3. TRAVEL-TIME ANALYSIS 33 This is the linear approximation of our non-linear problem. Its validity depends on the magni-tude of 6s(r); as we do not expect our slowness perturbations to be more than a few percent of the IASP91 values, this is a reasonable first approximation. By updating the starting model through successive iterations of linear inversion, a full non-linear inversion may be achieved. In order to solve the linear inverse problem stated above, we need to express it in matrix terms using the parameterization described in the previous section. The matrix expression of the above linearization, with no weighting or normalization included, is St = PSs, where St is the vector of measured travel-time perturbations, Ss is the vector of model parameters (values of the slowness function perturbation at grid nodes), and P is the matrix of partial derivatives PH = f £ l , = * ° (VanDecar, 1991). Calculating this matrix of partial derivatives is the most computationally intensive part of the inversion procedure. Since we begin with a standard radial Earth model, the takeoff angles and azimuths of the desired rays are easily computed. The next step is to determine how each ray is affected by each grid knot. This is done by "shooting" a ray through the model; as the effect of each node is quite local in this parameterization, it can safely be assumed that each ray is affected only by nearby nodes. Given sets of nodes likely to be affected by each ray, and sets of straight line segments approximating each ray's path, we can then slightly perturb each grid knot, recalculate the spline coefficients, integrate the perturbed slowness along the ray paths affected by the perturbed knot, and so calculate each partial derivative (VanDecar, 1991). As our problem is highly underdetermined, and we will in any case not want to fit the data exactly (due to the existence of errors in the data), the next step is to select some form of regularization for our model. A minimum-structure solution has the advantage of favoring simple, easy-to-interpret models, as well as avoiding the introduction of spurious structure into the model. We therefore minimize the Laplacian of the model (using a finite-difference matrix) in order to obtain a smoothest-model solution; a component of flattening is also included in the CHAPTER 3. TRAVEL-TIME ANALYSIS 34 regularization matrix (VanDecar, 1991). The final matrix to be inverted includes the travel-time equation, weighted by the errors in the data (as estimated by multi-channel cross-correlation), along with the regularization matrix and components to invert for the source and receiver corrections. It is very large, albeit sparse; sufficiently so that inverting it directly would require a prohibitive amount of available computer memory. Therefore, the matrix is inverted iteratively using a conjugate-gradient method (VanDecar, 1991). As this inversion method finds a least-squares solution to its matrix equation, a few data points with large errors (such as cycle skips) can greatly influence the solution. One way of dealing with this would be to reject measurements with large misfits after performing a preliminary inversion. A n alternative which does not require rejecting data is to downweight data points with large misfits, rather than simply rejecting them. After an inversion has been performed, outliers are downweighted to the equivalent of an LI norm, as opposed to the L2 norm inherent in the least-squares procedure (VanDecar et al., 1995; Scales, 1988). This process is then repeated (from 12 to 20 times in this case), adding a level of iteration to the calculation that reduces the effect of outliers without rejecting data. This linear inversion technique is iterated in order to perform a non-linear inversion. There are two techniques typically used for this purpose: the creeping method, where the regulariza-tion is applied to the model perturbation at each step, and the leaping method, where the entire model is regularized at each step. The leaping method, although not guaranteed to converge, tends to return a more regular model and so was used in this study (Parker, 1994; VanDecar, 1991). The linear numerical problem at non-linear iteration k is expressed as CHAPTER 3. TRAVEL-TIME ANALYSIS 35 ^ W P W R W E ^  V J V r(*) ^-AF(E&*.«>) J where W is a weighting matrix based on the estimated standard deviations of the data (Wy = ^f), R and E are index matrices (Rip = 1 if ray i was recorded at receiver p, = 0 otherwise; sim-ilarly, Eiq — 1 if ray i originated from event q), F is the difference matrix for the regularization used (we use F = A / F / + A , F , , where / denotes flattening and s denotes smoothing; the As are weights applied to each form of regularization), and P is the matrix of partial derivatives. The vectors to solve for are 6s(k\ the slowness model for iteration k; r(k\ the receiver corrections for iteration k; and e^k\ the event corrections for iteration k. The right-hand side of the equation contains the data £T(FC_1) (the travel-time residuals corresponding to the previous model) and a summation term using all of the previous slowness models (to ensure that the entire model, not just the new perturbation, is regularized). Solving this equation is the innermost step in this three-level procedure (VanDecar, 1991). Each non-linear iteration consists of an entire linear inversion, with the model from the previous iteration used as the reference model. As the reference model is no longer radial, an additional step is required: the 3-D ray tracing needed to find the correct launch angle and launch azimuth for each ray prior to calculating the partial derivatives. This is done using a shooting method (VanDecar, 1991). Chapter 4 Results 4.1 Residuals Before inverting travel-time data, it is useful to determine what conclusions can be drawn from arrival times; this can provide clues as to where to look for structure as well as providing a useful check for models generated by inversion. The relative travel time Uj of event i to station j may be decomposed into Uj — t\j + SUj, where t\j is the relative travel time as predicted by the IASP91 standard earth model, and SUj is the travel-time perturbation unaccounted for by the standard model. In order to account for crustal effects beneath the stations, we can further decompose the travel-time perturbation using the equation SUj = Stj + SUj, where Stj is the average travel-time perturbation to station j, the averaging being performed over all the recorded events. The remaining perturbation SUj is then approximately that component of the travel-time which the inversion maps into the upper mantle. A simple method for examining these data is to plot the travel-time residuals for a single event on a map. This can be quite valuable in experiments where the stations are finely spaced; however, due to the relatively small number of stations used in this experiment, the conclusions we can draw from such maps are limited. Figures 4.1 and 4.2 give examples of such maps for 36 CHAPTER 4. RESULTS 37 events arriving from widely varying back azimuths, the first figure being a plot of Stij and the second being Stij. There is a consistent pattern of westwardly-increasing delays in figure 4.1, much of which probably reflects the the younger mantle and crust of the Cordillera relative to the craton. The patterns in figure 4.2, which are more likely to reflect mantle structure, are far less systematic, and it is difficult to draw specific conclusions from them before inverting. 95/08/19 21.43.314.99N 75.65W 1263 6.2Mb A COLOMBIA 95/10/18 09J038 3634N 70J2E 2224 5.7Mb A HINDU KUSH REGION. AFGHANISTAN 220' 230' 240' 250' 220' 230' 240' 250' 95/10/20 19.21.2S 18JI7N 145.12E 225.4 5.3Mb B MARIANA ISLANDS 95/12/05 18.49.3139.19N 40.41E 33.0 5.6Mb B TURKEY 220' 230' 240' 250' 220" 230' 240* 250* Figure 4.1: Some sample contour maps of travel-time perturbations for individual events. The arrow indicates the direction of propagation of the wavefront; the contour interval is 0.5 s. It is perhaps more instructive to gather the perturbations by station rather than by event, CHAPTER 4. RESULTS 38 95/08/19 21.43 J l 4.99N 75.65W 126JI 6.2Mb A COLOMBIA 95/10/18 093038 3&34N 70J1E I 2 l » 5.7Mb A HINDU RUSH REGION, AFGHANISTAN 220' 230' 240' 250' 220' 230' 240' 260' Figure 4.2: Contour maps of travel-time perturbations after the station averages have been removed. The events shown are the same as in the previous figure; the contour interval is 0.25 s. CHAPTER 4. RESULTS 39 as there can be at most 17 data points per event, as compared to a possible 385 data points per station. Unfortunately, the perturbations for different events to the same station are not directly comparable; although the averaging to zero performed by using only relative arrival times eliminates any static event bias, it also biases data points differently depending on which particular set of stations recorded each event. Consequently, the variations in travel-time per-turbations to a given station between different events are only reflective of structure to the extent that the sets of stations recording the events were consistent. However, if we assume that this variation in the station set from event to event was largely independent of the location of the events, then we can suppose that the effect of the variation will only be to introduce scatter into the pattern of perturbations, and not to introduce spurious directional variations. With this in mind, the travel-time perturbations (after removal of the station average) for each station were plotted against the back azimuth and epicentral distance of the event (figure 4.3). This is somewhat analogous to the event-side technique of plotting residual spheres (Creager and Jordan, 1984). Not all of these plots display obvious structure; stations with relatively few events (e.g. L K S N and the Alaskan stations) were omitted. On the plots shown, the main conclusions that may be drawn are directional. For instance, on the D A W Y plot, we can state that there are regions of fast mantle to the northwest and southeast, and a slow region to the southwest; the plot for station W H Y suggests a substantial fast region to the north. This is consistent with the inversion results (see below). 4.2 Choice of inversion parameters In an inversion of this type, the choice of regularization parameters is crucial. Figure 4.4 demon-strates how the nature of the model varies depending on the choice of smoothing parameters; the most notable effect is the change in amplitude of the recovered anomalies, but there is also a considerable effect on the amount of structural detail recovered. The problem of choosing the CHAPTER 4. RESULTS 40 DAWY N WHY Figure 4.3: Sample plots of station-gathered corrected travel-time perturbations, on a radial projection. Crosses represent positive travel-time perturbations (slow arrivals) while circles represent negative perturbations (fast arrivals). CHAPTER 4. RESULTS 41 best regularization amounts to adjudicating a tradeoff between fitting the data and simplifying the model, quantified as a relationship between the travel-time misfit and the model norm. In our case, there are two regularization weights to be used, a smoothing weight (niinimizing model roughness) and a flattening weight (minimizing model gradients). The relative weight of these two regularizations is chosen based on the desired form of the recovered model; as we desire a minimum-structure solution, the flattening weight is held to ^ of the smoothing weight. The misfit-norm relationship for a Linear inversion may be plotted as a tradeoff curve. Two such curves are shown in figure 4.5, for the first and second Linear iterations. Choosing the correct point on a tradeoff curve can be done using knowledge of the standard error of the data, which should correspond to the target root-mean-square misfit. If the standard errors of the data are not well known, the optimum regularization may be picked by its proximity to the point of maximum curvature of the tradeoff curve (Parker, 1994). Due to both the presence of non-Gaussian outliers in the data and our uncertain knowledge of the measurement error, the standard error of the data is somewhat uncertain; therefore, the latter method was used to find optimum models for both iterations. In order to ease ray tracing for the non-linear step of the inversion, a somewhat smoother than optimum model from the first iteration was used as the starting point for the second. One problem in non-linear inversion is knowing when to stop iterating. The number of non-linear iterations required for convergence depends on the degree of non-linearity of the problem, and, due to the computationally-intensive nature of non-linear inversion, it is wasteful to pursue the inversion further than necessary. In figure 4.6, equivalent models from the first and second iterations are shown. The differences between the two are quite minor, much of the difference being due to the slight difference in regularization between the two models. A n exception is the strong negative slowness anomaly at the extreme east of the first model; however, as shall be shown in the next section, this feature lies in a region where the model resolution is very poor. CHAPTER 4. RESULTS 42 Figure 4.4: A n example of the effect of varying the weight given to smoothing in the linear inversion. These otherwise identical horizontal sections through the first-iteration model show the effect of smoothing weights ranging from 50000 (upper left corner) to 5000 (lower right), ranging from a severe underflt to a severe overfit of the data. The chosen fit lies between the values used for the second and third plots. The inversions used 12 downweighting iterations apiece; black areas are those without good ray coverage (i.e. no sampling rays). CHAPTER 4. RESULTS 43 Figure 4.5: Tradeoff curves for the two linear iterations performed. The root-mean-square travel-time misfit is plotted against the model roughness (the integrated Laplacian of the model). The values used here were calculated without any downweighting of outliers, so the magnitudes of the misfits are exaggerated. The optimal model for the initial inversion (left) lies between the two points in dashed circles; however, to ease ray tracing, a somewhat smoother model was used as a starting model for the next iteration (solid circle). The optimum model from the second iteration (circled, right) was used as the final model. CHAPTER 4. RESULTS 44 Considering the minor nature of the differences between the first-iteration and second-iteration models, we decided to stop the inversion after two iterations of linear inversion. Figure 4.6: Horizontal sections for models with approximately optimal regularizations from the first (left) and second (right) linear iterations. Note that the regularizations are not exactly equivalent; the left-hand model was slightly less heavily regularized, which may account for differences in amplitude. Remaining differences between the two can largely be attributed to non-linearity. 4.3 Resolution Calculating the resolution matrix for a linear inverse problem requires a singular-value decompo-sition of the matrix of partial derivatives. This is not possible in this case because the matrix is prohibitively large (VanDecar, 1991). Therefore, in order to evaluate the resolving power of this type of experiment, it is necessary to perform a resolution test. The way this is done is simple: a synthetic model is generated, the data that such a model would produce are calculated, these synthetic data are inverted, and the recovered model is compared to the input model. We per-formed only linear inversions of the synthetic data; therefore, the forward modeling amounted simply to a multiplication of the IASP91 model's partial derivatives matrix by the model vector Longitude West Longitude West CHAPTER 4. RESULTS 45 (St — PSs; see section 3.4), a very quick process involving no ray-tracing. Gaussian noise with a standard deviation of 0.015 second (approximating the estimated standard error of the real data after the downweighting of outliers) was added to the synthetic data; due to the Gaussian v nature of this noise, it was unnecessary to perform any downweighting of outliers, and so the inversion process took only about forty minutes of C P U time on a Sun SPARC 10. Two such tests were performed. The first test (figures 4.7 and 4.8) involved the recovery of a single large spike anomaly, in order to examine how the anomaly was deformed by the inversion given the available ray set. The most notable effect of the inversion is the reduction in amplitude of the anomaly; this is hardly surprising, as the smooth-model inversion used strongly penalizes spikes and tends to smooth out sharp changes in slowness. In addition to this reduction in amplitude, there is some downward and westward smearing of the anomaly; this reflects the paths of the rays intersecting the anomaly. Overall, however, this large anomaly was quite well recovered. The second, more detailed resolution test involved a three-dimensional checkerboard of evenly-spaced positive and negative anomalies (figures 4.9 through 4.12). This type of test evaluates the ability of the experiment to resolve small anomalies, as well as locating areas of good and bad resolution in the model, and so determining which areas in the model can safely be interpreted. We drew four main conclusions from the checkerboard resolution test. First, the resolution falls off quite rapidly to the east, particularly at shallower depths. In particular, resolution east of 1 2 2 ° W is very poor, except at great depth. Therefore, that part of the model that lies east of 122° will not be interpreted. Second, the magnitude of the anomalies was greatly underestimated. This is probably a consequence of the spike-like nature of the anomalies, as discussed previously for the large-spike test. As well, the peaks of the anomalies could only be sampled by a ray passing directly through their centers. For smoother structures, the magnitude CHAPTER 4. RESULTS 46 A A: ( 58 00N. 137 OOW ) A': ( 66.00N, 137.00W ) -4.0-3.0-2.0-1.0 0.0 1.0 2.0 3.0 4.0 SN096A1 cross section P-wave % slowness anomaly A: ( 60.00N, 145.00W ) A': ( 60.00N, 105.00W ) Figure 4.7: Three sections through the synthetic model used for the large-spike resolution test. The plots are: A) a horizontal section at 400 km depth, B) a vertical section oriented N-S along the 137° meridian, and C) a vertical section along an approximately E-W great circle intersecting the anomaly. Areas shaded black were not sampled by more than one ray and therefore cannot be resolved by the inversion. CHAPTER 4. RESULTS 47 40 30^20 10 00 10 20 30 40 A; ( 58 00N, 137.0OW ) A': ( 66 OON, 137 OOW ) -4.0-3.0-2.0-1.0 0.0 1.0 2.0 3.0 4.0 SN096AI cross section ^t^^^^^j^ P-wave % slowness anomaly A : ( 6 0 . 0 0 N , 1 4 5 . 0 0 W ) A ' : ( 6 0 . 0 0 N , 1 0 5 . 0 0 W ) Figure 4.8: Three sections through the recovered model from the large-spike resolution test. The sections correspond to those in the previous figure, and the color scale is the same. CHAPTER 4. RESULTS 48 Figure 4.9: Horizontal sections through the input (left) and recovered (right) models from the checkerboard resolution test. Note the difference in color scales between the input and recovered models. CHAPTER 4. RESULTS 49 -4.0-3-0-2.0-1 0 00 1 0 20 30 40 -1.0-0.8-0.5-0.2 0.0 0.2 0.5 08 1.0 Longitude West -4.0-3.0-2.0-1-0 0.0 1.0 2.0 3.0 4.0 Longitude West 1 0-0 5 0.5-0.2 0.0 0 2 0.5 0.8 1.0 P-wave % slowness anomaty Longitude West Longitude West Figure 4.10: Deeper horizontal sections from the checkerboard resolution test. (See the previous figure for details.) CHAPTER 4. RESULTS 50 SN096A1 cross section -4 0-3.0-2 0-1 0 0.0 1 0 2 0 3 0 4 0 P-wave % slowness anomaly SN096A1 cross section - 1 ^ ^ ^ ^ ^ 2 ^ ^ ^ 2 ^ ^ ^ ^ . 0 P-wave % slowness anomaly A: (62.50N, 144.00W) SN096A1 cross section A': (62.50N, 106.00W) -4.0-3.0-2.0-1.0 0.0 1.0 2.0 3.0 4.0 P-wave % slowness anomaly A: (62.50N, 144.00W) SN096A1 cross section A':(62.50N, 106.00W) -1.0-0.8-0.5-0.2 0.0 0.2 0.5 0.8 1.0 P-wave % slowness anomaly A: (60.50N, 144.00W ) SN096A1 cross section A':(60.50N, 106.00W) -4^^o^^^o^i)^o^o^^o P-wave % slowness anomaly A: ( 60.50N, 144.00W ) SN096A1 cross section A': ( 60.50N, 106.00W ) - 1 ^ ^ 8 ^ ^ ^ 2 ^ ^ ^ 2 ^ ^ ^ 8 ^ 0 P-wave % slowness anomaly A:(58.50N, 144.00W) A': (58.50N, 106.00W) A:(58.50N, 144.00W) A': (58.50N, 106.00W) Figure 4.11: East-west oriented vertical sections through the input (left) and recovered (right) models from the checkerboard resolution test. The slices correspond to the middle three rows of anomalies in the checkerboard. The sections are stacked from northernmost to southernmost. CHAPTER 4. RESULTS 51 Figure 4.12: North-south oriented vertical sections through the input (first and third rows) and output (second and fourth rows) models from the checkerboard resolution test. The sections are arranged from westernmost (upper left corner) to easternmost (lower right); each slice is oriented with south at the left. CHAPTER 4. RESULTS 52 of anomalies is likely to be better measured; however, we can expect some underestimation. The uppermost part of the model shows the greatest tendency for underestimation and poor resolution, probably as the consequence of structure being absorbed into the station corrections and the absence of ray crossings at shallow depths. Third, the resolution varies strongly with depth. The uppermost layer of anomalies was largely missed; in the deeper layers, there is considerable downward smearing. The best depth and lateral resolution seems to lie between 200 and 600 kilometers depth, in the western half the model; in that region, anomalies of about 150 km in all directions may be detected. Finally, there is considerable smearing of the anomalies along lines plunging southwestward. This reflects the uneven distribution of events in the data set (see figure 2.2), and we can probably expect similar smearing to occur in the model recovered from real data. The smearing is greatest below 500 km. In summary, we can expect reasonable recovery of structure below 200 km west of 122°W. Structure above 200 km may be missed and will probably be underestimated. We can expect structure below 500 km to be smeared somewhat southwestward. However, with all this in mind, we can interpret large-scale structure with a fair degree of confidence. 4.4 Observed features Figures 4.13 through 4.16 present the final model we have selected. It is the result of two itera-tions of linear inversion, each consisting of 20 downweighting iterations around 2000 conjugate-gradient iterations; the regularization parameters used correspond to the points marked on figure 4.5. The total execution time for the inversion was about 150 hours on a Sun Sparc 10 computer. The station static corrections are shown on the horizontal model slices (figures 4.13 and 4.14); the event static corrections are shown in figure 4.17. Before examining the structures located by the inversion, it is useful to examine the station CHAPTER 4. RESULTS 53 Figure 4.13: Horizontal cross-sections through the final tomographic model. The locations of the Tintina and Denali faults are marked in white; the cross-sections plotted in figures 4.15 and 4.16 are shown in grey on the 200 km section. Station statics are shown as triangles (negative) and squares (positive). CHAPTER 4. RESULTS 54 Figure 4.14: More horizontal cross-sections through the final tomographic model; see the pre-vious figure for details. CHAPTER 4. RESULTS 55 -3.0-2.3-1.7-1.0-0.30 3 1 0 1.7 2.33 0 A ( 62 OON, 144 OOW) A': ( 66.OON, 126.00W) B: ( 59.00N, 144.00W ) B': ( 64.00N, 122.00W ) -3 0-2 3-1.7-1.0-0 30.31 01.72 33 0 S N 0 9 6 A 1 cross section ^^^^^jnn^j^ P-wave % slowness anomaly C: (56.00N, 144.00W) C : (61.00N, 122.OOW) Figure 4.15: Southwest-to-northeast cross sections through the final tomographic model; see figure 4.13 for cross-section locations. CHAPTER 4. RESULTS 56 Figure 4.16: Northwest-to-southeast cross sections through the final tomographic model; see figure 4.13 for cross-section locations. Figure 4.17: A map of the event static corrections for the final tomographic model. CHAPTER 4. RESULTS 57 and event static corrections. The event statics (figure 4.17) are generally small in magnitude and do not display any discernible structure. As structure in the event corrections may indicate problems with the inversion, this is a reassuring sign. The station corrections (figure 4.13) show mostly small negative values for the Canadian stations, and positive values centered near 1 second for the Alaskan stations. This may be reflective of thin or high-velocity crust on the Alaska side, or, more likely, may be the result of a difference in response between the Canadian and Alaskan instruments. The strongest structure visible in the model is a large positive slowness anomaly, centered at about 60°N by 136°W at the 300 km level, ranging from 100 km to 600 km depth (although the vertical extent may be exaggerated by up to 100 km due to smearing). The anomaly appears to dip slightly to the southeast (see section D-D' in figure 4.16), and is elongated in a northwest-southeast direction. The dip is unlikely to be a smearing artifact, as it is oriented opposite to the expected direction of smearing. Its extent is greatest between 300 and 400 km depth, where it is about 700 km long (though it may extend beyond the model's edge) and 400 km wide. The northeastern edge of the anomaly, which approximately parallels the Tintina and Denali faults, is particularly sharp. The positive anomaly appears to extend beneath a negative anomaly (described below) to a depth of about 500 km (see, in particular, section B-B' , in figure 4.15). To the southwest of this anomaly, in the southwestern corner of the model, a smaller negative slowness anomaly is present. This feature extends from the top of the model (at 100 km) to a depth of about 250 km. It is not much bigger than the smallest feature this experiment was able to resolve (see previous section); however, it lies in the westernmost portion of the model, where resolution is sharpest. Other than these two features, the model displays few significant slowness anomalies. There are indications of structure in the eastern part of the model, particularly a positive anomaly at about 63°N by 126°W which appears to extend to 700 km depth; however, this is in the region CHAPTER 4. RESULTS where the resolution of the model deteriorates considerably. Chapter 5 Discussion of results 5.1 The nature of the observed anomalies In the previous chapter, two significant P-wave anomalies located in this study were described: a large positive slowness (low velocity) anomaly located between the Tintina and Denali faults and extending to considerable depth (400-600 km), and a negative slowness (high velocity) anomaly at the western edge of the model, penetrating to about 250 km depth. The high-velocity anomaly lies in roughly the expected horizontal position of the edge of the subducted slab from the Alaska subduction zone (figure 5.1). The depth and configuration of the slab edge are poorly constrained from previous studies, due to the decline in Benioff-zone seismicity east of 1 4 5 ° W (Stephens et al, 1984; Page et al., 1989). Extrapolation from the nearest available control on the slab location would suggest a depth to the top of the slab of about 50 km at 61°N by 143°N (Zhao et al., 1995). This extrapolation, combined with the slab thickness of 45-55 km from the tomographic model of Zhao et al. (1995), implies that the observed high-velocity anomaly is thicker and deeper than the actual slab, although the limited vertical resolution at upper levels and the general tendency toward downward smearing of structure in our inversion may account for much of this discrepancy. 59 CHAPTER 5. DISCUSSION OF RESULTS 60 In addition, extrapolating the minimal Benioff-zone information available from the Wrangell mountains region (which does not extend below 80 km depth) eastward into the Yukon may be a poor guide to the slab location (Stephens et al., 1995). The slab geometry in this region is likely to be complex, due to high strains in the subducted slab caused by the sharp corners in the North American plate at longitudes 149°W and 136°W, the collision of the Yakutat block (Bruns, 1983), and possible thermal erosion of the slab edge. Thus, it is possible that the slab lies lower than predicted by simple extrapolation of Benioff-zone contours. Figure 5.1: Depth contours of the upper boundary of the subducting Pacific plate in Alaska interpreted from Benioff-zone seismicity, from Stephens et al., 1984. With these factors in mind, we interpret the negative slowness anomaly to reflect the pres-ence of the edge of the descending slab. Although the poor vertical resolution in this area precludes us from drawing conclusions about the vertical position and dip of the slab, the ho-rizontal resolution is sufficient to place the edge of the the slab at approximately 1 4 1 ° W at CHAPTER 5. DISCUSSION OF RESULTS 61 60°N, trending approximately north-south. Based on reconstructions of past plate motions in this area (Engebretson et al., 1985), and estimates of current plate motions (Riddihough and Hyndman, 1991), the component of convergence between the North American and Pacific plates would place the edge of a rigid slab somewhat farther east than we have observed, although this statement is dependant on what assumptions are made about past plate boundary con-figurations. Therefore, our results suggest that the edge of the slab has not behaved rigidly. Additional support for the notion that the slab edge extends no farther than the 1 4 1 ° W merid-ian is provided by the absence of Neogene arc volcanism west of that line (Souther and Yorath, 1991). The large low-velocity feature is better resolved, as it lies in that portion of the model where, according to the resolution test, the horizontal and vertical resolutions are good. Considering its large size, and the 100 km resolution of the model in all directions in this region, the recovered shape of the anomaly is likely to be approximately correct (compare this to the large-spike resolution test, in figures 4.7 and 4.8). Therefore, this feature may be interpreted in more detail, and it will be important to ascertain its magnitude, and investigate its possible origin. The large low-velocity feature we observe reaches a maximum magnitude of 3%. This may be considered to be a,lower bound on the true anomaly, as the smoothing inherent in the inversion tends to reduce the amplitude of sharp features. The large-spike resolution test showed a drop from a 4% peak anomaly in the input model to a 2.5% peak anomaly in the recovered model, for the worst-case scenario of a spike-like feature. For smoother features the underestimation should be less. We can thus state with some confidence that the true anomaly is at least 3% of the IASP91 slowness, and may be as much as 5%. With this in mind, we can attempt to explain the magnitude of the anomaly in terms of compositional and thermal heterogeneity. If we assume a peridotite mantle composition, the magnitude of slowness anomaly expected due to compositional factors is limited. Lang (pers. comm.) calculated the difference in CHAPTER 5. DISCUSSION OF RESULTS 62 predicted P-wave velocity for two end-member mantle xenolith compositions from the northern Cordillera, and found that the maximum slowness perturbation obtainable was less than 1%. More generally, Sobolev et al. (1996) found that for peridotite mantle, velocity variations are quite small, seldom exceeding 1% even for fairly extreme peridotite compositions; similar results were obtained by Humphreys and Dueker (1994b). We therefore conclude that the low-velocity feature we have observed cannot be explained as a purely compositional feature, and that a thermal anomaly must be present. The interpretation of mantle images derived from seismic tomography in terms of temper-ature is an intricate problem. The most common method employed is to use experimentally derived temperature derivatives of seismic velocities. Such methods tend to produce unrealistic-ally large temperature anomalies ( « 4 0 0 ° C for a 3% anomaly) (Sobolev et al., 1996; Humphreys and Dueker, 1994a), which in most regions would indicate intersection with the solidus and the production of partial melt on a large scale. Sobolev et al. (1996) argued that this is due to the failure of such calculations to account for other temperature-dependent changes in the rock that occur over longer periods of time, particularly mineral reactions (e.g. the temperature-dependent partitioning of aluminum), anelasticity, and small degrees of partial melt of hydrous phases. Taking these factors into account, and employing chemical information obtained from mantle xenoliths, the authors determined that the 4% positive P-wave velocity perturbation measured in the experiment they examine (the French Massif Central) represents a 100-200° positive mantle temperature anomaly. Detailed mineral physics would be required to fully re-solve the matter, but we presume our results to be comparable to those of Sobolev et al. (1996), and so conclude that the 3% anomaly that we have observed requires less than 200° of thermal anomaly, and is probably closer to 100°, depending on the degree of compositional variability. CHAPTER 5. DISCUSSION OF RESULTS 63 5.2 Plume models for the observed low-velocity feature Deep-rooted low-velocity mantle anomalies have frequently been ascribed to mantle plumes. Mantle plumes, in general, are thought to consist of a narrow (100-200 km in diameter) con-duit of material which widens into a broader head only above 100-200 km depth (White and McKenzie, 1995; Griffiths and Campbell, 1990; Davies, 1994). Tomographic images of mantle plumes beneath the French Massif Central (Granet et al., 1995) and Yellowstone (Humphreys and Dueker, 1994a) have located anomalies of similar magnitude and shape (but lesser extent, both vertically and laterally) to the one described here, although neither of these experiments permitted the recovery of structure below 300 km. However, in both of the cases referenced above there are surface expressions of current plume activity, in the form of substantial hot-spot magmatism and uplift, which have not been observed in the northern Cordillera (Souther and Yorath, 1991). In addition, if a plume were currently present beneath the northern Cordillera, we would expect to see some indication of a plume track out to the northern B . C . coast, due to the westward motion of the North American continent. Although plume tracks are not necessarily visible above thick continental lithosphere (Davies, 1994), the plume should have affected the continental margin. We might expect such a track to continue onto the Pacific plate, curving northward due to plate motions, although the Pacific portion might well have been subducted (Engebretson et al., 1995). No such track is visible on either of the Pacific and North American plates. Finally, although resolution is limited, the general geometry of the observed anomaly does not match that expected for mantle plumes. The large, deep-rooted anomaly we have observed is considerably broader below 200 km than would be expected from such models. In particular, our anomaly is somewhat larger and considerably deeper than the feature observed at Yellowstone (Humphreys and Dueker, 1994a). Due to these objections, we do not favor the interpretation that the low-velocity anomaly we have observed represents a mantle plume beneath the northern Cordillera, although it cannot CHAPTER 5. DISCUSSION OF RESULTS 64 completely be ruled out. A second possibility is that this feature is a consequence of past plume activity. Precedent for this is given in Vandecar et al. (1995), who describe a 1.5-2% slowness anomaly located beneath the Parana Basin flood basalts of the Brazilian shield. These authors relate both the flood basalts and the anomaly to the eruption of the Tristan de Cunha plume in the Cretaceous. The anomaly they observe penetrates to 600 km (the bottom of their model), and is similar in appearance to the one located in this study. The Anahim Belt, a late Cenozoic chain of volcanic and plutonic rocks in central British Columbia, has been attributed to plume activity and so provides the nearest candidate plume; however, there are no surface rocks linked to this magma source in the northern Cordillera, and in any case, this plume would have been located a considerable distance out to sea at 20 M a , due to the westward motion of North America (Souther, 1986; Engebretson et al., 1995). The only nerby hotspot track is located offshore of the Queen Charlotte Islands, ending at the the Bowie seamount (Nataf and VanDecar, 1983; Batiza, 1989); however, it has probably never underlain the North American continent and so did not affect our region. Johnston et al. (1996) link the Carmacks group, a Late Cretaceous volcanic sequence, to the Yellowstone hotspot; the presence of flood basalts in the sequence suggests that it represents the initial effusion of the plume. The possibility then arises that the low-velocity anomaly we have detected is a preserved remnant of the Cretaceous Yellowstone plume head. There are, however, a number of serious objections to this hypothesis. First, the magnitude of the anomaly we observe is considerably greater than that described in Vandecar et al. (1995) (3% versus 1.5%), and is of comparable magnitude to those observed over currently active plumes (in the 2%-4% range) (Granet et al., 1995; Humphreys and Dueker, 1994a; Hoernle et al., 1995). In addition, while the Brazilian anomaly lies in a stable shield region, undisturbed by plate interactions since the Cretaceous, the northern Cordillera has experienced continuous CHAPTER 5. DISCUSSION OF RESULTS 65 activity over the last 100 Ma. In particular, it is difficult to see how any mantle anomaly could have survived the subduction of the Kula plate, as the Kula slab probably passed through the position of our anomaly and should have completely disrupted it. The anomaly we observe is therefore unlikely to be older than 45 or 50 Ma, and is probably much younger. It is primarily due to this last objection that we believe that the plume-remnant hypothesis can be rejected. 5.3 Plate-related models for the observed low-velocity feature There exist several possible sources for upper-mantle anomalies related to crust-mantle inter-actions and plate motions. The first stems from the widely accepted observation that orogenic belts are generally underlain by thinner lithosphere and hotter underlying mantle than other continental regimes. The "orogeny paradox" discussed by Silver and Chan (1991) refers to the difficulty in reconciling this observation with the expectation of an advectively thickened lithosphere in collisional zones. Delamination of continental lithosphere (Houseman et al, 1981) is a mechanism frequently called upon to resolve this apparent contradiction; however, the observed correlation between surface geologic fabric and lithospheric seismic anisotropy in oro-genic belts (Silver and Chan, 1991) implies that the mantle lithosphere beneath such belts is not removed. Kincaid and Silver (1996) attempt to account for lithospheric thinning and heating beneath orogenies as being the result of strain heating due to lithospheric collision, and find that such heating may produce considerable thermal anomalies above 100 km depth for mod-erately fast convergence rates, although such anomalies are found to be short-lived. Whatever the physical mechanism, the association between lithospheric zones of high deformation and low upper-mantle seismic velocities is widely accepted. In the case we are considering, lithospheric deformation is occurring between the Pacific and North American plates where they converge along the Alaska subduction zone, as well as to a lesser degree along the Fairweather-Queen Charlotte transform faults, where a component of CHAPTER 5. DISCUSSION OF RESULTS 66 east-west convergence exists (Engebretson et al., 1995; Riddihough and Hyndman, 1991; see also Russo and Silver, 1996). In this context, the eastern edge of the low-velocity anomaly, which lies subparallel to and between the Tintina and Denali faults, may, at shallower levels, mark the eastern boundary of deformation in the subcrustal lithosphere. Deformation is likely to be particularly severe at the corner in the North American plate boundary formed by the junction between the Alaska panhandle and mainland, as evidenced by the continuing uplift of the St. Elias Mountains (Gabrielse et al., 1991) above an underplating or subducting Yakutat block (Bruns, 1983). Termination of the positive slowness anomaly to the southeast between stations W H Y and D L B C would tend to support this, as it indicates concentration of the anomaly in this corner. However, it is unlikely that lithospheric heating, whatever its mechanism, is a sufficient explanation for the vertical extent of the low-velocity anomaly that we have detected, and therefore some other mechanism must be called upon to explain the heating of the upper mantle below 200 km depth. Our preferred model for the low-velocity feature is upwelling related to the presence of a slab window in the northern Cordillera. A slab window is generally denned as being the gap between two slabs formed by the subduction of a spreading ridge (Dickinson and Snyder, 1979). When the ridge encounters the trench, growth of the spreading plates ceases in the subducted portions. As the plates at the surface continue to separate, a widening gap opens in the descending slab, its geometry dependent on the spreading rate, subduction rate, angle of subduction and angle between the ridge and the trench (figure 5.2). The geometry of slab windows is discussed in some detail in Thorkelson (1996), while slab windows beneath the North American Cordillera are discussed in Thorkelson and Taylor (1989). Two are expected to exist in present times (figure 5.3), one beneath Californa south of the Mendocino Triple Junction, the other beneath B . C . and the Yukon between the northern end of the Explorer Plate and the Alaska subduction zone. Our area of interest lies in the northern half of the B.C.-Yukon slab window; the slab CHAPTER 5. DISCUSSION OF RESULTS 67 edge that we have imaged corresponds well with the expected western edge of the window. Figure 5.2: Geometry of a slab window, from Thorkelson (1996). It is argued that mantle upflow is to be expected along the edges of slab windows (Thorkel-son, 1994; Hole et al., 1991; Thorkelson, 1996). The retreat of the slab decompresses the underlying asthenosphere, which rises to fill the gap; this brings hotter material into the slab window, and the decompression may lead to partial melt and alkalic volcanism. This ties in well with our observations: the low-velocity anomaly parallels the slab edge, and peters out farther into the proposed window. As well, the recent low-volume volcanism in this region (Souther and Yorath, 1991; Souther, 1991) is suggestive. These volcanic rocks, generally asso-ciated with normal faults, occur throughout the northern Cordilleran slab window, and have alkalic compositions indicative of an undepleted mantle source (Souther and Yorath, 1991). They have generally been tied to young extension in the region, but could well be associated with slab-window decompression melting; however, such Neogene volcanism is not restricted to the edge of the slab window. Although little geophysical work on the topic of slab windows has been performed (in particular, the geodynamics of slab window upflow have not been worked out), there is support for this hypothesis in the form of a teleseismic tomography study of the northern end of the CHAPTER 5. DISCUSSION OF RESULTS 68 Figure 5.3: History of slab windows in the North American Cordillera from the Late Cretaceous to the present. From Thorkelson and Taylor (1989). CHAPTER 5. DISCUSSION OF RESULTS 69 southern Cordilleran slab window (Benz et al., 1992). Although the geometry of this window is considerably different, the authors noted a negative velocity anomaly of up to 5.5% along the window edge, reaching depths of up to 170 km (in a model with a maximum depth of 250 km). Thus, there is precedent for the tomographic observation of a low-velocity feature along the edge of a slab window, albeit not to the depths to which we have resolved this type of structure. The principal difficulty with this type of model is that the depth to which the withdrawal of a slab would produce a significant thermal anomaly is likely quite limited. Withdrawing the « 5 0 km thick slab imaged by Zhao et al. (1995) seems unlikely to be sufficient to account for a thermal anomaly below 150 or 200 km depth. This difficulty may perhaps be resolvable by an appeal to ridge dynamics. Consistent lateral motion of the slab due to the opening of the window could produce circulation in the upper mantle (figure 5.4); the scale of such circulation would be comparable to the width of the moving slab, and so could affect mantle below 600 km (Hager and O'Connell, 1979). This is borne out by Grand (1994), who imaged a low-velocity mantle anomaly beneath the Mid-Atlantic Ridge to a depth of 400 km. Due to the mild component of east-west convergence between North America and the Pacific plate, the slab window is unlikely currently to be widening at a significant rate; however, the slab window has opened considerably in the past 20 M a (Thorkelson and Taylor, 1989), requiring considerable lateral slab motion over that period, which may have been enough to produce the required circulation. Changes in slab dip may also have contributed to such mantle circulation, in that steepening of slab dip angles would require upflow to accomodate the enlarged mantle wedge above the slab (Thorkelson and Taylor, 1989). The hypothesis that the low-velocity anomaly we have observed represents slab window-related upflow is quite speculative and needs to be tested by further work; however, we favor it as being the model that best explains our the deep-seated nature of the positive slowness anomaly. Other aspects of plate-mantle interaction may affect the model we see. The unusually sharp CHAPTER 5. DISCUSSION OF RESULTS 70 northeastern edge of the low-velocity anomaly at shallow depths is a well-resolved feature that demands explanation. Its trend is roughly parallel to that of the Cordilleran deformation front, somewhat outboard of the edge of the ancestral North American miogeocline (figure 1.2). A possible hypothesis for the sharp edge we see is that it represents the edge of thicker, older, and colder lithosphere beneath ancestral North America, blocking the eastward flow of warmer material welling up along the slab edge. The more diffuse southeastern boundary of the anomaly in this context would then represent the limit of the slab's influence on upper-mantle circulation. Finally, the presence of a branch of the low-velocity anomaly extending beneath the inferred slab suggests that the flow in this region may be more complex than can be explained by simple geometric models. (" Overlying plate 3 ( •<— Slab ) ( Slab — • ) Figure 5.4: A conceptual model of flow due to the opening of a slab window. This pattern of flow would be superimposed on the downflow generated by subduction. 5.4 Summary We have located two significant upper-mantle P-wave velocity anomalies beneath the northern Cordillera. We have interpreted the first, a high-velocity, relatively shallow region at the extreme west of our model, as reflecting the presence of the edge of the subducting Pacific slab, its position having been smeared downward due to the poor resolution at the edges of the model. CHAPTER 5. DISCUSSION OF RESULTS 71 The second anomaly we have located, a large deep-rooted low-velocity zone in the best-resolved portion of our model, is more difficult to interpret. We suggest that it is a positive thermal anomaly of less than 200°C, possibly with some associated chemical anomaly as well. The most obvious source for such an anomaly is a mantle plume; however, their are no magmatic indications of a plume currently in this region, nor is there the plume track we would expect to find leading to it. Rather, the late Cenozoic intraplate volcanics in this area are widely scattered over much of the Yukon and British Columbia. The possibility that this anomaly represents the remnant of an older plume carried away by plate motions has been considered; however, it is improbable that such a deep-rooted anomaly could have survived the subduction of the Kula plate, and remained coupled to a crustal sliver which would have to have been transported northward by over 1000 km. We therefore favor the hypothesis that this low-velocity feature represents a combination of upwelling along the edge of the northern Corclilleran slab window, perhaps due to circulation produced by past lateral motion of the Pacific slab or other interactions between the slab edge and the asthenosphere, and a contribution at the upper levels from strain heating beneath the St. Elias Mountains. Such a combination of effects would explain the southwest-northeast orientation of the anomaly, as well as its position. Such upwelling has been previously detected in the southern Cordilleran slab window (near the Mendocino triple junction) (Benz et al., 1992); the presence of Neogene volcanic rocks of intraplate character throughout the area of the northern Cordilleran slab window as defined in Thorkelson and Taylor (1989) is suggestive of window-related extension and melting. However, in order to confirm this hypothesis, there re-mains both theoretical and experimental work to be done. The geodynamics of such slab-edge upwelling has not been worked out; although there are strong qualitative arguments for the existence of such upwelling, we know little about its magnitude and depth penetration. Exper-imentally, heat-flow data from above the anomaly may better constrain its thermal magnitude; CHAPTER 5. DISCUSSION OF RESULTS 72 as well, the location and configuration of the edge of the Pacific slab is poorly constrained, and could be determined through a more detailed tomographic experiment. Bibliography Atwater, T . (1989). Plate tectonic history of the northeast Pacific and western North America, in The Eastern Pacific Ocean and Hawaii, edited by E . L . Winterer, Donald M . Hussong, and Robert W . Decker, 21-72. Batiza, R (1989). Seamounts and seamount chains of the eastern Pacific, in The Eastern Pacific Ocean and Hawaii, edited by E . L . Winterer, Donald M . Hussong, and Robert W . Decker, 289-306. Benz, H . M . , G . Zandt, and D . H . Oppenheimer (1992). Lithospheric structure of northern California from teleseismic images of the upper mantle, J. Geophys. Res 97, 4791-4807. Bruns, T .R . (1983). Model for the origin of the Yakutat block, an accreting terrane in the northern Gulf of Alaska, Geology 11, 718-721. Creager, K . C . and T . H . Jordan (1984). Slab penetration into the lower mantle, J. Geophys. Res 89, 3031-3049. Davies, G . F . (1994). Thermomechanical erosion of the lithosphere by mantle plumes, J. Geophys. Res 99, 15,709-15,722. Dickinson, W.R. and W.S. Snyder (1979). Geometry of subducted slabs related to the San Andreas transform, J. Geol. 87, 609-627. Edwards, B.R., T.S. Hamilton, J . Nicholls, M . Z . Stout, J . K . Russell, and K . Simpson (1996). Late Tertiary to Quaternary volcanism in the Atlin area, northwestern British Columbia, Current Research 1996-A, Geolo-gical Survey of Canada, 29-36. Engebretson, D . C . , A . Cox, and R . G . Gordon (1985). Relative motions between pceanic and continental plates in the Pacific basin, Geological Society of America Special Paper 206, 59 pp. Erdmer, P. and J . K . Mortensen (1993). A 1200-km-long Eocene metamorphic-plutonic belt in the northwestern Cordillera: evidence from southwest Yukon, Geology 21 1039-1042. Gabrielse, H . , J . W . H Monger, J .O. Wheeler, and C.J . Yorath (1991). Tectonic framework, in Geology of the Cordilleran Orogen in Canada, edited by H . Gabrielse and C.J . Yorath, 15-28. 73 BIBLIOGRAPHY 74 Gorman, A .R. , R . M . Clowes, R . M . Ellis, M . J . A . Burianyk, E . R . Kanasewich, Z. Hajnal, G.D. Spence, G.R. Keller, and A . R . Levander (1995). Seis-mic refraction imaging of the crust and upper mantle in the Archean Hearne Craton, Southern Alberta, Canada, for 1995 Fall meeting of the American Geophysical Union. Grand, S.P. (1994). Mantle shear structure beneath the Americas and surround-ing oceans, J. Geophys. Res 99, 11,591-11,621. Granet, M . , M . Wilson, and U . Achauer (1995). Imaging a mantle plume be-neath the French Massif Central, Earth Planet. Sci. Let. 136, 281-296. Griffiths, R . W . and I.H. Campbell (1990). Stirring and structure in mantle starting plumes, Earth Planet. Sci. Let. 99, 66-78. Hager, B . H . and R . J . O'Connell (1979). Kinematic models of large-scale flow in the Earth's mantle, / . Geophys. Res 84, 1031-1048. Hoernle, K , Y . Zhang, and D. Graham (1995). Seismic and geochemical evid-ence for large-scale mantle upwelling beneath the eastern Atlantic and western and central Europe, Nature 374, 34-39. Hole, M . J . , G . Rogers, A . D . Saunders, and M . Storey (1991). Relation between alkalic volcanism and slab window formation, Geology 19, 657-660. Houseman, G . A . , D.P. McKenzie, and P. Molnar (1981). Convective instability of a thickened boundary layer and its relevance for the thermal evolution of continental convergent belts, J. Geophys. Res 86, 6115-6132. Humphreys, E . D . and K . G . Dueker (1994a). Western U.S. upper mantle struc-ture, J. Geophys. Res 99, 9615-9634. Humphreys, E . D . and K . G . Dueker (1994b). Physical state of the western U.S. upper mantle, J. Geophys. Res 99, 9635-9650. Irving, E . , D .J . Thorkelson, P . M . Wheadon, and R.J . Enkin (1995). Paleomag-netism of the Spences Bridge Group and northward displacement of the Intermontane Belt, British Columbia: a second look, J. Geophys. Res 100, 6057-6071. Johnston, S.T., P.J. Wynne, D. Francis, C.J.R. Hart, R .J . Enkin, and D . C . En-gebretson (1996). Yellowstone in Yukon: the Late Cretaceous Carmacks group, Geology, in press. Kennett, B . L . N , and E .R . Engdahl (1991). Traveltimes for global earthquake location and phase identification. Geophys. J. Int 105, 429-465. Kincaid, C . and P. Silver (1996). The role of viscous dissipation in the orogenic process, Earth Planet. Sci. Let. 142, 271-288: BIBLIOGRAPHY 75 Lowe, C , R.B. Horner, J . K . Mortensen, S.T. Johnston, and C F . Roots (1994). New geophysical data from the northern Cordillera: preliminary inter-pretations and implications for the tectonics and deep geology, Can. J. Earth Sci. 31, 891-904. Lowe, C. and J . F . Cassidy (1995). Geophysical evidence for crustal thickness variations between the Denali and Tintina Fault Systems in west-central Yukon, Tectonics 14, 909-917. Monger, J . W . H . and W . J . Nokleberg (1996). Evolution of the North Amer-ican Cordillera: generation, fragmentation, displacement and accretion of successive North American plae-margin arcs, for 1995 Geological Soci-ety of Nevada Symposium on Geology and Ore Deposits of the American Cordillera. Nataf, J . C . and J . VanDecar (1993). Seismological detection of a mantle plume?, Nature 364, 115-120. Page, R .A . , C D . Stephens, and J . C . Lahr (1989). Seismicity of the Wrangell and Aleutian Wadati-Benioff zones and the North American plate along the Trans-Alaska Crustal Transect, Chugach Mountains and Copper River Basin, southern Alaska, J. Geophys. Res 94, 16,059-16,082. Parker, R . L . (1994). Geophysical Inverse Theory, Princeton University Press, 386 pp. Riddihough, R..P. and R.D. Hyndman (1991). Modern plate tectonic regime of the continental margin of western Canada, in Geology of the Cordilleran Orogen in Canada, edited by H . Gabrielse and C . J . Yorath, 435-455. Russo, R . M . and P .G. Silver (1996). Cordillera formation, mantle dynamics, and the Wilson cycle, Geology 24, 511-514. Ryberg, T . , F . Wenzel, J . Mechie., A . Egorkin, K . Fuchs, and L . Solidov (1996). Two-dimensional velocity structure beneath northern Eurasia derived from the super long-range seismic profile Quartz, Bull. Seis. Soc. Am. 86, 857-867. Scales, J .A . and A . Gersztenkorn (1988). Robust methods in inverse theory, Inverse Problems 4, 1071-1091. Silver, P .G. and W . W . Chan (1991). Shear wave splitting and subcontinental mantle deformation, / . Geophys. Res 96, 16,429-16,454. Skulski, T . , Francis, D. , and Ludden, J . (1991). Arc-transform magmatism in the Wrangell volcanic belt, Geology 19, 11-14. BIBLIOGRAPHY 76 Sobolev, S.V., H . Zeyen, G . Stoll, F . Werling, R. Altherr, and K . Fuchs (1996). Upper mantle temperatures from teleseismic tomography of French Mas-sif Central including effects of composition, mineral reactions, anhar-monicity, anelasticity and partial melt, Earth Planet. Sci. Let. 139, 147-163. Souther, J . G . (1986). The western Anahim Belt: root zone of a peralkaline magma system, Can. J. Earth Sci. 23, 895-908. Souther, J . G . (1991). Volcanic regimes, in Geology of the Cordilleran Orogen in Canada, edited by H . Gabrielse and C.J . Yorath, 457-490. Souther, J . G . and C . J . Yorath (1991). Neogene assemblages, in Geology of the Cordilleran Orogen in Canada, edited by H . Gabrielse and C . J . Yorath, 373-401. Stephens, C D . , K . A . Fogleman, J . C . Lahr, and R . A . Page (1984). Wrangell Benioff zone, southern Alaska, Geology 12, 373-376. Stock, J . and P. Molnar (1988). Uncertainties and implications of the late Cre-taceous and Tertiary position of North America relative to the Farallon, Kula, and Pacific plates, Tectonics 7, 1339-1384. Tapley, W . C and J . E . l u l l (1992). SAC Users Manual, Revision 4, Lawrence Livermore National Laboratory, 52 pp. Thorkelson, D . J . (1994). Ridge subductidn: kinematics and implications for the nature of mantle up welling: discussion, Can. J. Earth Sci 31,1486-1489. Thorkelson, D .J . (1996). Subduction of diverging plates and the principles of slab window formation, Tectonophysics 255, 47-63. Thorkelson, D . J . and R.P. Taylor (1989). Cordilleran slab windows, Geology 17, 833-836. Umhoefer, P.J. , J . Dragovich, J . Cary, and D . C Engebretson (1989). Refine-ments of the "Baja British Columbia" plate-tectonic model for north-ward translation along the margin of western North America, in Deep Structure and Past Kinematics of Accreted Terranes, edited by John W . Hillhouse, 101-111. VanDecar, J . C . (1991). Upper-Mantle Structure of the Cascadia Subduction Zone from Non-Linear Teleseismic Travel-Time Inversion, Ph.D. thesis, Univ. of Washington, Seattle, 165 pp. VanDecar, J . C . and R.S. Crosson (1990). Determination of teleseismic relat-ive phase arrival times using multi-channel cross correlation and least squares, Bull. Seis. Soc. Am. 80, 150-169. BIBLIOGRAPHY VanDecar, J . C , D . E . James, and M . Assumpcao (1995). Seismic evidence for coherent flow of the crust and upper mantle below South America since the breakup of Gondwana, Nature 378, 25-31. Von Huene, R. (1989). Continental margins around the Gulf of Alaska, in The Eastern Pacific Ocean and Hawaii, edited by E . L . Winterer, Donald M . Hussong, and Robert W . Decker, 383-401. White, R.S. and D . McKenzie (1995). Mantle plumes and flood basalts, J. Geophys. Res 100, 17,543-17,585. Zhao, D . , D . Christensen, and H . Pulpan (1995). Tomographic imaging of the Alaska subduction zone, J. Geophys. Res 100, 6487-6504. Appendix A List of earthquakes used in this experiment The following is a list of the teleseismic earthquakes recorded during this experiment that were used in the travel-time inversion. The event times and locations given are preUminary. 94/01/05 13-24: 09 39.10N 15. 10E 273 0 5 7 77.29 23.42 94/05/23 06 46: 16 35.60N 24. 70E 76 0 6 0 82.34 16.62 94/05/26 08 26:52 35.30N 4. 10W 10 0 5 7 75.95 39.57 94/05/31 17 41- 55 7.40N 72. OOW 12 0 6 3 70.48 110.58 94/07/01 10 12 41 40.20N 53.40E 41 0 6 0 78.88 353.58 94/07/01 19 50 04 40.20N 53.40E 44 0 5 6 78.89 353.57 94/08/18 01 13 05 35.50N 0 10W 9 0 5 7 76.96 36.38 95/03/08 03 45 58 16.60N 59 60W 8 0 6 3 68.44 94.45 95/04/17 07 14 35 33.80N 38 60W 10 0 5 8 63.91 66.94 95/05/13 08 47 12 40.20N 21 70E 14 0 6 2 77.38 18.15 95/06/15 00 15 48 38.40N 22 30E 14 0 6 .0 79.19 18.05 95/07/21 22 44 08 36.40N 103 30E 33 5 .6 GANSU, CHINA [MCBB] 95/07/24 19 13 21 55.51N 35 18W 10 .0 5 .4Mb A NORTH ATLANTIC OCEAN 95/07/25 15 13 25 10.72N 40 99W 10 .0 5 .5Ms B NORTHERN MID-ATLANTIC RIDGE 95/07/26 23:41 58 2.50N 127 80E 33 6 .1 NORTHERN MOLUCCA SEA [MHBE] 95/07/28 14 29 04 21.00S 175 40W 33 6 .2 TONGA ISLANDS [MJAS] 95/07/29 16 18 44 30.40N 138 30E 433 5 .4 SOUTH OF HONSHU, JAPAN 95/07/30 05 11 21 23.40S 70 10W 33 7 .8 NEAR COAST OF NORTHERN CHILE 95/07/30 11:51 18 29.30N 129 40E 33 5 .5 RYUKYU ISLANDS [MLAY] 95/07/31 08 48 33 10.40S 77 80W 93 5 .7 NEAR COAST OF PERU [MMAK] 95/08/03 08 .18 53 28.10S 68 90W 103 5 .8 LA RIOJA PROVINCE, ARGENTINA 95/08/06 11 :59 29 44.42N 147 .22E 33 .0 5 .4Mb KURIL ISLANDS 95/08/14 04 :37 05 4.64S 151 .10E 33 .0 6 .5Mb A NEW BRITAIN REGION, P.N.G. 95/08/16 10 : 27:28 5.73S 154 .09E 33 .0 7. 8M B B SOLOMON ISLANDS 95/08/16 16 :24 :28 5.27S 153 .76E 33 .0 6 .8M B B NEW IRELAND REGION, P.N.G. 95/08/16 20 :58 :57 29.13N 128 .80E 33 .0 5 .5Ms B NORTHWEST OF RYUKYU ISLANDS 95/08/16 23 :10 :27 5.75S 154.09E 70 .6 7 .2MB A SOLOMON ISLANDS 95/08/17 00 :15 :52 5.88S 153 .95E 33 .0 6 .6Ms B NEW IRELAND REGION, P.N.G. 95/08/17 00 :59 :57 41.57M 88 .73E 0 .0 6 .1Mb A SOUTHERN XINJIANG, CHINA 95/08/17 10:01 :27 5.01S 153 .32E 33 .0 6 .4M B B NEW IRELAND REGION, P.N.G. 95/08/17 23 :14 :19. 36.40N 71 .30E 238 .0 5 .5 AFGHANISTAN-TAJIKISTAN BORD RE 78 APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 79 95/08/18 19: 07: 28 18. 94N 144. 89E 500 0 5. 0Mb B MARIANA ISLANDS 95/08/19 21: 43: 31 4. 99N 75. 65W 126 3 6. 2Mb A COLOMBIA 95/08/23 07: 06: 02 18.89N 145. 16E 597 0 6. 1Mb A MARIANA ISLANDS 95/08/24 01: 55: 34 18. 96N 144. 89E 587 4 5.7Mb A MARIANA ISLANDS 95/08/24 07:54: 41 18. 92N 144. 77E 600 0 5. 2Mb A MARIANA ISLANDS 95/08/25 14: 25: 25 20. 04S 178. 33W 540 7 5. 1Mb A FIJI ISLANDS REGION 95/08/25 16: 51:46 18. 51S 175. 61W 226 6 4. 9Mb B TONGA ISLANDS 95/08/26 06:57 18 5. 64S 153. 53E 33 0 6. 1MB B NEW IRELAND REGION, P.N.G. 95/08/26 14 47 50 8. 33N 127. 34E 33 0 5. 4M B B PHILIPPINE ISLANDS REGION 95/08/28 10:46 10 25. 92N 110 32W 10.0 6. 3Ms A GULF OF CALIFORNIA 95/08/29 08 51 33 20.79S 174 55W 33 0 5. 7Ms B TONGA ISLANDS 95/08/29 13 06 40 2 33N 127 51E 33 0 5. 3Mb B NORTHERN MOLUCCA SEA 95/08/31 17 10 36 15 87S 166 16E 33 0 6. 5Ms A VANUATU ISLANDS 95/09/01 05 18 04 13 33S 74. 61W 109 3 5. 1Mb B CENTRAL PERU 95/09/01 06 30 41 0 07S 123 51E 190 9 5 2Mb B MINAHASSA PENINSULA, SULAWESI 95/09/06 22 48 51 14 96N 94 09W 33 0 5 4M B A OFF COAST OF CHIAPAS, MEXICO 95/09/08 17 25 49 14 93N 94 15W 33 0 5 5Mb A OFF COAST OF CHIAPAS, MEXICO 95/09/11 04 22 52 0 97N 101 36W 10 0 5 2Mb A EAST CENTRAL PACIFIC OCEAN 95/09/11 11 40 30 53 79N 160 40E 18 2 5 2Mb B NEAR EAST COAST OF KAMCHATKA 95/09/14 14 04 31 16 79N 98 61W 21 3 7 2M B NEAR COAST OF GUERRERO, MEXICO 95/09/15 23 52:03 43 03K 143 73E 106 2 5.0Mb B HOKKAIDO, JAPAN REGION 95/09/16 01 03 37 6 14S 154 80E 159 6 5 7Mb B SOLOMON ISLANDS 95/09/18 20 22 13 20 33S 179 10W 617 7 4 9Mb B FIJI ISLANDS REGION 95/09/19 03 31 56 20 71S 68 74W 108 .2 5 8Mb C CHILE-BOLIVIA BORDER REGION 95/09/20 23 27 35 35 71H 117 63W 5 .0 5 5Mb A CENTRAL CALIFORNIA 95/09/22 05 39 28 5.92S 146 42E 33 .0 6 OMs A E NEW GUINEA REG., P.N.G. 95/09/23 02 34 13 5 80S 146 30E 33 .0 5 9Ms A E NEW GUINEA REG., P.N.G. 95/09/23 20 56 03 24.36S 128 OOW 10 .0 5 4Mb A SOUTH PACIFIC OCEAN 95/09/23 22 31 56 10 64S 78 25W 70 .0 6 2Mb A NEAR COAST OF PERU 95/09/26 04 39 07 41 89N 81 54E 50 .9 5.4Mb A SOUTHERN XINJIANG, CHINA 95/09/26 07 14:37 41 79N 143.38E 33 .0 6.0Mb A HOKKAIDO, JAPAN REGION 95/09/26 18 •24:12 13 15S 166 81E 186 .6 5.4Mb B VANUATU ISLANDS 95/10/01 15 :57 :15 38 03N 30 15E 33 .0 6 OMs A TURKEY 95/10/01 17 :06:03 29 .30N 139.04E 423 .6 5.6Mb A SOUTH OF HONSHU, JAPAN 95/10/01 23 :29 :57 22 .30S 138 .83W 0 .0 5 5Mb A TUAMOTU ARCHIPELAGO REGION 95/10/02 23 :48 :22 15 .28S 175 .01W 33 .0 5 9Ms A TONGA ISLANDS 95/10/03 01:51 :25 2 .70S 77 86W 33 .0 6 9Ms A PERU-ECUADOR BORDER REGION 95/10/03 12 :44 :59 2 .78S 77 79W 33 .0 6 .1MB A PERU-ECUADOR BORDER REGION 95/10/06 05 :13 :24 18 .88N 104 .18W 33 .0 5 .2MB A NEAR COAST OF JALISCO, MEXICO 95/10/06 11 :39 :36 19 .82S 176 .14W 209 .7 5 .5Mb A FIJI ISLANDS REGION 95/10/07 21 :28 :05 2 .73S 77 .70W 33 .0 5 .5Mb B PERU-ECUADOR BORDER REGION 95/10/08 08 :55:48 40 .96N 72 .06E 33 .0 6 .0Mb A KYRGYZSTAN 95/10/09 13:43:42 21 .38S 169 .86E 117 .8 5.6MS B LOYALTY ISLANDS REGION 95/10/09 15 :35 :51 18 .86N 104 .14W 33 .0 7 .6Ms A NEAR COAST OF JALISCO, MEXICO 95/10/12 16 :52 :56 18 .92N 103 .97W 33 .0 5 .5Ms A NEAR COAST OF MICHOACAN, MEX 4.6Mb 95/10/14 08:00 :41 25 .73S 177 .60W 71 .6 5 .9Mb A SOUTH OF FIJI ISLANDS 95/10/15 15 :04 :13 6.49S 154 .38E 50 .9 6 .OMs B SOLOMON ISLANDS 95/10/18 09 :30 :38 36 .34N 70 .32E 222 .8 5 .7Mb A HINDU KUSH REGION, AFGHANISTAN 95/10/18 10 :37:24 28 .14N 130.38E 10 .0 6.9Ms A RYUKYU ISLANDS 95/10/19 00 :32 :03 28 .37N 130 .25E 10 .0 6 .4M B A RYUKYU ISLANDS 95/10/19 02:41 :36 28 .41N 130 .64E 10 .0 6 .9Ms A RYUKYU ISLANDS APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 95/10/20 19: 21: 28 18. 87N 145. 12E 225.4 5. 3Mb B MARIANA ISLANDS 95/10/21 02: 38: 49 16. 57N 93. 38W 100. 0 6. 3Mb A CHIAPAS, MEXICO 95/10/23 22: 46: 54 25. 88N 102. 25E 33. ,0 6. 4Ms A YUNNAN, CHINA 95/10/26 04: 23:25 39. 17N 72.04E 33. ,0 5. 4Mb B KYRGYZSTAN 95/10/27 21:59:57 21. 95S 139. 19W 0. ,0 5. 5Mb A TUAMOTU ARCHIPELAGO REGION 95/10/28 14: 38: 33 6. 33S 154. 33E 33. .0 5. 8Mb B SOLOMON ISLANDS 95/10/29 06: 27: 20 39. 64N 51. 79E 33. .0 5. 7Mb A CASPIAN SEA 95/10/29 18: 44: 21 0. 78N 126. 07E 33. .0 5. 6Mb B NORTHERN MOLUCCA SEA 95/10/29 19: 24: 29 0. 77N 125. 99E 33. .0 6. 1Mb A NORTHERN MOLUCCA SEA 95/10/29 19:40: 56 21.58S 179. 68W 600. ,0 5.5Mb A FIJI ISLANDS REGION 95/11/01 00: :35: 33 28; 95S 71. 18W 33. .0 6. 3Ms A NEAR COAST OF CENTRAL CHILE 95/11/01 09: :35: 59 28. 89N 130. 06E 33. .0 5. 6Mb B RYUKYU ISLANDS 95/11/01 12: :29: 28 42. 91N 80. 29E 33. .0 5. ,5Mb B KYRGYZSTAN-XINJIANG BDR REG. 95/11/02 22: :13: 45 6. 74S 130. 42E 100, .0 5. ,7Mb B BANDA SEA 95/11/08 16:01: 20 1. 33N 121. 67E 33, .0 5. 6Ms B MINAHASSA PENINSULA, SULAWESI 95/11/09 05: :10: 31 35. 57N 59. 90E 33 .0 5. 4Mb B NORTHERN IRAN 95/11/13 02: :17: 51 3. 60N 126. .77E 33 .0 6. 2M B B TALAUD ISLANDS, INDONESIA 95/11/13 07: :38: :45 14. 96S 173. ,60W 33 .0 5. 9Ms B SAMOA ISLANDS REGION 95/11/13 08: :43: :14 55. 96N 114. 58E 24 .3 5. 9Mb A EAST OF LAKE BAYKAL, RUSSIA 95/11/14 04:01: :45 5. 92S 150. ,43E 33 .0 5. 6Mb A NEW BRITAIN REGION, P.N.G. 95/11/14 15: :14: :02 18. 99N 144. .69E 600 .0 4. .5Mb B MARIANA ISLANDS 95/11/17 02: :12: :36 6. 78N 72. 05W 156 .1 4. 8Mb B NORTHERN COLOMBIA 95/11/18 16: :24: :32 46. 52N 150. .16E 100 .0 5. ,0Mb B KURIL ISLANDS 95/11/22 04: :15: i l l 28. .54N 34. ,75E 10 .0 7. .2MB B EGYPT 95/11/23 04: :41: :44 41. .50N 142, .51E 33 .0 5. .0Mb A HOKKAIDO, JAPAN REGION 95/11/24 17: :24: :12 44. ,39N 149, .13E 33 .0 6. .4MB A KURIL ISLANDS 95/11/25 13: :24: :01 44.39N 149 ,17E 33 .0 5. .0Mb B KURIL ISLANDS 95/11/26 03: :04:04 12. • 87S 166.27E 33 .0 5, .8Mb C SANTA CRUZ ISLANDS 95/11/27 15: :52: :58 44. 52N 149 .27E 33 .0 6 .OMs A KURIL ISLANDS 95/11/29 18:40: :36 16. • 72S 176 .62W 371 .9 5 .1Mb B FIJI ISLANDS REGION 95/11/30 15:09: :23 44. .22N 145 .66E 145 .9 6 .1Mb A HOKKAIDO, JAPAN REGION 95/11/30 23 :37: :36 44.17N 149 .37E 33 .0 6.0MS A KURIL ISLANDS 95/12/01 05 :20 :27 10.06N 104 .04W 10 .0 6 .2MB A OFF COAST OF MEXICO 95/12/01 10 :30 :12 44. .42N 149.50E 33 .0 5 .0Mb B KURIL ISLANDS 95/12/02 17 :13 :21 44. .81N 149 .22E 33 .0 6 .5Ms A KURIL ISLANDS 95/12/02 17 :28 :16 44, .50N 149 .23E 33 .0 5 .4Mb B KURIL ISLANDS 95/12/02 19 :40 :10 44, .32N 149 .57E 33 .0 5 .4Mb A KURIL ISLANDS 95/12/03 18 :01 :09 44 .85N 149 .42E 33 .0 8 .OMs A KURIL ISLANDS 95/12/03 18 :14:27 45 .01N 150.66E 33 .0 6 .3Mb B KURIL ISLANDS 95/12/03 21 :38:36 44 .60N 150 .16E 33 .0 5 .5Mb B EAST OF KURIL ISLANDS 95/12/05 14 :54 :44 1 .65N 127 .31E 100 .0 5 .6Mb B HALMAHERA, INDONESIA 95/12/05 18 :49 :31 39 .19N 40.41E 33 .0 5.6Mb B TURKEY 95/12/06 23 :17 :20 44 .23N 149 .39E 33 .0 5 .6Mb B KURIL ISLANDS 95/12/07 03 :22 :03 44 .36N 149 .38E 33 .0 5 .7Mb A KURIL ISLANDS 95/12/07 05 :12 :22 44 .43N 149 .38E 33 .0 5 .8Mb A KURIL ISLANDS 95/12/07 10 :04 :13 19 .94S 168 .55E 33 .0 5 .7Ms A VANUATU ISLANDS 95/12/07 18 :00 :55 34 .81N 24 .06E 33 .0 5 .3Mb A CRETE 95/12/07 19:30 :24 44.68N 149 .57E 33 .0 5 .9Mb A KURIL ISLANDS 95/12/08 07 :41 :14 72.53K 2 .38E 10 .0 5.3Mb B NORWEGIAN SEA 95/12/10 03 :27 :49 34 .78N 24 .HE 33 .0 5 .3Mb A CRETE 95/12/10 22 :23 :15 44 .42N 149 .76E 33 .0 6 .4M B B KURIL ISLANDS APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 81 95/12/10 22: 48: 08 44.27N 149 89E 33 0 5.5Mb B 95/12/10 23: 47: 00 21.24S 178 40W 412 4 5.8Mb B 95/12/11 05:22: 46 64.47N 17.92W 10 0 4.9Mb B 95/12/11 06: 17:22 51.08N 157 69E 33 0 5.2Mb B 95/12/11 14:09: 21 18.50N 105 46W 33 0 6. 1MB A 95/12/17 23- 48:31 52.62N 32 08W 10 0 5.3Mb B 95/12/18 02 05: 58 52.65N 142 75E 33 0 5.1Mb B 95/12/19 23 28: 13 3.53S 140 22E 71 3 6.3MB B 95/12/20 09:08: 43 42.88N 145 63E 33 0 5.2Mb B 95/12/20 11 39:20 27.85N 128 26E 33 0 5.3Mb B 95/12/21 07 54 46 12.38S 166 67E 241 8 5.0Mb B 95/12/21 20 38 28 11.67K 86 35W 100 .0 5.2Mb B 95/12/22 22 54:19 15.45S 68 85W 244 .4 5.2Mb A 95/12/25 03 19 44 36.41N 70 41E 228 .0 5.4Mb B 95/12/25 04:43 26 6.36S 129 23E 150 .0 6.7Ms B 95/12/26 20 10 35 47.42N 148 25E 365 .4 4.9Mb B 95/12/26 21:55 15 5.80S 150 43E 33 .0 5.5Mb B 95/12/29 13 01:40 9.91N 70 13W 33 .0 5.5Mb A 95/12/29 14 36 50 14.34N 59 89W 33 .0 5.2Mb A 95/12/30 03 26 08 4.64S 104 66W 10 .0 5.7MB B 95/12/30 12 11 07 40.96N 143 18E 33 .0 6.2MB B 95/12/30 12 17 37 40.89N 143 28E 33 .0 5.5Mb B 95/12/30 12 23 15 25.80N 125 79E 33 .0 5.6Mb C 95/12/30 16 15 32 31.42N 140 10E 100 .0 5.5Mb B 96/01/02 06:41 04 18.69S 69 08W 107 .0 5.3Mb B 96/01/02 18:09 10 8.96N 126 22E 33 .0 5.5Mb C 96/01/06 15 48 32 4.08S 151 96E 100 .0 5.3Mb B 96/01/07 13 14 29 6.81S 155 55E 33 .0 5.5Mb B 96/01/08 09 20 17 16.22N 98 11W 33 .0 5.0Mb B 96/01/08 10:04 51 53.22N 142 .76E 33 .0 5.5Mb A 96/01/08 11 52 12 20.29S 68 53W 140 .0 5.0Mb B 96/01/08 13 40 57 17.86S 178 .70W 595 .5 4.8Mb B 96/01/08 13 .59 44 45.36N 150 .14E 28 .1 5.3Mb B 96/01/10 15:57 47 6.72N 73 .01W 170 .0 4.6Mb B 96/01/10 22 :35:58 6.18S 133 .49E 10 .0 5.9Mb B 96/01/11 03 :51 34 8.36S 158 .61E 94 .9 5.6Mb B 96/01/13 05:26:29 51.13N 157.70E 33 .0 5.4Mb A 96/01/14 06 :28 20 44.52N 148 .96E 33 .0 5.4Mb A 96/01/16 05 :15 .23 18.68S 177 .58W 300 .0 5.3Mb A 96/01/18 09:33:50 41.66N 77 .57E 33 .0 5.5Mb B 96/01/19 19 :01 :59 10.37S 78 .83W 35 .7 5.8Mb A 96/01/20 08 :43 :04 3.30S 151 .84E 300 .0 5.0Mb B 96/01/22 00 :10 :23 44.56N 147 .91E 33 .0 5.2Mb B 96/01/22 08 :59 :53 2.82S 141 .33E 33 .0 6.0Mb A 96/01/22 13 :14 :56 40. UN 142 .18E 33 .0 5.1Mb B 96/01/25 12 :45 :02 18.35N 101 .94W 33 .0 5.2Mb A 96/01/26 02 :21 :11 30.90N 91 .41E 33 .0 5.2Mb B 96/01/27 17 :48 :08 9.01N 126.48E 33 .0 5.6Ms B 96/01/27 21:29:57 22.30S 138.94W 0 .0 5.4Mb A 96/01/28 00 :28 :29 1.99S 77.51W 152 .3 4.6Mb B 96/01/29 13 :06 :21 11.22N 125 .23E 47 .9 5.4Mb B KURIL ISLANDS FIJI ISLANDS REGION ICELAND NEAR EAST COAST OF KAMCHATKA OFF COAST OF JALISCO, MEXICO NORTH ATLANTIC OCEAN SAKHALIN ISLAND IRIAN JAYA, INDONESIA HOKKAIDO, JAPAN REGION RYUKYU ISLANDS SANTA CRUZ ISLANDS NEAR COAST OF NICARAGUA CENTRAL BOLIVIA HINDU KUSH REGION, AFGHANISTAN BANDA SEA NORTHWEST OF KURIL ISLANDS NEW BRITAIN REGION, P.N.G. VENEZUELA WINDWARD ISLANDS CENTRAL EAST PACIFIC RISE OFF E COAST OF HONSHU, JAPAN OFF E COAST OF HONSHU, JAPAN SOUTHWESTERN RYUKYU ISLANDS SOUTH OF HONSHU, JAPAN NORTHERN CHILE MINDANAO, PHILIPPINE ISLANDS NEW BRITAIN REGION, P.N.G. SOLOMON ISLANDS NEAR COAST OF GUERRERO, MEXICO SAKHALIN ISLAND CHILE-BOLIVIA BORDER REGION FIJI ISLANDS REGION KURIL ISLANDS NORTHERN COLOMBIA ARU ISLANDS REGION, INDONESIA SOLOMON ISLANDS NEAR EAST COAST OF KAMCHATKA KURIL ISLANDS FIJI ISLANDS REGION KYRGYZSTAN-XINJIANG BDR REG. NEAR COAST OF PERU NEW IRELAND REGION, P.N.G. KURIL ISLANDS NR N COAST OF NEW GUINEA, PNG. NEAR E COAST OF HONSHU, JAPAN GUERRERO, MEXICO XIZANG MINDANAO, PHILIPPINE ISLANDS TUAMOTU ARCHIPELAGO REGION ECUADOR SAMAR, PHILIPPINE ISLANDS APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 96/01/30 02: 28: 30 47.05N 151.61E 33 0 5.4Mb B 96/01/30 21: 14: 57 36.47N 135. 22E 366 3 5. 1Mb A 96/01/31 19: 21: 26 44.50N 149. 48E 33 0 5. 3Ms A 96/01/31 20: 30: 44 44.49N 149.47E 33 0 5. 9Ms A 96/02/01 04:21: 19 17.88S 178. 61W 570 4 4.7Mb B 96/02/01 07: 18: 05 44.93N 146. 26E 180 0 5.8Mb A 96/02/01 17: 57: 56 37.72N 19. 86E 10 0 5. 3Mb A 96/02/02 00: 41: 05 32.67N 137. 32E 300 0 4. 8Mb B 96/02/02 18: 36 10 11.62N 141. 31E 33 0 5. 5Ms B 96/02/03 11: 14 22 27.25N 100 46E 33 0 6. 4Ms A 96/02/03 12 26 48 2.99N 79. 18W 33 0 5. 0Mb B 96/02/04 11 57 18 45.01N 149.45E 33 0 5. 4Mb A 96/02/04 16 58 09 27.04N 100 48E 33 0 5 6Mb B 96/02/05 06 45 24 43.93N 28 37W 10 0 5 0Mb B 96/02/05 21 20 40 6.19S 154 22E 54 .2 5.5Mb B 96/02/07 01 33 16 35.85N 136 30E 33 .0 5 2Mb B 96/02/07 21 36:44 45.17N 149.96E 33 .0 7 OMs A 96/02/09 17 33 49 5.86S 146 50E 33 .0 5 8M B A 96/02/11 09 28:49 45.37N 150 46E 33 .0 5 1Mb B 96/02/12 02 58:52 45.16N 150 15E 33 .0 5 3Mb B 96/02/14 20 31 05 45.29K 150 37E 33 .0 5 9Mb A 96/02/14 21 26 55 29.28N 140 37E 133 .1 5 9Mb B 96/02/16 11 34 30 15.08S 173 63W 33 .0 5.6MS B 96/02/16 15 22 57 37.26N 142 54E 33 .0 6 2Mb A 96/02/17 03:26 42 3.20N 147 39E 33 .0 5.4Mb B 96/02/17 05:59:33 0.50N 135 83E 33 .0 8. OMs C 96/02/17 08:42 08 0.93S 136 02E 33 .0 5 6Mb B 96/02/17 10 18:03 6.94S 125 30E 541 .0 5 9Mb B 96/02/17 13 25 36 6.16S 154 49E 56 .1 5.6Mb B 96/02/17 14 21 24 0.48S 135 91E 33 .0 6 7Ms B 96/02/17 20 17:50 0.76S 136 01E 33 .0 6 6Ms B 96/02/18 02 12:19 0.69S 136.45E 33 .0 6 3Ms B 96/02/18 02:25 :37 1.40S 136 39E 33 .0 6 4Ms C 96/02/18 09:57 :16 13.84N 120.60E 243 .3 5 5Mb B 96/02/18 12 :02 :49 0.69S 136 .40E 33 .0 5 5Mb C 96/02/18 13 :58 :16 0.55S 136 .42E 33 .0 5 5Mb C 96/02/19 12 :14 :18 40.40N 142 .40E 33 .0 5.2Mb B 96/02/19 23 :48 :36 20.42S 169 .05E 33 .0 5 .6Mb B 96/02/21 04 :59 :51 28.75N 34 .85E 10 .0 5 .3Mb A 96/02/21 12:51:04 9.66S 79.75W 33 .0 6 .7Ms B 96/02/21 13 :47 :19 9.52S 80 .18W 33 .0 5 .6Mb C 96/02/22 08 :38 :36 8.57N 83 .17W 33 .0 5 .0Mb A 96/02/22 13 :40 :53 33.58S 71 .29W 43 .7 5.9Hb A 96/02/22 14 :59:09 45.28N 148.52E 133 .3 5 .9Mb A 96/02/24 15 :52 :58 0.87S 137 .05E 33 .0 5 .9Ms B 96/02/25 03 :08 :16 16.OON 97 .90W 33 .0 6 .8 96/02/25 04 :17 :09 22.20S 176 .40W 33 .0 5.6 96/02/25 09 :17 :59 16.09N 97 .88W 33 .0 5 .7Mb A 96/02/25 14 :17 :20 12.92N 91 .01W 33 .0 5 .5Mb A 96/02/25 14 :27 :30 16.07N 97 .66W 33 .0 5 .3Mb A 96/02/25 16 :14 :10 35.72N 56 .92E 33 .0 5 .2Ms B KURIL ISLANDS SEA OF JAPAN KURIL ISLANDS KURIL ISLANDS FIJI ISLANDS REGION KURIL ISLANDS IONIAN SEA SOUTH OF HONSHU, JAPAN WESTERN CAROLINE ISLANDS YUNNAN, CHINA SOUTH OF PANAMA KURIL ISLANDS YUNNAN, CHINA NORTHERN MID-ATLANTIC RIDGE SOLOMON ISLANDS WESTERN HONSHU, JAPAN KURIL ISLANDS E NEW GUINEA REG., P.N.G. KURIL ISLANDS KURIL ISLANDS KURIL ISLANDS SOUTH OF HONSHU, JAPAN TONGA ISLANDS OFF E COAST OF HONSHU, JAPAN E. CAROLINE ISL, MICRONESIA IRIAN JAYA REGION, INDONESIA IRIAN JAYA REGION, INDONESIA BANDA SEA SOLOMON ISLANDS IRIAN JAYA REGION, INDONESIA IRIAN JAYA REGION, INDONESIA IRIAN JAYA REGION, INDONESIA IRIAN JAYA REGION, INDONESIA MINDORO, PHILIPPINE ISLANDS IRIAN JAYA REGION, INDONESIA IRIAN JAYA REGION, INDONESIA NEAR E COAST OF HONSHU, JAPAN VANUATU ISLANDS EGYPT OFF COAST OF NORTHERN PERU OFF COAST OF NORTHERN PERU COSTA RICA NEAR COAST OF CENTRAL CHILE KURIL ISLANDS IRIAN JAYA REGION, INDONESIA NEAR COAST OF OAXACA, MEXICO SOUTH OF FIJI ISLANDS [UOA] OAXACA, MEXICO OFF COAST OF CENTRAL AMERICA OAXACA, MEXICO NORTHERN IRAN APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 83 96/02/26 01: 37: 32 15. 78N 97. 67W 33. 0 5. 1Mb A 96/02/26 07: 17: 24 28. 27N 35. 54E 10. 0 5. 1Mb B 96/02/26 08: 08: 18 28. 27N 57. 21E 33. 0 5. 7Mb B 96/02/27 18: 03: 03 14. 12S 167. H E 33. 0 5. 7Mb B 96/02/28 09: 44: 00 1. 69N 125. 93E 33. 0 6. 3Mb B 96/02/28 11: 22: 01 29. 03N 104. 78E 33. 0 5. 3Mb B 96/02/29 19: 39: 57 24. 04S 66. 65W 200 0 4. 9Mb B 96/03/01 06: 48: 55 34. 26N 25. 98E 33 0 5. 1Mb A 96/03/02 01: 50: 04 6. 11S 146 36E 63 2 6. 1MB A 96/03/03 02: 41: 25 0. 56S 135 96E 21 0 5. 7Ms C 96/03/03 14: 55: 05 10. 83N 86. 64W 33 0 6. 4Ms B 96/03/03 16: 37: 26 11. 12N 86. 68W 33 0 6. 6Ms B 96/03/04 15: 59-05 2. 62N 125 44E 146 2 5. 9Mb B 96/03/05 14: 52 32 24. 75N 122 28E 33 0 6. 4M B B 96/03/05 17: 32 13 24 07N 123 27E 33 0 5. 6Ms B 96/03/06 01: 35 02 18 SOS 175 02W 133 3 5. 3Mb A 96/03/06 14: 35 30 35 84N 139 07E 33 0 5 2Mb B 96/03/07 01: 58 57 29 02N 138 34E 350 0 4 6Mb B 96/03/07 08 38 58 22 97S 70 08W 30 9 5 4Mb A 96/03/07 12 41 04 18 37N 64 51W 76 2 4 7Mb A 96/03/09 16 15 37 43 60N 147 99E 33 0 5 9Ms B 96/03/09 22 35 38 37 03N 24 33W 10 0 5 3Mb A 96/03/10 08 56 21 12 95S 69 25W 32 4 5 9Mb A 96/03/12 03 44 25 8 93N 126 24E 55 8 5 4Mb B 96/03/12 18 43 46 48 76N 87 90E 33 0 5 7Mb B 96/03/13 16 26 31 6 23S 126 29E 33 0 5 5Mb C 96/03/15 09 43 33 52 16N 30 11W 8 0 5 2Mb A 96/03/16 22 04 06 29 06N 138 91E 476 3 5 9Mb A 96/03/17 14 48 56 14 74S 167 23E 164 4 6 OMs A 96/03/17 17 58 20 5 90S 147 42E 33 .0 6 OMs B 96/03/18 03 15 17 23 92S 66 89W 173 .8 5 1Mb B 96/03/18 09 04 48 21 44S 178 .50W 400 .0 4 9Mb B 96/03/18 10 30 .14 6 02S 147 .36E 33 .0 5 6Ms A 96/03/18 22 05 :06 43 26N 147 .05E 33 .0 5 1Mb A 96/03/19 15 00 :26 39 92N 76 .69E 33 .0 5 9Mb A 96/03/19 17 12 :42 15 .80N 97 .20W 33 .0 5 8Mb A 96/03/20 04 53 :25 15 .77N 97 .23W 33 .0 5 .3Mb A 96/03/20 18 :08 :40 15 .85N 97 .22W 33 .0 5 .3Mb A 96/03/20 22 :22 :41 51 .09N 29 .56W 10 .0 5 .0Mb B 96/03/22 08 :26 :38 40 .ION 76 .72E 33 .0 5 .1Mb B 96/03/23 03 :05 :53 8 .32N 72 .68W 199 .0 4 .9Mb B 96/03/23 07 :10 :36 51 .03N 156 .93E 33 .0 5 .0Mb B 96/03/24 03 :19 :34 10 .58N 62 .59W 65 .2 5 .0Mb B 96/03/27 12 :34 :49 16 .43N 97 .97W 33 .0 5 .5Mb 96/03/27 20 :52 :06 11 .68N 87 .99W 33 .0 5 .6Mb A 96/03/28 09 :52 :49 43 .27N 147 .06E 33 .0 5 .2Mb B 96/03/28 23 :03 :49 1 .06S 78 .60W 33 .0 5 .7Mb A 96/03/29 03 :28 :57 24 .25N 122 .39E 33 .0 5 .5Ms B 96/03/29 04 :59 :56 10 .31N 126 .HE 33 .0 5 .3Mb B 96/03/30 23 :16 :37 44 .50N 149 .23E 33 .0 5 .2Mb A 96/03/31 01 :39 :31 9 .99S 160 .69E 64 .3 5 .2Mb B NEAR COAST OF OAXACA, MEXICO WESTERN ARABIAN PENINSULA SOUTHERN IRAN VANUATU ISLANDS NORTHERN MOLUCCA SEA SICHUAN, CHINA SALTA PROVINCE, ARGENTINA CRETE E NEW GUINEA REG., P.N.G. IRIAN JA7A REGION, INDONESIA OFF COAST OF COSTA RICA NEAR COAST OF NICARAGUA TALAUD ISLANDS, INDONESIA TAIWAN REGION SOUTHWESTERN RYUKYU ISLANDS TONGA ISLANDS NEAR S. COAST OF HONSHU, JAPAN SOUTH OF HONSHU, JAPAN NEAR COAST OF NORTHERN CHILE VIRGIN ISLANDS KURIL ISLANDS AZORES ISLANDS REGION CENTRAL PERU MINDANAO, PHILIPPINE ISLANDS NORTHERN XINJIANG, CHINA BANDA SEA NORTHERN MID-ATLANTIC RIDGE SOUTH OF HONSHU, JAPAN VANUATU ISLANDS E NEW GUINEA REG., P.N.G. JUJUY PROVINCE, ARGENTINA FIJI ISLANDS REGION E NEW GUINEA REG., P.N.G. KURIL ISLANDS SOUTHERN XINJIANG, CHINA NEAR COAST OF OAXACA, MEXICO NEAR COAST OF OAXACA, MEXICO NEAR COAST OF OAXACA, MEXICO NORTHERN MID-ATLANTIC RIDGE KYRGYZSTAN-XINJIANG BDR REG. VENEZUELA KAMCHATKA NEAR COAST OF VENEZUELA OAXACA, MEXICO NEAR COAST OF NICARAGUA KURIL ISLANDS ECUADOR TAIWAN REGION PHILIPPINE ISLANDS REGION KURIL ISLANDS SOLOMON ISLANDS APPENDIX A. LIST OF EARTHQUAKES USED IN THIS EXPERIMENT 84 96/03/31 23: 41: 43 10. 97S 165. 41E 33 0 6. lMs B 96/04/01 03: 43: 02 16. 59N 95. 71W 58 1 5. 1Mb A 96/04/01 05: 06: 08 14. 57N 93. 51W 33.0 5. 2Mb A 96/04/01 06: 10: 52 14. 50N 93. 36W 33 0 5. 3Mb A 96/04/01 08: 08: 02 31 51N 73. 52E 43 4 5. 6Mb A 96/04/02 07: 59: 23 37 83N 27. 07E 10 0 5. 2Mb A 96/04/02 18: 50: 36 3.27N 125 87E 33 0 5. 7Mb B 96/04/03 23: 00 48 14 66N 93. 44W 33 0 5. 1Mb A 96/04/04 11: 11 19 3 03N 126 28E 33 0 5 7Mb B 96/04/06 09- 18 51 9 97S 75. 42W 33 0 5 1Mb C 96/04/06 22- 04 41 44 31N 148 99E 33 0 5 3Mb A 96/04/07 00 07:25 44.40N 149.34E 33 0 5 3Mb A 96/04/07 14 18:59 53 29N 159 84E 51 0 5 3Mb B 96/04/08 16 49:42 6 18S 154.56E 57 2 5 6Ms B 96/04/08 20:46 38 10 61S 161.45E 61 2 5 3Mb B 96/04/10 23:24 13 13 02S 76 04W 70 3 5 0Mb B 96/04/11 10 51 14 56 96K 33 51W 10 0 5.0Hb B 96/04/11 11 24 25 10 77S 161 22E 33.0 5 8Mb B 96/04/12 18 45 49 6 04S 154 53E 33 .0 5 9Ms C 96/04/12 22 31 06 56 97N 33 37W 10 .0 4 9Mb A 96/04/15 14 55 32 6 07S 154 42E 55 .3 5 7Mb A 96/04/16 00 30 53 24 09S 177 28W 109 .8 6 8M B A 96/04/18 06 12 59 12 93N 125 00E 44 .6 5 7Ms B 96/04/19 00 19 31 23 85S 69 99W 49.6 6 1MB A 96/04/19 02 30 08 17 60S 179 77E 600 .0 5 2Mb A 96/04/20 19 17 06 23 88S 66 70W 195 .5 5 3Mb B 96/04/20 23 03 31 22.02S 179 94E 610 .0 5 1Mb B 96/04/22 11 27 55 29 80N 129 03E 196 .9 5 1Mb B 96/04/22 14 42 33 39 15N 47 31E 33 .0 5 2Mb B 96/04/23 04:08 00 39 13N 141 33E 73 .9 5.2Mb B 96/04/23 06 53 35 17 26N 101 .28W 33 .0 5 3Mb A 96/04/24 09 36 23 17 .84S 178 .75W 500 .0 5 1Mb A 96/04/24 17:06 36 8 .10S 74 27W 150 .3 5 6Mb A 96/04/24 18 56:22 18 .81N 70.39W 79 .6 5 2Mb A 96/04/25 04 51 .15 30 .07S 71 02W 60 .3 5 5Mb A 96/04/25 05 :50 :08 21.97S 178 .89E 641.6 5 .1Mb A 96/04/26 07 :01 :27 36 .43N 28 .13E 71 .8 5 .4Mb A 96/04/26 16 :30:59 28 .18N 87 .65E 33 .0 5 .1Mb B 96/04/26 17 :03 :44 44 .51N 150 .22E 33 .0 5 .3Mb B 96/04/26 18 :28 :55 20 .60S 179 .13W 500 .0 4 .9Mb B 96/04/27 00 :16 :19 54 .05N 162 .03E 33 .0 5 .3Mb C 96/04/27 03 :01 :04 52.57N 30 .42W 10 .0 4.6Mb B 96/04/27 08 :40 :45 3 .02N 79 .29W 10 .0 5 .9Ms B 96/04/29 14:40 :39 6 .52S 154 .80E 33 .0 7 .5Ms B 96/04/29 22 :31 :15 8 .06N 39 .02W 10 .0 5 .0Mb B 96/04/30 05 :27 :40 6 .29S 154 .04E 33 .0 5 .5Mb B 96/05/01 09:21 :23 6 .59S 154 .64E 33 .0 6. OMs A 96/05/01 10:05 :09 6 .67S 154 .69E 33 .0 6 .OMs B 96/05/02 02 :32:34 6 .32S 154.33E 33 .0 6 .OMs B 96/05/02 05:45 :12 31 .92N 131 .35E 33 .0 5 .1Mb B 96/05/02 06 :38 :24 6 .61S 154 .72E 33 .0 5 .5Mb B SANTA CRUZ ISLANDS OAXACA, MEXICO NEAR COAST OF CHIAPAS, MEXICO NEAR COAST OF CHIAPAS, MEXICO PAKISTAN TURKEY TALAUD ISLANDS, INDONESIA NEAR COAST OF CHIAPAS, MEXICO TALAUD ISLANDS, INDONESIA CENTRAL PERU KURIL ISLANDS KURIL ISLANDS NEAR EAST COAST OF KAMCHATKA SOLOMON ISLANDS SOLOMON ISLANDS NEAR COAST OF PERU NORTH ATLANTIC OCEAN SOLOMON ISLANDS SOLOMON ISLANDS NORTH ATLANTIC OCEAN SOLOMON ISLANDS SOUTH OF FIJI ISLANDS SAMAR, PHILIPPINE ISLANDS NORTHERN CHILE FIJI ISLANDS JUJUY PROVINCE, ARGENTINA SOUTH OF FIJI ISLANDS RYUKYU ISLANDS ARMENIA-AZERBAIJAN-IRAN BORD EASTERN HONSHU, JAPAN NEAR COAST OF GUERRERO, MEXICO FIJI ISLANDS REGION PERU-BRAZIL BORDER REGION DOMINICAN REPUBLIC REGION NEAR COAST OF CENTRAL CHILE SOUTH OF FIJI ISLANDS DODECANESE ISLANDS XIZANG EAST OF KURIL ISLANDS FIJI ISLANDS REGION NEAR EAST COAST OF KAMCHATKA NORTHERN MID-ATLANTIC RIDGE SOUTH OF PANAMA SOLOMON ISLANDS CENTRAL MID-ATLANTIC RIDGE SOLOMON ISLANDS SOLOMON ISLANDS SOLOMON ISLANDS SOLOMON ISLANDS KYUSHU, JAPAN SOLOMON ISLANDS APPENDIX A. LIST OF EARTHQUAKES USED LN THIS EXPERIMENT 96/05/02 13: 34: 19 4. 40S 154. 79E 400 0 5. 7Mb A SOLOMON ISLANDS 96/05/03 03: 32:48 40. 83N 109. 73E 33 0 5. 7Ms A WESTERN NEI MONGOL, CHINA 96/05/04 16- 13: 04 13.89N 146 33E 33.0 5. 6Ms A SOUTH OF MARIANA ISLANDS 96/05/04 16 49:24 13. 90N 146 22E 33 0 5. 7Ms A SOUTH OF MARIANA ISLANDS 96/05/07 08 44:36 1 56N 126 61E 33 0 6. 0Mb B NORTHERN MOLUCCA SEA 96/05/07 21 43:40 14 93S 69.69W 242 0 5. 1Mb B PERU-BOLIVIA BORDER REGION 96/05/07 23 19: 59 43 67N 147 58E 50 0 6.2MS A KURIL ISLANDS 96/05/10 10 19: 38 13 88S 74. 25W 101 1 5. 3Mb A CENTRAL PERU 96/05/11 02 18 45 19 28N 64. 95W 37 6 5.4Mb A VIRGIN ISLANDS 96/05/11 04 38 36 80 52N 2 20W 10 0 5 6Ms A NORTH OF SVALBARD 96/05/11 13 43 44 6.52S 154 85E 33 0 6 4Ms A SOLOMON ISLANDS 96/05/11 16 41 44 52 UN 30 02W 10 0 4 8Mb B NORTHERN MID-ATLANTIC RIDGE 96/05/13 09 59 20 4 73S 145 16E 195 2 5.2Mb B NR N COAST OF NEW GUINEA, PNG. 96/05/14 04 34 44 20 32S 69 19W 118 4 4 9Mb B NORTHERN CHILE 96/05/14 12 36 59 17.80S 178 74W 606 1 5 5Mb A FIJI ISLANDS REGION 96/05/14 17 34 10 24 28S 177 75W 167 8 5 2Mb B SOUTH OF FIJI ISLANDS 96/05/14 23 08 55 18 86S 174 12W 54 0 5 3Mb B TONGA ISLANDS 96/05/15 16 28 49 0 53S 135.73E 33 0 5 7Mb B IRIAN JATA REGION, INDONESIA 96/05/18 03 30 19 54 95N 168 21E 33 .0 5 1Mb A KOMANDORSKT ISLANDS REGION 96/05/18 07 42 25 23 U S 68 91W 96 .5 5 3Mb B NORTHERN CHILE 96/05/19 23 21 41 1 30N 122.63E 33 .0 5 6Mb B MINAHASSA PENINSULA, SULAWESI 96/05/22 20 26 32 20 99S 178 30W 435 .0 5 0Mb B FIJI ISLANDS REGION 96/05/23 01 57 22 5 84N 77 47W 33 .0 5 4Mb A NEAR WEST COAST OF COLOMBIA 96/05/24 06 35 55 27 79N 53 69E 10 .0 5 2Mb B SOUTHERN IRAN 96/05/26 01 43 43 22 20S 171 23E 107 .6 5.5Mb A LOYALTY ISLANDS REGION 96/05/26 07 31 00 22 58N 45 08W 10 .0 4 6Mb B NORTHERN MID-ATLANTIC RIDGE 96/05/26 08 58 27 53 65N 154 08E 460 .0 4 9Mb B SEA OF OKHOTSK 96/05/31 01 31 38 5 45S 147 86E 253 .8 5 2Mb A E NEW GUINEA REG. , P.N.G. 96/06/02 00 .50 37 9 62S 79 52W 33 .0 5 4Mb A OFF COAST OF NORTHERN PERU 96/06/02 02 :19 32 30 51N 41 83W 33 .0 5 1Mb A NORTHERN MID-ATLANTIC RIDGE 96/06/02 02:48 46 30 53N 41 73W 10 .0 5 2Mb B NORTHERN MID-ATLANTIC RIDGE 96/06/02 02 :52 09 10 .64N 42 29W 10 .0 6 8M B A NORTHERN MID-ATLANTIC RIDGE 96/06/02 09:37 :47 27.51N 128 .53E 44.3 5 9Mb B RYUKYU ISLANDS 96/06/03 08 :15:38 9 .10S 156 .88E 33 .0 6 .1MB B SOLOMON ISLANDS 96/06/03 10:46 :00 9 .08S 157 .00E 33 .0 5 .3Mb B SOLOMON ISLANDS 96/06/03 10 :50 :14 9 .09S 157.01E 33 .0 6 .2Ms B SOLOMON ISLANDS 96/06/03 11 :55 :22 17 .62N 94 .23W 160 .0 5 .1Mb A CHIAPAS, MEXICO 96/06/04 23 :22:03 18 .57N 146 .81E 33 .0 5 .5Ms B MARIANA ISLANDS 96/06/06 06 :26:51 21 .53S 169.03E 33 .0 5 .5Mb C LOYALTY ISLANDS REGION 96/06/06 09 :04 :59 6 .72N 73 .16W 160 .1 4 .5Mb B NORTHERN COLOMBIA 96/06/07 05 :09 :23 45 .39N 26.97E 150 .0 4 .5Mb B ROMANIA 96/06/07 14 :21 :39 12 .72N 88 .09W 64.1 4 .9Mb A OFF COAST OF CENTRAL AMERICA 96/06/08 02 :55 :57 41 .62N 88 .65E 0 .0 6 .0Mb A SOUTHERN XINJIANG, CHINA ' 96/06/08 05 :26 :08 58 .31N 31.69W 10 .0 5 .1Mb A NORTH ATLANTIC OCEAN 96/06/09 01 :12 :17 17 .50N 145 .74E 146 .6 6 .0Mb A MARIANA ISLANDS 96/06/11 07 :29 :20 21 .74S 176 .34W 140 .9 5 .0Mb B FIJI ISLANDS REGION 96/06/11 16:52 :12 17 .26N 68 .23W 33 .0 5 .2Mb A MONA PASSAGE 96/06/11 18 :22:55 12 .71N 125 .00E 33 .0 7 .OMs B SAMAR, PHILIPPINE ISLANDS 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items