Statistical Interpretation of Webnet Seismograms By Artificial Neural Nets

We employed multilayer perceptrons (MLP), self organizing feature maps (SOFM), and learning vector quantization (LVQ) to reveal and interpret statistically significant features of different categories of waveform parameter vectors extracted from three-component WEBNET velocigrams. In this contribution we present and discuss in a summarizing manner the results of (i) SOFM classification and MLP discrimination between microearthquakes and explosions on the basis of single-station spectral and amplitude parameter vectors, (ii) SOFM/LVQ recognition of initial onset polarities from PV'-waveforms, and (iii) a source mechanism study of the January 1997 microearthquake swarm based on SOFM classification of combined multi-station PV-onset polarity and SH/PVamplitude ratio (CPA) data. Unsupervised SOFM classification of 497 NKC seismograms revealed that the best discriminants are pure spectral parameter vectors for the recognition of microearthquakes (reliability 95% with 30 spectral parameters), and mixed amplitude and spectral parameter vectors for the recognition of explosions (reliability 98% with 41 amplitude and 30 spectral parameters). The optimal MLP, trained with the standard backpropagation error method by one randomly selected half of a set of 312 mixed (7 amplitude and 7 spectral) single-station (NKC) microearthquake and explosion parameter vectors and tested by the other half-set, and vice versa, correctly classified, on average, 99% of all events. From a set of NKC PV-waveform vectors for 375 events, the optimal LVQ net correctly classified, on average, 98% of all up and 97% of all down onsets, and assigned the likely correct polarity to 85% of the onsets that were visually classified as uncertain. Optimal SOFM architectures categorized the CPA parameter vector sets for 145 January 97 events individually for each of five stations (KOC, KRC, SKC, NKC, LAC) quite unambiguously and stable into three statistically significant classes. The nature of the coincidence of these classes among the stations that provided most reliable mechanism-relevant information (KOC, KRC, SKC) points at the occurence of further seven statistically significant subclassses of mechanisms during the swarm. The ten neural classes of focal mechanisms coincide fairly well with those obtained by moment tensor inversion of P and SH polarities and amplitudes extracted from the seismograms interactively. The obtained results, together with those of refined hypocenter location, imply that the focal area consisted of three dominant faults and at least seven subfaults within a volume of not more than 1 km in diameter that likely were seismically activated by vertical stress from underneath.     

Spatio-temporal variations of seismicity in the southern Peru and northern Chile seismic gaps

The spatio-temporal variation of seismicity in the southern Peru and northern Chile seismic gaps is analyzed with teleseismic data (m _b 5.5) between 1965 and 1991, to investigate whether these gaps present the precursory combination of compressional outer-rise and tensional downdip events observed in other subduction zones. In the outer-rise and the inner-trench (0 to 100 km distance from the trench) region, lower magnitude (5.0m _b <5.5) events were also studied. The results obtained show that the gaps in southern Peru and northern Chile do not present compressional outer-rise events. However, both gaps show a continuous, tensional downdip seismicity. For both regions, the change from compressional to tensional regime along the slab occurs at a distance of about 160 km from the trench, apparently associated with the coupled-uncoupled transition of the interplate contact zone. In southern Peru, an increase of compressional seismicity near the interplate zone and of tensional events (5.0m _b 6.3) in the outer-rise and inner-trench regions is observed between 1987 and 1991. A similar distribution of seismicity in the outer-rise and inner-trench regions is observed with earthquakes (m _b <5.5). In northern Chile there is a relative absence of compressional activity (m _b 5.5) near the interplate contact since the sequence of December 21, 1967. After that, only a cluster of low-magnitude compressional events has been located in the area 50 to 100 km from the trench. The compressional activity occurring near the interplate zone in both seismic gaps represents that a seismic preslip is occurring in and near the plate contact. Therefore, if this seismic preslip is associated with the maturity of the gap, the fact that it is larger in southern Peru than in northern Chile may reflect that the former gap is more mature than the latter. However, the more intense downdip tensional activity and the absence of compressional seismicity near the contact zone observed in northern Chile, may also be interpreted as evidence that northern Chile is seismically more mature than southern Peru. Therefore, the observed differences in the distribution of stresses and seismicity analyzed under simple models of stress accumulation and transfer in coupled subduction zones are not sufficient to assess the degree of maturity of a seismic gap.     

A radiometeorological study of the Chios-Lemnos line of sight radio link

Summary This study deals with the conditions of radio wave propagation in the link between the islands Chios and Lemnos and especially with the influence of weather situations upon this propagation. The field fading forms have been classified into three typesA, B andC which have an annual course of large amplitude. TypesA andB show their maximum frequency during the cold months of the year and their minimum during the warm months, while the typeC shows an annual course inverse to the courses of the typesA andB. This can be explained if it is taken into consideration that in the eastern Mediterranean the weather situations during the warm season are absolutely different from those of the cold season. Finally for the months January, April, July and October a correlation between the prevailing weather situations and the occurence of theA, B, C fading forms has been attempted, the main results of which were that the typeC occurs with weak winds, or with clear and calm weather as well as under the blowing of the Etesian winds, while the typeA is almost certainly due to the anticyclonic extensions during the cold season of the year and generally to the strong winds of the north sector.     

Seismic source-parameter relations for earthquakes in Greece

A data set of nineteen, mainly shallow, moderate to large earthquakes, which occurred in the Aegean and the surrounding area, has been used to derive empirical relations for kinematic fault parameters. Thus the relations between seismic momentM ₀ and magnitudeM _s andm _b and betweenM ₀ andM _s and fault dimensionsS andL have been determined. From these relations and theoretical ones it was deduced that earthquakes in the Aegean and the surrounding events, chiefly interplate, are characterized by low average stress drop values. Values of ranging from 1 to 30 bar are consistent with the data. It was also found that, in general terms, most of the data obey the geometrical similarity conditionL=2w, whereL is the fault length measured along the strike andw is its width measured along the dip. For strike-slip faults, however, the conditionL=4w seems to hold.     

Upper air mean dry-bulb and dew-point temperature profiles over Gauhati airport

Summary The best fit curves for upper air mean dry-bulb and dew-point temperatures over Gauhati airport (26°05N, 91°43E, 49 metres a.m.s.l.), for the month of April, have been calculated with the equation,x=A+By+Cy ²,y being the log value in mb of the isobaric level under consideration andx, the mean dry-bulb or dew-point temperature as the case may be, at the isobaric level under consideration. The values of constantsA, B andC for morning dry-bulb and dew-point curves come to be –29.54559, –93.65766 and +37.35048 and –118.84791, –31.15503 and +25.63585 respectively and values of these constants for evening curves come to be –35.86214, –94.15694 and +38.61870 and –127.55970, –29.97192 and +26.36538 respectively. These best fit curves help in finding out mean desired temperatures at any isobaric level in forecasting of thunderstorms and hailstorms, at a station, by dry-bulb and dew-point temperature anomaly technique propsed earlier by the authors.     

Analysing individual recession events: sensitivity of parameter determination to the calculation procedure

ABSTRACT

This study assesses the sensitivity to model fitting methods and segment selection of the estimated parameters A and B of the model dQ/dt = ?AQ^B for individual events. We investigated about 750 recession events observed at 25 US Geological Survey gauges in the Iowa and Cedar river basins in the United States, with drainage areas ranging from 7 to 17 000 km². The parameters of these recession events were estimated using three commonly adopted methods and recession segments with different extraction criteria. The results showed that the variations of the parameter estimates for the same recession event were comparable to the variations of parameters between different events due to using different model fitting methods and recession segments. This raises cautions for comparative analysis of individual recessions. The result also implies that the nonlinear direct fitting method is the most robust among the three model fitting methods compared.

EDITOR Z.W. Kundzewicz ASSOCIATE EDITOR T. Okruszko     

Analysis of State of Stress During The 1997 Earthquake Swarm In Western Bohemia

Focal mechanisms of 70 events of the January 1997 earthquake swarm were analysed using the Gephart and Forsyth method to determine the state of stress in the West Bohemia/Vogtland region and to reveal possible stress variations during the swarm activity. The method was applied to the whole set of 70 focal mechanisms and to several subsets distinguishing magnitudes of events and the spatial and temporal distribution of the swarm. The three representative stress models A ( ₁ ,A: 0/45 (az/dip), ₂ A: 134/35, ₃ A: 243/25), B ( ₁ ,B: 162/38, ₂ B: 263/14, ₃ B: 10/49), and C ( ₁ C: 135/30, ₂ C: 1/50, ₃ C: 239/23) which could characterise the state of stress in the region were found. Model C can be considered to be the most probable stress model because of its consistency with the European stress field. The results of the stress analysis applied to the individual subsets of focal mechanisms indicate that the state of stress was more uniform during the first phase of the swarm. To distinguish between the fault plane and the auxiliary plane of fault plane solutions a statistical approach was used. The results showed that the fault planes striking NW-SE and dipping 45° to NE were activated during the swarm.     

Regional characteristics of coda attenuation after the lancang-Gengma earthquake

In the light of the single scattering model of coda originating from local earthquakes, and based on the aftershock coda registered respectively at the 4 short period stations installed near the foci shortly after theM7.6 Lancang andM7.2 Gengma earthquakes, this paper has tentatively calculated the rate of amplitude attenuation and theQ _c-value of the coda in the Lancang and Gengma areas using a newly-founded synthetic determination method. Result of the study shows the rate of coda amplitude attenuation demonstrates remarkable regional differences respectively in the southern and northern areas. The southern area presents a faster attenuation (Q _c=114), whereas the northern area shows a slower attenuation (Q _c=231). The paper also discusses the reasons causing such differences. Result of the study also suggests a fairly good linear relation between the coda source factorA _o(f) and the seismic moment and the magnitude. Using the earthquake scaling law, the following formulas can be derived: lgM ₀=lgA ₀(f)+17.6,M _D=0.67lgA ₀(f)+1.21 and logM ₀=1.5M _D+15.79. In addition, the rates of amplitude attenuationβ _s andβ _m are respectively calculated using the single scattering and multiple scattering models, and the ratioβ _s/β_m=1.20−1.50 is found for the results respectively from the two models. Finally, the mean free pathL of the S-wave scattering in the southern and northern areas are determined to be 54 km and 122 km respectively by the relations which can distinguish between the inherentQ _i and scatteringQ _s, testify to this areas having lowQ-values correspond to stronger scatterings. The Chinese version of this paper appeared in the Chinese edition ofActa Seismologica Sinica,14, 71–82, 1992. This study is partly supported by the Seismological Science Foundation of the State Seismological Bureau of China, and the present English version of the paper is translated from its Chinese original by Wenyi Xia, Seismological Bureau of Yunnan Province.     

Temporal variations in the fractal properties of seismicity in the North Anatolian Fault Zone between 31°E and 41°E

We investigate the nature of temporal variations in the statistical properties of seismicity associated with the North Anatolian Fault Zone between longitudes 31°–41°E during the instrumental period 1900–1992. Temporal variations in the seismicb value and the fractal (correlation) dimensionD _c of earthquake epicenters are examined for earthquakes of magnitudeM _S 4.5, using sliding windows of 100 consecutive events.b varies temporally between 0.6 and 1.0, andD _c between 0.6 and 1.4, both representing significant fluctuations above the errors in measurement technique. A strong negative correlation (r=–0.85) is observed betweenb andD _c , consistent with previous observation of seismicity in Japan and southern California. Major events early in this century (M _S 7) are associated with lowb and highD _c , respectively consistent with greater stress intensity and greater spatial clustering of epicenters—both implying a greater degree of stress concentration at this time.     

Depression in the low-latitude horizontal intensity due to quiet-time ring current

Summary The external field due to plasma within the magnetosphere has been computed as a function ofA _p, which is a measure of solar wind velocity, for very quiet to slightly disturbed conditions using mean daily horizontal intensity from 1932 to 1968 at Alibag. The intensity, corrected for secular change and reduced to a common epoch, showed initially a small increase withA _p followed by a steady depression with further increase in the index. ForA _p7.5, which is representative of conditions over the 33-hour interval during which data relating to low-energy protons were acquired and used byHoffman andBracken [4]²) to compute current distributions, the decrease, computed here from surface data, is 6 . This is in goodagreement with the southward directed field of the quiet-time proton belt 9±5 obtained byHoffman andBracken.     

Some comparisons between mining-induced and laboratory earthquakes

Although laboratory stick-slip friction experiments have long been regarded as analogs to natural crustal earthquakes, the potential use of laboratory results for understanding the earthquake source mechanism has not been fully exploited because of essential difficulties in relating seismographic data to measurements made in the controlled laboratory environment. Mining-induced earthquakes, however, provide a means of calibrating the seismic data in terms of laboratory results because, in contrast to natural earthquakes, the causative forces as well as the hypocentral conditions are known. A comparison of stick-slip friction events in a large granite sample with mining-induced earthquakes in South Africa and Canada indicates both similarities and differences between the two phenomena. The physics of unstable fault slip appears to be largely the same for both types of events. For example, both laboratory and mining-induced earthquakes have very low seismic efficiencies where _a is the apparent stress and is the average stress acting on the fault plane to cause slip; nearly all of the energy released by faulting is consumed in overcoming friction. In more detail, the mining-induced earthquakes differ from the laboratory events in the behavior of as a function of seismic momentM ₀. Whereas for the laboratory events 0.06 independent ofM ₀, depends quite strongly onM ₀ for each set of induced earthquakes, with 0.06 serving, apparently, as an upper bound. It seems most likely that this observed scaling difference is due to variations in slip distribution over the fault plane. In the laboratory, a stick-slip event entails homogeneous slip over a fault of fixed area. For each set of induced earthquakes, the fault area appears to be approximately fixed but the slip is inhomogeneous due presumably to barriers (zones of no slip) distributed over the fault plane; at constant , larger events correspond to larger _a as a consequence of fewer barriers to slip. If the inequality _a / 0.06 has general validity, then measurements of _a =µE _a /M ₀, where is the modulus of rigidity andE _a is the seismically-radiated energy, can be used to infer the absolute level of deviatoric stress at the hypocenter.     

Research on the time changes of transfer functions at Heze observatory before and after Heze earthquake of magnitude 5.9 A discussion with Qi Wang

In this paper, the coefficients of Parkinson vectora, b and the in-phase partA _r, B_r and quadrature-phase partA _i, B_i of transter functions are calculated by using two kinds of data processing methods, respectively. The results of both methods are close to each other. All these six parameters above are low in value, showing that the electric structure in Heze area is rather uniform. It seems that the anomalous changes appeared in both results, and the values of parameters decrease before and after Heze earthquake. The Chinese version of this paper appeared in the Chinese edition ofActa Seismologica Sinica,13, 113–120, 1991.     

Non-adiabatic motion of charged particles traversing a weak magnetic field: Pitch angle scattering

Summary A charged particle moves with velocityv in a constant non-uniform magnetic fieldH, spiralling with Larmor radiusR. IfR is small compared with the scale lengthL of the field, the magnetic moment associated with the Larmor motion of the particle is nearly constant. Consequently , the (pitch) angle betweenv andH, varies as arcsinH ^1/2. Hence in such adiabatic motion is approximately the same at points on the path whereH has the same value. But the magnetic moment and the pitch angle may differ materially at two such points, each in the region whereR/L is small, if between them the particle traverses a region whereR/L is not small. This region of non-adiabatic motion scatters the pitch angles.Such scattering is investigated for regions of weak field (R large), with and without the presence of a neutral line along whichH=0. Either type of region, it is found, can scatter the pitch angles. This gives support to the theory proposed byAkasofu andChapman to explain why auroral arcs and bands are very thin.The scattering here examined is of interest also in connection with magnetic mirror devices for nuclear energy transformation. It may also have applications to phenomena of solar and stellar atmospheres.     

Chemistry and metal contents of magmatic gases: the new Tolbachik volcanoes case (Kamchatka)

The gaseous products of new Tolbachik volcanoes were studied during 1975 to 1977 throughout all eruptive stages and during the post eruptive activity. In investigations the northern break-out gases emitted during the eruption from the moving and consolidated lava flows there have been detected H₂O (the main component), H₂, HF, HCl, SO₂ and H₂S, CO₂, CO, NH₃, CH₄ and other hydrocarbons, NH₄Cl predominated in compositions of condensates and subtimates on lava flows and the most characteristic microcomponents were Zn, Cu, Pb, Sn, Ag and others. Sampling of gases and condensates at the southern break-out was conducted immediately from the flowing melt. In gases there have been detected H₂O (98 mol. %). HCl and H₂ (0.9 mol. % each) as well as HF, SO₂, H₂S, CO₂ and in small quantities O₂ and N₂, Gases reached the equilibrium state atT andP sampling and were characteristic of gas composition of the southern break-out magma. HCl, HF and H₂SO₄ were predominant during condensate and sublimate mineralization. The major raicrocomponents were represented by Pt, Sb, As, Zn, Cu, Pb, Ni, Co and others. Comparison of compositions of gases and of products of their reactions at the northern and at the southern break-outs allows us to assume the presence of the deeper magma source at the northern break-out and of shallow magma source at the southern break-out.     

High-time resolution conjugate SuperDARN radar observations of the dayside convection response to changes in IMF B y

We present data from conjugate SuperDARN radars describing the high-latitude ionospheres response to changes in the direction of IMF B_y during a period of steady IMF B_z southward and B_x positive. During this interval, the radars were operating in a special mode which gave high-time resolution data (30 s sampling period) on three adjacent beams with a full scan every 3 min. The location of the radars around magnetic local noon at the time of the event allowed detailed observations of the variations in the ionospheric convection patterns close to the cusp region as IMF B_y varied. A significant time delay was observed in the ionospheric response to the IMF By changes between the two hemispheres. This is explained as being partially a consequence of the location of the dominant merging region on the magnetopause, which is 8/12R_E closer to the northern ionosphere than to the southern ionosphere (along the magnetic field line) due to the dipole tilt of the magnetosphere and the orientation of the IMF. This interpretation supports the anti-parallel merging hypothesis and highlights the importance of the IMF B_x component in solar wind-magnetosphere coupling.     

The Krafla fissure swarm, Iceland, and its formation by rifting events

Fissure swarms at divergent plate boundaries are activated in rifting events, during which intense fracturing occurs in the fissure swarm accompanied by intrusion of magma to form dikes that sometimes lead to eruptions. To study the evolution of fissure swarms and the behaviour of rifting events, detailed mapping was carried out on fractures and eruptive fissures within the Krafla fissure swarm (KFS). Fracture densities of dated lava flows ranging from 10,000?years bp to ~30?years old were studied, and the fracture pattern was compared with data on the historical Myvatn rifting episode (1724–1729) and the instrumentally recorded Krafla rifting episode (1975–1984). Additionally, the interaction of transform faults and fissure swarms was studied by analysing the influence of the Húsavík transform faults on the KFS. During the historical rifting episodes, eruptions on the fissure swarm occurred within ~7?km from the Krafla central volcano, although faults and fractures were formed or activated at up to 60–70?km distance. This is consistent with earlier rifting patterns, as Holocene eruptive fissures within the KFS are most common closer to the central volcano. Most fractures within the central Krafla caldera are parallel to the overall orientation of the fissure swarm. This suggests that the regional stress field is governing in the Krafla central volcano, while the local stress field of the volcano is generally weak. A sudden widening of the graben in the northern KFS and a local maximum of fracture density at the junction of the KFS and the extrapolation of the Húsavík transform fault zone indicates possible buried continuation of the Húsavík transform fault zone which extends to the KFS. Eruptive fissures are found farther away from the Krafla central volcano in the southern KFS than in the northern KFS. This is either due to an additional magma source in the southern KFS (the Heiearsporeur volcanic system) or caused by the Húsavík transform faults, transferring some of the plate extension in the northern part. Fracture density within particular lava flow fields increases with field age, indicating that repeated rifting events have occurred in the fissure swarm during the last 10,000?years bp. The fracture density in the KFS is also generally higher closer to the Krafla central volcano than at the ends of the fissure swarm. This suggests that rifting events are more common in the parts of the fissure swarm closer to the Krafla central volcano.     

Historical seismicity of the southeastern Caribbean and tectonic implications

We have relocated the twenty-eight largest magnitude (4.3M _s 7.3) historical (1922–1963) earthquakes of the southeastern Caribbean. We also present new focal mechanisms for seven of these events. The relocations are based on reported ISSP andS arrival times that we analyzed using generalized linear inversion techniques. The new focal mechanisms were constrained by first motionP polarities as reported by the ISS and as picked by us where records were available, and by the polarities and ratios ofSH andsSH, andSV andsSV arrivals that we determined from seismograms. The results of the relocations are commensurate with the distribution of seismicity observed in the recent era: hypocenters are shallow and intermediate in depth (0–200 km), and the events occur almost exclusively in areas known to be currently seismic. The frequent seismic activity in the vicinity of the Paria Peninsula, Venezuela, is clearly a persistent feature of the regional earthquake pattern; intermediate depth earthquakes indicative of subduction beneath the Caribbean plate occur here and along the Lesser Antilles arc. The Grenadines seismic gap is confirmed as an area of low seismic moment release throughout the historical era. Trinidad and the eastern Gulf of Paria were also largely quiescent.The new focal mechanisms, despite being a sparse data set, give significant insight into both subduction processes along the Lesser Antilles arc and into the shallow deformation of the Caribbean-South America plate boundary zone. The largest earthquake to have occurred in this region, the 19 March 1953 event (M _m =7.01), is a Lesser Antilles slab deformation event, and another earthquake in this region of the Lesser Antilles is probably a rarely-observed interplate thrust event. Shallow deformation in the plate boundary zone is complex and, near the Paria Penninsula, involves mixed southeastward thrusting and dextral strike-slip on east-striking faults, and secondarily, normal faulting. Bending of the subducting Atlantic-South American plate also seems to generate seisms. The rather high ratio of intraplate deformation to interplate deformation observed along the Lesser Antilles subduction zone in the more recent era seems to have been operative in the historical era as well.     

Analysis of aeromagnetic anomalies over the central part of the Narmada-Son lineament

The Narmada-Son lineament (NSL) forms a major tectonic feature on the Indian subcontinent. The importance of this lineament lies in its evolution as well as its tectonic history. The lineament seems to have been active since Precambrian times. In order to understand the history of its evolution, it is necessary to know what igenous activity has been taking place along this lineament, and how the Deccan trap volcanics, which cover large areas along this lineament, have erupted.For the study of this problem an analysis of the aeromagnetic anomaly map lying between 76°15 to 77°30E and 21°45 to 22°50N has been carried out. Four different profiles (B ₁ B ₁,B ₂ B ₂,B ₃ B ₃ andB ₄ B ₄) have been drawn in N-S direction over this area and interpreted in terms of the intrusive bodies present within or below the surface of Deccan trap exposures. Inversion and forward modelling techniques have been adopted for interpretation purposes. An analysis of frequency spectra along the profiles has also been carried out to estimate the average depth of the different magnetic bodies. These results have been correlated with the available geological information. It has been found that most of the small wavelength anomalies are caused by dyke-like bodies within or below the Deccan trap at a depth of less than 0.5 km.     

Lateral structure of the subducting pacific plate beneath the Hokkaido corner from intermediate and deep earthquakes

We present a study of the lateral structure and mode of deformation in the transition between the Kuril and Honshu subduction zones. We begin by examining the source characteristics of the January 19, 1969, intermediate depth earthquake north of Hokkaido in the framework of slab-tearing, which for the December 6, 1978 event has been well documented by previous studies. We use a least-squares body wave inversion technique, and find that its focal mechanism is comparable to the 1978 event. To understand the cause of these earthquakes, which in the case of the 1978 event occurred on a vertical tear fault but does not represent hinge faulting, we examine the available International Seismological Centre [ISC] hypocenters and Harvard centroid-moment tensor [CMT] solutions to determine the state of stress, and lateral structure and segmentation in the Kuril and northern Honshu slabs. These data are evaluated in the framework of two models. Model (A) requires the subducting slab at the Hokkaido corner to maintain surface area. Model (B) requires slab subduction to be dominated by gravity, with material subducting in the down-dip direction. The distribution of ICS hypocenters shows a gap in deep seismicity down-dip of the Hokkaido corner, supporting model (B). From the CMT data set we find that three types of earthquake focal mechanisms occur. The first (type A) represents dip-slip mechanisms consistent with down-dip tension or compression in the slab in a direction normal to the strike of the trench. These events occur throughout the Honshu and Kuril slabs with focal mechanisms beneath Hokkaido showing NNW plungingP andT axes consistent with the local slab geometry. The second (type B) occurs primarily at depths over 300 km in the southern part of the Kuril slab with a few events in the northern end of the Honshu deep seismicity. These earthquakes have focal mechanisms with P axes oriented roughly E-W, highly oblique to the direction of compression found in the type A events, with which they are spatially interspersed. The third (type C) group of earthquakes are those events which do not fit in either of the first two groups and consist of either strike-slip focal mechanisms, such as the tearing events, or oddly oriented focal mechanisms. Examination of the stress axes orientations for these three types reveals that the compressional axes of the type C events are consistent with those of type B. The slab tearing events are just differential motion reflecting the E-W compressive states of stress which is responsible for the type B family of events. There is no need to invoke down-dip extension which does not fit the slab geometry. We conclude that these two states of stress can be explained as follows: 1) The type A events and the seismicity distribution support model (B). 2) The type B and C events upport model (A). The solution is that the slab subducts according to model (B), but the flow in the mantle maintains a different trajectory, possibly induced by the plate motions, which produces the second state of E-W compressive stress.