Harvard Centroid Moment Tensor Solutions Research Articles

Towards a homogeneous moment magnitude determination for earthquakes in South Africa: Reduction of associated uncertainties

New MW–ML magnitude relations valid for the periods January 1971–March 1997, April 1997–September 2012 and October 2012–present are presented for South Africa. Three MW–ML relations were derived for the respective periods by applying a general orthogonal regression using 85 mining-related and tectonic events. Moment magnitudes (MW) were calculated using spectral analysis of the vertical component seismograms. A Brune model for source radiation was assumed. Attenuation functions previously derived from local seismic events were used. For each event, the final MW was computed by taking the average of the values determined at four or more stations. These MW values were used to compute new MW –ML regression equations for three previous ML relations that had been used to compile the South African earthquake catalogue. The new MW–ML relations were then compared with the MW determined from the Harvard Centroid Moment Tensor (CMT) project. The MW computed using the relation derived for the most recent period (October 2012–present) produces MW values that are similar to those reported by the Harvard CMT solutions. The relations can be used in the homogenization of the South African earthquake catalogue. Application of the new relations for the three periods will reduce uncertainty in the homogenized catalogues, which are important when calculating seismic hazard parameters (i.e. when estimating activity rates, b-values and Mmax).

W phase source inversion for moderate to large earthquakes (1990-2010)

Rapid characterization of the earthquake source and of its effects is a growing field of interest. Until recently, it still took several hours to determine the first-order attributes of a great earthquake (e.g. M_w ≥ 7.5), even in a well-instrumented region. The main limiting factors were data saturation, the interference of different phases and the time duration and spatial extent of the source rupture. To accelerate centroid moment tensor (CMT) determinations, we have developed a source inversion algorithm based on modelling of the W phase, a very long period phase (100–1000 s) arriving at the same time as the P wave. The purpose of this work is to finely tune and validate the algorithm for large-to-moderate-sized earthquakes using three components of W phase ground motion at teleseismic distances. To that end, the point source parameters of all M_w ≥ 6.5 earthquakes that occurred between 1990 and 2010 (815 events) are determined using Federation of Digital Seismograph Networks, Global Seismographic Network broad-band stations and STS1 global virtual networks of the Incorporated Research Institutions for Seismology Data Management Center. For each event, a preliminary magnitude obtained from W phase amplitudes is used to estimate the initial moment rate function half duration and to define the corner frequencies of the passband filter that will be applied to the waveforms. Starting from these initial parameters, the seismic moment tensor is calculated using a preliminary location as a first approximation of the centroid. A full CMT inversion is then conducted for centroid timing and location determination. Comparisons with Harvard and Global CMT solutions highlight the robustness of W phase CMT solutions at teleseismic distances. The differences in M_w rarely exceed 0.2 and the source mechanisms are very similar to one another. Difficulties arise when a target earthquake is shortly (e.g. within 10 hr) preceded by another large earthquake, which disturbs the waveforms of the target event. To deal with such difficult situations, we remove the perturbation caused by earlier disturbing events by subtracting the corresponding synthetics from the data. The CMT parameters for the disturbed event can then be retrieved using the residual seismograms. We also explore the feasibility of obtaining source parameters of smaller earthquakes in the range 6.0 ≤_Mw < 6.5. Results suggest that the W phase inversion can be implemented reliably for the majority of earthquakes of Mw= 6 or larger.

Plate convergence at the westernmost Philippine Sea Plate

To understand the convergent characteristics of the westernmost plate boundary between the Philippine Sea Plate (PSP) and Eurasian Plate (EP), we have calculated the stress states of plate motion by focal mechanisms. Cataloged by the Harvard centroid moment tensor solutions (Harvard CMT) and the Broadband Array in Taiwan (BATS) moment tensor, 251 focal mechanisms are used to determine the azimuths of the principal stress axes. We first used all the data to derive the mean stress tensor of the study area. The inversion result shows that the stress regime has a maximum compression along the direction of azimuth N299°. This result is consistent with the general direction of the rigid plate motion between the PSP and EP in the study area. In order to understand the spatial variation of the regional stress pattern, we divided the study area into six sub-areas (blocks A to F) based on the feature of the free-air gravity anomaly. We compare the compressive directions obtained from the stress inversion with the plate motions calculated by the Euler pole and the Global Positioning System (GPS) analysis. As a result, the azimuth of the maximum stress axis, σ 1, generally agrees with the directions of the theoretical plate motion and GPS velocity vectors except block C (Lanhsu region) and block F (Ilan plain region). The discrepancy of convergent direction near the Ilan plain region is probably caused by the rifting of the Okinawa Trough. The deviation of the σ 1 azimuth in the Lanhsu region could be attributed to a southwestward extrusion of the Luzon Arc (LA) block between 21°N and 22°N whose northern boundary may be associated with the right-lateral NE–SW trending fault (i.e. Huatung Fault, HF) along the Taitung Canyon. Comparing the σ 1 stress patterns between block C and block D, great strain energy along HF may not be completely released yet. Alternatively, the upper crust of block C may significantly have decoupled from its lower crust or uppermost mantle.

Fault plane solutions in Sichuan-Yunnan rhombic block and their dynamic implications

Harvard Centroid Moment Tensor (CMT) solutions for earthquakes from 1977 to 2004 showed that the stress fields are obviously different in northwestern Sichuan sub-block (NWSSB), western parts of Central Yunnan sub-block (CYSB) and eastern part of CYSB. The characteristics of the mean stress fields in these three regions are obtained by fitting to CMT solutions. The stress state in NWSSB is characterized by its sub-horizontal tensile principal axis of stress (T axis) in roughly N-S direction and west dipping compressive principal axis of stress (P axis); the one in western part of CYSB is characterized by its ENE dipping T axis and sub-horizontal medium principal axis of stress (B axis) in roughly N-S direction; the one in eastern part of CYSB is characterized by its sub-horizontal P axis in roughly NNW-SSE direction and sub-horizontal T axis in roughly WSW-ENE direction. Finite element method simulation clearly shows that the Indian Plate imposes great extrusion on Sichuan-Yunnan rhombic block (SYRB) near Assam massif. The value of the simulated compressive principal stress decreases with the distance from Assam massif. The simulated directions of the T axes in SYRB form annular distribution encircling Assam. For a homogeneous elastic medium with free boundary conditions on the top and bottom surfaces as well as the displacement boundary conditions derived from the GPS observations on the lateral boundaries, the computation results are consistent with the Harvard CMT solutions in NWSSB and western part of CYSB, while inconsistent with the Harvard CMT solutions in eastern part of CYSB. The inconsistency in eastern part of CYSB can be reduced when it includes inhomogeneous elastic media. The stress states in NWSSB and western part of CYSB revealed by the Harvard CMT solutions are not local, which are mainly controlled by the boundary force on the whole region. On the other hand, the stress state in eastern part of CYSB given by the Harvard CMT solutions is local, which may be affected by local topography, material inhomogeneity, and the drag force underneath.

Automated Receiver Function Processing

The aim of this is article is twofold: first to introduce an automated processing system based on SOD (Standing Order far Data , http://www.seis.sc.edu/SOD/) (Owens et al. , 2004), and second to give some initial results and pitfalls from our experience with using this system for automated receiver function calculations. EARS, the EarthScope Automated Receiver Survey, aims to calculate bulk crustal properties for all stations within the continental United States that are available from the IRIS DMC using receiver functions and to distribute these results to interested research and education communities as a “product” of the USArray component of EarthScope. Because of the high level of automation that this requires, a natural extension of this effort is to calculate similar results for all stations in the IRIS DMC. To do this we have employed SOD , a FISSURES/DHI-based software package (Ahern, 2001a, b) that we have developed to do automated data retrieval and processing. We introduce EARS as an example of what we term “receiver reference models” They are standard analysis techniques applied to stations. These are analogous to the Harvard Centroid Moment Tensor solutions (which could be termed source reference models). An RRM need not be a definitive result but rather provides a standard, well known, globally consistent reference. This may be sufficient for many users, just as CMT's are sufficient for many, but may also be a starting point for more in-depth, focused studies. The key features of an RRM are that it be generated from a well known and widely accepted technique, produce results for most or all stations, and provide updated results as new sources of data become available. In addition, it must produce results that are of interest in a form usable by the community. Because automated processing proceeds, by definition, without a large …

Seismotectonic pattern and the source region of volcanism in the central part of Sunda Arc

The seismotectonic pattern in the central part of the Sunda Arc (Java, Nusa Tenggara) was studied in relation to the distribution of active calc-alkaline volcanoes, using global seismological data. Hypocentral determinations of the International Seismological Centre from the period 1964–1999, as relocated by Engdahl, and Harvard Centroid Moment Tensor Solutions from the period 1976–2003 were used. The following phenomena, which could assist the location of the source region of primary magma for island arc calc-alkaline volcanism, were observed: (1) An aseismic gap without any strong teleseismically recorded earthquakes was found in the Wadati-Benioff zone of the subducting slab along the whole investigated region of the Sunda Arc, forming a continuous strip of laterally variable depth and shape, at depths between 100 and 200 km. The absence of strong earthquakes (with m b>4.0) indicates a significant change in the mechanical properties of the subducting slab at intermediate depths. All active calc-alkaline volcanoes in the Sunda Arc are located above this gap. (2) The majority of earthquakes occurring in the lithospheric wedge of the Eurasian Plate above the subducted slab could be attributed to several deep-rooted seismically active fracture zones of regional extent. All delineated active fracture zones display a thrust tectonic regime as shown by the available fault plane solutions. (3) Clusters of earthquakes were found beneath active volcanoes of western Java, Bali and Nusa Tenggara in the lithospheric wedge above the slab and identified as seismically active columns. These clusters occur only beneath the volcanoes that are located at the outcrops of seismically active fracture zones. We interpret the earthquakes in these clusters beneath volcanoes as events induced by magma transport through the medium of the lithospheric wedge that has been subcritically pre-stressed by the process of plate convergence. (4) Beneath the volcanoes of central Java no seismically active columns were observed. The latter volcanoes are not located at the outcrop of any seismically active fracture zone. The presence of the intermediate-depth aseismic gap in the Wadati-Benioff zone and the presence of seismically active columns above the gap beneath some active volcanoes support the concept that the subducted oceanic lithosphere is the source region for the primary magma for calc-alkaline volcanoes at convergent plate margins. However, for volcanoes without any seismically active columns and fracture zones beneath them, a source region for the primary magma in the lithospheric wedge above the subduction zone cannot be excluded.

Source Parameters of Eastern California and Western Nevada Earthquakes from Regional Moment Tensor Inversion

The moment tensors and centroid depths are estimated for eastern California and western Nevada earthquakes from regionally recorded long-period seismograms using the moment-tensor inversion method. We compiled the moment tensor solutions into a catalog complete for earthquakes with M W > 4 since 1990. The earthquakes in this study are mostly located within the aftershock zones of Eureka Valley, Double Spring Flat, Coso, Ridgecrest, Fish Lake Valley, and Scotty's Junction, with the remaining earthquakes distributed across the region. We validated the moment tensor solutions by a comparison with Harvard centroid moment tensor (CMT) and P -wave first-motion focal mechanism solutions. The mean difference in strike, dip, and rake between our moment tensors and either Harvard CMT solutions or first-motion focal mechanisms were less than approximately 15° with a standard deviation less than 10°. We also examine the solution mean and variance by inverting resampled datasets using the delete- j Jackknife resampling method. Based on two well-recorded sample events, with more than 11 and 14 associated recording stations, the P - and T -axis trend and plunge can be determined to within ±10° when at least three or four well-distributed stations are used in the inversion. We separated the T -axis trends based on the approximate boundaries for three tectonic regions: Sierra Nevada, Walker Lane-eastern California shear zone, and Basin and Range. Relative to the mean T -axis trend for the Sierra Nevada, the mean T -axis trend for the Walker Lane region is rotated clockwise by 25° and the mean T -axis trend for the Basin and Range is rotated clockwise by 40°. We separated the earthquakes into strike-slip and normal-slip focal mechanisms based on a simple rake angle criterion. About 70% of modern earthquakes and 73% of the 11 large historical earthquakes since 1860 are in the strike-slip category. The fraction of seismic moment released as strike-slip earthquakes is approximately 50% for modern earthquakes and 75% for historical earthquakes. The difference is well within the variability expected from short catalogs. These results emphasize the importance of the component of right-lateral shear within the region and are qualitatively consistent with recent geodetic results. Manuscript received 11 January 2002.

Seismotectonics of the Explorer region

The Explorer region offshore western Canada is a tectonically complex area surrounded by the Pacific, North America, and Juan de Fuca plates. Existing tectonic models for the region differ fundamentally. Proposed plate configurations range from multiple independent plate fragments to an Explorer plate now fused to North America along the continental margin and cut by Pacific–North America transform faults in the west. We present new seismological data constraining the region's current tectonics. We use three‐component regional waveforms to determine the source parameters of 84 earthquakes with magnitude greater than 4. Combined with 34 Harvard centroid moment tensor solutions, they represent the region's largest earthquake source parameter data set obtained by robust waveform modeling techniques. In addition, we perform joint epicenter determination to relocate larger earthquakes recorded since 1918. The source parameters and improved locations provide a consistent tectonic picture. Earthquake slip vector azimuths along the Pacific plate boundary change smoothly and are significantly less northerly oriented than the Pacific‐North America plate motion direction, requiring an independent Explorer plate. The present‐day Pacific‐Explorer boundary is formed by transform faults subparallel to the Revere‐Dellwood‐Wilson fault. Plate motion vectors indicate that the Winona block is part of the Explorer plate. Current Explorer motion is more northerly than indicated by magnetic anomalies prior to 2 Ma, implying a recent change, possibly coinciding with a northwestward ridge jump near Explorer plate's northern end transferring the Winona block from the Pacific to the Explorer plate. In response to these plate motion changes the region north of the western Sovanco fracture zone was assimilated into the Pacific plate. The region around the eastern Sovanco fracture zone, characterized by broadly distributed seismicity, is composed of well‐defined sets of conjugate faults bounding rotating crustal blocks. Earthquake fault strikes agree with the dominant northwest‐southeast fault sets; however, the conjugate sets must be also active to fully accommodate present‐day Explorer plate motion. The SW portion of the strike‐slip Nootka fault zone, the Explorer‐Juan de Fuca plate boundary, is well defined by focused seismicity; however, its full extent under Nootka Island remains unresolved. The Explorer–North America boundary shows sporadic low‐magnitude seismicity. Our Explorer–North America rotation pole predicts convergence varying from negligible at the boundary's northwest end to ∼2 cm/yr at the SE end. This convergence can be accommodated either by subduction or by crustal thickening extending to the North American continent. We favor subduction based on low deformation rates observed by onshore GPS sites. The present Explorer plate system configuration is a result of stepwise reorientation of the Explorer ridge system, each step successively reducing the subduction rate relative to North America.

The October 15, 1997 Punitaqui earthquake (Mw=7.1): a destructive event within the subducting Nazca plate in central Chile

The 1943 Illapel seismic gap, central Chile (30–32°S), was partially reactivated in 1997–1998 by two distinct seismic clusters. On July 1997, a swarm of offshore earthquakes occurred on the northern part of the gap, along the coupled zone between Nazca and South American plates. Most of the focal mechanisms computed for these earthquakes show thrust faulting solutions. The July 1997 swarm was followed on October 15, 1997 by the Punitaqui main event (Mw=7.1), which destroyed the majority of adobe constructions in Punitaqui village and its environs. The main event focal mechanism indicates normal faulting with the more vertical plane considered as the active fault. This event is located inland at 68-km depth and it is assumed to be within the oceanic subducted plate, as are most of the more destructive Chilean seismic events. Aftershocks occurred mainly to the north of the Punitaqui mainshock location, in the central-eastern part of the Illapel seismic gap, but at shallower depths, with the two largest showing thrust focal mechanisms. The seismicity since 1964 has been relocated with a master event technique and a Joint Hypocenter Determination (JHD) algorithm, using teleseismic and regional data, along with aftershock data recorded by a temporary local seismic network and strong motion stations. These data show that the 1997 seismic clusters occurred at zones within the Illapel gap where low seismicity was observed during the considered time period. The analysis of P and T axis directions along the subduction zone, using the Harvard Centroid Moment Tensor solutions since 1977, shows that the oceanic slab is in a downdip extensional regime. In contrast, the Punitaqui mainshock is related to compression resulting from the flexure of the oceanic plate, which becomes subhorizontal at depths of about 100 km. Analog strong motion data of the Punitaqui main event show that the greatest accelerations are on the horizontal components. The highest amplitude spectra of the acceleration is in the frequency band 2.5–10 Hz, in agreement with the energy band responsible for the collapsed adobe constructions. The isoseismal map derived from the distribution of observed damage show that a high percentage of destruction is due to the proximity of the mainshock, the poor quality of adobe houses and probably local site amplification effects.

Travel times of P and S from the global digital seismic networks: Implications for the relative variation of P and S velocity in the mantle

We present new data sets of P and S arrival times which have been handpicked from long‐period vertical and transverse component recordings of the various global seismic networks. Using events which occurred from 1976 to 1994 results in ∼38,000 globally well‐distributed measurements of teleseismic P and ∼41,000 measurements of S. These data are particularly useful for looking at the relative variation of S and P velocities in the lower mantle. We describe both the measurement techniques and the gross characteristics of the data sets. The size of our data sets allows us to exploit the internal consistency of the data to identify outliers using a summary ray analysis. Since the polarity of each arrival is also known, we can construct fault plane solutions and/or compare with polarities predicted by the Harvard centroid moment tensor solutions to further diagnose phase misidentification. This analysis results in ∼5% of the data being identified as outliers. An analysis of variance indicates that the S residual travel times are dominated by the effects of three‐dimensional structure but the P data have comparable contributions from noise and source mislocation effects. The summary ray analysis reveals the basic character of lower mantle structure, and there are large‐scale patterns in both the S and P data sets that correlate quite well with each other. This analysis suggests that on average, d ln vs/d ln vp is an increasing function of depth in the mantle going from a value of ∼1.7 at the top of the lower mantle to an apparent value of 4 near the base of the mantle. This latter extreme value of R seems to result mainly from data which sample one region in the lowermost mantle under the central Pacific, where large positive S residuals are associated with very small P residuals. Such an anomaly cannot be thermal in origin.

Compound rupture of the great 1998 Antarctic plate earthquake

The March 25, 1998, Antarctic plate earthquake ruptured a portion of the Antarctic plate more than 200 km west of its boundary with the Australian plate. The Harvard centroid moment tensor solution indicates that the earthquake was primarily a strike‐slip event, but the large non‐double‐couple component of the moment tensor suggests considerable complexity in the rupture process. We use a finite fault inversion method to determine details of the rupture process from teleseismic body waves recorded by the Global Seismic Network. The P waves are poorly fit by one or more subevents having only a strike‐slip mechanism. We find that the presence of a large oblique‐normal faulting subevent located to the east of the hypocenter is necessary to improve the fit. This subevent combines with a larger strike‐slip subevent to the west to comprise the main moment release in the earthquake and is the cause of the large non‐double‐couple component in the long‐period focal mechanism. The earthquake exhibited very high slip and high stress drop compared with most interplate strike‐slip events and the rupture was largely confined to the upper 15–20 km of the lithosphere. Both constituent focal mechanisms indicate that this part of the Antarctic plate is under NW‐SE oriented tension, although the origin of these stresses is unknown.

Delimitation of domains with uniform stress in the subducted Nazca plate

This work concentrates on the determination of detailed variations of stress in the descending Nazca plate and is based on the distribution of seismic activity and the orientation of P- and T-axes of fault plane solutions. International Seismological Centre (ISC) and Harvard Centroid Moment Tensor Solutions (HCMT) data were used. The method of Gephart and Forsyth [J. Geophys. Res. 89 (1984) 9305–9320] was applied to determine the state of stress. It is shown that the state of stress varies along the whole Wadati–Benioff zone, where 16 different domains with uniform state of stress were delimited. The state of stress tends to change more systematically with the depth of the Wadati–Benioff zone than laterally, and is connected with the morphology of the Wadati–Benioff zone. A decreased stability of the state of stress in the Wadati–Benioff zone was found in the depth range 50–300 km. The state of stress in the depth ranges 0–100 km and deeper than 500 km appears to be very stable. The physical interpretation of the observed phenomena is not the subject of this investigation.

Significant trends related to the slab seismicity and tectonics in the Burmese arc region from Harvard CMT solutions

A careful examination of 50 centroid-moment tensor (CMT) earthquake focal mechanisms in an east-west depth section in the Burmese arc region reveals some interesting trends. The P-axis azimuths of most of the mechanisms in the region of the east-dipping Indian lithospheric slab are, on average, oriented NNE. This indicates that the principal compressive stress direction in the Burmese arc region is commensurate with the direction of convergence of the Indian plate with respect to the Eurasian plate, rather than with respect to the Burmese plate. Another interesting observation is that the strike-slip and thrust type focal mechanisms along the slab do not coexist, but are distinctly segregated, the predominantly strike-slip type occurring in the upper half of the slab down to a depth of about 90 km, and the thrust type in the lower hair, below that depth level. The former possibly indicates shearing of the Indian plate past the Burmese plate, along the zone of contact, which is inferred down to a depth of 90 km. The latter is suggestive of a NNE-directed stress on the lower portion of the subducted slab, as it is dragged against the asthenosphere by the Indian plate.

The 1977 Three Kings Ridge earthquake ( Ms = 6.7): broad-band aspect of the source rupture

The 1977 Three Kings Ridge earthquake ( M s = 6.7) is an isolated event which took place in the South Fiji Basin. The centroid moment tensor (CMT) solution by the Harvard group shows a left-lateral strike-slip motion with a dip angle of 55° while the P-wave first motions require a nearly vertical fault plane. The centroid depth 17.4 km is deeper than the bottom of ordinary oceanic crust. Using long-period P and SH waveforms from the Worldwide Standardized Seismograph Network, we model the entire source rupture process with three subevents which span a 42 s duration. The best-fit double-couple focal mechanism obtained from the tensorial sum of the subevents is primarily a strike-slip with a nearly vertical fault plane subparallel to the Cook Fracture zone. The total moment release is (2.2–2.3) × 10 26 dyn cm, which is about 40% lower than the CMT estimate. The subevents' depth distribution is similar to the centroid depth (17 km), which is deeper than a typical oceanic Moho. Precise modeling of the waveform is also incompatible with a standard oceanic crustal structure, lending support to previous interpretations of the Three Kings Ridge as a fossil island arc. The discrepancy between our body waveform moment tensor solution and the Harvard CMT solution is dominated by the tensor components M rθ and M rθ . In the case of a shallow source, these components are notoriously unresolvable because of their vanishing excitation of seismic waves in the low-frequency range ( T ⩾ 45 s). However, the low-pass filtered body waves can be better modeled with a dip angle similar to that of the CMT solution. This observation suggests that the source process may include some level of evolution of the rupture mechanism with time. We interpret the earthquake as expressing a moderate level of internal deformation of the Australian plate, in the context of significantly different spreading rates in the Lau and Havre Basins.

Surface deformation and tectonic setting of the 13 March 1992 Erzincan earthquake, Eastern Turkey

AbstractThe earthquake of 13 March 1992 (Ms = 6·8) in the densely populated Erzincan region of eastern Turkey resulted in about 500 deaths, at least 2000 people injured and widespread destruction of buildings. The results are presented of field observations carried out within the Erzincan Basin immediately after the earthquake. The main area of surface deformation is about 55 km2 in extent and lies 15 km south‐east of Erzincan. Surface fractures with significant displacements were observed with orientations as follows: strike‐slip 095–110° (dextral), 040° (sinistral) and 090° (sinistral); extensional 150–180°. Most of the observed fractures are consistent with the Harvard centroid moment tensor solution for the main shock, which indicates north‐south compressional and east‐west extensional principal horizontal stress axes. It is suggested that the earthquake probably occurred as a result of dextral strike‐slip motion on a segment of the North Anatolian Fault Zone.

Downdip tensional earthquakes beneath the Tonga Arc: A double seismic zone?

Characteristics of seismicity at intermediate depth beneath the Tonga arc are studied. Special attention is paid to newly identified downdip tensional events because the dominance of downdip compression has been known in this area. We reexamine the relatively shallow (&lt;70 km) downdip tensional events, reported in the Harvard centroid moment tensor solutions, by modeling bodywave waveforms of World‐Wide Standard Seismograph Network seismograms and conclude that they are in fact intraplate downdip tensional events. Furthermore, these downdip tensional events are consistently located below and seaward of downdip compressional events and appear to constitute a double seismic zone. The result of a relative relocation technique applied to these earthquakes also supports this observation, which was originally based on routinely determined epicenters. The subducting slab beneath the Tonga arc has been considered as a prototype of a compressionally loaded slab. The existence of a double seismic zone in Tonga, therefore, would suggest that double seismic zones, located at intermediate depth in subducting slabs, are more common than previously thought.