Seventeen ocean-bottom seismometers recorded useful horizontal dis- ...... 2000], consistent with a reduction in the horizontal align- ..... WorkshopReports.html).
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. B7, 2352, doi:10.1029/2002JB002217, 2003
Imaging the transition between the region of mantle melt generation and the crustal magma chamber beneath the southern East Pacific Rise with short-period Love waves Robert A. Dunn Department of Geological and Geophysics, SOEST, University of Hawaii at Manoa, Honolulu, Hawaii, USA
Donald W. Forsyth Department of Geological Sciences, Brown University, Providence, Rhode Island, USA Received 22 September 2002; revised 19 March 2003; accepted 3 April 2003; published 30 July 2003.
[1] Beneath the southern East Pacific Rise resides a crustal magma chamber confined to
the immediate vicinity of the rise axis, yet the mantle melt production region is broad and asymmetric. Short-period (4–17 s) Love waves propagating along the rise provide a means of probing the transition from the asymmetric melt production region to the risecentered crustal magma chamber. Low velocities beneath the rise act as a waveguide for Love wave energy. We model this complex wave propagation with a Gaussian beam representation as part of our nonlinear inversion for velocity structure. Our results show that the asymmetry exists at the crust-mantle interface, or Moho, and increases downward; the lowest velocities are located beneath the rise at the Moho; beneath the rise there is an asymmetric gap in the high-velocity ‘‘lid’’ at the top of the mantle; and there are lower asthenospheric velocities beneath the Pacific plate than the Nazca plate. The transition from the broad upwelling region to the narrow crustal magmatic system occurs over a 15-km-depth interval at the top of the mantle. Our observations indicate higher temperatures and greater melt production beneath the Pacific plate. East of the rise, melting appears to abruptly shut off at shallow levels, perhaps due to a component of downward mantle flow. We suggest that melt percolates upward, increasing in concentration, until reaching a permeability barrier beneath the lithosphere, it then moves up along the lithosphere to accumulate at the Moho beneath the rise. West of the rise, higher temperatures and upward percolation of melt delay growth of the INDEX TERMS: 3035 Marine Geology and Geophysics: Midocean ridge processes; 7218 lithosphere. Seismology: Lithosphere and upper mantle; 7255 Seismology: Surface waves and free oscillations; 8120 Tectonophysics: Dynamics of lithosphere and mantle—general; 8180 Tectonophysics: Tomography; KEYWORDS: MELT experiment, Love waves, Gaussian beam, mid-ocean ridge, mantle Citation: Dunn, R. A., and D. W. Forsyth, Imaging the transition between the region of mantle melt generation and the crustal magma chamber beneath the southern East Pacific Rise with short-period Love waves, J. Geophys. Res., 108(B7), 2352, doi:10.1029/2002JB002217, 2003.
1. Introduction [2] The generation of new oceanic lithosphere and crust at mid-ocean ridges involves four key elements: the mantle flow pattern, the thermal structure, the distribution of melting, and melt migration in the system. A long debated question is whether the flow pattern is passively driven by viscous drag of the overlying plates, resulting in a pattern of broad mantle upwelling and melting [e.g., Reid and Jackson, 1981; Phipps Morgan and Forsyth, 1988], or whether there is a dynamic component of upwelling arising from low viscosities and buoyancy due to melt, resulting in narrow mantle upwelling and melting [e.g., Buck and Su, 1989; Copyright 2003 by the American Geophysical Union. 0148-0227/03/2002JB002217$09.00
ESE
Scott and Stevenson, 1989]. In addition, there is the issue of how melt gets from the broad mantle source region to the narrow crustal magma chamber and whether it is drained off quickly, leaving only a small melt fraction present at any time within the mantle. Understanding these processes requires direct measurement of mantle structure beneath spreading centers. [ 3 ] The Mantle Electromagnetic and Tomography (MELT) experiment was carried out in 1995– 1996 [MELT Seismic Team, 1998] to distinguish between competing models for mantle flow, thermal structure, and melt distribution by producing seismic and electromagnetic images of the upper mantle beneath the fast spreading, southern East Pacific Rise (EPR) near 17S. Initial results from body wave and surface wave studies reveal a 150 – 200-km-wide region of low P and S wave velocities in the mantle that
8-1
ESE
8-2
DUNN AND FORSYTH: IMAGING TRANSITION REGIONS
is consistent with a broad zone of melting and a passive flow pattern; less than 2% melt is estimated to reside in the upwelling zone [Forsyth et al., 1998; Hammond and Toomey, 2003] indicating that melt is drained off relatively quickly. Most striking in almost all geophysical measurements is a strong asymmetry across the ridge. To the west of the ridge, the low-velocity region is more pronounced [Forsyth et al., 1998; Hammond and Toomey, 2003], shear wave splitting delays are larger [Wolfe and Solomon, 1998], the seafloor subsides more slowly [Cochran, 1986], seamounts appear in greater abundance, the mantle Bouguer gravity anomaly is larger [Scheirer et al., 1998], and the mantle has higher electrical conductivity [Evans et al., 1999]. The transition across the ridge from a relatively conductive mantle west of the ridge to a highly resistive mantle east of the ridge is sharp, taking place over at most a few tens of kilometers [Evans et al., 1999]. The cause of the asymmetry is uncertain, but may be related to asymmetric mantle flow and higher temperatures and greater melt production in the mantle to the west of the rise coming from the Pacific superswell region [MELT Seismic Team, 1998; Toomey et al., 2002; Conder et al., 2002]. Although crustal accretion is roughly symmetric in this region, the ridge axis migrates rapidly westward (32 mm/yr) in the hot spot coordinate frame, thus providing a possible explanation. However, numerical models of mantle flow are unable to produce significant asymmetry in the upwelling region through ridge migration alone [Toomey et al., 2002; Conder et al., 2002]. [4] One approach to furthering our understanding of the ridge system is through refinement of the seismic models. For example, the uppermost 30 km of mantle is not well constrained by previous studies and lateral velocity gradients are much less certain than the general distribution of low velocities in the upwelling region. By further refining our seismic models, we can begin to address such issues as the thickness of the oceanic lithosphere near the ridge (Is it asymmetric across the ridge?), the transition from broad mantle upwelling to the narrow crustal magmatic system (Is it abrupt? At what depth does the asymmetry in structure begin?), and the distribution and concentration of melt in the uppermost mantle (Is there an abrupt change across the ridge axis as suggested by the conductivity structure?). [5] Here we develop a method to solve for the shear wave structure of the upper mantle and crust using short-period Love waves recorded by the MELT seismic array. Shortperiod surface waves have the advantage of being sensitive to relatively fine-scale (10– 20 km) variations in upper mantle and crustal structure. In the presence of lateral velocity variations, surface wave energy travels along paths that can deviate strongly from great circle paths and amplitudes can be severely distorted. Rather than avoid these effects by using longer period waves, as is usually the case, we develop a method that explicitly includes the effects of lateral refraction and multipath interference. This allows us to investigate the small-scale variations in subridge structure and to exploit a unique feature of the southern EPR: beneath this long straight section of ridge the low velocities in the upwelling zone produce a waveguide that traps and guides surface wave energy. Waveforms of seismic energy propagating in this guide are very sensitive to the width, velocity, and velocity gradients
within the guide. In this study we use the phase, group arrival time, and amplitude of waveforms to investigate the shear velocity structure beneath the EPR and surrounding mantle to 50 km depth beneath the seafloor and 100 km to either side of the rise with higher resolution near the ridge axis than previously possible. The Love waves also supply some lower resolution constraints on structure at greater depths and distances from the rise.
2. Experiment Design and Data Processing [6] To investigate the nature of the uppermost mantle beneath the southern EPR, we model the waveforms of short-period Love waves and solve for the phase velocity and shear velocity structures. The seismic data were collected as part of the MELT seismic experiment in 1995– 1996 during a 6-month deployment of 51 ocean-bottom seismometers [MELT Seismic Team, 1998]. Seventeen ocean-bottom seismometers recorded useful horizontal displacement information necessary for measuring Love wave arrivals. These instruments form a network that extends 200 km to either side of the EPR near 17S (Figure 1). During recording the data were sampled at 16 samples/s and an antialiasing filter was applied before digitization. After the instruments were recovered, the time base was corrected for clock drift and the instrument responses from different types of ocean-bottom seismometers [MELT Seismic Team, 1998] were corrected to match that of the single, most common instrument [Webb et al., 2001]. We solved for the orientations of the horizontal components of each instrument and corrected the data for instrument tilt. If the instrument is slightly tilted, it changes the transfer function of the instrument response, not just the relative amplitude of the two horizontals. However, at long periods (greater than the natural period of the instrument) the effect of instrument tilt is a DC shift in gain of the tilted horizontal component, with no significant phase shift [Webb et al., 2001]. We calibrate the stations relative to one another by assuming that horizontal amplitudes of nearby stations, averaged over a variety of events or phases, are constant. Some instruments automatically releveled every 2 weeks to account for settling, but there may remain on the order of a degree deviation from horizontal. Therefore to determine an instrument’s tilt, we used large earthquakes that occurred just after the regional events of interest, thereby capturing an instrument’s orientation at the time in which it recorded the data used in our analysis; we used a large teleseismic event (Ms 8.0) that occurred to the northwest in the Kuriles and another large event that occurred on the Clipperton Fracture Zone to the north. We solved for the orientation and relative amplitudes of the horizontal components to achieve orthogonal long-period Love and Rayleigh waves, uniform amplitudes of SS waves from the Kurile event and S waves from the Clipperton event across the array, and constant relative amplitude of radial and vertical components of long-period Rayleigh waves (the amplitude of vertical components is not affected by small tilt). We estimate that we can correct the amplitude response of the horizontal components to within about 25% of a standard response for a truly horizontal, uniform sensor. [7] During the experiment a swarm of earthquakes occurred south of the experiment array along transform
DUNN AND FORSYTH: IMAGING TRANSITION REGIONS
ESE
Figure 1. The southern East Pacific Rise between the Garrett Fracture Zone and the Easter Microplate. Curves are isobaths at 200 m intervals; the spreading axis is the linear feature oriented approximately north-south through the center of this figure. An array of 17 ocean-bottom seismometers (indicated by the numbered triangles), deployed as part of the MELT experiment, recorded Love wave data generated by earthquakes located along the northern boundary of the Easter microplate. Seafloor spreading along this section of the EPR is asymmetric. Relative to the plate boundary the Pacific plate moves away at 69 mm/yr and the Nazca plate moves away at 76 mm/yr. In the hot spot reference frame, however, the plate boundary is moving at 32 mm/yr in a WNW direction.
8-3
ESE
8-4
DUNN AND FORSYTH: IMAGING TRANSITION REGIONS
Figure 2. Transverse component seismograms of Love waves at three stations from the larger source event. These seismograms have been corrected for instrument response and prewhitened with a common filter. Time is relative to origin time of the event. Note differences in waveform amplitudes at different periods due to focusing and interference as well as progressively greater travel times for comparable periods from site 43 (bottom) to site 40 (middle) to site 12 (top). These time differences are much greater than that would be expected for changes in epicentral distance (631, 643, 662 km, respectively), indicating slower group velocities near the ridge axis (see Figure 1). faults that bound the Easter microplate. These crustal-level earthquakes have left-lateral strike-slip mechanisms on a vertical plane [Forsyth et al., 2003]. For this study, we chose the two largest events, which were well separated in time from other events in the swarm. These events are uniquely situated such that surface wave energy traveling from the source to the instrument array samples 700 km of the EPR over a region up to 400 km wide (Figure 1). [8] Love wave energy generated by these events has a useful range of 4 – 17 s. At these periods, the energy samples the oceanic crust and upper 150 km of the mantle. At longer periods, ocean currents on the seafloor generate noise that prevents measurement of useful signal. At shorter periods, observations are limited by source excitation, scattering, and attenuation; these events have only incoherent arrivals at periods 3 s. These very short periods are prone to extreme scattering effects due to variability in the upper crust and overlying sediments, which makes their analysis difficult. Everywhere in the study area sediments are
