Distinct regional differences in crustal thickness ... - Wiley Online Library

Geochemistry Geophysics Geosystems

3

G

Article Volume 7, Number 4 19 April 2006 Q04011, doi:10.1029/2005GC001119

AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society

ISSN: 1525-2027

Distinct regional differences in crustal thickness along the axis of the Mariana Trough, inferred from gravity anomalies Kazuya Kitada Graduate School of Science and Technology, Kobe University, 1-1 Rokkodai, Nada, Kobe 657-8501, Japan ([email protected])

Nobukazu Seama Research Center for Inland Seas, Kobe University, 1-1 Rokkodai, Nada, Kobe 657-8501, Japan Also at Graduate School of Science and Technology, Kobe University, Nada, Kobe, Japan

Toshitsugu Yamazaki Institute of Geology and Geoinformation, Geological Survey of Japan, AIST, 1-1-1 Higashi, Tsukuba 305-8567, Japan

Yoshifumi Nogi National Institute of Polar Research/Sokendai, 1-9-10 Kaga, Itabashi, Tokyo 173-8515, Japan

Kiyoshi Suyehiro Japan Agency for Marine-Earth Science and Technology, 2-15 Natsushima, Yokosuka 237-0061, Japan

[1] We have compiled extensive gravity and bathymetry data for the whole Mariana Trough, which were collected during several Japanese scientific cruises over the last few years. This study aims to clarify the lateral distribution of the local differences in geochemical signatures, which have been observed locally in the Mariana Trough. Shipboard free-air gravity anomaly data from eight Japan Agency for Marine-Earth Science and Technology (JAMSTEC) cruises were compiled with those crossover errors of 2.85 mgal. Mantle Bouguer anomalies (MBA) were calculated by subtracting the predictable gravity signal due to the seawater/crust and crust/mantle density boundaries. The crustal thickness variation along the spreading axis was estimated from the MBA. Different features in crustal thickness, its variation, and segment length for each segment, allow us to identify four distinct regional differences in magmatic activity along the spreading axis of the Mariana Trough. Segment in region A (to the north of 20°350N) shows the largest sectional dimensions of crust along the axis and it is probably affected by an additional supply from island arc magma sources. A variety of crustal thickness values and of along-axis crustal thickness variations in region B (between 15°380N and 20°350N) suggests two types of segments. One is similar to a slow spreading ridge segment that has a plume-like mantle upwelling under the spreading axis, and the other is a magma-starved segment. Region C (between 14°220N and 15°380N) is a less magmatic region (individual crustal thickness averages of 3.4–4.1 km). Region D (to the south of 14°220N) has higher individual crustal thickness averages of 5.9–6.9 km, suggesting higher magmatic activity with a sheet-like mantle upwelling under the spreading axis. Different features in the MBA for off-axis areas suggest that these four regions have existed since the Mariana Trough started spreading. Moreover, comparison between our results of crustal thickness and previous geochemical results indicates that less magmatic spreading segments with thin crust, which are locally distributed in both regions B and C, probably result from mantle source depleted of water and incompatible elements. This suggests that lateral compositional variation of water and incompatible elements exists on a segment scale in the mantle source beneath the spreading axis of the Mariana Trough. Components: 8627 words, 8 figures. Keywords: Mariana Trough; crustal thickness; gravity anomaly; back-arc basin; spreading center; melt production. Copyright 2006 by the American Geophysical Union

1 of 14


3

G

kitada et al.: differences in crustal thickness

10.1029/2005GC001119

Index Terms: 1219 Geodesy and Gravity: Gravity anomalies and Earth structure (0920, 7205, 7240); 3001 Marine Geology and Geophysics: Back-arc basin processes; 3035 Marine Geology and Geophysics: Midocean ridge processes. Received 14 August 2005; Revised 29 December 2005; Accepted 21 February 2006; Published 19 April 2006. Kitada, K., N. Seama, T. Yamazaki, Y. Nogi, and K. Suyehiro (2006), Distinct regional differences in crustal thickness along the axis of the Mariana Trough, inferred from gravity anomalies, Geochem. Geophys. Geosyst., 7, Q04011, doi:10.1029/2005GC001119.

1. Introduction [2] The Mariana Trough is a typical active, backarc basin that has been well studied. The crescent shaped Mariana Trough is bounded by the West Mariana Ridge (remnant arc) and the Mariana Island Arc (active arc to the east). The tectonics of the Mariana Trough was revealed by several studies of geomagnetic anomalies [e.g., Martıńez et al., 2000; Yamazaki et al., 2003; Seama et al., 2002], and several different spreading features were reported. In the northern Mariana Trough, the trough develops southward from incipient rifting to seafloor spreading, with the rifting-tospreading transition occurring at 22°N, proposed from evidence of clear magnetic lineations, gravity anomalies, and seafloor-spreading fabrics in the bathymetry [Yamazaki et al., 2003]. The seafloor spreading started prior to 5 Ma at latitudes between 19°N and 20°N, and at 4 Ma at latitudes between 20°N and 21°300N [Yamazaki et al., 2003]. The spreading half rates in the western side of the spreading center of the northern Mariana Trough were 20 to 30 km/Myr before 2.58 Ma south of 21°30 0N, and also 20 to 30 km/Myr during Matuyama Chron north of 21°300N, with an average spreading half rate of 10 km/Myr or less during the Brunhes Chron for the same region [Yamazaki et al., 2003]. In the central Mariana Trough, an axial valley commonly observed at a slow spreading ridge is well developed [Hussong and Uyeda, 1982; Kong et al., 1992]. In the central Mariana Trough between 16°N and 19°N, seafloor spreading started at 6 Ma (being the oldest in the Mariana Trough) with spreading half rates of 20 km/Myr [Iwamoto et al., 2002]. The spreading half rate has decreased with time, from 25 km/Myr during Gauss Chron to 9 km/Myr during Brunhes Chron in the central Mariana Trough at 18°N [Yamazaki and Stern, 1997]. In the southern Mariana Trough, the seafloor morphology has a shallow bathymetry (less than 3000 m) and a fast spreading type axial high feature [Martıńez et al., 2000; Hasegawa et al., 2000]. South of 14°N,

seafloor spreading started at 3 Ma with a spreading half rate of 35 km/Myr [Seama et al., 2002]. Between 11°500N and 13°400N, Martıńez et al. [2000] reported that the spreading half rate is less than 32 km/Myr. [3] Local differences in geochemical signatures have also been observed in the Mariana Trough. Gribble et al. [1996] conducted a geochemical survey along the axis of the Mariana Trough between 15°N and 17°N, and they reported that higher melting along the axis has occurred between 16°N and 17°N than between 15°N and 16°N. They also suggested that the degree of melting has a strong correlation with higher abundances of water as well as relative abundances of large ion lithophiles elements (LIL) and light rare earth elements (LREE). Stolper and Newman [1994] reported that an increase in the degree of melting in the vicinity of 18°N in the Mariana Trough which is due to the lowering of the solidus temperature of mantle peridotite, caused by an increase in the water component. Moreover, Gabbros and ultramafic rocks that have experienced low partial melting, were recovered from the Central Graben (20°N, 144°E), leading to the suggestion of a magma-starved condition under the spreading axis in the Central Graben [Stern et al., 1996, 1997; Ohara et al., 2002]. However, the lateral distributions of such local point features in the Mariana Trough are still unknown. [4] In this paper, we will show the comprehensive crustal structure along the axis of the Mariana Trough through the analysis of gravity data in order to clarify the lateral distribution of different melting process. The Mariana Trough shows several different spreading features, but an indication of the general crustal structure was obtained by some gravity analyses [e.g., Ishihara and Yamazaki, 1991; Yang et al., 1992; Yamazaki et al., 2003] and seismic experiments [e.g., Ambos and Hussong, 1982; Takahashi et al., 2004] only in limited areas of the Mariana Trough. The local differences were reported only from point geo2 of 14


3

G


chemical samples [Gribble et al., 1996; Stolper and Newman, 1994; Stern et al., 1996, 1997; Ohara et al., 2002]. We will first show the data acquisition. Japanese science groups have conducted several surface geophysical surveys over the Mariana Trough since 1990s. The almost complete multinarrow beam bathymetry and gravity data for this study were collected aboard the Japanese research vessels Kairei and Yokosuka, operated by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC). Secondly, we will present results of the free-air and mantle Bouguer gravity anomalies (MBA) for the whole of the Mariana Trough, calculated from extensive gravity and bathymetry data. Then, we will estimate the crustal thickness variation along the spreading axis from the calculated MBA in the Mariana Trough. Finally, the individual crustal thickness and its variation will reveal the distinct regional differences in magmatic activity and mantle upwelling pattern along the axis of the Mariana Trough.

2. Data Acquisition and Analysis [5] Free-air gravity anomaly grid data was derived from two sources, namely shipboard gravity anomaly data and data from satellite altimetry [Sandwell and Smith, 1997]. Shipboard free-air gravity anomalies were calculated by subtracting the normal gravity field data from observed absolute gravity field data, with a correction applied for the Eo¨tvo¨s effect using Differential Global Positioning System (DGPS) data. The gravity data for this study were collected during the KR98-12, KR00-03, KR02-01, KR02-14 and KR03-13 cruises by the R/V Kairei [Yamazaki et al., 1999; Seama et al., 2003; Yamazaki et al., 2003], and the YK99-11, YK01-11 and YK03-09 cruises by the R/V Yokosuka [Mitsuzawa et al., 2000; Goto et al., 2002; Yamazaki et al., 2003]. Sea surface gravity field measurements were obtained every 60 seconds using KSS-31 (BODENSEEWERK Perkin-Elmer GmbH) onboard the R/V Kairei, and every 10 seconds using S-63 type (LaCoste & Romberg Gravity Meters, Inc.) onboard the R/V Yokosuka, in conjunction with simultaneously recording DGPS, ship heading, and speed information. The survey line spacing during all cruises ranged from 5 to 6 arc min. We compiled these calculated free-air gravity anomaly data, and the instrumental drift correction was applied to these data. The crossover errors were also estimated, the standard deviation of these errors being 2.85 mgal.

10.1029/2005GC001119

The gravity data along the survey lines were interpolated on an equally spaced grid using a minimum curvature algorithm [Smith and Wessel, 1990] with a 1 arc min grid interval. Free-air gravity anomalies derived from satellite altimetry [Sandwell and Smith, 1997] with 2 arc min spacing, were used to fill the gravity data gap outside the region surveyed. We merged shipboard free-air gravity anomaly grid data with free-air gravity anomaly grid data from satellite altimetry. The method used for merging the two data sets was fourfold: (1) standardizing the grid intervals by transforming the free-air gravity grid data from satellite altimetry (2 arc min) into grid data with a 1 arc min grid interval, using a minimum curvature algorithm [Smith and Wessel, 1990], (2) calculating planar trends of shipboard and satellite free-air gravity grid data in the region surveyed, by means of the least squares method, (3) subtracting the shipboard planar trend from the satellite planar trend in this region, and adding this difference to the shipboard free-air gravity data, resulting in a new shipboard free-air gravity data set, and (4) combining the new shipboard free-air gravity anomaly grid data in the region surveyed, with grid data from satellite altimetry (transformed in the first procedure) from outside the region surveyed. By performing this merging procedure, the average of the differences between shipboard and satellite free-air gravity grid data (for the merged region) is improved from 8.7 mgal to 0.0 mgal. The free-air gravity anomaly map resulting from this procedure is shown in Figure 1. [6] Mantle Bouguer anomalies (MBA) were calculated by subtracting the predictable gravity signal (due to the seawater/crust and crust/mantle boundaries) from the obtained free-air gravity anomalies (Figure 1). The bathymetric data for this calculation were collected during the KR97-11, KR98-12, KR00-03, KR02-01, KR02-14 and KR03-13 cruises by the R/V Kairei [Yamazaki et al., 1999; Seama et al., 2003; Yamazaki et al., 2003], the Y96-13, YK99-11, YK01-11 and YK03-09 cruises by the R/V Yokosuka [Honsho et al., 1997; Mitsuzawa et al., 2000; Goto et al., 2002; Yamazaki et al., 2003], and KH92-1 by the R/V Hakuho-Maru [Kong et al., 1992]. Most of the Mariana Trough between 12°N and 22°300N was mapped using the multinarrow beam systems SeaBeam2112 (SeaBeam, Inc., USA) and HS-10 (Furuno electric co., ltd., JAPAN). The navigation system used for each cruise was DGPS. The bathymetric data outside the surveyed area is based on the predicted bathymetry [Smith and Sandwell, 3 of 14


3

G


10.1029/2005GC001119

1997] with a 2 arc min grid interval. The predicted bathymetry grid data were interpolated on an equally spaced grid with a 1 arc min grid interval, using a minimum curvature algorithm [Smith and Wessel, 1990]. The bathymetric map for the Mariana Trough is illustrated in Figure 2. The MBA computation was carried out, using the merged free-air gravity anomalies (Figure 1) and the swath bathymetric data (Figure 2), by applying the Fast Fourier Transform (FFT) method of Parker [1972]. Assumptions for the MBA calculations are (1) the crustal thickness is a constant of 6 km, (2) the seawater, crust, and mantle densities are 1030, 2700, and 3300 kg/m3, respectively, and (3) the mean water depth in the Mariana Trough is 3.8 km. The input bathymetry grid data was mirrored at the boundaries of the calculation areas to prevent discontinuities at the edge of these areas, because the structure is assumed to be periodic in this applied FFT method. The calculated MBA map is displayed in Figure 3. It is important to note that predicted bathymetry [Smith and Sandwell, 1997] was only used to avoid the data gap for the area outside the region surveyed in order to reduce the errors in the calculation of MBA. Therefore we did not use this MBA result from the outside area for the interpretation, because the predicted bathymetry is derived from satellite altimetry [Sandwell and Smith, 1997], and this renders the MBA meaningless as a gravity anomaly. Moreover, the MBA analysis assumes that the seafloor represents the top of the oceanic basement. This is a good approximation for the axis and most of the western basin in the Mariana Trough, but it becomes poorer to the east where volcaniclastic sedimentation increases. Thus we do not discuss the MBA results from the eastern off-axis areas. The MBA map (Figure 3) was masked by a 5 arc min grid interval to show which grid cells were valid.

Figure 1. Free-air gravity anomaly map of the Mariana Trough contoured at 5 mgal intervals. The grid interval is 10 10. The gravity data are from the shipboard gravimeter with satellite altimetry [Sandwell and Smith, 1997]. The white dashed lines and the red lines represent the locations of the ridge segmentation boundaries and present spreading axes, respectively. The black thin dashed lines are ship tracks during all cruises. The inset shows the location of the Mariana Trough and adjacent regions. The red rectangle within illustrates the study area.

[ 7 ] The crustal thickness variation along the spreading axes was estimated from the calculated MBA (Figure 3) for the Mariana Trough. The calculations were based on the method of Kuo and Forsyth [1988]. The MBA were downward continued to a depth of 9.8 km below sea level (mean water depth of 3.8 km plus the estimated mean crustal thickness of 6 km), and then converted to crustal thickness using the density contrast between the crust and mantle of 600 kg/m3. We only used the crustal thickness data along the spreading axes for the interpretation, because the crustal age along axes are all the same, indicating density anomalies associated with thermal cooling of the lithosphere are also the same. Consequently, 4 of 14


3

G


Figure 2. Multinarrow beam bathymetric map of the Mariana Trough. The grid interval is 10 10. The bathymetric data outside the area surveyed are based on the predicted bathymetry [Smith and Sandwell, 1997]. The white dashed lines, the red lines, and the black thin lines correspond with those in Figure 1. The bathymetric data were used for calculating the mantle Bouguer anomalies (Figure 3).

10.1029/2005GC001119

Figure 3. Mantle Bouguer gravity anomaly map of the Mariana Trough contoured at 5 mgal intervals. The grid interval is 10 10. The computation was carried out based on the merged free-air gravity anomaly (Figure 1) and the swath bathymetric data (Figure 2) by using the fast Fourier transform method of Parker [1972]. The map was masked by 50 50 grid interval to show only grid cells which are valid. The white dashed lines, the red lines, and the black thin lines correspond with those in Figure 1. 5 of 14


3

G


10.1029/2005GC001119

Figure 4. Along-axis variations of crustal thickness and mantle Bouguer gravity anomalies (MBA). The seafloor and the Moho depth along the spreading axes are plotted against the latitude in the Figure 4 (top), and the MBA along the spreading axes are plotted in the Figure 4 (bottom). The crustal thickness variations were estimated from the MBA by the method of Kuo and Forsyth [1988]. The red dashed lines show the boundaries of the spreading segments. The numbers between the Figures 4 (top) and 4 (bottom) show the labeled segment numbers from north to south, plotted in Figure 5 – 7. The red and blue arrows show the slow spreading ridge segment type and magma-starved segment type in region B, respectively.

the relative variation in crustal thickness is available. The spreading axes and segmentation boundaries for each segment were defined by shaded relief images of the bathymetric map, analyses of the geomagnetic anomalies and the side-scan images [Yamazaki et al., 2003; Seama et al., 2002]. The seafloor depth, the crustal thickness, and the MBA along the spreading axes were plotted against the latitude in Figure 4. Sixteen spreading segments exist along the axes of the Mariana Trough, labeled in succession from 1 in the north to 16 in the south.

3. Results [8] Four distinct regions which have different features in crustal thickness were identified along the axis of the Mariana Trough (regions A, B, C and D from north to south). Region A (segment 1) is north of 20°350N, region B (segments 2–11) is between 15°380N and 20°350N, region C (segments 12 – 14) is between 14°220N and 15°380N, and region D (segments 15 and 16) is south of 14°220N, as illustrated in Figure 4. Features were identified for each segment by normalized varia-

tion in crustal thickness (crustal thickness variation is normalized by its segment length) versus crustal thickness (Figure 5), and by variation in crustal thickness versus segment length (Figure 6). Region A has an average crustal thickness of 5.0 km (Figure 5), a large pronounced value for the variation in crustal thickness along its axis (7500 m) (Figure 6), and normalized variation in crustal thickness of 46 m/km (Figure 5). Ten segments in region B were represented in terms of their various crustal thickness values, and their various along-axis crustal thickness variations normalized by segment length. Having calculated the average crustal thickness within an individual segment, these individual crustal thickness averages range from 2.9 to 5.8 km (Figure 5). The variations in crustal thickness along the ten axes range from 900 m (minimum variation) to 6600 m (maximum variation) (Figure 6). Normalized variations in crustal thickness are 26–120 m/km per segment in region B (Figure 5). Three segments in region C have individual crustal thickness averages that range from 3.4 to 4.1 km (Figure 5), while the variations in crustal thickness along the three axes range from 1300 m (minimum variation) to 2100 m 6 of 14


3

G


10.1029/2005GC001119

Figure 5. Normalized variation in crustal thickness versus crustal thickness. Crustal thickness is represented by average crustal thickness within an individual segment (circles) with the maximum and minimum crustal thickness values within each segment (associated bars). The crustal thickness variation is calculated as difference between the maximum and minimum crustal thickness values within a segment, and the variation is normalized by its segment length. The green, red, blue, and yellow symbols show regions A, B, C, and D, respectively. The attached symbol numbers represent the segment number. The result of segment 2 is plotted as a triangle because an anomalous value was calculated at 20°N along the axis. The gray dashed lines show the upper limit of the average crustal thickness in region C and the lower limit of the average crustal thickness in region D.

(maximum variation) (Figure 6), these two respective indices being lower than those of other regions in the Mariana Trough. The normalized variations in crustal thickness are 34–60 m/km per segment in region C (Figure 5). The two segments of region D have individual crustal thickness averages that range from 5.9–6.9 km, rendering this the highest region in the Mariana Trough (Figure 5). The variations in crustal thickness along the two axes range from 2100 to 3000 m (Figure 6), and the normalized variations in crustal thickness are 28–30 m/km in region D (Figure 5), these two respective indices being lower than those of other regions in the Mariana Trough.

4. Discussion [9] Segment 1 in region A shows the largest sectional dimensions of crust along the axis (variation in crustal thickness of 7500 m and segment length of 160 km; Figure 6) compared to the other segments. We interpret that this segment is probably affected by an additional abundant magma supply from an active island arc magma sources. The crustal thickness of the segment 1 does not simply have a maximum at the center of its axis, but has two relative maxima (both 8 km

thick) at the points 21°200N and 21°550N (Figure 4) corresponding to the locations of discrete island arc volcanoes which exist only 40 km away from the spreading axis (Figure 2). This suggests that the spreading axis of this segment is affected by an additional magma supply from magma sources of the active Mariana Island Arc. Yamazaki et al. [2003] originally proposed that by analyzing freeair gravity data from satellite altimetry [Sandwell and Smith, 1997] in the northern Mariana Trough between 18°300N and 24°N, and our further analysis agrees well with their proposal. Their proposal is based on the following evidences that (1) the mantle Bouguer anomalies (MBA) of this segment is much lower than other segments in the aforementioned latitudes, (2) the spreading axis of this segment is located nearer to the active island arc in comparison with other segments in this latitude range, and (3) the MBA low continues to the island arc. Moreover, our interpretation is also supported by a correlation between amount of magma supply under the spreading axes and distance from the arc volcanic front in the Lau basin [Martıńez and Taylor, 2002]. They suggested the Valu Fa Ridge that is 40–60 km away from the island arc, is affected by an additional magma supply from island arc magma sources, based on high magmatic 7 of 14


3

G


10.1029/2005GC001119

Figure 6. Variation in crustal thickness within a segment versus segment length along the axis. The gray dashed line shows the trend that has a relatively smaller variation of 30 m/km in crustal thickness. Symbol numbers and colors are the same as for Figure 5.

activity features such as a relatively shallower depth, a lower MBA value, and greater crustal thickness of 7.5–9 km. Segment 1 in region A is only 40 km away from the island arc. Therefore this segment 1 should be affected by the additional magma supply from the active island arc magma sources if amount of magma supply under the spreading axes similarly correlates with the distance from the arc volcanic front. [10] Ten segments in region B showed a variety of crustal thickness values and of along-axis crustal thickness variations, which is classified into two different types of segments. To characterize this feature, we compared the plot of along-axis variation in MBA versus segment length in region B with the same plot for bulls-eye gravity anomalies reported from the studies of other slow spreading ridges [Detrick et al., 1995; Lin et al., 1990] (Figure 7). The calculation of the MBA variation is based on the method of Lin et al. [1990]. The grey dashed trend line in Figure 7 was determined from the data for a slow spreading ridge [Detrick et al., 1995, Lin et al., 1990] by means of the least squares method. Most segments in region B represented similar axial variation in MBA to the trend for slow spreading ridge segments, while segments 2, 5, 9 and 11 are dissimilar. One type of segment displays similar axial variation in MBA to the trend for bulls-eye features from slow spreading ridge segments [Kuo and Forsyth, 1988; Lin et al., 1990], suggesting the existence of a plume-like mantle

upwelling under the spreading axis in these segments. The other segment type displays that the axial variation in MBA are out of trend for slow spreading ridge segments and that the maximum crustal thickness (5.0–5.5 km) in the vicinity of the midpoint along the axis are lower than other segments from region B (Figure 5), suggesting that these four are magma-starved segments. The values of MBA among these less magmatic segments show a difference. Segments 5, 9 and 11 display relatively small values of axial variations in MBA (0.08 to 0.14 mgal/km), while segment 2 displays relatively high values (0.61 mgal/km) (Figure 7), which results from a extremely high MBA value of 95 mgal at the end of segment known as the Central Graben (20°N, 144°E). This high MBA in this Central Graben suggests the existence of a very thin oceanic crust and/or highdensity mantle material under the spreading axis, leading negative values for the estimated crustal thickness in the calculation (Figure 4). This interpretation is supported by rock sample analyses: Gabbros and ultramafic rocks that have experienced low partial melting were recovered from the Central Graben, leading to the suggestion of a magmastarved condition under the spreading axis in this graben [Stern et al., 1996, 1997; Ohara et al., 2002]. [11] Three segments in region C have individual crustal thickness averages that range from 3.4 to 4.1 km (Figure 5). This leads to the interpretation 8 of 14


3

G


10.1029/2005GC001119

Figure 7. Relationship between the along-axis variation in mantle Bouguer anomalies (MBA) within a segment and segment length along the spreading axis. The calculation of the MBA variation is based on the method of Lin et al. [1990]. The colors of the symbols show the data sources: red, blue, and green symbols are from region B in this study, Lin et al. [1990], and Detrick et al. [1995], respectively. The gray dashed line was determined from the data for a slow spreading ridge [Detrick et al., 1995; Lin et al., 1990] by means of the least squares method. Numbered symbols are the same as for Figure 5.

that region C is a less magmatic region with a low magma supply. Moreover, the maximum crustal thickness at the vicinity of the midpoint along the axis ranges from 4.4 to 4.7 km for the three segments, which is also much lower than those of other segments in the Mariana Trough. In addition, the normalized variations in crustal thickness represented relatively small values of 34–60 m/km per segment in region C (Figure 5). These results, namely lower maximum value of crustal thickness and small variation in crustal thickness, indicate a low magma supply, additionally suggesting that magmatic activity under the whole spreading axes in region C is lower than other segments in the Mariana Trough. [12] We interpret higher individual crustal thickness averages (5.9–6.9 km) and smaller normalized variations in crustal thickness (28–30 m/km) in region D as the existence of a sheet-like mantle upwelling with a relatively higher magmatic activity under the spreading axes. Such a sheet-like mantle upwelling has been interpreted for fast spreading ridges [Madsen et al., 1990; Lin and Phipps Morgan, 1992]. Overall, the relatively low MBA (25 mgal) and axial high morphological feature are very similar to the features for a fast spreading ridge [Wang and Cochran, 1993; Wang et al., 1996]. Moreover, the axial variations in MBA within a segment in region D are

0.09 mgal/km (segment 15) and 0.10 mgal/km (segment 16), which are similar to values of 0.1 to 0.2 mgal/km for the northern fast spreading East Pacific Rise (EPR) (9°N to 13°N) reported by Lin and Phipps Morgan [1992], and to the value of 0.1 mgal/km at the southern fast spreading EPR (7°S to 9°S) reported by Wang and Cochran [1995]. Martıńez et al. [2000] also conducted a geophysical survey in the southern Mariana Trough between 11°450N and 13°450N, and reported a 0.2 mgal/km axial variation in MBA, which correlates with typical values from a fast spreading ridge. This contributed to suggesting that the spreading process of the southern Mariana Trough is similar to that of a fast spreading ridge [Martıńez et al., 2000]. [13] The four distinct features of crustal thickness still exist in the Mariana Trough, even if the assumption of the constant crustal density (2700 kg/m3) for the calculation of crustal thickness has a different actual value. Two possible cases of deviations from the assumed crustal density value are presented to demonstrate this point. The first case is that the density has a constant value which is different from the assumed density (2700 kg/m3) in the Mariana Trough. In this case, the crustal thickness features still remain, because the estimated crustal thickness only shifts systematically from our results to higher or lower values according to 9 of 14


3

G


10.1029/2005GC001119

simple 1-D model, the relatively low crustal thickness in region C (3.9 km on average) requires crustal density of 2830 kg/m3 to become an average crustal thickness (4.9 km) of the Mariana Trough. The higher density implies magma-starved condition with low partial melting under the spreading axis, because high density requires existence of mafic or ultramafic rocks containing Fe- and Mg-bearing minerals [Rudnick and Fountain, 1995] and these rock were reported only from other less magmatic regions containing slow and ultraslow less magmatic spreading ridges [e.g., Gra`cia et al., 2000; Lagabrielle et al., 1998; Bach et al., 2002]. Therefore the less magmatic features remain in region C, even if the crust has the average thickness but higher density. On the contrary, the relatively high crustal thickness in region D (6.3 km on average) requires crustal material which has a density of 2540 kg/m3 to become an average crustal thickness (4.9 km) of the Mariana Trough. This density value is too low to be explained by an igneous rock, because even felsic rocks that consist of continental crust enriched in Si- and Al-bearing minerals, have densities of higher than 2570 kg/m3 [Rudnick and Fountain, 1995]. The existence of the rocks with a high porosity such as a sedimentary rock could explain the low density value, but it is unrealistic to suppose the existence of such low density rock material with thickness of 4.9 km under the spreading axis in this region. Thus the relatively thicker crustal thickness remains in Region D.

Figure 8. Four distinct regions (A, B, C, and D) shown over mantle Bouguer gravity anomaly map (Figure 3) with their boundaries (white thick lines). Contours of the subducted slab (black solid and dashed lines) [Gudmundsson and Sambridge, 1998] are also shown, with labels in km. The white dashed lines and the red lines correspond with those in Figure 1.

the actual density. The second case is that local deviations in the assumed density (2700 kg/m3) of the crustal material exist. In this case, the different features in magmatic activity will still remain between regions in the Mariana Trough. In the

[14] Our gravity analysis did not account for the gravitational influence of the subducted slab, but this additional effect does not invalidate our results because of two reasons. First, the geometry of the subducted slab [Gudmundsson and Sambridge, 1998] shows that the distance from the slab to the spreading axis varies so gently (Figure 8), leading that the slab would not affect the alongaxis variation in MBA (and in crustal thickness) within each segment. Secondly, the gravitational influence of the subducted slab may produce systematic variations along the axis in a regional scale, but the effect only emphasizes our results. In the central basin, the slab lies entirely to the east of the axis (Figure 8), where the gravitational effect becomes small away from the slab [Watts and Talwani 1975]. On the other hand, the slab underlies the spreading axes in the northern and southern ends of the basin (Figure 8). Since the subducted slab increases gravity anomaly above the slab [Watts and Talwani, 1975], the correction for the gravitational effect of the slab results in an increase of the crustal thickness estimation in the northern 10 of 14


3

G


and southern ends of the basin, where thicker crust was estimated from our gravity analysis without the correction (i.e., regions A and D) compared to the central basin (i.e., regions B and C). Therefore the gravitational influence of the subducted slab does not invalidate our results. On the contrary, it may cause underestimation of the regional difference in the crustal thickness, which emphasizes our results. [15] The MBA for off-axis areas suggest that the same subdivision (i.e., the four regions) can be determined separately from along- and off-axis areas for the Mariana Trough, similarly as alongaxis differences in crustal thickness and in its variation were used to identify the four regions. The MBA features of off-axis regions (Figure 8) are as follows: (1) for region A, the MBA value increases westward into the off-axis area far from the spreading axis, (2) for region B, off-axis areas which have a relatively high MBA (more than 50 mgal) and off-axis areas which have a relatively low MBA (less than 40 mgal) are found in coexistence, (3) for region C, the highest MBA area (60 mgal) is found predominantly off-axis, while the MBA value for the region decreases more gradually toward its western margin (West Mariana Ridge) than all other region margins, and (4) for region D, the lowest MBA area (25 mgal) is found predominantly off-axis. The off-axis MBA features reflect both of the differences of the crustal thickness and the thermal cooling effect of the lithosphere with increasing distance from the axis. However, thermal cooling effect makes low contribution to the MBA, because the gravity anomaly due to the thermal cooling effect is estimated as only 15 mgal for the largest offset (36 km) of the spreading axis. The offset that corresponds to the 1.8 Ma of crustal age difference calculated by using minimum estimate for spreading half rate of 20 km/Myr, and the estimation of the gravity anomaly value is based on the calculation of Kuo and Forsyth [1988]. Therefore the off-axis MBA features mainly reflect the difference of the crustal thickness between the four off-axis regions in the Mariana Trough, suggesting that the four distinct regions identified along the axes of the Mariana Trough are extended off-axis and that these four regions (A-D) have existed since the Mariana Trough started spreading. This implies that magmatic stages are established early in basin evolution and persist through time as the basin opens. [16] This study has clarified that spreading segments with lower magmatic activity are locally

10.1029/2005GC001119

distributed in both regions B and C. We propose two models to explain these less magmatic spreading segments, based on the local differences of chemical and physical conditions. The first chemical condition model involves the local existence of a compositional variation of the mantle source that changes the mantle solidus temperature, and this affects the depth of initial melting of the mantle and degree of melting in turn affecting the crustal thickness. This model is strongly supported by a geochemical survey along the axis of the Mariana Trough between 15°N and 17°N [Gribble et al., 1996]. They reported that higher degree of mantle melting has occurred along the axis between 16°N and 17°N than along the axis between 15°N and 16°N, and this degree of melting is strongly correlated with our result of crustal thickness (crustal material production), that is higher crustal thickness values along the axis between 16°N and 17°N than between 15°N and 16°N (Figure 4). They also showed that the degree of melting correlates with higher abundances of water and relative abundances of large ion lithophiles elements (LIL) and light rare earth elements (LREE). Further, Stolper and Newman [1994] reported that an increase in the degree of melting in the vicinity of 18°N in the Mariana Trough is caused by an increase in the water component through the lowering of the solidus temperature of mantle peridotite. This report is also consistent with our results that show higher crustal thickness values in the vicinity of 18°N. Moreover, Niu et al. [2001] proposed that the degree of mantle melting and crustal material production are controlled by fertile mantle source compositional difference of alkalis, water and incompatible elements, based on geochemical analysis of rock samples from Mid Atlantic Ridge between 33°N and 35°N. The whole evidence supports that the crustal thickness is a strong indicator for degree of mantle melting that correlates with water and incompatible elements in this chemical condition model. The less magmatic spreading segments with thin crust, locally distributed in both regions B and C, probably result from mantle source depleted of water and incompatible elements. This suggests that lateral heterogeneity of these elements exists on a segment scale in the mantle beneath the spreading axis of the Mariana Trough. [17] The second physical condition model involves a lower local temperature anomaly in the upper mantle. An upper mantle temperature low requires lower pressures (lower crustal thickness) before intersection of the solidus curve and initial melting, 11 of 14


3

G


and consequently it is more difficult to generate a large degree of melting. McKenzie and Bickle [1988] proposed that different crustal thicknesses can be explained by difference of the mantle potential temperature. The 4 km average crustal thickness of region C is 2 km thinner than the typical oceanic crust [Chen, 1992; Reid and Jackson, 1981], and this can be explained by a mantle temperature reduction of the 20°C estimated from numerical experimental work [Su et al., 1994]. This reduction could be brought about by the existence of roll-like small-scale convection as one possibility along the spreading center in the Mariana Trough, thereby causing these lowtemperature anomalies. A numerical experiment showed that such small-scale convection whose axis is perpendicular to strike of the spreading axis occurs in the low-viscosity wedge [Honda and Saito, 2003]. Their results suggested that the small-scale convection is caused by the instability of the top cold thermal boundary, and that a shear caused by the movement of subducting slab aligns the axes of convection roll perpendicular to the strike of the spreading axis. However, the model of Honda and Saito [2003] appears to require a low dip slab to generate a low-viscosity wedge extending beneath the back-arc basin. In the Mariana Trough, the slab is steeply dipping [Isacks and Barazangi, 1977; Chiu et al., 1991; Gudmundsson and Sambridge, 1998] and it does not underlie the spreading axis in regions B and C, suggesting that this mechanism could not work here. Therefore the chemical condition model is more plausible than the physical condition model. [18] Martıńez and Taylor [2003] and Taylor and Martıńez [2003] discussed the effects of mantle wedge hydration and chemical depletion effects in controlling magmatism and crustal thickness, and our chemical condition model is generally consistent with these previous studies. An important new finding of our study is that the crustal provinces defined by the gravity analysis appear to persist in time during essentially all the basin opening. This is important because it differs from models of a progressive evolution with basin opening. In such models, a magmatically robust stage of basin opening evolves with time to less magmatic stages as the spreading centers separate from the arc volcanic front [Martıńez and Taylor, 2003; Taylor and Martıńez, 2003]. This may not be the case from our results. The northern and southern axes (regions A and D) have always been more magmatically robust, where the subducted slab underlies the spreading axes (Figure 8); that could lead

10.1029/2005GC001119

an interpretation that these regions might still be in a magmatically robust stage of basin opening. However, region B and C have persisted the less magmatic stages since the basin started opening, indicating that these regions did not experience a magmatically robust stage. This implies that magmatic stages are established early in basin evolution and persist through time as the basin opens and that mantle wedge chemistry, perhaps controlled by the slab geometry may determine these patterns without a progressive evolution with basin opening. Moreover, spreading segments with lower magmatic activity are locally distributed in regions B and C, suggesting that lateral heterogeneity of mantle source on a segment scale is required. Since the cause of the heterogeneity on this scale cannot be explained by the slab geometry, the mantle itself probably has this lateral heterogeneity on a segment scale. The lateral heterogeneity is prominent in the overall cooler mantle of the Mariana Trough that can be inferred from its geochemical position within global trends [Martıńez and Taylor, 2003; Taylor and Martıńez, 2003], suggesting that magmatic activity could be sensitive to mantle lateral heterogeneity only when the mantle is cool.

5. Conclusions [19] This study shows the comprehensive crustal structure along the whole spreading axis of the Mariana Trough through the analysis of gravity data and clarifies the lateral distribution of the different melting process, which is locally suggested by previous geochemical studies. Distinct regional differences in crustal thickness were revealed along the axis of the Mariana Trough, and these regional crustal thickness features were identified for each segment by normalized variation in crustal thickness versus crustal thickness, and by variation in crustal thickness versus segment length. The following three points are concluded from this study: [20] 1. Four distinct regions were identified along the axis of the Mariana Trough (regions A, B, C, and D from north to south). Along the spreading axes, region A is north of 20°350N, region B is between 15°380N and 20°350N, region C is between 14°220N and 15°380N, and region D is south of 14°220N. These four regions have existed since the Mariana Trough started spreading. This implies that magmatic stages are established early in basin evolution and persist through time as the basin opens. 12 of 14


3

G


[21] 2. We have interpreted the different features in crustal thickness among four distinct regions individually. The segment from region A receives an additional magma supply from island arc magma sources. Two types of segments exist in region B. One is similar to a slow spreading ridge segment, which has a plume-like mantle upwelling under the spreading axis, and the other is a magma-starved segment. Region C is a less magmatic region in the Mariana Trough. Region D has a sheet-like mantle upwelling under the spreading axis and has experienced higher magmatic activity than other segments in the Mariana Trough. [ 22 ] 3. We have clarified that less magmatic spreading segments are locally distributed in both regions B and C, indicating that the source mantle is depleted of water and incompatible elements with reducing the degree of mantle melting and crustal material production. This suggests that lateral compositional variation of water and incompatible elements exists on a segment scale in the mantle source beneath the spreading axis of the Mariana Trough.

Acknowledgments [23] We would like to thank all who assisted in the collection of the geophysical data during Japanese scientific cruises by JAMSTEC, including Captains, crews, and scientific parties. K.K. would like to thank Hiroaki Sato, Toshiya Fujiwara, Motoyuki Kido, and Hisanori Iwamoto for their useful suggestions and discussions. Valuable comments by Wilfried Jokat and Matthew Brayshaw are also greatly acknowledged. We would also like to thank Fernando Martinez and editor Peter van Keken for their comments and suggestions that significantly improved the paper. This work was supported by ‘‘The 21st Century COE Program of Origin and Evolution of Planetary Systems’’ in Ministry of Education, Culture, Sports, Science, and Technology (MEXT), and by Grantin-Aid for Scientific Research (A)(2)(14253003), Japan Society for the Promotion of Science (JSPS). The GMT software package [Wessel and Smith, 1998] was used for the figures in this study.

References Ambos, E. L., and D. M. Hussong (1982), Crustal structure of the Mariana Trough, J. Geophys. Res., 87(B5), 4003 – 4018. Bach, W., N. R. Banerjee, H. J. B. Dick, and E. T. Baker (2002), Discovery of ancient and active hydrothermal systems along the ultra-slow spreading Southwest Indian Ridge 10° – 16°E, Geochem. Geophys. Geosyst., 3(7), 1044, doi:10.1029/2001GC000279. Chen, Y. J. (1992), Oceanic crustal thickness versus spreading rate, Geophys. Res. Lett., 19, 753 – 756. Chiu, J.-M., B. L. Isacks, and R. K. Cardwell (1991), 3-D configuration of subducted lithosphere in the western Pacific, Geophys. J. Int., 106, 99 – 111.

10.1029/2005GC001119

Detrick, R. S., H. D. Needham, and V. Renard (1995), Gravity anomalies and crustal thickness variations along the MidAtlantic Ridge between 33°N and 40°N, J. Geophys. Res., 100(B3), 3767 – 3787. Goto, T., et al. (2002), Geophysical experiments in the Mariana Region: Report of the YK01-11 cruise, InterRidge News, 11(1), 22 – 24. Gra`cia, E., J. L. Charlou, J. Radford-Knoery, and L. M. Parson (2000), Non-transform offsets along the Mid-Atlantic Ridge south of the Azores (38°N – 34°N): Ultramafic exposures and hosting of hydrothermal vents, Earth Planet. Sci. Lett., 177, 89 – 103. Gribble, R. F., R. J. Stern, S. H. Bloomer, D. Stu¨ ben, T. O’Hearn, and S. Newman (1996), MORB mantle and subduction components interact to generate basalts in the southern Mariana Trough back-arc basin, Geochim Cosmochim Acta, 60(12), 2153 – 2166. Gudmundsson, O., and M. Sambridge (1998), A regionalized upper mantle (RUM) seismic model, J. Geophys. Res., 103(B4), 7121 – 7136. Hasegawa, Y., N. Seama, K. Nakase, K. Mitsuzawa, H. Masuda, and T. Yamazaki (2000), Two distinctly different features in the Mariana Trough, Eos Trans. AGU, 81(48), Fall Meet. Suppl., Abstract T61B-23. Honda, S., and M. Saito (2003), Small-scale convection under the back-arc occurring in the low viscosity wedge, Earth Planet. Sci. Lett., 216, 703 – 715. Honsho, C., K. Tamaki, and H. Fujimoto (1997), Bathymetric and geomagnetic survey of the Mariana Trough 16°N, JAMSTEC J. Deep Sea Res., 13, 21 – 29. Hussong, D. M., and S. Uyeda (1982), Tectonic processes and the history of the Mariana arc: A synthesis of the results of Deep Sea Drilling Project Leg 60, Initial Rep. Deep Sea Drill. Proj., 60, 909 – 929. Isacks, B. L., and M. Barazangi (1977), Geometry of Benioff Zones: Lateral segmentation and downwards bending of the subducted lithosphere, in Island Arcs, Deep Sea Trenches, and Back-Arc Basins, Maurice Ewing Ser., vol. 1, edited by M. Talwani and W. C. Pitman, pp. 99 – 114, AGU, Washington, D. C. Ishihara, T., and T. Yamazaki (1991), Gravity anomalies over the Izu-Ogasawara (Bonin) and northern Mariana Arcs, Bull. Geol. Surv. Jpn., 42(12), 687 – 701. Iwamoto, H., M. Yamamoto, N. Seama, K. Kitada, T. Matsuno, Y. Nogi, T. Goto, T. Fujiwara, K. Suyehiro, and T. Yamazaki (2002), Tectonic evolution of the central Mariana Trough, Eos Trans. AGU, 83(47), Fall Meet. Suppl., Abstract T72A1235. Kong, L. S. L., N. Seama, H. Fujimoto, J. Kasahara, and KH92-1 Shipboard Scientific Party (1992), Segmentation of the Mariana Trough back-arc spreading center at 18°N, InterRidge News, 1(1), 2 – 5. Kuo, B.-Y., and D. W. Forsyth (1988), Gravity anomalies of the ridge-transform system in the South Atlantic between 31 and 34.5°S: Upwelling centers and variations in crustal thickness, Mar. Geophys. Res., 10, 205 – 232. Lagabrielle, Y., D. Bideau, M. Cannat, J. A. Karson, and C. Me´vel (1998), Ultramafic-mafic plutonic rock suites exposed along the Mid-Atlantic Ridge (10°N and 30°N): Symmetrical-asymmetrical distribution and implications for seafloor spreading processes, in Faulting and Magmatism at Mid-Ocean Ridges, Geophys. Monogr. Ser., vol. 106, edited by W. R. Buck et al., pp. 153 – 176, AGU, Washington, D. C. Lin, J., and J. Phipps Morgan (1992), The spreading rate dependence of three-dimensional mid-ocean ridge gravity structure, Geophys. Res. Lett., 19, 13 – 16. 13 of 14


3

G


Lin, J., G. M. Purdy, H. Schouten, J.-C. Sempe´ re´ , and C. Zervas (1990), Evidence from gravity data for focused magmatic accretion along the Mid-Atlantic Ridge, Nature, 344, 627 – 632. Madsen, J. A., R. S. Detrick, J. C. Mutter, P. Buhl, and J. A. Orcutt (1990), A two- and three-dimensional analysis of gravity anomalies associated with the east pacific rise at 9°N and 13°N, J. Geophys. Res., 95(B4), 4967 – 4987. Martıńez, F., and B. Taylor (2002), Mantle wedge control on back-arc crust accretion, Nature, 416, 417 – 420. Martıńez, F., and B. Taylor (2003), Controls on back-arc crustal accretion: Insights from the Lau, Manus and Mariana basins, in Intra-oceanic Subduction Systems: Tectonic and Magmatic Processes, edited by R. D. Larter, and P. T. Leat, Geol. Soc. Spec. Publ., 218, 19 – 54. Martıńez, F., P. Fryer, and N. Becker (2000), Geophysical characteristics of the Mariana Trough, 11°500N – 13°400N, J. Geophys. Res., 105(B7), 16,591 – 16,607. McKenzie, D., and M. J. Bickle (1988), The volume and composition of melt generated by extension of the lithosphere, J. Petrol., 29, 625 – 679. Mitsuzawa, K., H. Masuda, N. Seama, Y. Hasegawa, M. Miyamoto, N. Togashi, A. So, and H. Yamanobe (2000), Preliminary report of Deep-tow/Yokosuka cruise at hydrothermal areas in the mid and southern Mariana, JAMSTEC J. Deep Sea Res., 17, 73 – 87. Niu, Y., D. Bideau, R. He´kinian, and R. Batiza (2001), Mantle compositional control on the extent of mantle melting, crust production, gravity anomaly, ridge morphology, and ridge segmentation: A case study at the Mid-Atlantic Ridge 33 – 35°N, Earth Planet. Sci. Lett., 186, 383 – 399. Ohara, Y., R. J. Stern, T. Ishii, H. Yurimoto, and T. Yamazaki (2002), Peridotites from the Mariana Trough: First look at the mantle beneath an active back-arc basin, Contrib. Mineral. Petrol., 143, 1 – 18. Parker, R. L. (1972), The rapid calculation of potential anomalies, Geophys. J. R. Astron. Soc., 31(4), 447 – 455. Reid, I., and H. R. Jackson (1981), Oceanic spreading rate and crustal thickness, Mar. Geophys. Res., 5, 165 – 172. Rudnick, R. L., and D. M. Fountain (1995), Nature and composition of the continental crust: A lower crustal perspective, Rev. Geophys., 33, 267 – 309. Sandwell, D. T., and W. H. F. Smith (1997), Marine gravity anomaly from Geosat and ERS 1 satellite altimetry, J. Geophys. Res., 102(B5), 10,039 – 10,054. Seama, N., T. Yamazaki, H. Iwamoto, K. Kitada, M. Yamamoto, T. Fujiwara, Y. Nogi, K. Okino, and K. Suyehiro (2002), Tectonic features of the Mariana Trough, Eos Trans. AGU, 83(47), Fall Meet. Suppl., Abstract T71F-08. Seama, N., et al. (2003), Preliminary report of KR02-14 Kairei cruise, JAMSTEC J. Deep Sea Res., 22, 89 – 106. Smith, W. H. F., and D. T. Sandwell (1997), Global sea floor topography from satellite altimetry and ship depth soundings, Science, 277, 1956 – 1962.

10.1029/2005GC001119

Smith, W. H. F., and P. Wessel (1990), Gridding with continuous curvature splines in tension, Geophysics, 55, 293 – 305. Stern, R. J., S. H. Bloomer, F. Martıńez, T. Yamazaki, and T. M. Harrison (1996), The composition of back-arc basin lower crust and upper mantle in the Mariana Trough: A first report, Island Arc, 5, 354 – 372. Stern, R. J., T. Yamazaki, S. Danishwar, and C.-H. Sun (1997), Back-arc basin lower crust and upper mantle in the northern Mariana Trough studied with ‘‘Shinkai 6500,’’ JAMSTEC J. Deep Sea Res., 13, 47 – 61. Stolper, E., and S. Newman (1994), The role of water in the petrogenesis of Mariana trough magmas, Earth Planet. Sci. Lett., 121, 293 – 325. Su, W., C. Z. Mutter, J. C. Mutter, and W. R. Buck (1994), Some theoretical predictions on the relationship among spreading rate, mantle temperature, and crustal thickness, J. Geophys. Res., 99(B2), 3215 – 3227. Takahashi, N., S. Kodaira, A. Ito, S. L. Klemperer, Y. Kaneda, and K. Suyehiro (2004), Crustal growth of oceanic island arc inferred from seismic structure of Mariana arc-backarc system, Eos Trans. AGU, 85(47), Fall Meet. Suppl., Abstract T41D-1250. Taylor, B., and F. Martıńez (2003), Back-arc basin basalt systematics, Earth Planet. Sci. Lett., 210, 481 – 497. Wang, X., and J. R. Cochran (1993), Gravity anomalies, isostasy, and mantle flow at the East Pacific Rise crest, J. Geophys. Res., 98(B11), 19,505 – 19,531. Wang, X., and J. R. Cochran (1995), Along-axis gravity gradients at mid-ocean ridges: Implications for mantle flow and axial morphology, Geology, 23, 29 – 32. Wang, X., J. R. Cochran, and G. A. Barth (1996), Gravity anomalies, crustal thickness, and the pattern of mantle flow at the fast spreading East Pacific Rise, 9° – 10°N: Evidence for three-dimensional upwelling, J. Geophys. Res., 101(B8), 17,927 – 17,940. Watts, A. B., and M. Talwani (1975), Gravity effect of downgoing lithospheric slabs beneath island arcs, Geol. Soc. Am. Bull., 86, 1 – 4. Yamazaki, T., and R. J. Stern (1997), Topography and magnetic vector anomalies in the Mariana Trough, JAMSTEC J. Deep Sea Res., 13, 31 – 45. Yamazaki, T., K. Okino, Y. Hasegawa, H. Saitake, and M. Ito (1999), Geophysical mapping of Mariana Trough and West Philippine Basin: A preliminary report of Kairei KR9812 cruise, JAMSTEC J. Deep Sea Res., 15, 63 – 72. Yamazaki, T., N. Seama, K. Okino, K. Kitada, M. Joshima, H. Oda, and J. Naka (2003), Spreading process of the northern Mariana Trough: Rifting-spreading transition at 22°N, Geochem. Geophys. Geosyst., 4(9), 1075, doi:10.1029/ 2002GC000492. Yang, C. S., J. Segawa, and Y. Fukuda (1992), Density structure of the Mariana Arc and its vicinity obtained from successive inversion of the gravity anomaly, Tectonophysics, 206, 325 – 339.

14 of 14