In 2003, we conducted a survey of 11 serpentinite mud volcanoes in the Mariana fore arc. Here we report sediment pore water data from navigated gravity and piston cores and from push cores collected by the ROV system Jason2-Medea. Systematic variations in profiles of pore water chemical compositions from these mud volcanoes are consistent with models that include active upflow of pore water relative to the surrounding serpentinite matrix. The speed of upwelling, based on fits of an advection-diffusion model to observed data (K, Na, Rb, and Cs), reaches a maximum of 36 cm/yr at Big Blue Seamount. Results from these simulations constrain the pore water composition at depth and the degree of additional alteration as the pore water ascends through the sampled section. For example, the transition metals (e.g., Mn, Fe, Co, Ni, Cu, and Mo) are mobilized under conditions of low upwelling speeds and microbial activity. Similarly, the rare earth elements (REE) show evidence of near-surface alteration. In addition to these surficial reactions, distinctive pore water compositional patterns exist as a function of the distance from the trench axis, which is a proxy for the depths of water generation from the downgoing plate below each seamount. Systematic trends in the chemical composition of these slab-sourced fluids are consistent with increasing temperature and pressure at depth west of the trench. These trends include an increase in K, sulfate, carbonate alkalinity, Na/Cl, B, Mn, Fe, Co, Rb, Cs, Gd/Tb, Eu, and light REE (LREE) and a decrease in Ca, Sr, and Y with increasing distance from the trench. Mg and U are universally depleted in the upwelling water. We constrain the thermal conditions along the décollement using concentrations of fluid mobile elements (K, B, Cs, and Rb) and the mobilization of LREE relative to heavy REE (HREE). The 80°C isotherm is estimated at a depth of 15 km between Blue Moon Seamount and Cerulean Springs. At slab depths of 17 to 24 km, pore waters lack significant Rb and K enrichments relative to seawater, suggesting an upper bound near 150°C. There is an observed enrichment in LREE relative to HREE at Big Blue Seamount (slab depth 25 km) indicating that the décollement at this site is ∼200°C. The relative mobilization of Cs outpaced that of Rb at all seamounts sampled in this survey. On the basis of laboratory experiments, this observation sets an upper limit of ∼350°C at a depth of 30 km below the seafloor.
 Within subduction zones the lithospheric slab is subject to increases in temperature and pressure as it sinks, remobilizing elements from its crustal sections and transferring them upward either along the décollement or into the overlying plate [Fryer et al., 1985, 1999, 2000, 2006; Silver, 2000]. Water plays a critical role in the transfer of elements by several distinct processes: dissolution of material in the sediment [Benton et al., 2001]; serpentinization of the overlying mantle [Fryer and Fryer, 1987]; and metasomatism during serpentinization and arc magma generation [Peacock, 1990; Girardeau and Lagabrielle, 1992]. Examining the flux of elements through subduction zones is integral to models of elemental recycling [e.g., Stern et al., 1991; Mottl, 1992; Peacock, 1993; Domanik and Holloway, 1996; You et al., 1996; Bebout, 1995; Fryer et al., 1999; Chan and Kastner, 2000; Spivak, 2002]. Because processes that govern elemental recycling occur at depths unreachable within the present limits of the Integrated Ocean Drilling Program (IODP), direct observation of P-T conditions, mineral phase changes, and physical properties at the décollement are not possible. To understand these processes researchers rely upon seismic and tomographic imaging, geochemical balances of subduction inputs and outputs, and the examination of exposed paleosubduction zones when constructing models of elemental cycling and subduction processes.
 Within the outermost fore arc of an accretionary margin, pore water is released at low temperatures through sediment compaction, which can influence the geometry of the accreting wedge [Moore and Vrolijk, 1992]. As the slab continues to subduct, dewatering of clays and mafic minerals within the oceanic crust contribute water to the fore arc and subarc mantle through prograde metamorphism at depths approaching 260–360 km below the seafloor [Domanik and Holloway, 1996]. During dehydration, removal and fractionation of mobile elements occurs (e.g., B, N, As, Be, Cs, Li, Pb, Rb, and the rare earth elements (REEs)) [You et al., 1996; Bebout et al., 1999]. However, varying pressure and temperature (P-T) conditions along the broad source zone and the exchange of elements along the path of fluid migration are coupled with potential hydration-dehydration or recrystallization reactions, thus obscuring many potential geochemical tracers.
 The focus of this study is to explore processes within the nonaccreting Mariana fore arc that distill components from the subducting slab through a succession of dehydration/decarbonation reactions well before reaching the zone of arc magmatism. Within the active Mariana fore arc, deep-seated materials ascend from the décollement and underlying mantle, forming mud volcanoes. These volcanoes provide the only means to sample directly the subducting slab as it evolves while migrating into the mantle. These conditions allow us to collect pore waters and mineral parageneses to elucidate in situ P-T conditions at the source region. This manuscript builds upon previous research in the Mariana fore arc [Mottl, 1992; Fryer et al., 1999; Benton et al., 2001; Mottl et al., 2004; Alt and Shanks, 2006; Fryer et al., 2006] by presenting new and more inclusive chemical data from navigated sediment cores within areas of active pore water upwelling on serpentinite mud volcanoes. These data are used to constrain the pore water composition at depth, which differs systematically as a function of the distance to the trench. On the basis of this systematic difference in composition, we constrain the extent of element mobilization, possible reaction mechanisms, and temperature along the décollement. These data, coupled with mineralogical data from deep ocean drilling, form the foundation for future modeling efforts to further constrain physical and chemical mechanisms and conditions along the décollement in the Mariana fore arc and subduction zones globally.
2. A Geological Description of the Mariana Fore Arc
 Pore waters that upwell through serpentinite seamounts do so within a mud matrix that consists of (1) medium blue-green to dark blue serpentine; (2) veins or precipitates of chlorite, brucite, magnetite and Cr-spinel; (3) moderately to completely serpentinized ultramafic harzburgite and dunite clasts containing olivine and orthopyroxene with lesser amounts of clinopyroxene; and (4) calcite and aragonite carbonates [Salisbury et al., 2002]. Results from chemical analyses of pore waters and the serpentinite host have concluded that the pore waters are not significantly altered on their migration to the surface through serpentinized peridotite conduits, and that their composition represents conditions deep within the subduction zone [Mottl, 1992; Fryer et al., 1999; Benton et al., 2001; Mottl et al., 2003, 2004; Savov et al., 2005; Fryer et al., 2006].
 Previous direct seafloor explorations by submersibles, ROV, and gravity coring discovered active pore water venting at the summit regions of three mud volcanoes. The submersible Alvin surveyed active mud volcanism on Conical Seamount, prior to drilling ODP Leg 125, and discovered carbonate chimneys [Fryer et al., 1990; Haggerty, 1991]. The submersible Shinkai 6500 was deployed in 1995 to investigate S. Chamorro Seamount near ODP Site 1200 and discovered the only known benthic megafaunal community associated with serpentinite mud volcanism [Fryer and Mottl, 1997]. Pacman Seamount was surveyed with the Shinkai 6500 submersible and with the ROV Jason-Medea. Active seeps along the SE flank of Pacman Seamount, the Cerulean Springs, produce thin brucite chimneys almost 1 m in length [Fryer et al., 1999].
 A multidisciplinary survey of the Mariana fore arc, from 23 March through 4 May 2003, utilized acoustic swath-mapping surveys and sediment gravity/piston coring operations to locate active pore water seeps on 11 of the Mariana fore-arc serpentinite mud volcanoes (Figure 1).
 Sediment from gravity and piston cores was immediately split and sampled. Sediment was scooped into acid-washed polycarbonate centrifuge tubes using Teflon-coated stainless steel spatulas. Sediment within 0.5 cm of the PVC liner and on the bisected surfaces was discarded to avoid sample smearing and contamination from the core wall. Sediment-filled centrifuge tubes were cooled to 1–4°C, placed in a cooled rotor to eliminate warming during centrifuging, and spun in a centrifuge for 5 min at ∼10,000 rpm. Pore water was siphoned off the top of the sediment using acid-washed high-density polyethylene (HDPE) syringes and filtered (0.45 micron) into acid-leached HDPE bottles and glass ampules. Some aliquots were immediately acidified using subboiled (6 N) HCl to lower the pH below 2.
 Select push cores collected during ROV operations were immediately transported to a walk-in refrigerator (4°C) and placed in a nitrogen-filled glove bag. These push cores were sectioned in the glove bag into ∼3 cm intervals, depending on the water content and the presence of rocks or precipitates. Because of the high volume of cores recovered on some dives, only the highest-priority cores were sampled within the freezer. The remaining cores were sectioned at ambient temperatures and atmospheric conditions. Only the centers of the extruded core sections were collected to avoid sample smearing and contamination artifacts. Sediment samples were placed in acid-washed polycarbonate centrifuge tubes and capped within the nitrogen-filled glove bag. These samples were subsequently processed in the same method as the gravity and piston core samples.
 Several chemical analyses were conducted immediately at sea. Chlorinity was measured by titration with silver nitrate [e.g., Knudsen et al., 1902] using an automated electrochemical endpoint. For those fluids with high dissolved sulfide content, samples were dried, rehydrated, and analyzed. Pore water pH was measured on 2 ml aliquots that were then titrated with 0.1 N hydrochloric acid for a measure of the alkalinity by the Gran function method [Stumm and Morgan, 1981]. Concentrations of hydrogen sulfide were measured by the photocolorimetric method developed by Cline .
 The major elements (Ca, Mg, Na, K, S, Sr, and Li) and minor elements (Sr, Li, Ba, B, Mn, Fe, and Si) were measured using standard inductively coupled plasma atomic emission spectrometer (ICP-AES) techniques. High-resolution inductively coupled plasma mass spectrometry (HR-ICP-MS) was used to measure a selection of diluted (1% in 0.5 N QHNO3) samples for Rb, Mo, Cs, Ba, and U. A standard addition technique was used to analyze V, Cr, Co, Ni, Cu, and Zn with 10% sample dilutions. For this analysis, a metal-free seawater solution was produced by pumping bottom seawater through an 8-hydroxyquinoline (8-HQ) column to provide a matrix-matched standard addition procedural blank. Because of their minimal concentrations and susceptibility to interferences from a seawater matrix, the REEs, Y and Cd were extracted from pore water samples using a combination of pumps, valves, and 8-hydroxyquinoline (8-HQ) columns (for more detail see Hulme  or Hulme et al. ). H and O stable isotopic measurements were made using a mass spectrometer at the University of California, Berkeley [DePaolo et al., 2004]. A total of 42 chemical species were analyzed within the highest-priority samples.
 Eleven geologic features related to serpentinite mud volcanism were sampled with 15 gravity cores, 14 piston cores, and 51 push cores (Unless otherwise noted, data that are discussed are from this 2003 expedition.). Pore water chemical profiles of gravity, piston and ROV push cores from the 2003 expedition with the highest speed of pore water upward flow are presented in Figures 2 and 3. Pore water gradients at all of the active sites display decreasing (approaching 0 mmol/kg) Mg concentrations and increasing pH (up to 12.3 at Big Blue) down-core. Ca and Sr concentrations exhibit increases with depth relative to seawater at seamounts near the trench (e.g., E. Quaker; maximum of 78 mmol/kg Ca and 914 μmol/kg Sr) and decreases with depth at the more distal seamounts (e.g., Big Blue; minimum 0.1 mmol/kg Ca and 10 μmol/kg Sr). K and B concentrations generally behave inversely to Ca and Sr across the Mariana fore arc with maxima of 14 mmol/kg K at Conical and 1800 μmol/kg B at Big Blue and minima of 2.2 mmol/kg K and 15 μmol/kg B at Cerulean Springs.
 Concentrations of Rb show internally consistent trends down-core at each site (Figure 3). Pore waters that seep from three seamounts, Blue Moon, Cerulean Springs and NE Quaker, are depleted in Rb relative to seawater. For example, measured Rb values of 0.45 μmol/kg at Cerulean Springs are less than half the seawater concentration of 1.37 μmol/kg. In contrast, the highest pore water concentration of Rb is 6.65 μmol/kg at Big Blue. Unlike Rb, Cs concentrations within the deepest-sampled intervals are greater than those of the surface pore water concentration at all of the seamounts, with the exception of Blue Moon (Figure 3).
 Other trace elements measured in pore water, Mo and U, show a marked degree of mobilization and irregular variation in the uppermost sediment but approach asymptotic concentration profiles deeper in the cores (Figure 3). For example, U concentrations within the upper tens of centimeters of the sediment column exceed 40 nmol/kg in a push core from the summit of Big Blue and are significantly elevated at E. Quaker and Blue Moon. Despite the surface enrichments, down-section pore water concentrations approach zero at all of the sampled sites with active pore water upwelling, similar to Mg profiles.
 Two other elements, Ba and Li, are apparently affected during ascent from depths below the deepest sampled interval, but this can only be confirmed at a single mud volcano for each element (E. Quaker for Ba and Quaker for Li). For example, Ba concentrations are elevated in borehole samples from S. Chamorro [Wheat et al., 2008], but the lack of a pore water profile precludes determining if the mobilization occurs at depth or as a result of microbial redox reactions. Mobilization of Ba at Cerulean Springs does fit an asymptotic profile, so the results are inconclusive there as well. Regardless of these uncertainties, concentrations of Ba and Li in deep-sourced pore waters do not follow any trend with increasing distance from the trench axis.
 The complex behavior of the fourth period transition metals is attributed to differing redox potentials within the sediment column (Figures 4 and 5). The most abundant transition metals in pore water, Fe and Mn, behave nearly identical to one another. Often there are two distinct zones of enrichment as well as depletion. Other transition metals show similar zones of enrichment. For example, the Co enrichment is up to 22 nmol/kg in pore waters from Big Blue. These upwelling pore waters have high concentrations of sulfate and high alkalinities. In general, faster upward pore water flow compresses the vertical zone of transition metal reactivity, in contrast to slower upwelling seeps where mobilization of these elements occurs over a greater depth range. Cr, however, is mobilized in pore waters from the upper sediment section at Big Blue independent of flow speed, and high flow speeds correlate with the greatest degree of V enrichment.
 The entire suite of rare earth elements (REE) and Y were measured in pore waters from 5 of the Mariana fore-arc seamounts and water sampled from the venting borehole at S. Chamorro [Wheat et al., 2008]. Given the minimal volumes extracted from the sediment during centrifuging, it was not possible to measure pore water concentrations of REE from all of the sampled cores. Concentrations of the REE at Big Blue were studied in greater detail and show a progressive depletion with faster pore water upwelling speeds (Figure 6). Examination of the REE concentrations from the deepest pore water sample from the core with the highest seepage speed reveals the elemental fingerprint for each seamount (Figure 7). Some observed trends across the fore arc include: Cerulean Springs and Blue Moon have the highest abundances of REEs, distributed in a slightly LREE enriched pattern; E. Quaker has the lowest measured concentrations of La and the only positive Ce anomaly; Big Blue, E. Quaker, and Baby Blue have LREE-enriched pore waters and a positive Gd anomaly; S. Chamorro has LREE-enriched pore waters and a positive Eu anomaly.
 Selected pore water profiles of the stable isotopes ∂18O (SMOW) and ∂D (Figure 8) show considerable variation and no clear relationship with distance from the trench. The highest measured value of ∂18O (+1.8‰) occurs at Cerulean Springs and the lowest measured value (−1.5‰) occurs at Blue Moon. The highest measured ∂D within these samples is from 1 m depth at Big Blue (+27‰), but values below this horizon converge with values from nearby cores (at around +12 to +16‰). This anomalous sample also has a significant decrease in ∂18O (−1.3‰) compared with the bracketing samples. The lowest values of ∂D (−3‰) are from Blue Moon. The range of typical surface pore water values from all of the seamounts is −0.9 to −0.5‰ ∂18O and −2 to −1‰ ∂D.
5.1. Estimating Depth to the Subducting Slab
 Previous results from the Mariana serpentinite volcanoes show a trend in the chemical composition of the upwelling pore water as a function of distance from the Mariana Trench axis [Fryer et al., 1999; Mottl et al., 2004]. This distance was then used as a proxy for the depth to the décollement from which a temperature at depth was estimated. We refine these previous estimates by first quantifying the distance from the trench for each sample location using a vectorized trace of the Mariana Trench axis and the commercial GIS software package ArcView8 (trace shown in Figure 1). After combining the subducting slab slope with distances from the trench, we applied a correction to account for bathymetric variations up to 3 km to for each sample location (Table 1).
Table 1. Results of the Distance-From-Trench Calculations and Estimates of the Depth to the Subducting Slab Under the Serpentinite Seamountsa
Distance to Mariana Trench Axis (km)
Depth to the Subducting Slab (km)
The estimates are based on combining the distance from the trench axis with estimates of the slab dip angle under the fore arc [Fryer et al., 1985]. Corrections based on the bathymetric depth of each seamount were made to account for up to 3 km variations in seamount summit heights.
 The slab-depth estimate beneath Celestial produced by the trench distance-bathymetry model (19 km) fell within the range of depths (16–20 km) predicted from seismic imaging of the subducting slab by Oakley et al. . Depth estimate errors include an additional ±2 km resulting from the presence of subducting seamounts that alter the shape of the trench axis and reduce the depth to the décollement. An additional source of uncertainty is the differing slab dip angles across the Mariana Trench [Fryer et al., 2003; Miller et al., 2004], although this effect is minimal at shallow depths near the trench. These estimates are used in section 5.4 to discuss geochemical and thermal conditions within the subducting slab and upper mantle.
5.2. Estimates of Pore Water Upwelling Speeds
 Quantifying the speed that pore water upwells at each site provides a benchmark for establishing the degree to which the pore water composition is altered during ascent. Slower upwelling speeds result in longer residence times, allowing chemical exchange with the surrounding matrix and possible alteration by microbial populations [e.g., Wheat and Mottl, 2000]. In addition, the downward diffusion of seawater in such areas can result in the formation of secondary minerals such as carbonates. The degree of pore water sediment/rock alteration becomes apparent by quantifying the upward speed of flow at multiple locations within a seamount and assessing systematic variations in pore water chemical profiles. Cores that sample the highest speed of pore water upwelling are the primary basis for estimates of the composition of formation water at depth, presumably at or near the décollement.
 The general equation for pore water flow is described by Berner  and simplified for this application. In the simplified model the concentration [C] of an element within the sediment column is
where; v is the velocity of pore water upwelling, z is the depth within the sediment, Ds is the sediment diffusion coefficient of the modeled ion, C0 is the concentration of the ion in bottom seawater, and β is a function of the composition (C) at a particular depth (zf),
The variable Ds is a function of the conditions in the sediment,
where Dw is the molecular diffusion coefficient of the element in seawater at 4°C [Li and Gregory, 1974], v is the viscosity of the water (0.95 at 0°C [Li and Gregory, 1974]), f is the sediment formation factor, and p is the porosity. Values for the formation factor and porosity are 3.5 and 0.55, respectively, based on previous measurements of serpentinite mud [Shipboard Scientific Party, 2002]. We chose an end-member depth of 60 m for these models to ensure the theoretical profiles were far below the sampled depth. We quantify upwelling speeds using this model by iteratively calculating the sum of the residuals between the observed and theoretical concentrations.
 We apply this model to all of the sampled sites across the Mariana fore arc. K, Rb, and Cs concentration profiles are modeled because the source of these elements is the subducting slab, not the overlying mantle, suggesting that these elements are the most conservative elements during pore water ascent in this system [e.g., Mottl et al., 2004]. We rely more on the Rb results, because Rb appears to be the most conservative element in these systems and there is a large quantifiable difference between pore water and seawater concentrations. However, for three of the seamounts (Pacman Summit, Quaker, and Baby Blue) Rb concentrations in the upwelling pore waters are nearly indistinguishable from bottom seawater concentrations and thus, we rely mostly on the Cs concentration profiles to model flow, as pore water concentrations of Cs are readily distinguishable from seawater. Pore waters from cores with the highest calculated upwelling speeds from each seamount are used to assess the composition of the upwelling pore water from depth (Tables 2 and 3).
Table 2. Major Element Composition of Samples Most Representative of Deep-Sourced Fluids That Were Sampled in the 2003 Expeditiona
Maximum Flow (cm/yr)
Minimum Mg (mmol/kg)
Maximum Sulfide (μmol/kg)
The maximum flow rates at each site as well as the minimum Mg are given to show the proximity to an end-member composition (i.e., high flow, low Mg). Where the end-member compositions were not reached, the minimum or maximum values are shown based on the down-core asymptotic trends in composition. Because upwelling pore water was not sampled at S. Chamorro during the 2003 cruise, values from ODP Leg 195 are presented [Shipboard Scientific Party, 2002; Mottl et al., 2004]. The Conical Seamount 2003 pore water composition of the most altered interval is shown along with measured end-member values from ODP Leg 125 drilling results [Mottl, 1992].
Table 3. Minor, Trace Element, and Stable Isotopic Composition of Samples Most Representative of Deep-Sourced Fluids Sampled During the 2003 Expeditiona
Where the end-member compositions were not reached, the minimum or maximum values are shown based on the down-core asymptotic trends in composition. Because upwelling pore water was not sampled at S. Chamorro during the 2003 cruise, values from ODP Leg 195 are presented [Shipboard Scientific Party, 2002; Mottl et al., 2004]. Because U was not measured in the ODP Leg 195 samples, the concentration of U in water venting from the Site 1200 borehole is presented (Mg = 26.3 mmol/kg [Wheat et al., 2008[). Conical Seamount estimates are shown along with measured end-member values from ODP Leg 125 drilling results [Mottl, 1992].
 We address the role of pore water/matrix reactions by assuming the deep-seated composition of upwelling pore water is based on cores from areas with the fastest pore water upwelling speeds. By examining a wide variety of chemical species and their systematic differences in cores from areas with different pore water upwelling speeds, it is possible to estimate the degree of alteration as water upwells and to determine which elements are conservative during ascent. Such characterizations are important because they provide a measure of confidence in our estimates for the fluid composition at depth, which is thought to reflect slab dehydration reactions that are constrained by P-T conditions along the décollement [Mottl, 1992; Fryer et al., 1999; Mottl et al., 2004].
 In a natural seafloor setting, however, mobilization of some elements is not limited to abiotic reactions, microbially mediated redox reactions with readily available oxidants, such as oxygen and nitrate from bottom seawater, affect chemical compositions [e.g., Berner, 1980]. Other microbial mediated diagenetic processes are present in sediment pore waters (e.g., reduction of metallic oxyhydroxides and sulfur species [Van Cappellen and Wang, 1996; Thullner et al., 2005]). These abiotic and microbially mediated reactions affect chemical compositions of pore waters within surface sediments of the Mariana fore-arc serpentinite mud volcanoes [Peacock, 1990; Mottl et al., 2003; Takai et al., 2005]. Such reactions can mask the composition of upwelling deep-sourced water, if chemical fluxes from reaction are a significant portion of the advective transport portion of the chemical flux. Thus, we need to recognize these reactions and their affects to elucidate the chemical composition of the deep-seated fluid from which we constrain the P-T conditions within the downgoing slab.
5.3.1. Major Elements
 Upwelling pore water is depleted in Mg relative to seawater at each seamount, approaching sub-mM concentrations in the deepest cores. The depleted mantle beneath the Mariana fore arc is primarily harzburgite and dunite containing up to 99% olivine [Fryer, 1996; Saboda et al., 1992; Parkinson et al., 1992; Parkinson and Pearce, 1998]. Serpentinization of these protoliths produces secondary mineral phases such as brucite or magnesite, which can incorporate dissolved Mg from the pore water.
 Typically, in hydrothermal systems, Mg depletion of the end-member fluid is the basis for estimating the degree of mixing with seawater and is used for constraining the end-member composition; conservative elements when plotted together produce a line joining the two end-member values. However, for the slower pore water upwelling speeds observed in the Mariana cores, Mg is reactive in the sediment column, presumably being removed in brucite. We highlight Mg-Ca systematics at two seamounts of differing pore water compositions, indicating significant nonconservative behavior in the cored sediment (Figure 9a). Ca in pore waters also is reactive. For example, pore water from Cerulean Springs has high concentrations of Ca and low alkalinities, whereas pore water from Big Blue is depleted in Ca but contains appreciable levels of alkalinity (>68 meq/kg). Near the surface of Big Blue, Ca from seawater diffuses into the sediment and reacts with this high-alkalinity pore water, forming carbonates [Mottl, 1992; Fryer et al., 1999; Mottl et al., 2003].
 Concentrations of K in the ascending pore waters vary dramatically across the fore arc, being depleted relative to bottom seawater near the Mariana Trench (e.g., Cerulean Springs) and enriched farther from the trench (e.g., Big Blue).This trend is consistent with increasing temperatures in the subducting slab and is not related to serpentinization reactions [Mottl, 1992; Janecky and Seyfried, 1986]. By plotting the conservative component K against the molar ratio of Mg/Ca (Figure 9b), reactions in the surface sediment noted above become more evident. For example, the loss of Ca to carbonate formation at Big Blue correlates with a dramatic rise in Mg/Ca molar ratio. At Cerulean Springs, which has low-alkalinity pore water, the shallow slope of K versus the Mg/Ca molar ratio in the near surface is the result of rapid Mg loss because K versus Ca are conservative (linear mixing relationship) at this site. The uptake of Mg in the upwelling pore water is probably results from brucite precipitation.
 Experiments by Janecky and Seyfried  showed that while Na and Cl can vary during serpentinization reactions, the Na/Cl molar ratio is conserved. A plot of Na versus Cl at Big Blue and Cerulean Springs shows two contrasting trends (Figure 10). There is a high linear correlation (0.9968) between Na and chlorinity at Cerulean Springs. This trend is consistent with these two species being conservative in the upwelling pore water. In contrast, the behavior of these two components at Big Blue is nonlinear (Figure 10), indicating that at least one of these elements is not conservative in the sampled section.
 One possible explanation is that dissolved chloride is removed in, or added to, the sampled serpentinite mud as a product of precipitation or dissolution of chloride-bearing minerals such as iowaite (Mg3Fe3+(OH)8Cl·2H2O) [Sharp and Barnes, 2004]. However, the presence of iowaite has yet to be documented. Another, less likely, possibility is that Na is substituted for Mg and/or Ca in surficial sedimentary serpentinite, thus maintaining charge balance [e.g., Mottl et al., 2004]. Magnesium is universally depleted in the upwelling pore water and is not the likely cation involved with Na exchange. In contrast, Ca-Na exchange has been observed in marine sediment pore water [e.g., Wheat et al., 2000], but the composition of sedimentary serpentinite vastly differs from that of typical pelagic sediment. Even though Ca varies dramatically with increasing distance from the Mariana Trench axis, the Ca concentration is too low at Big Blue for the required molar equivalent exchange with Na. Furthermore, the lack of any significant deviation in linear trends between Na and carbonate alkalinity versus K reinforces the possibility of pore water Cl concentrations being altered by Iowaite formation in the sampled section.
5.3.2. Minor and Trace Elements
 In general, pore water concentrations of Rb, Cs, and K are conservative in the uppermost sediment section, as concentrations are governed by reactions with potassic minerals in the subducting slab [Sadofsky and Bebout, 2003; Barr et al., 2002]. The products of differing P-T conditions at depth result in differing Rb, Cs, and K concentrations in the upwelling fluid. Rubidium concentrations at Big Blue and Cerulean Springs differ in a manner similar to K. In contrast, Cs is enriched in the upwelling pore water at both localities (Figure 11). The linear relationship between Rb and Cs reaffirms our assumption that these elements are conservative within the upper serpentinite mud. The only observation for the mobilization of Cs in the sediment matrix is in the uppermost (<10 cm) of serpentinite mud/pelagic sediment at Big Blue.
 Because K, Rb, and Cs are generally conservative in the pore waters from our cores, especially those from below the upper tens of centimeters, we can use any of these elements to ascertain the behavior of other elements. We used K in the example above to establish a relationship between Mg and Ca alteration reactions at Big Blue and Cerulean Springs. We chose Rb to constrain the degree of alteration of pore water constituents across the entire Mariana fore arc because of its large dynamic range in concentration, and because Rb does not suffer from the same susceptibility to near-surface reactions as Cs. Therefore nonlinear element/Rb relationships provide a measure of how reactive an element is in this regional setting, and in the sampled sections.
 Uranium is the only element other than Mg that uniformly approaches sub-nM (<0.1 nmol U/kg) concentrations in upwelling pore water throughout the Mariana fore arc, but unlike Mg, pore water U enrichments are observed near the seafloor. A plot of U versus Rb at Big Blue and Cerulean Springs clearly illustrates this behavior (Figure 12a). Uranium in surface pore water from Big Blue is enriched by a factor of 3.5 over bottom seawater concentrations. In contrast, no significant mobilization of U occurs at Cerulean Springs where carbonate alkalinity is low. These distributions are related to microbially mediated dissolution of metallic oxyhydroxides present in the shallow sediments [Morford et al., 2005, 2007, 2009]. Although U is released during Mn and/or Fe oxidation at both sites, it is kept in solution at Big Blue by the formation of uranyl-carbonate complexes.
 Other trace elements show differing degrees of mobilization and are likely linked to “major” dissolved ions present in the pore water. Sulfate and sulfide concentrations at Big Blue and Cerulean Springs have a marked affect on the degree of Ba and Mo mobilization, respectively (Figure 12b). At Cerulean Springs, depleted sulfate concentrations (8.5 mmol/kg) allow Ba mobilization that is normally inhibited by the formation of barite (BaSO4). Elevated concentrations of Ca (49.5 mmol/kg) and Sr (0.32 mmol/kg) saturate the solution with respect to their sulfate minerals gypsum (CaSO4·2(H2O)) and celestite (SrSO4), and the precipitation of these minerals increases Ba mobilization [You et al., 1996]. Mottl et al.  also suggested gypsum saturation, but these minerals have not been found, in spite of a search for them. In contrast, low Ba concentrations occur at all depths in cores recovered from Big Blue, which has elevated concentrations of sulfate relative to seawater (30 versus 28 mmol/kg).
 Molybdenum mobilization occurs dramatically (up to 6,420 nmol/kg) at Big Blue associated with intervals that have high concentrations of dissolved Mn. Previous studies described the removal mechanism of Mo from solution by adsorption onto Mn oxyhydroxides [e.g., Bertine and Turekian, 1973; Emerson and Huested, 1991]. Microbially mediated dissolution of Mn oxyhydroxides was determined as a significant source of Mo mobilization in near-surface pore water by Morford et al.  and is likely responsible for Mo mobilization at Big Blue. Minimal, or much less than that observed at Big Blue, Mo enrichment is apparent at Cerulean Springs even though the surface interval includes Mn mobilization. Concentrations of Ba and Mo in the deepest core sections at Big Blue converge to concentrations of 47 and 220 nmol/kg, respectively. In contrast, concentrations of Ba and Mo at Cerulean Springs converge to concentrations of 297 and 20 nmol/kg, respectively (Figures 3 and 12b).
5.3.3. Transition Metals
 Mobilization patterns in pore water profiles for Mn and Fe show a large degree of symmetry. Many of the remaining transition metals (Co, V, Cr and Cu) have profiles that appear to behave opposite to or independent of Fe and Mn (Figures 4 and 5). The underlying reason for the observed profiles in the shallow serpentinite mud is related to microbially mediated redox states that, in part, result in the consumption of Fe-Mn oxyhydroxides and the precipitation of metal sulfides [Poulton et al., 2004; Morford et al., 2009]. The reduction of oxyhydroxides results in the mobilization of Fe and Mn, which can then react with dissolved sulfide to precipitate pyrite and metal sulfides. The presence of high concentrations of carbonate and hydroxide ions further complicates the shape of the profiles by increasing metal carbonate and metal hydroxide complexing, and possibly precipitation [Zachara et al., 2001].
 Reactions among Fe, Mn, carbonates, oxyhydroxides and sulfur species impact the degree of mobility for a variety of other transition metals. For example, at Big Blue, Co enrichment occurs within the zone of Fe and Mn depletion (Figure 4). Experiments by Zachara et al.  demonstrated the preferred dissolution of Co(II) over Fe(II) from oxyhydroxides through bacterial metabolism. Nickel shows the same behavior as Co in this zone, and shows significant mobilization at Big Blue. In contrast, pore waters from Cerulean Springs do not show the same magnitude of Co enrichment observed at Big Blue, and samples that are enriched in Co do not exhibit significant Ni mobilization. Instead, a correlation between Co and Cu exists at Cerulean Springs (Figure 5). This suggests the possibility that chemical species incorporated into serpentinite at depth under reducing conditions can be mobilized at the seafloor. Serpentinization at depth has a significant effect on the transition metals because the water is simultaneously reducing and hyperalkaline. Water reacting with basaltic crust will become acidic causing metals to mobilize, but serpentinization consumes metals by sequestering them into native metal and sulfur species depending on water-rock ratios [Palandri and Reed, 2004]. High water-rock ratios release additional sulfur into the system, which is reduced, leading to the formation of metallic sulfides. At low water-rock ratios, serpentinizing fluids also are reducing, but, in the absence of available sulfur, dissolved metals will precipitate as native metals and/or oxyhydroxide species [Palandri and Reed, 2004].
5.3.4. Rare Earth Elements
 Concentrations of REEs in pore water from Mariana fore-arc mud volcanoes are uniformly depleted relative to bottom seawater, with the exception of slightly elevated Ce at S. Chamorro (Figure 7). One possibility is that carbonates coprecipitate REEs from solution (e.g., Ancylite-(La): Sr(La,Ce)(CO3)2(OH)•(H2O)). Evidence for carbonate coprecipitation is clear in push core J36–4 (Figure 6), which was taken from a region of fast pore water upwelling at the summit of Big Blue with the highest alkalinity (68 mmol/kg). Significant levels of LREE depletion are present in the uppermost pore waters of the serpentinite mud where the carbonate-rich pore water reacts with Ca-rich seawater to form carbonates (including hydrated hydroxycarbonates). Scavenging by Fe oxyhydroxides has been documented as producing positive La, Gd, Lu and Y anomalies [Bau, 1999], and must not be overlooked in this system. Another possibility is that REEs are not mobilized in such settings, consistent with bulk chemical analyses of land-based blueschist formations [Bebout et al., 1999], and require much warmer temperatures for mobilization to occur at the high pressures expected along the décollement [e.g., Klinkhammer et al., 1994]. The origin of REE signatures are discussed in more detail later.
5.3.5. Stable Isotopic Effects
 The O and H stable isotopic composition of pore waters indicates considerable alteration of the initial deep-sourced fluid composition. On the based of previous studies of Mariana fore-arc serpentinites, pore water ∂18O values should respond to variations in temperature along the décollement with increasing temperature resulting in higher ∂18O [Alt and Shanks, 2006]. This trend is violated in the case of Cerulean Springs, which has a relatively high ∂18O value (+1.76‰). Yet, the other chemical data do not support such a high projected temperature (e.g., Cs, Rb, and K data). If serpentinization in the near surface was responsible for altering the ∂18O values, then one would expect a shift in the opposite direction toward lighter ∂18O because of cooler temperatures at the seafloor. The cause of this anomalous shift toward heavier ∂18O may be attributed to isotopic fractionation during brucite formation, which is more abundant at Cerulean Springs than at the other seamounts.
 Similar anomalous behavior is observed in the ∂D pore water data, which differ inconsistently with projected depth to the subducting slab. The highest values from the 2003 survey were measured at E. Quaker and Big Blue, similar to results reported from ODP operations at Conical and S. Chamorro [Benton, 1997; Mottl et al., 2003] (Figure 13). There is a possibility that our lower ∂D values in gravity cores from Conical, compared to ODP drilling data, are the result of serpentinization reactions at shallow depths, but this does not account for the elevated ∂D at E. Quaker. It is more likely that there is a significant isotopic effect related to the production of H2 during serpentinization that results in a heavier ∂D in the complementary pore water as proposed by Früh-Green et al. . The lack of a correlation with ∂18O and chlorinity precludes the possibility that fractionation results from hydrate formation [e.g., Hensen et al., 2004].
 Additional evidence, correlated minima and maxima of the ∂18O and ∂D, respectively, in the profile from Big Blue, points toward significant alteration of pore water near the seafloor. These changes may result from biogenic uptake of low ∂D-containing methane and the precipitation of heavy ∂18O into silicates. Although little is currently known about the mechanisms responsible for the observed isotopic values, these observations should be considered when constraining the formation history of serpentinites based on stable isotope compositions [e.g., Sakai et al., 1990; Früh-Green et al., 1996; Alt and Shanks, 2006].
5.4. Deep-Seated Temperature and Pressure Implications
 The ultimate goal of developing a comprehensive mass balance for the Mariana fore arc will require a coupled chemical and physical model. Although faults in the Mariana fore arc provide conduits for fluid and solid materials from the décollement that can then be collected using existing sampling techniques, the means to measure physical parameters along the décollement is nonexistent. These physical constraints are critical to models of seismic activity within subduction zones, not only in the Mariana region, but in all subduction systems. Temperature is one of the key physical parameters required by all models, not only affecting the physical state of the décollement but also the chemical state. Here, we use chemical data from seafloor fluid seeps to constrain thermal conditions along the décollement below the Mariana serpentinite mud volcanoes by comparing our results with laboratory experiments and land-based studies.
5.4.1. Major Elements
 Concentrations of the major elements in end-member pore waters show clear differences with distance from the trench that are useful in predicting thermal and pressure conditions at the décollement (Table 2). The three most monotonic, with respect to predicted depth to subducting slab (Table 1), major element patterns across the fore arc are the Na/Cl molar ratio and K and Na concentrations. The Na/Cl molar ratio increases with depth to the slab (Tables 1 and 2), resulting from the increased mobilization of Na relative to Cl with increasing temperature [Mottl et al., 2004]. By using this ratio as a conservative indicator of increasing depth to the slab, errors from modeling the slab depth, based on distance to the trench axis, thickness of the overriding plate, and the subduction angle, are examined. The most striking disparity in the predicted slab depth is Quaker where the depth was estimated to be 26 km, based solely on the geometry of the subducting slab and distance from the trench axis. A better approximation for the depth is 23 km, which is the average predicted depth below Baby Blue and Pacman Summit, with similar Na/Cl molar ratios. The cause for this expected difference could be the presence of a large seamount that has been subducted below Quaker.
 The Na trend across the fore arc offers additional constraints for temperature and pressure based on the dramatic decrease in Na concentration at Conical. End-member concentrations of Na steadily increase with depth to the slab from 220 mmol/kg to over 600 mmol/kg at Cerulean Springs and Big Blue, respectively. This increase likely results from exchange with Ca in the subducted slab, combined with balancing increases in the anions carbonate and sulfate. A drop of >200 mmol Na/kg in pore water from Conical may indicate that the P-T conditions at the source are greater here than at S. Chamorro. Such a drop in Na concentration could result from greater partitioning into sodic amphiboles and pyroxene minerals, which were recovered on ODP Leg 125. Yet, there was evidence of retrograde metamorphism, and only incipient blueschist facies rocks were recovered [Maekawa et al., 1992, 1993, 1995]. High-pressure sodic amphiboles were found in cores from S. Chamorro. On the basis of the coexistence of epidote and magnesiorebeckite/barroisite amphibole at S. Chamorro, P-T estimates of 0.4–0.5 GPa and 250–300°C at the décollement were calculated [Fryer et al., 2006]. Alternatively, the shift to lower concentrations of Na in the pore water at Conical may be explained by deposition into albite, which agrees with the observation that most of the plagioclase in metabasalts at Conical has been albitized [Johnson, 1992].
 Given the source of K is from the deep slab [Mottl et al., 2004], experiments involving seawater and basalt provide additional thermal constraints. These experiments illustrate that below ∼150°C K is removed from solution, whereas at warmer temperatures basalt is a source of K to solution [Seyfried and Bischoff, 1979]. This suggests that the slab temperature below Blue Moon, Cerulean springs, E. Quaker, NE Quaker, and Quaker is <150°C (Table 2) with the remaining serpentinite mud volcanoes having a warmer temperature at slab depths.
 The increase in alkalinity at greater distances from the trench is not the result of decarbonation from thermal sources through the mechanism described by Kerrick and Connolly . The low-temperature isotherms in the Mariana fore arc inhibit thermal devolatization to depths possibly exceeding the depth of arc magmatism. The source of the carbonate in the Mariana serpentinite mud volcanoes is attributed to the dissolution of carbonates [Mottl et al., 2003, 2004]. This may occur as a by-product of mineral phase transformations upon reaching critical water/rock ratios. The formation of lawsonite through the dehydration of zeolite minerals was described for medium-grade Franciscan metasedimentary rocks, and was thought to be associated with slight decarbonation of carbonate cements [Sadofsky and Bebout, 2003]. A minimum P-T boundary for carbonate mobilization is estimated at 0.3 GPa and ∼100°C, based on the occurrence of unaltered carbonate cements in the Coastal Belt of the Franciscan Metamorphic Complex [Sadofsky and Bebout, 2003].
5.4.2. Minor and Trace Nonmetallic Elements
 Estimates for compositions of minor and trace elements within the upwelling pore water of Mariana fore-arc serpentinite mud volcanoes do not always show the consistent patterns across the fore arc that many of the major elements show (Table 3). The elements that lack systematic trends include Li, Si and Ba. Uranium is depleted in the upwelling pore water at all the seamounts. Strontium and B vary consistently with predicted slab depth across the fore arc. Despite a lack of consistent trends for some of these elements, there are still some insights gained using the pore water data to constrain physical conditions at the décollement.
 Silica and Ba end-member concentrations fluctuate widely, which we attribute to secondary reactions at the slab-mantle interface and in the sampled sediment. Laboratory experiments show increased Ba concentrations in water contacting décollement sediment at elevated temperature and limited pressure, but Ba is rapidly removed in the presence of sulfate [You et al., 1996]. This might explain the elevated concentrations in pore water at E Quaker, which contains low sulfate (Table 2). These data indicate a temperature that is greater than 150°C below E. Quaker, based on the aforementioned experiments at 0.08 GPa.
 Estimates for Li concentrations in deep-seated fluids are significantly higher at Pacman Summit, Quaker, and Baby Blue than in seawater, but are depleted at the other seamounts. Lithium was measured at elevated concentrations within serpentinized peridotite clasts and the serpentinite mud matrix at Conical despite near-zero concentrations in upwelling pore waters [Mottl, 1992; Benton et al., 2004]. These enrichments, however, are highly variable (0.76 to 26.4 ppm), pointing toward initial removal of Li from the pore water during serpentinization of conduit wallrock, and subsequent serpentinization of the remaining peridotite by Li-depleted pore water [Benton et al., 2004]. If this were the reason for the Li depletion in upwelling pore water, then the presence of Li in upwelling pore water at the three aforementioned seamounts could be interpreted as the result of little to no active serpentinization along the path of pore water upwelling, consistent with the low pHs of these pore waters.
 Given this interpretation, one explanation for elevated concentrations of Li in pore water at only Pacman Summit, Quaker, and Baby Blue could be that these seamounts have not erupted for a relatively longer period than the remaining active seamounts sampled in this survey; and therefore the pore water conduits are completely serpentinized due to a lack of freshly exposed peridotite surfaces. An alternative to this theory is that a mineral phase transformation, limited to a narrow P-T range dictated by the depth to the décollement beneath these seamounts, is responsible for mobilizing Li. A third mechanism for the Li enrichment could be the presence and reaction of a high-lithium source at the décollement or in the fore-arc mafic crust. The subduction of a seamount can potentially incorporate a large mass of sediment in its wake [e.g., Lallemand et al., 1989], and when the sediment begins to dehydrate at depth it could release excess Li to the pore water. The dehydration of volcaniclastic material within the subducting seamount could also release copious amounts of Li. As in the case for K, seawater-basalt reactions show a distinct Li pattern where Li is removed from solution at low temperatures and added to solution at high temperatures. The lack of any systematic Li trend across the fore arc and the observed reactivity of Li in a serpentinite matrix preclude us from using the concentration of Li as a proxy for slab temperature.
 Strontium concentrations mimic Ca across the fore arc, consistent with reactions with carbonates and mobilization as a result of clay dehydration through the collapse of the 17 Å smectite layer. Uranium is universally depleted at all the seamounts, which agrees with studies of the Catalina Schist that show limited to no U mobilization at depths of 50 km and temperatures as high as 600°C [Bebout et al., 1999]. This is attributed to the reduction of U into the mineral-compatible U4+ state under reducing conditions [Langmuir, 1978]. Boron concentrations in pore water increase continuously with increasing depth to the décollement as a result of increased mobilization as the temperature rises, but the lack of an inflection point, as with K and Li, preclude using the B data to constrain thermal conditions at depth [Benton et al., 2001; You et al., 1996].
5.4.3. Rb and Cs
 Experiments with hydrothermal fluids in contact with décollement sediment at different temperatures show the progressive mobilization of Cs and Rb with temperature [You et al., 1996]. In these experiments, the molar ratio of Rb/Cs decreases continuously to temperatures up to 350°C. At higher temperatures, Rb mobilization outpaces Cs, producing higher Rb/Cs molar ratios. Given that the measured percent change in Cs is always greater than Rb, temperatures at the slab must be less than 350°C across the entire fore arc. Subducting sediment and ocean crust are the primary sources for these elements, as these elements are not mobilized from and are largely depleted in peridotite. This temperature maximum is consistent with temperature ranges predicted from stable isotopic analyses of serpentinites from Conical [Alt and Shanks, 2006] and from metamorphic mineral assemblages recovered from drilling at S. Chamorro [Fryer et al., 2006].
 The relationship between Rb and Cs in the shallow (prevolcanic arc) subduction setting has been a subject of detailed investigations under varying protolith and P-T conditions. These studies have demonstrated the increased mobility of Cs over Rb during dehydration of the subducting slab by observing increases in Rb/Cs molar ratio of low-grade versus high-grade metamorphic facies in the Catalina Schist [Bebout, 1995; Bebout et al., 1999; Sadofsky and Bebout, 2003], laboratory experiments [You et al., 1996], and thermodynamic calculations [Busigny et al., 2003]. While it is evident that Cs is preferentially mobilized into the liquid phase relative to Rb, the Rb/Cs molar ratio is conserved to depths of 90 km and temperatures of 600°C in metasediment of the Schistes Lustrés nappe, which was attributed to closed system behavior [Busigny et al., 2003]. Given that the fractionation of these elements in metasediment at shallow depths within the high P-T Franciscan Complex has been observed [Bebout et al., 1999], a mechanism for transporting fluid mobile elements away from the slab likely existed. Elevated concentrations of these elements and decreasing Rb/Cs molar ratios in upwelling pore water across the Mariana fore arc both verify the observations of increasing Rb/Cs molar ratios in paleosubduction zones and provide a potential mechanism for removal of these components from the system.
 In contrast to the Rb/Cs molar ratio, the Rb/K molar ratio remains relatively constant in both high- and low-grade metasedimentary rocks [Sadofsky and Bebout, 2003]. The Rb/K molar ratios in fluids at depth are generally equal to that of seawater (1.4 × 10−4) in the serpentinite mud volcanoes that are close to the trench. This ratio increases to 4.8 × 10−4 at Big Blue from values measured at Quaker (Rb/K 1.7 × 10−4), as Rb increases from 1.57 μmol/kg to 6.65 μmol/kg with only a 4.4 mmol/kg increase in K (Figure 14). The remaining western serpentinite volcanoes have similar ratios of Rb/K, with the Rb/Cs ratio steadily increasing to the west (Figure 14). This suggests separate lithologies are responsible for producing these fluid compositions, rather than thermal conditions alone. Differences have been demonstrated in volcaniclastics and zeolites with low Rb/K molar ratio compared with higher Rb/K molar ratios of ancient and weathered sources [Plank and Langmuir, 1998]. The shift from a low Rb/K molar ratio to higher values in pore water at Big Blue may therefore be the result of a transition between dehydrating low Rb/K molar ratio subducted sediment closer to the trench to dehydration of the higher Rb/K molar ratio basaltic ocean crust at greater slab depths.
5.4.4. Transition Metals
 Transition metals in the Mariana samples show strong redox behavior and are not conservative. The best approximation of their concentrations at depth within the seamount conduits can therefore only be given as the concentrations within the deepest sections of the cores with the highest pore water upwelling speeds (Table 4). Greater mobilization is evident at seamounts with low seepage speeds (e.g., Quaker and Baby Blue), and the pore water there is enriched in almost all transition metals. Faster upwelling speeds, such as those at Cerulean Springs and Big Blue, produce pore water with lower concentrations of transition metals, except Fe and Mn. The high Fe and Mn concentrations at these locations could indicate a source at depth, based on the rapid flow rates and low levels of the remaining transition metals. The lower concentrations of the remaining transitions metals at Cerulean Springs and Big Blue could indicate scavenging by oxyhydroxides, sulfides, or carbonates that are catalyzed by the presence of high Fe and Mn. Given the significant mobilization of Co and Ni at Big Blue, the pore water at greater depths could also be enriched in these components. Blue Moon, with a high upwelling speed, is uniformly depleted in transition metals, which could also reflect the composition at the source of pore water generation. Nevertheless, these data do not provide additional thermal constraints even though such elements are generally mobilized in seawater-basalt reactions above ∼300°C.
Table 4. Transition Metal Concentrations of Samples Most Representative of Deep-Sourced Fluids Sampled During the 2003 Expeditiona
Asymptotic values were not reached in any of the cores as complex gradients are caused by redox reactions in the sediment column; therefore the values given are from the deepest section of the core with the highest upwelling speed at each site. ODP Leg 125 (Conical [Mottl, 1992]) and Leg 195 (S. Chamorro [Shipboard Scientific Party, 2002; Mottl et al., 2004]) drilling results are given where available. Also included are measured values of the Site 1200 borehole water sampled in the 2003 expedition [Wheat et al., 2008].
 Pore water concentrations of the REE within all of the serpentinite mud volcanoes are significantly depleted relative to bottom seawater (Table 5 and Figure 7). This may result from either the consumption of REEs by precipitation of carbonates and Fe oxyhydroxides in the near-surface, through the depletion of REE at depth, or the initial fluid composition at depth is depleted of REE. If the former is the case, one might expect a correlation between the concentration of dissolved iron and/or alkalinity and REE concentrations. This is clearly not the case with respect to concentrations of Fe and the more than two orders of magnitude difference in the observed alkalinity. For example, Cerulean Springs and Big Blue are significantly enriched in dissolved Fe; but the REE patterns and concentrations from these two sites are distinctly different. Depletion of REE at depth would be required if the source of fluids is pore water, which typically have concentrations that are greater than bottom seawater values [Hulme et al., 2008]. Alternatively, the starting fluid could be basaltic formation water, which is generally depleted of REE relative to bottom seawater in ridge flank settings [e.g., Wheat et al., 2002] or from the dehydration of clays. Pore water REE data from 7 of the serpentinite mud volcanoes are generally consistent with a warm (50–150°C) basaltic formation water. For example, basaltic formation fluid from Baby Bare (∼64°C) lacks a seawater Ce anomaly and elevated Y/Ho ratio, similar to results from Mariana fore-arc pore water. In contrast, Baby Bare fluid lacks the slight LREE enrichment that is present in the Mariana fore-arc pore water [Wheat et al., 2002].
Table 5. REE Composition of the Samples Most Representative of Deep-Sourced Fluids Collected in the 2003 Mariana Fore-Arc Expeditiona
All the upwelling pore waters are depleted relative to bottom seawater. Mud volcano pore water near the trench contained higher concentrations of the REE than the deeper-sourced pore water (i.e., greater depth to décollement). Big Blue concentrations show a slight LREE-enriched pattern, which could reflect a slab component and elevated temperatures.
 A positive Gd anomaly occurs in all but S. Chamorro, Cerulean Springs and Blue Moon. If the Gd anomaly were associated with the formation of Fe oxyhydroxides as described by Bau , one would expect corresponding La, Ce, and Lu anomalies, in contrast to the data. Alternatively, the Gd anomaly could be related to reactions with plagioclase, which contains significant amounts of Gd. Because the ionic radius of Gd is similar to that of Ca, a major component in plagioclase [e.g., Goldberg et al., 1963], Gd and Ca will be exchanged with Na during albitization, producing a positive Gd anomaly [e.g., Wheat et al., 2002]. There is no correlation between Ca and Gd in the upwelling pore water, however, possibly because of subsequent Ca removal in carbonates.
 The Gd anomaly could also result from complexing. Gd has lower complexation constants with solution and surface ligands than its neighboring elements because of the tetrad effect [Masuda and Ikeuchi, 1979]; and, therefore, whenever complexing by surface ligands far exceeds complexing by solution ligands a relative Gd enrichment in the solution occurs [Kim et al., 1991; Zhang and Nozaki, 1998]. Experiments by Quinn et al.  predict similar Gd anomalies in solutions with low carbonate ion concentrations (<1 μM) as surface complexation dominates; but the anomaly is less apparent with increasing carbonate concentration.
 Carbonate complexing often dominates REE concentrations in high pH solutions [Lee and Byrne, 1992; Millero, 1992; Tang and Johannesson, 2003]; but pressure also has a strong influence on REE complexing. Gd carbonate experiments at 500 atm revealed a fivefold decrease in carbonate complexing relative to laboratory conditions [Lee and Byrne, 1994]. Pressures within the slab below the Mariana mud volcanoes are more than 20 times greater than this experimental pressure, implying an even greater diminished capacity for carbonate complexing. This complexing argument may offer insight into the process responsible for the overall REE depletions, by suggesting that the decrease of REEs in the deeper-sourced pore water is a result of scavenging processes rather than a lower degree of mobilization.
 There is a positive Eu anomaly in the REE pattern at S. Chamorro, which is a common occurrence in hydrothermal vent fluids [Klinkhammer et al., 1994; Douville et al., 1999]. Initial hypotheses for the source of Eu involved the dissolution and alteration of plagioclase crystals; but after analyzing REEs in hydrothermal fluids collected at the ultramafic Rainbow vent field, and detecting the same anomaly in the absence of plagioclase, Cl complexing was suggested as causing the Eu anomaly [Douville et al., 2002]. This theory was tested and confirmed by laboratory experiments conducted by Allen and Seyfried ; However, at high (up to 12.5) pH conditions Cl complexing should have little effect on REE mobilization because of the prevalence of OH– complexing [Schijf and Byrne, 2004; Gammons et al., 1996].
 Given there is a low degree of chloride complexing at higher pH, there are two possible explanations for the positive Eu anomaly in S. Chamorro pore fluids. The first is that the pH at depth is significantly lower during initial serpentinization reactions, and the REE pattern produced at depth is preserved as the fluid ascends. Lower pH is common in ultramafic-hosted high-temperature vent fluids, but can vary with the degree of phase separation and water-rock ratios [Douville et al., 2002]. Experiments by Janecky and Seyfried  demonstrated the control of Mg concentration on the pH of serpentinizing water. While seawater exposed to peridotite rapidly decreased in pH, once Mg was removed from the solution the pH rose to 11.5. The pore water at S. Chamorro is free of Mg, so the pH during serpentinization should remain high, thus preventing chloride complexing.
 Another possible mechanism for producing the Eu anomaly is equilibrium with plagioclase during alteration of the mafic basement, as described by Klinkhammer et al. . This is consistent with evidence of albitization at Conical and the K versus Rb/Cs of upwelling pore waters across the Mariana fore arc as discussed in sections 5.4.1 and 5.4.3. Equilibrium with the ultramafic basement was also considered to explain the presence of positive Eu anomalies in serpentinites analyzed from S. Chamorro and Conical seamounts [Savov et al., 2005, 2007]. However, the REE patterns of these serpentinites are U-shaped, have elevated HREE/LREE, and no Gd anomalies. It is possible that the source of LREE, including Eu, at S. Chamorro is serpentinite, but the mobilization of these elements is still thermally controlled and significantly altered by complexing and mineral dissolution reactions.
 In addition to the positive Eu anomaly, pore water at S. Chamorro is enriched in LREE relative to HREE. REE patterns at Big Blue are depleted, but they also have a slight LREE enrichment, You et al.  observed LREE mobilization during reactions between hydrothermal fluids and sediment at temperatures above 200°C. LREE depletion also is found in blueschist grade metasediment that formed at pressures of 0.6 to 0.9 GPa and temperatures between 150 and 300°C [Sadofsky and Bebout, 2003]; but they interpreted this result to indicate that the protolith REE had the same pattern and, as such, was not the result of LREE mobilization. Given the overall REE depletion of the formation pore water, as predicted by the pore water profiles, the presence of increased LREE relative to HREE in the upwelling pore water could be significant and would constrain the pore water formation temperature beneath Big Blue to >200°C. This recent observation supports the possibility of LREE mobilization within subducted sediment undergoing lawsonite-albite to lawsonite-blueschist metamorphism. Temperature constraints, based on the best chlorite geothermometers, of 180 to 280°C were made for the top of the subducting slab beneath S. Chamorro [Gharib, 2006]. Given the increase in the degree of LREE enrichment at this site, together with the onset of a positive Eu anomaly, the actual temperature is likely to be near the middle of this range.
 The Mariana fore arc is an ideal environment for examining physical and chemical properties of a shallow subduction zone because of (1) the lack of an accretionary prism, which normally obfuscates the composition of slab-derived pore waters as they ascend through a thick sedimentary sequence, and (2) the presence of active serpentinite mud volcanoes, which provide a direct conduit for slab-derived water and altered mantle matrix to rise to the seafloor [e.g., Fryer et al., 2006]. Active serpentinite mud volcanism is currently known to exist only in the Mariana fore-arc region, where a combination of extensional tectonic forces and subducting seamounts fracture the overriding fore arc, allowing serpentinite and associated pore water to escape from mantle depths. Serpentinite mud volcanoes resulting from this process are distributed along the outer fore arc at up to ∼90 km from the Mariana Trench axis. By examining the pore water composition within these seamounts we constrain the chemical and thermal properties of the subducting slab as it descends toward the zone of arc magmatism.
 Systematic variations in pore water chemical profiles are evident across the fore arc as a function of distance from the Mariana Trench, which acts as a proxy for the depth to the slab (Figure 15). At Blue Moon, the seamount nearest the trench, only Ca and Sr concentrations are elevated above that of seawater, reflecting a pore water source associated with the collapse of smectite. The immobility of Cs at Blue Moon requires a temperature <80°C at the source of pore water generation. At Cerulean Springs, which is farther from the trench axis, Cs mobilization is evident, consistent with a temperature in excess of 80°C at 13 to 16 km below the seafloor. Further from the trench axis, representing slab depths of 17 to 24 km, pore water lacks significant Rb and K enrichments relative to seawater, suggesting temperatures <150°C at the source. The low Rb/K and Rb/Cs at all of these intermediate seamounts suggest the upwelling pore water is generated from dehydration of altered subducted clays [Plank and Langmuir, 1998]. The depth to slab at Big Blue is ∼25 km below the seafloor, here we estimate the temperature is ≥200°C, based on elevated concentrations of LREE relative to HREE. In addition, the ratios of Rb/K are higher at Big Blue, reflecting a basaltic source for these pore water. Trends of Rb/Cs across the fore arc place upper limits on slab temperatures beneath S. Chamorro and Conical, seamounts farthest from the trench, at 350°C.
 These new insights into the thermal conditions and elemental exchange within an actively subducting slab will not only aid in modeling P-T conditions within active subduction zones, but also in interpreting exposures of paleosubduction zones around the world. For example, a series of exposed metamorphic terrains exist along the western North American Continent where subduction occurred through much of the Mesozoic; and the thermal history was thought to be similar to that of the Mariana subduction zone. Sequences of previously subducted sediment and oceanic crust are now exposed as a consequence of tectonic processes associated with transform motion along the San Andreas Fault. Considerable data exist regarding formation histories of these units [e.g., Coleman, 2000; Bebout et al., 1999; Sadofsky and Bebout, 2003], but some of the details of the system are unknown because of retrograde reactions that occur during exhumation. By combining the information gained through studying active Mariana fore-arc serpentinite mud volcanoes with the results of mineral assemblages of exposed metamorphic sequences, constraints on the closed versus open system behavior of subducting slabs can be defined. For example, if no conduits exist for serpentinite and pore water to vent, fluid mobile elements will be subducted to the depth of arc magmatism, and consequently there will be no mass fractionation in the fore arc [Busigny et al., 2003]. Studies in the Catalina Schist of southern California show that fractionation does occur within oceanic crust subducted to fore-arc depths [Bebout, 1995; Bebout et al., 1999]. Given this result and the ones presented herein, open subduction systems, such as the active Mariana mud volcanoes, existed in recent geologic history.