Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Hydrothermal systems at mid-ocean spreading centers play a fundamental role in Earth's geothermal budget. One underexamined facet of marine hydrothermal systems is the role that permeability of the uppermost seafloor veneer plays in the distribution of hydrothermal fluid. As both the initial and final vertical gateway for subsurface fluid circulation, uppermost seafloor permeability may influence the local spatial distribution of hydrothermal flow. A method of deriving a photomosaic from seafloor video was developed and utilized to estimate relative surface permeability in an active hydrothermal area on the Endeavour Segment of the Juan de Fuca Ridge. The mosaic resolves seafloor geology of the axial valley seafloor at submeter resolution over an area greater than 1 km2. Results indicate that the valley walls and basal talus slope are topographically rugged and unsedimented, providing minimal resistance to fluid transmission. Elsewhere, the axial valley floor is capped by an unbroken blanket of low-permeability sediment, resisting fluid exchange with the subsurface reservoir. Active fluid emission sites were restricted to the high-permeability zone at the base of the western wall. A series of inactive fossil hydrothermal structures form a linear trend along the western bounding wall, oriented orthogonal to the spreading axis. High-temperature vent locations appear to have migrated over 100 m along-ridge-strike over the decade between surveys. While initially an expression of subsurface faulting, this spatial pattern suggests that increases in seafloor permeability from sedimentation may be at least a secondary contributing factor in regulating fluid flow across the seafloor interface.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The surface veneer of volcanic rock outcrops and sediment cover at the seafloor represents a critical gateway for fluids entering and exiting the subsurface hydrothermal reservoir at mid-ocean ridges. Although the uppermost few hundreds of meters of igneous crust at the Juan de Fuca spreading center has a high porosity and presumably a high permeability [Gilbert and Johnson, 1999], previous, nonsystematic observations of the axial valley seafloor at Endeavour indicate the presence of large areas with unbroken basalt flows or nearly complete sediment cover that may limit fluid transmission between the ocean and crust. This study attempts to interpret relative seafloor permeability from a systematic seafloor imaging survey of a portion of the axial valley. Surface permeability throughout the axial valley floor is an important but largely unevaluated boundary condition for modeling fluid circulation of the axial hydrothermal systems at the Endeavour Segment and at other spreading centers.
 Hydrothermal fluid circulation through oceanic crust is estimated to provide ∼25% of the Earth's surface heat flux [Stein and Stein, 1994]. Understanding marine hydrothermal circulation is required to constrain the planetary heat budget and to define a system that provides energy and nutrients to chemosynthetic communities, Earth's oldest surviving ecosystem [Childress and Fisher, 1995]. Among the remaining principal undefined parameters for hydrothermal systems are the circulation pathways, depth of penetration, velocities, and residence times of fluid traveling through the subsurface. Seafloor hydrothermal systems vary widely on a global scale, and models of fluid circulation for a given system must necessarily be based on local conditions.
 There are currently three generalized models proposed to describe the circulation pathways of fluid in hydrothermal systems at medium-to-fast-spreading mid-ocean ridges: (1) seawater enters the subsurface along deep, normal faults at one or both edges of the axial valley and then flows across-axis to emerge as high temperature outflow in the center of the valley or on the side opposite the recharge zone [Williams et al., 1979; Johnson et al., 2010], (2) fluid circulation takes place primarily within individual faults that are oriented along-strike and parallel to the ridge axis in distinct slot convection cells [Rabinowicz et al., 1999; Wilcock, 1998; Tolstoy et al., 2008], and (3) fluid convects in a basement layer with uniformly isotropic permeability as annular rings with broad circular recharge areas surrounding narrow up-flow regions [Coumou et al., 2008; Tivey and Johnson, 2002: Johnson et al., 2010]. Each of these models appears to be applicable for specific hydrothermal vent systems located on spreading centers within a different geological environment. While these models have utilized a wide range of data types and boundary conditions, the potential for the possible influence of the permeability of the surficial seafloor has not been considered, due largely to the lack of appropriate data.
 A novel method for estimating seafloor surface permeability is the use of high-resolution photomosaics constructed from video footage taken during remotely operated vehicle (ROV) transects over a large area of the seafloor. Although multibeam swath bathymetry and side-scan sonar data are useful for identifying features on a scale of tens to hundreds of meters, visual observations from submersibles and photomosaics from remote cameras are required for accurately identifying the location and lateral extent of hydrothermal and volcanic features on the scale of 10−1 to 10+2 m [Robigou et al., 1993; Escartín et al., 2008]. As examples, Lessard-Pilon et al.  used photomosaic images of cold seep communities in the Gulf of Mexico to describe changes in environmental conditions over small spatial and temporal scales and described how the composition of associated biological communities responded to these changes. Mittelstaedt et al.  quantified the heat flux from diffuse and discrete venting using photomosaics and video velocity analysis of outflow at the Lucky Strike hydrothermal field on the Mid-Atlantic ridge. Temporal changes in heat flux in this study were quantified using analysis of the spatial extent of bacterial mats and proximity to faults. The Lucky Strike study utilized specialized software to create extensive photomosaics of the seafloor with a precision sufficient for accurate comparison of images obtained separated by several years [Barreyre et al., 2012]. The present project utilizes photomosaic interpretive techniques to study the seafloor of the Raven portion of the Endeavour Segment of the Juan de Fuca spreading ridge (Figure 1).
1.2. Endeavour Background
 The axial magma lens beneath the Endeavour Segment has been imaged seismically as a midcrustal reflector 2.1–2.4 km below the seafloor that underlies all of the known hydrothermal vent fields [Van Ark et al., 2007; Wilcock et al., 2009; Kelley et al., 2012]. The zone of partial melt beneath the Endeavour Segment extends approximately 24 km along-axis and 0.4–1.2 km across-axis, with the nonhorizontal upper surface dipping to the east with slopes that vary from 8° to 36° [Van Ark et al., 2007]. Episodic replenishment of melt occurs along the Endeavour magma lens, providing the periodically renewed heat source previously suggested as a requirement for long-lived, extensive hydrothermal circulation systems [Wilcock et al., 2009; Liu and Lowell, 2009].
 Seismic velocities for the axial portion of Layer 2A at the Endeavour Segment are exceptionally low, in some areas approaching 1.8 km/s [Van Ark et al., 2007], suggesting a very high porosity of up to 30% for the uppermost 350 m of basalt [Nedimović et al., 2008]. The immediately underlying 2B layer on axis has an average seismic velocity near 5.2 km/s [Newman et al., 2011]. Upper crustal densities based on seafloor gravity measurements at the axis are also quite low, supporting a high subsurface porosity of >30% for the uppermost crust at the Juan de Fuca Ridge [Holmes and Johnson, 1993; Gilbert and Johnson, 1999]. For fluid flow estimates, we used the established correlation between seismic velocity and permeability for upper crust from Carlson 
along with average measured seismic velocities of 2.5 km/s in Layer 2A at Endeavour from Nedimović et al.  to estimate uppermost crustal permeability ranging from 2 × 10−10 to 2.5 × 10−12 m2 assuming no additional uncertainty in seismic velocity. Similarly estimated permeability for the underlying Layer 2B was lower on axis at 7.9 × 10−13 to 4.0 × 10−15 m2 also assuming no additional uncertainty in seismic velocity from Newman et al. . The resulting general model for circulation includes a highly permeable upper layer overlying a relatively low permeability layer of dikes or altered transition zone material. In contrast to the upper crustal layers of extrusive lavas, the composite hemipelagic sediment and fine-grained turbidites common to the Juan de Fuca Ridge flank have a significantly lower permeability, ranging from 10−15 to 10−18 m2 [Giambalvo et al., 2000]. Comparatively, at the Mid-Atlantic ridge flank, in situ sediment permeability at shallow burial depths (<3 m) was measured at 7.55 × 10−16 [Langseth et al., 1992]. Due to the small thermal buoyancy forces driving fluid circulation on the ridge flanks, and the low permeability of sediments on the ridge flanks, only minimal fluid seepage has been reported for sediment layers less than tens of meters thick [Giambalvo et al., 2000; Spinelli et al., 2004; Hutnak et al., 2006].
 Additional constraints for hydrothermal fluid circulation pathways at the axis of the Endeavour Segment were proposed by Johnson et al.  using a series of bare-rock conductive heat flow measurements across the axial valley. This heat flow data suggested that multiple circulation modes may exist simultaneously at the nearby Main Endeavour Field. The high temperature vents were supplied by a deeply circulating pathway that extended from the initial seawater recharge sites at the base of the eastern valley wall to the high temperature discharge sites at the base of the western wall. A secondary system is located in shallow crust above the deep high temperature circulation system, and that seawater recharge zone is located far closer to the low temperature diffuse outflow zones that surround the high temperature emission sites [Lowell et al., 2007, 2012]. In the interest of improving hydrothermal circulation models at mid-ocean ridge systems, additional parameters for the control of fluid flow should be considered. The purpose of this paper is to investigate the effect of regional sediment cover and intact basalt flows over a portion of the axial valley in the Endeavour Segment of the Juan de Fuca ridge.
2.1. Project Goals
 The goal of this project was to generate a map of the surface geology of a portion of the Endeavour axial valley with compiled photomosaic images using inexpensive commercially available software and then to interpret it in terms of the relative permeability of the seafloor. The photographic data were collected during two ROV JASON II dives on the Raven hydrothermal field (see supporting information1, Figure B). The southern boundary of this area is located ∼100 m north-east along axis from the northern boundary of the well-studied Main Endeavour hydrothermal field [Johnson et al., 2010; Kelley et al., 2012] (Figure 1). These dives were conducted primarily for research goals other than video imaging, including the acquisition of bare-rock conductive heat flow, crustal magnetization, and bottom water temperature data [Salmi et al., 2012]. The video images were acquired continuously during the dives, standard geometric transformations were applied, and ArcGIS was used to align, warp, and spatially coregister images into a photomosaic of a 0.8 km × 1.0 km area of the axial valley, largely centered on the small Raven hydrothermal field. Bottom water temperature data from a ROV-mounted CTD (conductivity, temperature, depth) were acquired simultaneously and used to identify sites of active fluid emission within the study area. These data were compiled into a high resolution map of seafloor geology and near-bottom water temperatures that were interpreted in terms of the permeability of the uppermost surface of the crustal reservoir.
 The project intended to identify spatial patterns in upper seafloor surface permeability within the Endeavour axial valley at the Raven vent field through semiquantitative analysis of ROV video images. However, the surface relative permeability, inferred from seafloor geological classification and degree of sediment cover, is unlikely to be representative of the subsurface crust of Layer 2A. Geological features which are visible at the surface may not extend to depth below the surface; an example being high permeability talus piles which could overlie intact basaltic sheet flows of low permeability. Nevertheless, fluid that either enters or leaves the subsurface fluid reservoir must pass through the observable seafloor veneer. Thus, while our study provides no information directly relating to fluid flow within the subsurface crustal reservoir, estimates of surface permeability may provide useful localized constraints on the sites of potential inflow and outflow, since fluid either entering or leaving the subsurface must pass through this veneer.
2.2. Data Acquisition
 Raw image sets were derived from the video recording system mounted on the ROV JASON II and acquired in 2011 during dives 586 and 590. The JASON II navigation system utilizes an ultrashort baseline transponder. Absolute navigation errors after postprocessing are conservatively estimated at less than 10 m, although our ability to easily relocate instruments, even at sites with poor visibility, suggests a precision roughly a factor of two better than this. Comparison of cross-track image feature alignment also suggests that both the accuracy and precision errors may be closer to 1–2 m. Standard resolution (non high definition) video recordings were maintained for three onboard cameras throughout all ROV dives on the cruise, providing multiple simultaneous viewpoints. For the purposes of constructing the primary photographic mosaic, the video camera on the brow of the vehicle was used exclusively for the following reasons: (1) the camera position provided the most uniformly continuous coverage of the seafloor, (2) obstructions to the field of view were minimal, and (3) the camera orientation relative to the vehicle and zoom settings were held constant throughout all dives, allowing a standardized image transformation method to be used for all obtained images. The camera equipped for these dives was an Insite Pacific MINI ZEUS high definition color camera with an 85° horizontal by 64° vertical viewing angle when held at the lowest zoom setting, with a 5.1 mm lens diameter. The brow camera was located above the center of the leading edge of the vehicle (see supporting information1, Figure A). The camera's point of view was 36.5 cm forward, 227.3 cm above, and directly in line with the vehicle's navigational nodal point along the long axis. All navigation data points were adjusted for this discrepancy. The ROV was tasked entirely with instrument deployment and recovery during the dives and did not perform a regular survey pattern. The altitude was not held constant during transit, although an effort was made to keep seafloor features in view for logging purposes.
2.3. Image Registration and Preparation
 Video files were processed using Freestudio's “Video to JPG Converter” to extract approximately 300,000 frame-grab images from the continuous video at 3 s intervals. The oblique camera angle and position required all images to be cropped at consistent pixel distances from the frame edges. This editing removed areas of the images that were poorly illuminated or obscured by the vehicle frame or static instrument positions. Movable structures, including manipulator arms and the extendable equipment basket, occasionally entered the frame and required additional manual processing. Although camera orientation was fixed, the ROV lighting system configuration changed depending on dive objectives, altering seafloor illumination from dive to dive. However, these changes were not logged, and an image lighting correction was not attempted.
2.4. Image Selection
 An automated image selection process was designed to provide optimal seafloor area coverage while reducing the distortion and blurring caused by minor misalignment of multiple images in areas of high image density. To reduce image clustering, a minimal horizontal separation of 3 m in vehicle position was required for consecutive images. Images were sampled frequently when vehicle transit speeds over the bottom approached the maximum of 0.5 m/s, and sampled less often when the vehicle progress was slow or stopped. An altitude filter of 12.2 m was used to remove images where the lighting and visibility were inadequate for feature recognition. Over areas of smooth terrain, this filtering process resulted in optimal photographic coverage and feature clarity. In contrast, regions of rough topography or extremely varied terrain produced image clustering or gaps in the mosaic.
 The ROV was equipped with fiber optic north-seeking gyro, solid state flux-gate compass, and 300 kHz Benthos altimeter with a 30 m range. The altimeter had a poor signal-to-noise ratio over rough terrain. Given the need for high quality navigation and attitude information for accurate image projection and alignment, substantial smoothing of the altimeter data was required. A 13 s moving average was used to reduce misalignment between consecutive images and was forward-stepped to remove artificial lag times.
2.6. Image Projection
 An algorithm was constructed to convert raw navigation data and camera configuration into accurate image position, orientation, and scaling within a 2-D geospatial framework (see supporting information, Figure A). This process utilized trigonometric transformations to account for variations in vehicle position, heading, altitude, pitch, and roll. The algorithm calculated the coordinate points for the four corners of the projected image using an idealized flat projection surface. This spatial data was merged with subsampled images to supply the GIS (geographic information system) software with all requisite parameters for mosaic production. Within the ArcGIS Data Management toolbox, the “Warp” function was used to transform images which had been photographed at oblique angles into horizontally projected and georeferenced 2-D representations of the seafloor (Figure 2).
2.7. Mosaic Construction
 After projection onto a horizontal plane, all images were subjected to final manual filtering to remove images based on the following criteria: (1) images showing portions of vehicle or attached equipment, (2) images containing a large amount of disturbed sediment in the water column that obscured the seafloor, and (3) images where illumination, contrast, or distance prevented recognition of geological features on the seafloor. Once selected for quality, the individual images were merged into a data set as a continuous 1-D track-line image within ArcGIS. A distance-weighted blend function was chosen to reduce contrasting edges between overlapping images. Since physical dimensions and resolution for a fixed image are inherently inversely proportional, smaller images were given priority in any overlap with larger images.
2.8. Creation of Geologic Map
 A manual interpretive method was used in the creation of geologic maps from the linear image mosaics. Each individual track-line mosaic was analyzed independently to improve consistency and reduce regional bias. Classification of seafloor features utilized an along-track approach designed to identify small selections of the seafloor with homogenous features. Polygonal shapes representing specific geological classes were drawn over each feature in the track-line image, and all borders between polygons were artificially forced to be contiguous and without overlap (Figure 2). Viewing scale of the mosaic was maintained uniformly throughout the process at 1:100. Classification of features fell into five main geological categories (Figure 3), with separate designations for rare or specialized features. The classification scheme was based on the following criteria: (1) unbroken sediment: uniform covering of 100% sediment, underlying geologic features not visible or not identifiable, no outcroppings, no cracks, (2) sedimented flows: light to thick sediment draping over recognizable flow or talus morphology, occasional basalt outcroppings, small cracks, (3) broken flows: sediment-free or only light sediment draping, loosely assorted material of varying shapes and sizes, numerous small voids and small gaps between basalt rocks, (4) sedimented talus: light sediment draping over recognizable talus, substrate has consistently small particle size, and (5) talus: no sediment draping, substrate has consistently small particle size. Special categories were created for less common or challenging bottom types such as fissures, cracks, sheet flows, vent biology, and fault scarps. Defining the extent of sheet flows within the valley was difficult given the level of sediment coverage. Even a light covering of sediment, on the estimated order of tens of centimeters, obscured the edges of sheet flows and prohibited accurate registration. Thus, while these features were identified regularly and assumed to be common throughout flat portions of the valley, the extent of this bottom type was inclusively classified as either unbroken sediment or sedimented flows. Faults and fissures are not included on the figures due to the relatively fine scale of the individual features. The entire mosaic data set was classified for geological type by a single observer, and the examination order of individual track-line sections for classification was kept random to reduce any regional bias.
2.9. Sources of Error
 Error accumulation for the geospatial position of pixels throughout the mosaic process limited confidence in the absolute location of submeter scale seafloor features. Propagated absolute pixel errors were conservatively estimated as high as 15 m, while relative errors between consecutive images were less than 2 m. The ∼200 seafloor instrument placements and recoveries provided extensive and redundant coverage of the axial valley and walls; regional trends and substantial features are easily identifiable, and agreement between both consecutive and cross-track images is high (see supporting information, Figure B). Additionally, a high degree of correlation exists between topographic features visible in previously collected SM2000 swath bathymetric data [Johnson et al., 2010] and classified bottom types. As an example, all areas classified as talus are colocated with areas of high seafloor slope.
 The linear photomosaics and resulting geologic maps were limited in coverage to the specific track lines taken by the vehicle, resulting in some data gaps between image sets. For purposes of trend identification and statistical evaluation, an interpolation technique was used for gap-filling of geologic classification data. To maintain objectivity and consistency, an iterative nearest neighbor majority interpolation process [Cover and Hart, 1967] was applied to the entire classification data set (Figure 4). The vector classification polygons were rasterized at 0.5 m cell size, and cells were given numeric values based on the source polygon's geological classification. The interpolation function consulted all cells within a three cell range and determined the majority class value and then reassigned the subject cell with a matching value. This process added new classified cells to the perimeter of the existing observed survey area. This method projected classification values into unobserved space while preserving 100% of the original observations by ensuring that source cells were reassigned to their original class values after each iteration. After 15 sequential iterations, the majority of the gaps in the survey area were filled (see supporting information1, Figure C for additional information). Each cell was also assigned a confidence value between 0 and 100 based on the iteration step that was used for classification; the manually identified cells were given the greatest confidence of 100, decreasing outward to those classified on iterative step 15, which were given the lowest level of zero confidence (Figure 4).
2.11. Permeability Estimated From Interpreted Seafloor Geology
 In order to create a map of relative surface permeability, each of the five major bottom-type classifications was assigned a relative permeability value based on the observed frequency of cracks, voids, openings, and extent of sediment cover. Because of the semiquantitative nature of our geological interpretations, relative surface permeability was conservatively binned into only three broad categories using the following criteria: (1) high permeability: talus and sedimented talus, (2) medium permeability: broken flows, and (3) low permeability: sedimented intact pillow basalt flows and unbroken sediment cover without outcrops. In our model, as relative sediment cover for each bottom class increases, and the occurrence of cracks, fissures, and gaps between basaltic rocks decreases, the relative permeability of the surface seafloor in our model decreases. We make this key assumption based on estimations of the vertical hydraulic impedance of a layer of thin sediment [Karato and Becker, 1983]. Estimations of sediment depth within regions classified as “low” permeability were made by assessing the vertical relief of unsedimented and partly sedimented extrusive rock features through comparison with scale references in the images. The vertical relief range of the visible layer of pillow lavas and sheet flows in flat areas is 0.2–2.1 m in over 100 observations, averaging just over 1 m. In order for this layer of rock to be entirely covered with sediment, as is the case within the valley floor and northern terrace, a blanket of sediment with a depth of over 1 m would be required. A layer of local sediment that is 1 m thick, with an in situ estimated permeability of 10−16 m2, would have a vertical hydraulic impedance of 1.0 × 1016 (1/m). In contrast, an identical 1 m thick section of upper Layer 2A at Endeavour with a permeability of 2 × 10−10 to 2.5 × 10−12 m2 would have a vertical hydraulic impedance of 5.0 × 109 (1/m) to 4.0 × 1011 (1/m). At an average depth of 200 m on axis, the resulting hydraulic impedance of a vertical section of Layer 2A would be 2–3 orders of magnitude less than that of the overlying sediment layer. It is critical to stress that this interpretation is limited to only the uppermost visible layer of sediment and geologic structure, providing no ability to project permeability estimates downward into the subsurface.
2.12. Near-Bottom Water Temperature Data Acquisition
 The ROV-mounted SeaBird CTD sensor logged water properties throughout all dives at 1 s intervals. These water property data were filtered to retain only those measurements made between 0.6 and 25 m of the seafloor. During instrument deployments or other pauses in vehicle motion, heat generated by the ROV electrical systems would be transferred to the water and recorded by the CTD sensor. Filtering all CTD temperatures to remove data when the vehicle speed dropped below 0.1 m/s greatly reduced the influence of artificial heating by lack of ROV forward motion.
 The water temperature data were also corrected for the decrease in sea water temperatures associated with increasing depth. All ROV CTD water temperature and depth measurements taken during the dives were combined, sorted into 20 m depth bins, and averaged. A linear regression of water temperature versus depth was calculated from this compilation, since a more complex variation was not justified. All CTD measurements were then corrected for changes in vehicle depth by subtracting the linear regression value, assuming the residual is the temperature anomaly due to near-seafloor heating (see supporting information1, Figure D). Previous studies have shown that within the Endeavour axial valley, near-seafloor water parcels can migrate along-axis several hundred meters within a single tidal period [Garcia-Berdeal et al., 2006]. Removing this tidal signal for the several-day period of our survey was not possible since no current meter data were available.
2.13. Data Acquisition Over Survey Area
 The 2011 ROV JASON II dive area encompassed in a rectangular area 1 km wide (E-W) and 0.8 km along-strike (N-S) spanning the width of the axial valley and valley walls that included the Raven hydrothermal field (Figure 1). Initial photographic coverage of the area of the survey track lines was estimated at ∼48%, while images that passed the filtering processes and were included in the final mosaic represented ∼45% of the total area. Average camera altitude for the entire survey was 5.5 m, with the images used for geological interpretation ranging from 2.2 to 12.2 m. The bottom time for the two individual JASON dives at Raven was 96 and 48 h.
3.1. Geological Description of the Survey Area
 Several distinct regions of seafloor geology and sediment coverage exist within the axial valley (Figures 5 and 6). From west to east, the topography of the western wall is uneven and steeply sloped. Normal faults are observed, with exposed scarps extending subvertically nearly 50 m. The most common bottom geological classes on the west wall are talus and broken flows, with islands of scattered sedimented flows in isolated flat areas. A large terrace is located east of the steepest part of the western wall that gently slopes inward toward the center of the axial valley (Figure 5). Topography of this terrace structure ranges from moderately rough in the south to wide and flat at its northern extent. Sediment cover of the seafloor within the terrace is more extensive than on the western wall, and sedimented and broken flows are common with unbroken sediment found primarily to the north. East of the terrace, the seafloor drops steeply to the floor of the axial valley. The transition zone between the terrace and the flat valley floor consists almost entirely of broken flows and talus with only minimal sediment coverage. From the eastern boundary of the talus accumulation derived from the western wall, the flat axial valley floor extends 300 m to the east and has relatively complete sediment cover and few exposed rock outcrops. The only breaks in the continuous sediment cover of the axial valley floor were cracks, short fissures, and small exposed faults observed in rock outcrops on the western side of this region, although many of these gaps were at least partially filled with sediment. Across the valley floor, estimated vertical relief from geologic structures averages only 1 m, while fault scarps on the eastern and western ridges can commonly rise abruptly 10 or more meters.
 At the eastern edge of the valley floor, the eastern wall rises abruptly and seafloor geology is dominated by small successively higher terraces composed of talus and broken flows interleaved with steep slopes and only minimal sediment cover. Survey tracks are less dense to the east but show that topography flattens at the summit of the wall and sediment cover increases dramatically, with few exposed rock outcrops. For the entire survey area covered by the photographic images, sedimented flows were the largest single type by area (41%), followed by broken flows (29%), unbroken sediment (21%), and talus (8%), with sedimented talus being the least common (2%). The interpolation process altered the percentage coverage of individual classes by less than 1% from that of the original uninterpolated observations.
3.2. The 2011 Hydrothermal Vents
 Within the Raven survey area, a total of 40 individual inactive sulfide mound structures were identified. The majority of these inactive hydrothermal deposits were clustered in groups that were distributed along a roughly linear pattern running NW to SE on the western wall, across the lower terrace, and the western edge of the central valley floor (Figures 5-7). Sites of active fluid venting observed in 2011 were restricted entirely to the foot of the west wall. The 16 active sites were identified by the presence of vent biology, including tube worms and bacterial mats, shimmering water from diffuse flow, or by sulfide chimneys actively emitting cloudy hydrothermal fluid.
3.3. Hydrothermal Activity From a Previous 2001 Survey
 Compilation of video images from the 2001 tn129 JASON II cruise to the Raven area showed evidence of 18 sites actively venting diffuse low temperature fluid in close proximity to (<20 m) the area imaged in 2011 [Johnson et al., 2002]. In contrast, a single chimney emitting high temperature fluid was identified in 2001 with a temperature of 229°C. The only high temperature chimney observed in 2011 was located 104 m to the south of the 2001 high temperature vent site, a distance well outside of any navigational errors. Although the older 2001 high temperature site is still an area of low level fluid emissions, and the inactive 2001 sulfide spire was still visible in 2011, it is clear that the single high temperature fluid emission site within the Raven field has migrated over 100 m along-strike to the south in the 10 years between observations.
3.4. Near-Bottom Water Temperature Anomaly
 Water temperature anomalies from the processed ROV CTD data ranged from −0.034°C to +0.85°C over the survey area. For the duration of both dives, there was no statistically significant temperature anomaly trend with respect to vehicle altitude within hydrothermal outflow zones (r < 0.02). Warm water appeared to be accumulating in the western half of the valley for both dives at Raven (Figure 7). The average temperature anomaly for all measurements west of Universal Transverse Mercator (UTM) X = 493,300 was +0.007°, while all measurements to the east averaged −0.006°. All temperature anomalies above the arbitrary threshold value of +0.08°C were located above the west-wall foot and within 40 m of an active 2011 fluid emission site. The regions of minimum temperature anomalies were located at the summits of the eastern and western walls, averaging −0.02°C. Negative temperature anomalies are the result of our arbitrary baseline selection and have no physical significance.
4.1. Relative Surface Permeability
 The spatial distribution of the simplified three categories of relative permeability mapped over the axial valley show distinct patterns (Figure 8). High surface permeability regions include limited areas on both the eastern and western walls, and the intersection of the valley floor with the west wall. These areas had little to no sediment cover and consisted largely of bare talus blocks and lightly sedimented talus accumulations. The central valley floor, northernmost terrace, and summits of the eastern and western valley walls had low relative surface permeability, with nearly complete sediment cover, and few outcrops or surface discontinuities. The southern portion of the terrace and segments of the western-wall foot and valley walls had moderate permeability and were composed largely of broken flows (Figure 6). The northeast section of the valley floor and northern terrace were classified equally as medium and low surface permeability, with a moderate amount of sediment coverage, more frequent outcrops, and visible cracks present within the seafloor. The axial valley floor is covered with continuous, unbroken sediment cover and may represent a restrictive boundary layer for fluid transmission, as described earlier.
4.2. Bottom Water Temperature Anomalies
 An accumulation of anomalously warm bottom water is located over the western half of the axial valley, with the highest temperature anomalies, unsurprisingly, in close proximity to sites of active venting. The eastern half of the valley, including the central valley floor and eastern wall, is the area with the lowest temperature anomalies and has no active vents. The noticeable agreement between the locations of highest temperature anomalies and observed active vents strongly supports the conclusion that warm fluid is only being discharged on the western side of the axial valley within the Raven area, specifically along the foot of the western wall. While the shoaling of the magma lens on the western side of the valley seems the most likely cause of the position of the warm water pool, it is worthwhile to note that the highest water temperature anomalies also appear above a region mapped as having high seafloor permeability.
4.3. Location of Vents
 In 2011, all active and almost all fossil fluid vent sites were found at the foot of the west wall. Several previous studies have observed that the partial melt zone of the axial magma chamber along the Endeavour Segment of the Juan de Fuca ridge shoals on the western edge of the valley, rising up to 400 m closer to the seafloor than on the eastern side [Van Ark et al., 2007; Wilcock et al., 2009; Kelley et al., 2012]. The shoaling of the roof of the magma lens beneath the western portion of the Raven field may largely explain why the majority of active and fossil structures are located on the western half of the axial valley, as circulation upflow should be enhanced where the vertical thermal gradient is steepest. The observation that all current or recently active vents are located within the zones of relatively high upper surface permeability, while most fossil sulfide structures appear in zones classified as low permeability, may suggest that sediments that accumulate across low relief sections of the axial valley could eventually restrict the relatively low intensity fluid outflow zones.
4.4. Spatial Distribution of Relative Permeability
 Within the axial valley at Raven, our estimation of relative upper surface permeability is largely controlled by the extent of sediment cover, which appears correlated with the slope and roughness of the underlying seafloor. Regions of low topographic slope and smooth terrain capture and retain sediment cover that is observably thicker and more spatially continuous, while steeper and rougher regions of the seafloor accumulate little sediment. It is important to note that while the vertical nonhydrothermal sediment flux is assumed to be roughly uniform across the valley, the accumulation rate appears heavily biased by topography. In regions with rough and uneven topography with relief greater than several meters, the underlying extrusive rocks with high porosity and presumed high relative permeability are directly exposed to the seawater. In these uneven regions of low sediment accumulation, the uppermost veneer of the seafloor appears to provide little resistance for fluid entering or leaving the subsurface fluid reservoir. In contrast, even within the youngest crust of the central axial valley, seafloor with smooth basement topography and relief of the order of 1 m rapidly accumulates a complete and unbroken layer of sediment. This sedimentation is presumably partially derived from mineral precipitation and intense biological activity associated with hydrothermal fluid emission and originates from active vent sites located throughout the Endeavour axial valley. This sediment not only accumulates on the axial valley floor but is also redistributed by tidally driven bottom currents over the valley walls, terraces, and the top of the summit ridges [Hautala et al., 2005]. This migration of sediments both downslope and horizontally throughout the axial valley may also contribute to redeposition on the smallest scale; accumulating within local topographic depressions, open fissures, and in gaps between individual pillow basalt extrusions, consequently gradually reducing the permeability of the surface veneer locally with increasing time.
 The influence of surface permeability can also be examined in terms of predicted fluid flow velocity using a simple formulation of Darcy's law:
where v is the predicted fluid velocity, η is the fluid viscosity, ϕ is the porosity, k is the permeability, and is the pore pressure gradient. Taking the porosity of the lavas and sediment to be of the same order, and keeping both the fluid viscosity and pressure gradient forces constant, fluid flow velocity through the sediment layer should be far slower than through the lavas, regardless of direction. The weakly buoyant diffuse hydrothermal fluid (rather than the focused high-temperature upflow) could flow horizontally to the nearby openings in the surface veneer more readily than penetrating even the thin overlying sediment cap.
 Tectonic activity caused by continued seafloor spreading and magma injection from below will control the distribution of faults, fissures, and morphology within the axial valley [Wilcock et al., 2002]. Each new cycle of tectonic or magmatic activity could “reset” the spatial distribution of Layer 2A and surface layer permeability within the axial valley, redistribute portions of the mobile sediment cover, and relocate both fluid emission and seawater recharge sites for the subsurface crustal reservoir.
 The eastern and western valley walls and the talus accumulations at the foot of the western wall are the primary regions of high relative upper surface permeability, while the lowest surface permeability was restricted to the central valley floor and northern terrace (Figure 6). Within 40 m of active vent sites, relative surface permeability is high and sediment cover is presently sparse. At a distance of approximately 100 m from active vent sites, we observed a predominance of low surface permeability zones and an abundance of relatively thick sediment cover. Cracks and larger fissures are far more common in the areas adjacent to active fluid vent sites, although there is evidence that many of these are partially or completely filled with sediment (see supporting information1, Figure E).
 Our observations suggest a revised hydrothermal circulation model where the permeability of the upper surface veneer of the seafloor of the axial valley is a time-dependent and evolving primary boundary condition for fluid flux into and out of the subsurface hydrothermal reservoir (Figure 9). The spatial definition of this proposed boundary condition is dependent on photographic observations, and therefore interpretations are restricted to only the uppermost surface veneer of the seafloor. Almost all of the geological features and processes primarily responsible for driving and controlling ridge axis hydrothermal fluid circulation, such as magma chamber location and depth, normal and listric faults that penetrate deep within the crust, and chemical alteration and precipitation zones, are clearly hidden from the ROV video camera. However, the warm hydrothermal fluid that is emitted, or the cold seawater that is recharging the subsurface crustal reservoir, must still pass through this thin surface interface, making the visible seafloor the location of critical access ports for mid-ocean ridge hydrothermal fluid circulation. Additional experiments could be conducted at Raven to test the observations made in this study. Direct in situ measurements of sediment permeability and thickness would provide the additional critical parameters for numerical models. Direct measurements of fluid flux at the seawater/sediment/basement interface would quantify fluid flow rates across the veneer. Finally, mapping the surface conductive heat flow will help characterize the structure of any fluid circulation cells within Layer 2A, providing the link between surface permeability and fluid recharge or emission, a study that is presently underway.
 Support for this project was provided by NSF grant 1230102 to H. P. Johnson. The crew of the R/V Atlantis and operating crew of the ROV Jason II were integral to the success of this study. Personal communications with Maurice Tivey, Javier Escartín, Susan Hautala, Miles Logsdon, Mike Hutnak, and Marie Salmi were instrumental in the completion of this research. Special thanks are given to Baxter Hutchinson of the ROV Jason II for providing technical specifications for the vehicle and camera system and Scott McCue at WHOI for his patience and advice.