Geophysical Research Letters

Ocean circulation promotes methane release from gas hydrate outcrops at the NEPTUNE Canada Barkley Canyon node


Corresponding author: L. Thomsen, School of Engineering and Science, Jacobs University Bremen, Campusring 1, D-28759, Bremen, Germany. (


[1] The NEPTUNE Canada cabled observatory network enables non-destructive, controlled experiments and time-series observations with mobile robots on gas hydrates and benthic community structure on a small plateau of about 1 km2 at a water depth of 870 m in Barkley Canyon, about 100 km offshore Vancouver Island, British Columbia. A mobile Internet operated vehicle was used as an instrument platform to monitor and study up to 2000 m2of sediment surface in real-time. In 2010 the first mission of the robot was to investigate the importance of oscillatory deep ocean currents on methane release at continental margins. Previously, other experimental studies have indicated that methane release from gas hydrate outcrops is diffusion-controlled and should be much higher than seepage from buried hydrate in semipermeable sediments. Our results show that periods of enhanced bottom currents associated with diurnal shelf waves, internal semidiurnal tides, and also wind-generated near-inertial motions can modulate methane seepage. Flow dependent destruction of gas hydrates within the hydrate stability field is possible from enhanced bottom currents when hydrates are not covered by either seafloor biota or sediments. The calculated seepage varied between 40–400 μmol CH4 m−2 s−1. This is 1–3 orders of magnitude higher than dissolution rates of buried hydrates through permeable sediments and well within the experimentally derived range for exposed gas hydrates under different hydrodynamic boundary conditions. We conclude that submarine canyons which display high hydrodynamic activity can become key areas of enhanced seepage as a result of emerging weather patterns due to climate change.

1. Introduction

[2] Natural gas hydrates represent a large global reservoir of natural gas [Kleinberg and Brewer, 2001], mainly methane, and can play an important role in climate change [Xu et al., 2001] and seafloor stability [Suess, 2010]. They originate from methane-supersaturated pore fluids under conditions of elevated pressure (>60 bar) and low temperature (<4°C) found in the hydrate stability zones (HSZ) on continental slopes [Wallmann, 2001]. However the quantification of methane fluxes and their role in global elemental budgets and climate change remains a major challenge in oceanic research.

[3] The NEPTUNE Canada cabled ocean observatory network [Barnes et al., 2011] enables non-destructive, controlled experiments and time-series observations with mobile robots on gas hydrates and benthic community structure on a small plateau of about 1 km2 at a water depth of 870 m in Barkley Canyon, about 100 km offshore Vancouver Island, British Columbia. There the sea floor hydrates are of thermogenic origin and more complex [Lu et al., 2007; Chapman et al., 2004], being generated within the underlying Tofino Basin [Johns et al., 2012]. The gas hydrates are patchily exposed on thinly sedimented (silty mud) mounds of 1–3 m height that contain gas, quantities of light oil [Chapman et al., 2004] and show evidence of episodic gas emissions. The mounds support spatially distributed chemosynthetic communities [Boetius and Suess, 2004] (vesicomyid clams) and microbial mats that cover large portions of the hydrates and sediments (Figure 1). These gas hydrate mounds supply specific bacterial-consortia that generate hydrogen sulfide, which rises and oxidizes in the seafloor biota while calcium carbonate phases precipitate [Suess, 2010]. The intense spatial variability of the hydrate-system favors studies with mobile robots, capable of investigating the ≈20 m2 sized mounds with their patchily distributed hydrate outcrops (0.01–0.2 m2 in size).

Figure 1.

Photograph and map of the gas hydrate mound at 870 m water depth in Barkley Canyon about 100 km off the west coast of Vancouver Island, British Columbia. The figure shows the Neptune Canada network, and the small study site with the ≈20 m2sized gas hydrate mound with the patchily distributed clams, microbial mats, hydrate outcrops, number-markers for navigation and the short transect (red line) with four survey stations (A–D). Methane seepage at those four stations was studied with the Internet Operated Vehicle (IOV) during February 2010.

[4] The aim of the first robot mission at the site was to investigate the importance of oscillatory deep ocean currents on methane release at continental margins. Our hypothesis was that submarine canyons which display high hydrodynamic activity are key areas of enhanced erosion of gas hydrate outcrops.

2. Methods

[5] Our observations result from an experimental strategy combining spatial and temporal multi-sensor monitoring and imaging of benthic boundary layer (BBL) processes within the NEPTUNE Canada observatory network. This network of five subsea observatories is linked by 800 kilometers of powered electro-optic cable across the northern Juan de Fuca tectonic plate. The data flow of ≈50 terabytes per year is managed through a data management and archive system (DMAS). DMAS adopts Web services and service-oriented architecture and can subscribe to data streams of selected sensors and interact directly with the instruments [Barnes et al., 2011]. We concentrated our research on the temporal and spatial variability of methane fluxes and benthic community structure around gas hydrate outcrops, which clustered within a few hundred square meters. Our sensors were attached to an Internet operated vehicle (IOV) with caterpillar propulsion. The IOV is a mobile instrument platform (130 × 106 × 89 cm, LWH) which can monitor and study up to 2000 m2of sediment surface in real-time. Its caterpillar drive creates a footprint on the seafloor of 0.35 m2 with a weight of ≈10 g/cm2. It is connected to a junction box with a 70 m long floating cable and can be controlled from any computer logged into the Internet. The sensors suite for this study included CTD at 20 cm height above bottom (h.a.b) (ADM electronics, Germany), fluorescence at 30 cm h.a.b. (Seapoint, USA), optical backscatter at 50 cm h.a.b. (Seapoint, USA), flow velocity and direction at 100 cm h.a.b. (HS Engineers, Germany) and methane at 20 cm h.a.b. (Franatech, Germany). All sensors besides those for flow and direction were located at the front end of the IOV. Visual navigation was achieved through a pan/tilt webcam (Panasonic, Japan) with zoom function. Data sampling rate was ≥1 Hz.

[6] Time series analyses in both the time and frequency domains were carried out to evaluate the impact of previously documented types of oscillatory flow in the northeast Pacific [cf. Thomson et al., 1998, 2003] on the methane signal. In the frequency domain, standard (Fourier-based) spectral analysis was used to obtain estimates of the power spectral density for overlapping segments of the time series. These were then block-averaged to give statistically significant results for the relatively long summer and winter portions of the overall time series (many wave cycles are captured in this approach). To examine shorter segments of the time series in more detail (study site B–D, see below), we used a time domain approach based on local correlation functions 〈U(t)·M(t-Tlag)〉 and 〈V(t)·M(t-Tlag)〉 which link the methane signal M to the flow components U (directed along the axis of the canyon, positive in toward 65.2 degrees True) and V (cross-shore component, upslope to the NW toward 294.8 degrees True). Here, 〈…〉 denotes the time (t) average over a 12- hour window centered at time Tcen. Local lag-correlation functions indicate how strongly changes in the methane signal are coupled to changes in the current velocity as a function of time, Tcen; the phase offsets between the patterns for a given time are indicative of the period of motion dominating the correlation function.

[7] During the first 24 months of investigation, the IOV monitored in real-time an area of approximately 200 m2 around one gas hydrate mound.

3. Results

[8] From January to September 2010, flow velocities (u100) in the benthic boundary layer exceeded 0.1 ms−1 over 22% of the time, ranging between 0.01–0.22 ms−1. Temperature varied between 2.5–3.5°C, pressure between 870–875 dbar and methane between 0.8–6.9 μM. Although the observed temperature and pressure excursions were not of significant magnitude to destabilize the hydrate deposits, visual observations with the IOV around the mound indicated local destruction within periods of 4–8 months. Internet access to the IOV via NEPTUNE Canada allowed testing of a flow dependent seepage.

[9] Therefore a short transect with the IOV onto the plateau of the gas hydrate mound was carried out (A–D in Figure 1). This transect along spatial variations of methane seepage started at position A, a seafloor depression covered by muddy silt, with bioturbation marks and sporadic microbial mats where the methane concentration was low (Figure 2, top). The IOV was then moved linearly one meter to position B that was well inside the belt of vesicomyid clams covering the rim of the mound (Figure 1). There the methane concentrations were already elevated (Figure 2, top). After 15 hours of monitoring at this station, the IOV was transferred 3 m to location C at one flank of the mound where authigenic carbonates and microbial mats were abundant and a gas vent of ≈5 cm diameter was visible at a distance of 20 cm from the methane sensor. On February 11 a methane peak of >3.5 μM was recorded, probably due to venting. Then methane decreased to <2 μM between February 13 and 16. On February 16, the IOV was positioned for ∼4 days at location D on the plateau of the mound, in close proximity to several larger gas hydrate outcrops which were randomly spread over an area of ≈4 m2and where a clearer flow-dependent methane signal was expected.

Figure 2.

Temporal and spatial variability of dissolved methane at 20 cm height above the sediments and couplings with flow components. (top) Dissolved methane at the study site from February 6–22, 2010 at the four survey stations A–D. (top middle and bottom middle) Contour plots of local correlation functions 〈V(t)•M(t-Tlag)〉 and 〈U(t)•M(t-Tlag)〉 between the methane signal M and the flow components V (cross-shore component, upslope to the NW toward 294.8 degrees True) as well as U (directed along the axis of the canyon, positive in toward 65.2 degrees True). Here 〈…〉 denotes the time (t) average using a 12 hour time window at the center time given on the horizontal axes; the correlation time lag, Tlag, on the right-hand vertical axes ranges from −24 hours to +24 hours. Also displayed are also the velocity time series U and V (black lines, see annotations on left axes). Local correlation functions are a means to reveal time-dependent couplings between two variables that show a relative phase offset (Tlag). The vertical distance between peak values (e.g., correlations >0.9 between M and V on February 16–17) indicates recurrence tendencies (in one variable or in both) with quasi-periods of about 15–18 hours. (bottom) Flow speed.

[10] The local correlation function at this position revealed a positive correlation between methane seepage and the components of flow for time-lags close to zero and also at lag values around 15.3 hours (Figure 2, top middle and 2, bottom middle). The correlation function of flow component V with methane confirmed a strong link between the burst of near-inertial energy (∼15.3 hour period) beginning around February 16 and ending around February 19.5. This burst of energy was superimposed on strong downslope mean currents (toward the southeast at 155.2° True) which spanned the time of the oscillatory motions. The U-component of the motions (positive toward the northeast at 65.2° True) was much weaker than the V-component during this 3.5 day time span indicating that the motions were weakly polarized and may have been strongly rectified through reflection of the downward propagating inertial waves following interaction with the seafloor. The V signal ended around February 19.5 but was replaced by strong, roughly 16 hour near-inertial motions in the U-component with strong co-variability with the methane signal. These near-inertial motions and strong co-variability with the methane signal were also observed during February 14–16 and 20–22 when the IOV was located on the mound but away from the plateau of the mound. In addition, our investigation showed pronounced (up to 99%) coherence levels between current velocity and methane within the diurnal, near-inertial and semidiurnal bands from January to September at distances of 2–10 m from the mound.

[11] Spectral analyses of the time series measured 2–10 m around the mound over five weeks in winter and summer gave further indications for the impact of wind-forced inertial- and diurnal currents (Figure 3). Bottom currents during winter and summer were dominated by semi-diurnal currents (M2 ∼ 12 hr), diurnal currents (K1 ∼ 24 hr), and wind-forced near-inertial motions (13–16 hrs). The relative strength of the near-inertial currents was greater in the winter series. Along-channel currents dominated during the summer series. Strong high frequency motions were accompanied by pronounced downslope currents which then turned to out-canyon currents at the time of the inertial motions.

Figure 3.

Component spectra for along- and across channel flow in Barkley Canyon. Spectral peaks at the inertial frequencyfand the principal tidal frequencies are labeled. Currents were dominated by semi-diurnal (M2), diurnal (K1), and wind-forced near-inertial (13–16 hrs) motions. The relative strength of the near-inertial currents was greater in the winter series.

4. Discussion

[12] The IOV and NEPTUNE Canada infrastructure performed well during the period of investigation and allowed to precisely maneuver the IOV on the gas hydrate mound. A release of oily gas bubbles was only observed during a test drive when the caterpillars of the IOV were moved through repeated rotations of more than 90° on one spot. Such rotations were carefully avoided during the study period, with each change in IOV course direction requiring rotation of only a few degrees. We did not use an inverted barrel or benthic chamber inserted into the sediment [e.g., Sayles and Dickinson, 1991; Linke et al., 1994] to estimate methane seepage or even entrainment of bottom water into the sediments [Tryon et al., 2002]. Like other authors we assumed that gas hydrate outcrops will dissolve over time in a diffusion-controlled process and that proper hydrodynamical analyses are needed to determine mass transfer (dissolution) rates [Egorov et al., 1999; Rehder et al., 2004; Bigalke et al., 2009; Hester et al., 2009; Rehder et al., 2009; Lapham et al., 2010]. Our aim was to use their results from experiments with extracted gas hydrates in-situ or from pressure laboratory studies to estimate the seepage from gas hydrate outcrops “non-invasively” with our IOV (without extracting a gas hydrate from sediments or using a benthic chamber).

[13] We separated the data series into summer and winter series since previous multi-year observations of the flow structure in this region shows marked differences in the flow intensity and frequency between the two seasons and for example the 1 degC temperature range between January and September 2010 is a typical feature for canyons cutting across the continental margin in the presence of marked cross-slope temperature gradients [e.g.,Thomson and Krassovski, 2010; de Stigter et al., 2007].

[14] The IOV allowed us to not only focus on one location at the mound but also to carry out visual observations around the mound. Real-time video observations taken around the mound since 2010 have confirmed that the mound locally has decreased and increased in size, while small areas have collapsed and microbial mats disappeared or reappeared. These results support the idea of a varying supply of supporting flux from below [MacDonald et al., 1994; Egorov et al., 1999; Chapman et al., 2004; Lapham et al., 2010] and/or from the tidally driven oscillation of fluids through permeable sediments overlying the hydrate deposits [Tryon et al., 2002]. Open questions remain as to the dynamics of these seafloor hydrate outcroppings.

[15] The positive correlation of methane seepage and the components of flow on the top of the gas hydrate mound close to the hydrate outcrops revealed that the duration and intensity of oscillatory currents had a direct and dynamic impact on methane seepage. Specifically, bursts of strong bottom currents associated with wind-generated inertial motions, diurnal shelf waves, or internal semidiurnal tides appear capable of “scouring” the gas hydrate deposits. The inertial motions of interest have typical spatial scales of a hundred kilometers and durations of days to a week [Jordi and Wang, 2008]. It is likely that the recorded changes in the inertial signal were due to changes in the regional inertial wave field. Our data thus suggest that the oscillating methane signal observed on the gas hydrate mound was also the result of wind-generated inertial motions that were coincident with downslope bottom currents.

[16] Previously, dissolution rates for naturally formed hydrates at Barkley Canyon under flow conditions were only investigated with ship based ROVs using extracted gas hydrates which were exposed to the ambient flow via manipulator arms to obtain diffusional mass transfer coefficients; an average dissolution rate of 390 μmol CH4 m−2 s−1 was reported [Hester et al., 2009]. Bigalke et al. [2009]carried out experiments on gas hydrate dissolution under flow in a pressure laboratory. They demonstrated that methane dissolution rates correlate linearly with the friction velocity, u* and derived an equation to predict the flux of methane from gas hydrate outcrops into the bottom water. We used their equation for methane seepage under flow, our data on flow-velocity from the IOV and a methane saturation concentration of 54 mol m−3 previously determined for yellow hydrate at Barkley Canyon [Hester et al., 2009; Zhang et al., 2011] to estimate the flow dependent seepage via gas hydrate outcrops from the study site in Barkley Canyon. The calculated dissolution rate varied between 40–400 μmol CH4 m−2 s−1. This is 1–3 orders of magnitude higher than dissolution rates of buried hydrates through permeable sediments [Lapham et al., 2010] and well within the experimentally derived range for exposed gas hydrates under different hydrodynamic boundary conditions [Egorov et al., 1999; Rehder et al., 2004; Bigalke et al., 2009; Hester et al., 2009; Rehder et al., 2009; Lapham et al., 2010].

[17] Our results with real-time data and imagery of outcropping gas hydrates show a flow-dependent methane seepage and reveal that both during winter and summer seepage of the climate-gas methane can also be connected to atmospheric processes which generate inertial currents. The flow dependent sediment erosion in the highly dynamic canyon can probably also lead to a subsequent release of trapped oil coated gas bubbles of lower solubility [Lapham et al., 2010]. Whether these can ascend through the water column and reach the atmosphere remains under heavy discussion [Hester and Brewer, 2009; Hu et al., 2012]. It is therefore important to further study the effects of wind-induced inertial currents on methane release in canyons, particularly at these latitudes where the more intense and poleward cyclones of the northeast Pacific are present. These weather patterns can directly affect submarine canyons which often display high hydrodynamic activity [Kunze et al., 2002; Jordi and Wang, 2008]. Whether those storms are a result of natural decadal oscillations or human-induced climate change is under discussion [Schneider and Cornuelle, 2005; Salathé, 2006]. Although the current hypothesis is that gas hydrate mounds are generally widely distributed, they likely represent small resource volumes. However, there is still little information from gas hydrate mounds in submarine canyons and while almost 6000 large submarine canyons have been reported so far [Harris and Whiteway, 2011] they are still severely understudied and detailed ROV operations are needed to map possible gas hydrate outcrops.


[18] This work was supported by NEPTUNE Canada. The development of the IOV was supported by the European Commission (NoE ESONET, contract 036851), Titanium Solutions Bremen and Statoil. We thank Autun Purser (JUB) and Maxim Krassovski (IOS, Sidney) for helping process the current meter data and generate the associated figures.

[19] The Editor thanks two anonymous reviewers for assisting in the evaluation of this paper.