Where Are the Undiscovered Hydrothermal Vents on Oceanic Spreading Ridges?

Supplementary Information

Where are the undiscovered hydrothermal vents on oceanic spreading ridges?

Stace E. Beaulieua*, Edward T. Bakerb, and Christopher R. Germana

a Woods Hole Oceanographic Institution, Woods Hole, MA 02543 USA (* )

b Joint Institution for the Study of the Atmosphere and Ocean—PMEL, University of Washington, Seattle WA 98115 USA

29 January 2015

22pages (including cover page)

Supp. Methods

Supp. Discussion

Supp. References

1 Supp. Data file (separate file)

1 Supp. Figure

2 Supp. Tables

Supplementary Methods

Calculating strike length surveyed for hydrothermal activity

Estimates of total strike length for mid-ocean ridges (MORs) and back-arc spreading centers (BASCs) in the literature vary depending on the model used for divergent plate boundaries and whether or not to include certain categories of spreading such as continental rifts. The total spreading ridge strike length used in this global analysis (71,284 km) is calculated by assigning every submarine oceanic spreading ridge (OSR) “digitization step” (66,908 km) and a subset of continental rift boundary (CRB) steps (4,376 km) in the PB2002 model (Bird, 2003) to MOR or BASC regions (Supp. Data; note: plate identifiers are provided in the Supp. Data caption below).A digitization step is defined as “the short great circle arc between adjacent digitized plate boundary points”(Bird, 2003). We retain the word “step” when referring to our use of the PB2002 model, to distinguished from ridge “segment” and surveyed “portions” of ridges. The subset of CRB steps is mainly three regions with known vents (Bransfield Strait, Okinawa Trough, and Red Sea) and also smaller lengths contained within a series of OSR steps [e.g., in Lau Basin, theMangatolu Triple Junction (MTJ) portion of TO-NI boundary and Lau Extensional Transform Zone (LETZ) portion of NI-AU boundary]. The MOR strike length (60,139 km, 84% of total spreading ridge) matches the previous review (Baker and German, 2004), while the BASC strike length (11,145 km, 16% of total spreading ridge) is ~4,000 km longer. The MOR strike length includes the ultraslow-spreading OSR steps on NA-SA (992 km) and IN-AU (1059 km) (Supp. Table 1), new plate boundaries as determined by seismicity (Bird, 2003).We note that the more recent NNR-MORVEL56 model (Argus et al., 2011) includes four more plates,but plate boundaries “are merely a minor update to those of Bird (2003).”

To determine the total strike length within and outside exclusive economic zones (EEZs) (Supp. Table 2), we applied a “point in polygon” algorithm (modified from to the start and end positions of each submarine OSR and CRB step used from the PB2002 model (Bird, 2003) against EEZ polygons in the Maritime Boundaries Geodatabase(VLIZ, 2009).Positions given as high seas in this database translate to The Area (the seafloor beyond the limits of national jurisdiction) and extended continental shelves at the seafloor. Presently, applications for extended continental shelves outside of EEZs are still being submitted by states to the United Nations Commission on the Limits of the Continental Shelf; thus, we can only report our estimates within and outside EEZs in terms of national jurisdiction. For the few instances when either the start or end position fell within an EEZ and the other was high seas, the full step was counted in the EEZ strike length (thus, in a few cases there could be a slight over-estimate when a step straddled the EEZ boundary). For the few instances in which the start and end fell within different EEZs (e.g., Red Sea and Lau Basin Futuna Spreading Center), each step was assigned to national jurisdiction based on having a majority of its length within a particular EEZ.

To determine the strike length of spreading ridges surveyed for hydrothermal activity, we identified the subset of OSR and CRB steps corresponding to every published systematic survey (Supp. Data), with the exception of Mohns Ridge for which we used 10% of its total strike length (Pedersen et al., 2010). The original intention with cataloging hydrothermal surveys was to identify those surveys long enough and dense enough to contribute to aglobal correlation of plume incidence (fraction of ridge crest over which a plume anomaly is observed in the water column)or vent field frequency (at the seafloor) to spreading rate (e.g., > 200 km and densely or moderately surveyed; (Baker and German, 2004)). Survey density is described fully by Baker and German (2004); briefly: denserefers tocontinuous (e.g., by a towed instrument package) or closely-spaced(e.g., vertical profiles ~10 km apart) mapping of hydrothermal plumes, moderaterepresentsdiscrete mapping of plumes and/or seafloor deposits with less dense spacing (e.g., vertical profiles ~10 km apart), and sparse indicates discrete mapping with spacing too far apart to achieve a representative sampling of a continuous plume distribution in the water column or to assess the distribution of seafloor deposits. The re-analysis of surveyed strike lengths as of year 2001 using the PB2002 model (Bird, 2003) matched well with the previous review with a total of 14,585 km (Supp. Table 1; compare to 13,300 in (Baker and German, 2004), plus ~1000 km more sparsely surveyed in (Baker et al., 1995)). This review adds 8731 km of surveys in the decade from ~2001-2010, bringing the total strike length at least sparsely surveyed to 23,316 km (Supp. Table 1). Most of this total surveyed length is moderately or densely surveyed (20,229 km). We thank those private companies and national jurisdictions who shared their data from systematic surveys for hydrothermal plumes and seafloor massive sulfide deposits.

An asterisk in Supp. Table 1indicates that additional systematic surveys have been performed in the region since ~2010. In general publications are not yet available for these very recent surveys, listed here by region, year, and cruise: Chile Rise, 2010, INSPIRE; Galápagos Spreading Center:2009, R/V DayangYihaoDY115-21, and 2011, GALREX; Gulf of California, 2012, Monterey Bay Aquarium Research Institute; Red Sea, 2012, PELAGIA 64PE350; Southeast Indian Ridge (SEIR), 2011-2013, Korea Polar Research Institute; Southwest Indian Ridge (SWIR), 2010 and 2011, R/V DayangYihao; Lau Basin Futuna Spreading Center, 2010 and 2012, IFREMER (MNF-ss2012_v02); Lau Basin NELSC and Eastern Lau Spreading Center (ELSC) and Havre Trough, 2010, Korea Ocean Research and Development Institute; Mid-Atlantic Ridge13-33 S, 2013, R/V Merian MSM25.

Methods for new linear fit of vent field frequency to spreading rate

An equation for the linear relationship of spatial density,or frequency (Fs) of vent fields per 100 km spreading ridge, to weighted average full spreading rate (us) was published in the previous global compilation of hydrothermal surveys on spreading ridges(Baker and German, 2004) and revised by (Baker et al., 2004). More recent publications only included equations for the linear relationship of hydrothermal plume incidence to spreading rate or magma budget (e.g., Baker et al., 2008b). Here, we report as Eq. 1 a revisedlinear fit for vent field frequency to spreading rate including the previous 12 (with SEIR extended) and 9new non-hotspot, moderately-to-densely surveyed portions:

Fs = 0.95 + 0.020 us(Eq. 1)

(R2 = 0.47; “21-pt linear regression” in Fig. 3A). Although there is scatter, expected due to variability in survey effort and discrimination of vent field locations (Baker et al., 2004), the correlation for the 21 data points is significant (r = 0.68, p = 0.0007)We also report here as Eq. 2 the linear fit for all 27 surveys in Table 1A includinghotspot-influenced portions of MORs and the Lau ELSC/Valu Fa Ridge (VFR):

Fs = 0.85 + 0.023 us(Eq. 2)

(R2 = 0.27; “27-pt linear regression” in Fig. 3A).The correlation for the 27 data points is significant(r = 0.5209, p = 0.005). Hotspot-affected regions were excluded in previous publications for the linear fit of vent field frequency to spreading rate (Baker and German, 2004;Baker et al., 2004;Dyment et al., 2007;Baker et al., 2008b). The Lau ELSC/VFR has by far the highest observed Fs of any spreading ridge and is grossly affected by arc magma (Baker et al., 2006; Martinez et al., 2006). All listings in Table 1A were surveyed with sufficient density and length (> 200 km) and with data available for vent field locations; additional listings in Table 1B that were moderately or densely surveyed for > 200 km may be included in future iterations of the linear fit with publication of and/or access to more data.

In previous publications for this linear relationship, to achieve a minimum ridge length of > 500 km for each data point and reduce scatter(Baker and German, 2004;Baker et al., 2004), data were binned into five spreading rate categories: ultraslow (0-20 mm/yr), slow (20-55 mm/yr), intermediate (55-80 mm/yr), fast (80-140 mm/yr), and superfast (>140 mm/yr). For comparison, we binned the 21non-hotspot portions into the same five spreading rate categories (Fig. 3B) and report the binned fit:

Fs = 0.85 + 0.021 us(Eq. 3)

(R2 = 0.88; greater coefficient of determination than Eq. 1). The correlation for the binned data points is significant (r = 0.94, p-value 0.018). Binned values use reported and/or measured lengths (i.e., summing the vent fields in the non-contiguous portions and plotting against weighted average spreading rate).All three new equations have similar intercepts but slightly greater slope than the most recently published equation(Fs = 0.98 + 0.015 us;Baker et al., 2004). We note that at fast spreading rates the new binned fit (Eq. 3) just exceeds the 95% confidence limit in Baker et al.(2004; their Fig. 21b).

For vent field locationswe used the InterRidge Global Database of Active Submarine Hydrothermal Vent Fields (InterRidge Vents Database), Version 2.1, with updates as described in (Beaulieu et al., 2013). We excluded off-axis vent fields, and we combinedtwo original listings into one on SWIR 58.5-66° E. In addition, for the most recent surveys (vent fields not yet in the database) we added8 vent fields to Lau NELSC/MTJ (E. Baker, unpub. data), 9 to the Central Indian Ridge 8-17° S (Son et al., 2014) and 31 to the SEIR(Baker et al., 2014).We included confirmed and inferred active vent fields from the database, with “confirmed” meaning ground-truthed with observations at the seafloor and “inferred” usually meaning that a hydrothermal plume was detected in the water column and the location for venting at the seafloor was estimated. We note that confirmed active vent fields categorized as “low-temperature” or with measured fluid temperatures < 100° C contribute to the Fs value in several of the data points in the scatterplot (Fig. 3A). Vent fields that are inferred active from hydrothermal plume surveys using MAPRs and/or optical backscatter are likely to be high-temperature (i.e., black-smoker). Modern oxidation-reduction potential (ORP)-equipped MAPRs could detect low-temperature vent fields(Walker et al., 2007), especially if deployed close to the seafloor (e.g., in association with imaging survey).

In Table 1A we provide the vent field frequency for each survey length, using both the modeled step lengths (Bird, 2003) and reported survey lengths and/or measured lengths from bathymetry in the Global Multi-Resolution Topography (GMRT) synthesis (Ryan et al., 2009) accessed via GeoMapApp software ( In general the modeled length was slightly higher than the (contained) measured length, and thus, resulted in slightly less Fs.

The surveyed regions used in the “21-pt” and “binned” linear fits account for12313 km (modeled; 11763km measured), or 17% of global spreading ridge strike length. This subset is quite representative of global spreading ridges with 90% MOR and 10% BASC (compared to 84% MOR and 16% BASC globally) and 53 mm/yr weighted average spreading rate (compared to 46 mm/yrin this global analysis).To obtain weighted average spreading rate for each portion of surveyed ridge, we multiplied length and spreading rate for each step in the PB2002 model (Bird, 2003), summed those results and then divided by the total modeled length of the surveyed portion. The 21 data points in Fig. 3A were binned as follows: ultraslow (n=5, combined 2534 kmmodeled, 21% of total lengthused in linear fit), slow (20-55 mm/yr, although all data were within 20-40 mm/yr, n=7, combined 4120 kmmodeled, 33%), intermediate (55-80 mm/yr, n=4, combined 3253km modeled, 26%), fast (80-140 mm/yr, although all data were within 80-110 mm/yr, n=3, combined 1279 km, 10%), and superfast (>140 mm/yr, n=2, combined 1127 km, 9%). Again, we note that this is a representative subset of the global ridge crest, as the percentage contributions to the total strike length match well with the global model (% of strike length in this global analysis:ultraslow26%,slow 38%, intermediate 22%, fast 11%, superfast 3%). We note that as a consequence of using the PB2002 model for spreading rates, N EPR15.5-18.5° N is grouped into the intermediate spreading rate category [previously assigned to fast spreading in Baker and German (2004)].

Using the global weighted average full spreading rate (46 mm/yr), the binned fit Eq. 3 yields a global average Fs of 1.8(with 95% confidence limits 1.0-2.6), and multiplied by total strike length predicts atotalof 1305 active vent fields on spreading ridges (with 95% confidence limits 713-1853). Interestingly, this matches well with the range (~800-1600) proposed by German et al. (2011), inclusive of high and low temperature vent fields. Here we explain why our predicted total numberis consistent with (although exceeds) the prediction from the previous review (1060, with 95% confidence limits 992-1153; (Baker and German, 2004)). First, although we are using a value similar to the previous review for global average spreading rate, we are including approximately 6% more strike length, which would increase the predicted number of vent fields to 1120. Second, the equation in the previous review (Fs = 0.88 + 0.015us) was revised (Baker et al., 2004)(Fs= 0.98 + 0.015 us), and plugging global us into the revised equation yields a global average Fs of 1.7 (as reported in (Baker et al., 2008a)), and multiplied with the strike length in this global analysis predicts 1191 vent fields. Third, for the 12 surveyed regions binned in (Baker and German, 2004) and (Baker et al., 2004), our currentvalues for Fs are slightly higher because we now know of more vent fields in these regions as compared to a decade ago due to additional discoveries (e.g., SEIR). We binned and re-calculated the linear regression for just these 12 regions(Fs = 1.28 + 0.017 us),and the re-calculated prediction of total population of vent fields was1476 (higher than our current prediction using Eq. 3). We also note that the total and 95% confidence interval for our predicted population of vent fields on spreading ridges are consistent with the total (~900) and range (~500 - 5000) of high-temperature hydrothermal deposits estimated using a very different method independent of spreading rate (Hannington et al., 2011).

Supplementary Discussion

Additional regions intriguing for exploration

Other ultraslow- and slow-spreading MOR regions that are intriguing for exploration include the Red Sea, Ayu Trough, Sorol Trough, and American-Antarctic Ridge. Although the Red Sea was the first region to be known for deep seafloor hydrothermal activity, perhaps dating back to the 1880's (Miller et al., 1966), the first systematic hydrothermal survey was just completed in 2012 (cruise PELAGIA 64PE350). Of the 6 small plates in the PB2002 model (Bird, 2003) with no known vents on any boundaries (only 3 of which have submarine OSR steps), the Caroline (CL) plate is an interesting target for exploration, including the ultraslow (and perhaps extinct?) spreading ridges of the Ayu Trough and Sorol Trough, the latter of which is tantalizing as its location (in Micronesia EEZ) is about halfway between known vents in the Mariana Trough and Manus Basin (separated by almost 2000 km) and could be an important stepping stone in present-day biogeography of vent fauna. The ultraslow American-Antarctic Ridge (AN-SA plate boundary) has yet to be explored for vents which may confirm the new biogeographic province suggested for Antarctic vent fauna (Rogers et al., 2012). Some of the ultraslow and slow spreading OSR steps on otherwise transform boundaries are also intriguing to consider for future exploration in the context of biogeography, for example at high latitudes on the South Scotia Ridge (SC-AN, Antarctic 200NM zone) and Macquarie Ridge Complex (PA-AU, Australia - Macquarie Island EEZ).

In terms ofother BASC regions that are intriguing for exploration, at ultraslow spreading rateswe highlight the boundaries of Amur (AM) and Anatolia (AT), the two remaining small plates with submarine OSR steps but no known vents. These plate boundaries, AM-ON in Japan’s EEZ and AT-AS in the EEZs of Turkey and Greece, respectively, have relatively short lengths of spreading ridge with close access to ports. There is also a “jog” with ultraslow spreading in the otherwise convergent boundary at the North Scotia Ridge (SC-SA), of interest to explore for vent-endemic fauna similar to the new biogeographic province proposed for the East Scotia Ridge (Rogers et al., 2012). At faster spreading rates, the TO-NI, PA-NI, and NI-FT plate boundaries are targets in the Lau Basin, and in Manus Basin targets include NB-SB and MN-SB (the Southern Rifts), known for active volcanism albeit with observations suggesting a low magmatic budget (Sinton et al., 2003). In the North Fiji Basin plate boundaries with spreading rates 100-140 mm/yr are CR-NH and a short strike length at the easternmost (NH-AU) boundary, an incipient rift at the southern end of the Hunter Ridge (although MAPRs on dredges to that location did not indicate activity in 2009; L. Danyushevsky, pers. comm.). Atintermediate-spreading rates, a short length of the Woodlark-Birds Head (WL-BH) boundary, in a complicated region of extensional tectonism (see recent review (Baldwin et al., 2012)), also has no known vents (Indonesia EEZ).

Other EEZs lacking systematic hydrothermal surveys that also have no known vents include (in descending order of total strike length), Norway - Bouvet Island (mainly SWIR), Palau (Ayu Trough), South Africa - Prince Edward Islands (mainly SWIR), and Brazil (N MAR) (Supp. Table 2).

Supplementary References

Argus, D.F., Gordon, R.G., DeMets, C., 2011. Geologically current motion of 56 plates relative to the no-net-rotation reference frame. Geochemistry, Geophysics, Geosystems 12 (11), Q11001.

Baker, E.T., Edmonds, H.N., Michael, P.J., Bach, W., Dick, H.J.B., Snow, J.E., Walker, S.L., Banerjee, N.R., Langmuir, C.H., 2004. Hydrothermal venting in magma deserts: The ultraslow-spreading Gakkel and Southwest Indian Ridges. Geochemistry, Geophysics, Geosystems 5 (8), Q08002, doi:10.1029/2004GC000712.

Baker, E.T., Embley, R.W., Walker, S.L., Resing, J.A., Lupton, J.E., Nakamura, K.-i., de Ronde, C.E.J., Massoth, G.J., 2008a. Hydrothermal activity and volcano distribution along the Mariana arc. J. Geophys. Res. 113 (B8), B08S09, doi:10.1029/2007JB005423.

Baker, E.T., Haymon, R.M., Resing, J.A., White, S.M., Walker, S.L., Macdonald, K.C., Nakamura, K.-i., 2008b. High-resolution surveys along the hot spot–affected Galápagos Spreading Center: 1. Distribution of hydrothermal activity. Geochemistry, Geophysics, Geosystems 9 (9), Q09003, doi:10.1029/2008GC002028.

Baker, E.T., German, C.R., 2004. On the global distribution of hydrothermal vent fields. In: German, C.R., Lin, J., Parson, L.M. (Eds.), Mid-Ocean Ridges: Hydrothermal Interactions Between the Lithosphere and Oceans. American Geophysical Union, Washington, D.C., pp. 245-266, doi:10.1029/148GM10.