By continuing to browse this site you agree to us using cookies as described in About Cookies
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of February 24 & 25 and will be live on February 26, 2018. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/
 During RV Polarstern cruise ARK-XXIV/3 (2009), a geophysical study along the ultraslow spreading Knipovich Ridge was conducted. The survey, located in the rift valley south of the Logachev Seamount (∼76°36′N), provides a crustal thickness of ∼4.5 km in the amagmatic parts and 5.7 km underneath the seamount itself. The velocity-depth function indicates the presence of a thick oceanic layer 2 (4.5 km), but no indication for a thick oceanic layer 3. The only exception is the area underneath the Logachev Seamount, where velocities higher than 6 km/s are detected. This indicates a stronger and focussed melt supply underneath the seamount. Local seismicity was analysed for two days. In total, 191 quakes were identified in the vicinity of the rift valley with magnitudes up to ML = 2.6. At least 48 of them are located in the upper mantle (up to 18 km below sea level), supporting models predicting a cold mantle or conductive cooling underneath ultraslow ridges to explain the reduced melt supply.
 In this study, we concentrate on new deep seismic and earthquake data along the Knipovich Ridge, which is located between Svalbard and Greenland (Figure 1; 73°45′N–78°35′N). Its rift valley is, in general, between 2500 and 3500 m deep with 1000 to 2000 m high rift flanks. The Knipovich Ridge does not have any large-offset transform faults along its entire length. The seafloor spreading direction is highly oblique (φ = 35 − 50°, where φ is the angle between the spreading direction and the topographic expression of the ridge). The structure of the axial valley is segmented by 14 axial topographic highs (Figure 1) with 7 of them being stronger magmatic centres [Okino et al., 2002]. To be compatible with the original interpretation, we use the numbering of the segment centers in this study, but indicate with an abbreviation whether they were interpreted as strong (SMC) or weak (WMC) magmatic centers. The SMCs have an elevation of 500–1000 m above the seafloor and a spacing of 60–110 km [Okino et al., 2002]. Because of the lack of high quality magnetic data as well as a weak anomaly pattern [Ehlers and Jokat, 2009], the spreading history of the adjacent basins is controversial. One class of spreading models for the Knipovich Ridge predicts highly asymmetric spreading rates of 7 mm yr−1 towards the west and 1 mm yr−1 to the east [Crane et al., 2001; Kandilarov et al., 2008], while other models assume constant full spreading rates of 15–17 mm yr−1 [Okino et al., 2002; Dick et al., 2003; Hellevang and Pedersen, 2005].
 Published deep seismic profiles cross the Knipovich Ridge rift valley only perpendicular or oblique. Thus, only few instruments were able to record crustal thickness information in the rift valley unbiased from the rough topography. In contrast to these experiments, the new deep seismic line [Jokat, 2010] is located along the ridge axis between two SMCs. In the north, the Logachev Seamount (∼76°36′N, SMC #7) has an elevation of more than 1000 m above the rift valley, while in the south SMC #9 (∼77°54′N) rises only 500 m above sea floor (Figure 1). In between, the WMC #8 at ∼76°15′N (Figure 1) is situated in the rift valley. This WMC has only a small elevation of ∼200 m above rift valley floor. In addition to seismic and bathymetric experiments, a seismicity study was conducted along this line to better constrain the tectonic processes along this segment.
2. Experimental Setup and Methods
 In total 28 broadband ocean-bottom seismometers, provided by the “Deutsche Geräte-Pool für amphibische Seismologie (DEPAS)”, were used simultaneously for gathering seismicity and wide-angle seismic data (auxiliary material). The OBS were equipped with a Güralp CMG-40T broadband seismometer and a long-period hydrophone. The seismometer did not work well at OBS227, therefore only its hydrophone component could be used. The sampling rate was 100 Hz. Along line 20090250 (length 125 km), in total six OBS with a spacing of 15 km were deployed. Seismic refraction data were acquired using an airgun cluster of six 8-litre G-guns towed at 8–10 meter depth. Shooting started at 76°45.5′N/007°22.3′E at OBS 223 and terminated on 75°53.8′N/007°08.4′E south of OBS 228. Shots were fired every 60 s at a speed of 5 kn, providing a shot spacing of 150 m (in total 796 shots). Zplot [Zelt, 1994] was used for phase picking, and Rayinvr [Zelt and Smith, 1992] for raytracing the travel time branches. To estimate the velocity and Moho depth errors, crustal velocities were varied to see when they no longer fit the data; this procedure suggested typical uncertainties of ±0.3 km/s and ±0.6 km, respectively. A seismic reflection line across the rift valley at OBS 216 (Figure 3) indicates variable sediment thickness. The sediments are thickest at the flanks of the valley [Berger and Jokat, 2009]. In the central part the seismic reflection data indicate some deeper reflections. However, it is not clear if these are offline signals or top of the basement reflections. Since the variability of the sediments along the line is unknown, we have not incorporated any sediment layer, which might be only a few hundred meter thick and, thus, is not resolvable with our OBS survey.
 Earthquake detection for the seismicity study uses a short time average-long time average (STA/LTA) trigger, with a STA window length of 0.5 s and a LTA window length of 60 s. Two dates (Sep 3rd and 4th, 2009) are selected to study seismicity along the profile 20090250. We identify 77 and 114 events along the rift axis per day, respectively. Locating the hypocenters was done within SEISAN software [Havskov and Ottemoeller, 2000], using a modified version of the HYPOCENTER module [Lienert et al., 1986]. For the local earthquake location, a simplified 1-D velocity function for the rift valley, derived from the velocity model of profile 20090250, and all OBS stations deployed west of the rift valley were used (auxiliary material). The velocity function starts at a reference depth of 3.9 km, which sets the minimum earthquake depth allowed in the inversion. Starting depth of location procedure was chosen following a systematic RMS-versus-depth search with steps of one kilometer. Local magnitudes were calculated according toGeissler et al. . Most events (∼85%) have a local magnitude ML between 0.8 and 2.0. The general detection threshold of the array lies at a local magnitude ML of 1.3. However, some scattered events of lower magnitude, even down to ML = 0.0, could be located.
3. Crustal Model
 High amplitudes of the direct water wave overprint the near offset phases in all OBS except OBS 227. For the short near offset travel time branches, velocities ranging from 2.7 to 3.2 km/s were calculated. The first layer with this velocity range and a thickness of 0.7 km was then applied for the entire profile. Seismic reflection data [Berger and Jokat, 2009] (profile AWI20020700) show that in the central part of the rift valley sediments up to 200 m thick occur, thickening at the flanks to 400 m. There might, however, be a mixture of sediment and volcanic rocks present. Thus, we did not include any sediment unit in our model. The second layer displays velocities ranging from 3.2 to 6.5 km/s and has a thickness of up to 4 km. A velocity of 7.6 km/s was calculated from upper mantle phases, and a velocity gradient of 0.05 km/s/km was used for the upper mantle [Muller et al., 1999] (Southwest Indian Ridge). Most parts of the crustal model are well constrained by the crustal Pg2 refractions (Figure 2 and auxiliary material). However, ray coverage is not as good in the uppermost part of the crust, because the strong water wave amplitudes mask the shallow travel-time branches. Moho refractions Pn (Figure 2 and auxiliary material) are observed along the entire model except for the area between km 80 and 90, where the Moho depth considerably shallows (Figure 2). However, in this area reasonable good Moho reflections are observed, which provide sound constraints for the depth of the crust-mantle boundary.
 The average crustal thickness below the Knipovich Ridge rift valley is 4.5 km. At SMC #7 (Figures 1 and 2; Logachev Seamount), the crust thickens to 5.7 km. Between SMC #7 and #9 (amagmatic segment), the crust has a thickness of 4 km only. Just north of SMC #9, between OBS 216 and OBS 228 (Figure 2), the crust thins to 3.4 km, while below the SMC #9 it thickens again to 5.5 km. The uppermost 600 m (velocities from 2.7 km/s to 3.2 km/s) are interpreted as oceanic layer 2a, consisting of basaltic pillow lava and lava debris. Below, oceanic layer 2b is characterized by seismic velocities ranging from 3.8 to 6.6 km/s on average. The high velocity gradient of 0.7 km/s/km of layer 2b indicates that the velocity is mainly controlled by porosity. Seismic velocities above 6.6 km/s, typical for oceanic layer 3, are only sparsely observed between the SMC's #7 and #9. The only exception represents the area beneath the Logachev Seamount. We interpret the absence of oceanic layer 3 as indicative for a reduced melt supply. This observation coincides with a seismic profile of Ritzmann et al.  near SMC #7. In contrast, Ljones et al.  modelled a ∼2.5 km thick oceanic layer 3 at the rift valley along a seismic profile near SMC #9. However, we think that this result is not representative of the rift valley, as the line crossed the structure perpendicularly.
4. Seismicity Study
 In Figure 3, all events detected within two days by the OBS along the central rift axis and off-axis (auxiliary material) are displayed. Slight earthquake clustering can be observed near OBS 216 at the western and eastern flanks of Knipovich Ridge (Figure 3). One cluster at WMC #8 (Figure 3; OBS 227) is roughly arranged along the central spreading axis. In general, most events were located in the area between SMC's. The largest earthquakes with local magnitudes ML > 2 occurred in segments #8 (WMC) and #9 (SMC) (Figure 3; near OBS 227 and south of OBS 228).
 The depth distribution of earthquakes shows a large number of events occurring at shallow depths within the crust (Figure 4; 0–4 km below seafloor). These crustal events are mainly observed in the vicinity of OBS 216 and OBS 227. The second peak in the depth distribution is at a depth of 6 to 14 km below seafloor, in the upper mantle, even if we take the vertical error bars into account (Figure 4). The deepest earthquake (with location depth error <8 km) occurred at 14 km depth below the rift valley (18 km below sea level). Since we only analysed 2 days of data, this value gives us a minimum estimate for the maximum depth. Events located below OBS 227 should be interpreted with caution since only the hydrophone component could be analysed and, therefore, no S-wave arrivals could be used to calculate a good hypocenter solution. Depth estimates for events outside the seismological network are less well constrained, and are not discussed in this contribution.
 Large magmatic segment centers like the Logachev Seamount (SMC #7) can be clearly bathymetrically identified and show a minimum in the mantle Bouguer anomaly [Okino et al., 2002]. In addition, the velocity model exhibits significantly higher velocities and crustal thicknesses compared to the rest of the profile. The seismic refraction model shows a crustal thickening of up to 5.5 km at these SMCs. In the amagmatic part between the SMCs, no traveltime branches are observed, which could indicate the presence of oceanic layer 3.
 In contrast, the crustal structure below WMC #8 is rather uniform both in crustal thickness and seismic velocities. Okino et al. noted that the interpretation and significance of the WMCs is speculative. They might be fed by lateral melt migration from adjacent magmatically stronger segments or represent smaller, short-lived, discrete mantle upwelling centres [Okino et al., 2002]. Our results show that most of the focussed magmatism is concentrated at the SMC #7. We suggest that these observations are strong indicators of a focused melt supply, in case of our research area concentrated at the SMC #7. The structures are quite similar to findings along the Gakkel Ridge [Jokat et al., 2003].
 The velocity jump at the Moho to 7.6 km/s suggests a petrological change from mafic crustal rock (basalt or gabbro) to ultramafic mantle rock (peridotite). The low upper mantle velocity of 7.6 km/s, compared to the global average of 8.1 km/s, are explained along mid-ocean rift valleys by elevated temperatures, cracks or serpentinization of the upper mantle [Carlson and Miller, 2003]. Such low mantle velocities are frequently observed at slow [Planert et al., 2010] and ultraslow [Jokat and Schmidt-Aursch, 2007] spreading ridges, and are believed to have been caused by a mixture of higher mantle temperatures and serpentinization. A maximum estimate for the amount of serpentinization [Carlson and Miller, 2003] yields ∼15% serpentinization for the upper mantle, if a difference of mantle velocity of 0.5 km/s (8.1 vs 7.6 km/s) is applied.
 About two-thirds (97 of 145) of more or less well-constrained hypocenters (vertical error less than 8 km) are located within the crust (<4 km below the median valley seafloor). An earthquake cluster in the crust between OBS 216 and 227 might be evidence for a low angle fault zone. However, a significant number of earthquakes (48) are located in the upper mantle. Well constrained earthquakes (20 events with a vertical error less than 8 km and a horizontal error less than 10 km) occur down to 14 km below the rift valley. This large number of earthquakes occurring in the upper mantle argues against a widespread serpentinization, since this would weaken mantle peridotites sufficiently to exclude brittle failure [Escartin et al., 1997]. Taking both results into account, it remains difficult to quantify based on the current data set the amount of serpentinization and temperature effect on the seismic velocities and earthquake distribution. Focused melt supply, in general, predicts a variable mantle temperature, which might be supported here by an uneven distribution of earthquakes in the upper mantle, e.g., less earthquakes below the Logachev Seamount.
 Comparing the focal earthquake depths of the ultraslow-spreading Knipovich Ridge with slow- and fast-spreading mid-ocean ridges, their distribution shows remarkable differences to other studies.Tilmann et al. observed at the Mid-Atlantic Ridge, 5°S, a maximum focal depth of 8 km below the median valley seafloor.Cessaro and Hussong observed earthquakes with a maximum depth of 14 km below the median valley seafloor at the intersection of the Oceanographer Fracture Zone and the Mid-Atlantic Ridge. They explained their observations with the deepening of the brittle-ductile transition due to the proximity of the ‘cold’ adjacent plate margin to the truncated spreading center. For comparison,Bohnenstiehl et al. reported at the fast-spreading East Pacific Rise very shallow earthquakes with a maximum depth of ∼1.7 km below the seafloor. Thus, the observations, including this study, support previous suggestions [e.g.,Huang and Solomon, 1988] that the maximum depth of seismicity at mid-ocean ridges might depend on the spreading rate, and hence, on melt supply.
 The investigation of a short segment of the Knipovich Ridge provided some interesting results concerning the structure and tectonic processes along ultraslow spreading ridges:
 1. The thin oceanic crust in the amagmatic segments (∼4.5 km/s) and their velocity-depth function (upper crust: 700 m: 2.7–3.2 km/s; 4000 m: 3.8–6.6 km/s; no or only a thin lower crustal oceanic layer 3) is typical for ultraslow spreading systems. The study along the Knipovich Ridge confirms this global observation.
 2. The original segmentation of the Knipovich Ridge has to be revised. Only the SMCs seem to be able to produce significant amounts of magmatic material, which is reflected in a thicker crust. This is similar to observations along e.g. the Gakkel Ridge. Spacing between strong magmatic centers is approximately 100 km.
 3. Clustering of earthquakes occurred at WMC #8 (OBS 227), at the eastern and western rift flanks near OBS 216 and south of SMC #9. The largest earthquakes occurred between strong magmatic centres with magnitudes up to ML = 2.6. About one third of the events were located in the upper mantle.
 4. Crustal earthquakes mainly occurred at the rift flanks near OBS 216 and along the central axis near OBS 227. Seismic activity with a focal depth of up to 14 km below the median valley seafloor and moderately low velocities of the upper mantle preclude a large degree of serpentinization of the upper mantle along the investigated segment of the Knipovich Ridge.
 These observations support models, which predict a cooler upper mantle below ultraslow ridges. Here, as a response of the reduced magma supply, conductive cooling is the major process transferring heat to the seafloor. Earthquakes in the upper mantle confirm that the upper mantle below Knipovich Ridge is brittle and cold enough.
 We thank the master of RV Polarstern and its crew for the excellent support. We acknowledge the possibility to use instruments (OBS) from the DEPAS instrument pool. The comments of three anonymous reviewers are greatly appreciated.
 The Editor thanks Frederik Tilmann and an anonymous reviewer for their assistance in evaluating this paper.