Moho and basement depth in the NE Atlantic Ocean based on seismic refraction data and receiver functions

Abstract Seismic refraction data and results from receiver functions were used to compile the depth to the basement and Moho in the NE Atlantic Ocean. For interpolation between the unevenly spaced data points, the kriging technique was used. Free-air gravity data were used as constraints in the kriging process for the basement. That way, structures with little or no seismic coverage are still presented on the basement map, in particular the basins off East Greenland. The rift basins off NW Europe are mapped as a continuous zone with basement depths of between 5 and 15 km. Maximum basement depths off NE Greenland are 8 km, but these are probably underestimated. Plate reconstructions for Chron C24 (c. 54 Ma) suggest that the poorly known Ammassalik Basin off SE Greenland may correlate with the northern termination of the Hatton Basin at the conjugate margin. The most prominent feature on the Moho map is the Greenland–Iceland–Faroe Ridge, with Moho depths >28 km. Crustal thickness is compiled from the Moho and basement depths. The oceanic crust displays an increased thickness close to the volcanic margins affected by the Iceland plume.

The rifting in the NE Atlantic Ocean region ( Fig. 1) occurred in several episodes spanning from the Carboniferous Period to the early Cenozoic break-up (e.g. Ziegler 1988;Doré et al. 1999). Seafloor spreading propagated from the Central Atlantic Ocean northwards into the NE Atlantic Ocean (Srivastava & Tapscott 1986), where spreading between Greenland and NW Europe began in the Early Eocene (Storey et al. 2007). Large sections of the continental margins fringing the NE Atlantic Ocean are magma-rich margins with extensive flood-basalt volcanism and igneous intrusions (e.g. Eldholm & Grue 1994;Holbrook et al. 2001). The amount of break-up-related magmatism depends on the distance to the Iceland plume (Holbrook et al. 2001). Furthermore, the plume has profoundly influenced the creation of oceanic crust throughout the NE Atlantic region (Howell et al. 2014). The influence of the plume appears to extend further to the south along the Reykjanes Ridge than to the north along the Kolbeinsey Ridge (Hooft et al. 2006). The most outstanding feature on the bathymetric map is the Greenland -Iceland -Faroe Ridge (composed of the Greenland -Iceland Ridge, Iceland and the Iceland -Faroe Ridge: Fig. 1), the result of enhanced melting in the Iceland plume (White & McKenzie 1995;Staples et al. 1997). Most of the NE Atlantic Ocean has shallower bathymetry and hotter mantle when compared to other oceans (Parkin & White 2008;Artemieva & Thybo 2013).
The complex rifting and spreading history, as well as the interaction with the Iceland plume, have shaped the present configuration of the crust in the NE Atlantic realm. The first-order structure of the crust can be recognized by regional mapping of the Moho and basement depth. A number of maps already exist for the region, such as the global model CRUST1.0 of Laske et al. (2013) or the model for the North Atlantic region by Artemieva & Thybo (2013). In addition, large parts of the region are included in compilations that cover the European plate (Grad et al. 2009;Molinari & Morelli 2011). In contrast to these existing maps for the NE Atlantic Ocean, the new compilation is exclusively compiled from seismic refraction data supplemented by receiver functions along the bordering land areas. Most seismic refraction lines are acquired along pre-existing multichannel seismic (MCS) data. The interpretation of the MCS data down to the basement is generally used as the starting point for velocity modelling of the seismic refraction lines. Deep crustal seismic reflection data were not considered for two reasons. First, even though the reflection and refraction Moho generally correlate well, there can be deviations (cf. Mooney & Brocher 1987). Second, seismic reflection data require a conversion from time to depth, which is difficult to do when no velocity information is available. To allow for the best possible quality control and internal consistency, existing local compilations were not incorporated into our database. Examples of such compilations are the seismically constrained gravity inversion for Iceland (Kaban et al. 2002) or the model for the Barents Sea (Ritzmann et al. 2007) that is based on both seismic reflection and refraction data, as well as on gravity modelling.
By incorporating so far unpublished seismic refraction data, the compilation has an unparalleled data density (Fig. 2), even though there are still substantial data gaps. These unpublished datasets comprise lines that have only been presented at conferences or in internal reports. Knowledge of the basement and Moho depth allows the calculation of the crustal thickness from which stretching factors can be estimated (cf. Kimbell et al.,this volume,Fig. 2. Data coverage. Solid lines show the location of seismic refraction lines. The labels refer to Table 1 where line names and references are given. Circles mark the positions of receiver functions (see Table 2 for references). The background shows a shaded relief map (ETOPO1: Amante & Eakins 2009).  Breivik et al. (2002Breivik et al. ( , 2003Breivik et al. ( , 2005, Mjelde et al. (2002b), Ljones et al. (2004) 47 Horsted '05 Czuba et al. (2008) 48-49 Knipovich 02 lines 1 and 2 Kandilarov et al. (2008Kandilarov et al. ( , 2010

Dataset
The core study area for the mapping of the Moho and basement depth comprises the NE Atlantic Ocean, extending from the Bight Fracture Zone and the southern limit of the Edoras Bank, Rockall and Porcupine highs in the south, to the Fram Strait and western Barents Sea in the north (Fig. 1). While the focus was on the offshore region, Iceland was an integral part of this study. The primary database behind the compilation consists of velocity -depth models derived from seismic refraction data. Figure 2 shows the location of all lines that were used in this study, and Table 1 provides the line names and references. The 202 lines under consideration were acquired between 1969 and 2011, but only 10 of them prior to 1980. There are numerous other seismic refraction lines in the study area that were not included. The reason for dismissal was mainly associated with age, which frequently was associated with a limited resolution of the velocity models and often the published information would not allow for a proper quality assessment. The majority of lines that were considered in this study were experiments that used ocean-bottom seismometers (either equipped with geophones or hydrophones or both) and seismic land stations as receivers, and airgun arrays or explosives as the source. Occasionally, some additional sonobuoys were used to receive the seismic signals. Only two lines incorporated expanded spread profiles (ESP) Hatton Bank line A Scrutton (1970Scrutton ( , 1972, Bunch (1979) 149 W-reflector profile Warner et al. (1996) Kodaira et al. (1997Kodaira et al. ( , 1998a, Mjelde et al. SIST Bjarnason et al. (1993) and 12 lines from the Faeroes Large Aperture Research Experiment (FLARE) (e.g. Fliedner & White 2003) used long-offset seismic lines by towing two multichannel streamers with maximum offsets of 38 km. The secondary dataset used in this study consisted of receiver functions analysis. This method was based on the observation of teleseismic events at either permanent or temporary deployed stations. Receiver function analysis provides information on the depth of the Moho beneath the station and can fill in some gaps in the primary dataset. Regions of particular interest were Iceland, Greenland and Ireland. Table 2 provides information on the receiver function studies used in the compilation, while the station location is shown in Figure 2.

Compilation
The database consisting of the velocity models from the seismic refraction lines and the Moho depths from the receiver function studies was used to compile both the depth to the basement and the depth to the Moho in the NE Atlantic Ocean. The Moho is generally characterized by an increase in P-wave velocity to values greater than 7.6 km s 21 (White et al. 1992). This increase is often associated with a prominent wide-angle reflection (PmP). However, serpentinization processes in the mantle rock can reduce the seismic velocity to values as low as 4.8 km s 21 (Christensen 2004). In these cases, the Moho depth is measured at the top of the (partially) serpentinized mantle.
For the definition of basement, the top of the igneous crust is used in the oceanic domain. Landwards of the continent-ocean boundary, basement is measured at the top of the crystalline crust. That way, any volcanic rocks located above the continental crystalline crust are considered as part of the sedimentary column. Off NE Greenland, the top of the crystalline crust could not be resolved on some of the lines owing to the presence of consolidated Palaeozoic and Mesozoic sedimentary rocks with velocities exceeding 5 km s 21 and a lack of a clear seismic discontinuity at the basement. These lines include AWI 20030200 ), AWI 20030300 ), AWI 20030400 (Voss & Jokat 2007) and AWI 20030500 (Voss & Jokat 2007, and here the 5.7 km s 21 velocity contour was used as a proxy for the basement. This value differs from the commonly assumed 6.0 km s 21 velocity contour as the base of sediments, as the existing data indicate that the crystalline Caledonian basement both in East Greenland and in northern Svalbard can display P-wave velocities down to 5.5 km s 21 (Schmidt-Aursch Czuba et al. 2005). Therefore, we used the 5.7 km s 21 contour as a compromise to avoid an overestimation of the sedimentary thickness.
The original digital velocity models and navigation data were obtained for a majority of the seismic refraction lines, thus avoiding possible inaccuracies that result from digitizing figures from papers. For the compilation of the basement map, only the portions of the lines with seismic control on the basement depth were used. In the next step, the misfit at the cross-points was investigated. In the case of misfits, the original velocity models and the underlying documentation were revisited to check the basement constraints. Data in the vicinity of the cross-point was then excluded along the line with the poorer constraints.
In the case of Iceland, the basement is mostly equivalent to the topography, with exception of areas that are covered by glaciers and thin sediments (mostly beach sands). The existing seismic refraction lines on Iceland often use simplified versions of the topography as the large shot and receiver spacing makes it unnecessary to take the small-scale topographical variations into account. In addition, there are a number of shot and receiver positions that are projected onto the seismic lines, which introduces erroneous elevations. For this reason, the basement points in Iceland and the adjacent coastal zone (see the extent in Fig. 3) were removed from the dataset and replaced with the ETOPO1 Global Relief Model (Amante & Eakins 2009). Regions with glaciers and beach sands were not assigned a basement depth.
All data points were converted to a Lambert conformal projection with 408W as the central meridian, and 558N and 758N as standard parallels. Along the seismic lines, the basement depth was sampled at a spacing of 1 km. Together with the basement points around Iceland, a basement map was compiled employing the universal kriging technique (Stein 1999). Prior to this, a block mean with a point spacing of 10 km was applied to maintain lateral resolution in areas with dense line spacing, such as along the Norwegian margin. Universal kriging within the SAGA GIS tool (Böhner & Antonić  Owing to the lack of seismic constraints, the onshore areas -with the exception of Iceland, western Ireland, Svalbard, the Faroe Islands and some other smaller islands -were clipped from the resulting basement map shown in Figure 3. For the compilation of the Moho depth, velocity models along the seismic refraction lines were reviewed to reject portions of the profiles on which the Moho is not constrained by Moho reflections (PmP) or mantle refractions (Pn). Not all publications display the ray coverage to assess the model constraints. In these cases, the outer portions of the models were excluded. The consistency checks at the cross-points of the lines were performed the same way as for the basement depth. Two lines were completely removed from the Moho dataset. The first line is the IFR profile (Bohnhoff & Makris 2004) across the Iceland -Faroe Ridge, which has a substantially lower Moho depth (23 km) than the FIRE offshore line (Richardson et al. 1998) along the ridge (.30 km) (Fig. 4). Given that the SIGMA line 1 (Holbrook et al. 2001) on the conjugate Greenland-Iceland Ridge displays Moho depths greater than 30 km, the IFR profile was eliminated from the Moho dataset.
The second dataset to be dismissed was DLC 94 line 5 (Dahl-Jensen et al. 1998) along the SE coast of Greenland. Along this line, the Moho depth varies between 39 and 52 km, which is substantially more than the maximum Moho depth of 33 km on the two nearby lines 3 (Hopper et al. 2003) and 4 (Holbrook et al. 2001) of the SIGMA experiment. DLC 94 line 5 has a difficult geometry, with a crooked shot line along the coast and only three receivers onshore. Hence, there might be out-of-plane phases mimicking a Moho. Alternatively, the deep Moho could be an intra-mantle reflection similar to the one observed off West Greenland (Gerlings et al. 2009) in an area that was affected by the Iceland plume. The depth at the top of the high-velocity zone (7.5 km s 21 ) on DLC 94 line 5 varies between 26 and 34 km, and would, in fact, be in reasonable agreement with the interpreted Moho on SIGMA lines 3 and 4.
Moho depths obtained from receiver functions were reviewed critically before they were added to the database. Of particular concern were the measurements on Iceland, as Schlindwein (2006) pointed out that the P-S converted phases there are only weak, which is why the use of the receiver function technique may be limited. For this reason, the 53 available measurements by Kumar et al. (2007) were cross-checked with the seismic refraction data and the seismically controlled gravity inversion of Kaban et al. (2002). In this process, 14 receiver functions were rejected, as they were not consistent with the other methods. Most of the excluded stations are located close to the coast.
West of the main mapping area, the cleaned dataset of Moho points was supplemented with the global crustal model of Laske et al. (2013) at a resolution of 18 (Fig. 5). In the east, the dataset was extended with the European Moho map of Grad et al. (2009) at a resolution of 0.18 (Fig.  5). The available computational power required a resampling of the European Moho on a 50 km raster in the Lambert projection described above. To equalize the data distribution, a block average filter was applied to the entire dataset using a 10 km raster. The Moho map (Fig. 6) was then interpolated using the SAGA GIS tool Global Ordinary Kriging. The variance in the logarithmic form was used as a quality measure and the inverse distance was applied as an interpolation method. Finally, the crustal thickness ( Fig. 7) was calculated from the difference between the Moho and basement depth.

Results
The basement, Moho and crustal thickness maps compiled from the seismic data are shown in Figures  3, 6 and 7, respectively. A brief description of the main observations is given in this section, while some features of the maps are discussed in more detail in the following section.
On the basement map (Fig. 3), the Greenland -Iceland -Faroe Ridge stands out as a zone of elevated basement. On the Greenland-Iceland Ridge, the maximum basement depth along SIGMA line 1 is 1.6 km below sea level (Holbrook et al. 2001). In Iceland, the basement is above sea level and the adjacent mid-oceanic spreading ridges (the Fig. 4. Comparison of P-wave velocity models of the intersecting seismic refraction lines FIRE Offshore (Richardson et al. 1998) and IFR (Bohnhoff & Makris 2004) on the Iceland-Faroe Ridge. Velocities are specified in km s 21 .
Reykjanes and Kolbeinsey ridges) are also characterized by elevated basement. However, the basement depth on these ridges increases with increasing distance from Iceland.
Another prominent feature on the basement map (Fig. 3) is the continuous basin along the NW European margin that extends from the southern Rockall Trough to the Lofoten Basin. The maximum basement depth in this zone is 15 km in the Møre and Vøring basins off mid-Norway. The SW Barents Shelf is also characterized by a deep basement, often exceeding 8 km in depth and up to a maximum depth of 17 km. The sedimentary basins off East Greenland are not well covered by seismic refraction profiles, but there is an indication for basement depths exceeding 8 km off NE Greenland. A smaller basin in the south (the Ammassalik Basin) will be discussed below.
Similar to the basement map, the Greenland-Iceland and Iceland -Faroe ridges are very prominent on the depth to Moho map (Fig. 6). The minimum Moho depth on these ridges is 29 km, while the maximum Moho depth beneath Iceland is 39 km. South and north of Iceland, the Moho is generally shallowest along the spreading ridges (5-10 km) from where the Moho deepens towards the continent-ocean boundary (15 -20 km). This increase in Moho depth is related to the cooling of the lithosphere with age and to excess magmatism around the time of break-up, which resulted in anomalously thick oceanic crust close to the continent -ocean boundary (Holbrook et al. 2001). The basins along the NW European margins are associated with a relatively shallow Moho depth, best seen in the Rockall Trough (12 km) and the Porcupine Basin (10 km).
Onshore Greenland, the Moho is deepest in the southern part with a depth of 40-45 km, while the NW part displays depths of 35-40 km (Fig. 6). In NW Europe, the deepest Moho in the mapping area is found in Scandinavia (47 km: Grad et al. 2009). In Ireland and the UK, the Moho is substantially shallower, varying mainly between 30 and 36 km.
The crustal thickness map (Fig. 7) displays similar characteristics to the Moho map (Fig. 6). The Greenland-Iceland -Faroe Ridge has a minimum thickness of 28 km. South of the ridge, the thickness of the oceanic crust varies mostly between 5 and 8 km, not too dissimilar from the average oceanic crust thickness of 7 km (White et al. 1992). North of Iceland, the crust produced along the Kolbeinsey Ridge is around 8 km thick (e.g. Kodaira et al. 1997), while the crust that formed at the now extinct Aegir Ridge is mostly thinner than 6 km (e.g. Breivik et al. 2006). Further north along the Mohns Ridge, crustal thickness decreases to 4-5 km (Klingelhöfer et al. 2000), and along the Knipovich Ridge segment, the thickness varies between 3 km (Hermann & Jokat 2013) and 7 km (Kandilarov et al. 2010).
Within the Rockall Trough, the crust thickens from 5 km in the south to 10 km in the north (Fig. 7). Further to the north, within the Faroe -Shetland Trough, the thickness of the crystalline crust thins again to values as small as 7 km (Makris

Conjugate margin comparison
The maps displaying basement depth (Fig. 3), Moho depth (Fig. 6) and crustal thickness (Fig. 7) are compiled from seismic refraction data that are unevenly distributed across the study area (Fig. 2). In particular, the NW European margin is much better sampled with seismic data when compared to the East Greenland continental margin. This offers the opportunity to learn from NW Europe when studying the structures found along the East Greenland coast. At the same time, the conjugate margin comparison can reveal potential shortcomings of the compilation that result from the distribution or interpretation of data.
Along the SE Greenland margin, the Ammassalik Basin shows up as a 3 km-deep basement low (Fig. 3). The structure is not sampled by seismic refraction data, but was introduced by the kriging algorithm that used free-air gravity (Andersen et al. 2010) as a constraint. Hence, the basin outline correlates with the corresponding gravity low in that area. There are few seismic reflection data available that could independently confirm the shape and depth of the basin. Hopper et al. (1998) were the first to notice the presence of a sedimentary basin in this region: they interpreted the structure as a rift system with a sediment infill corresponding to 1 s two-way travel time (TWT). Reprocessing of the seismic data suggests a probable sediment thickness of at least 3 km (Gerlings et al., this volume, in review).
When the basement depth is reconstructed for Chron C24 (Fig. 8) using the rotation poles of , the Ammassalik Basin lies just to the north of the Anton Dohrn Lineament Complex at the conjugate NW European margin. Kimbell et al. (2005) identified three individual lineaments within this complex. The continent-ocean boundary is shifted landwards across the complex (when moving from south to north), while the axis of the Rockall Trough is offset seawards by some 200 km. The Hatton Basin does not continue northwards of the Anton Dohrn Lineament Complex. This is why the Ammassalik Basin could be a NW-shifted continuation of the Hatton Basin that was left on the Greenland side at the time of final break-up. This shift would essentially be similar to the one observed between the northern and southern Rockall Trough. Hence, knowledge of the poorly studied Ammassalik Basin can probably be increased by comparison to the Hatton Basin.
In the Hatton Basin, the total sediment thickness is more than 6 km in the northern part (Hopper et al., this volume, in prep) and this could be an indication Fig. 8. Reconstruction of the basement depth for Chron C24 using the rotation pole of . Europe is fixed in this reconstruction and the rotation of Greenland is carried out with GMT (Generic Mapping Tools) software (Wessel et al. 2013). White lines show the contours with a 2 km interval. The solid and dashed red lines mark the continent -ocean boundary (COB) of Greenland and NW Europe, respectively  of the possible depth of the Ammassalik Basin. Hence, the kriging technique with gravity as a constraint may underestimate the basement depth if seismic data are lacking.
Such an underestimation of basement depth also occurs on the NE Greenland continental shelf. The basement reconstruction for Chron C24 (Fig. 8) shows that the deep basins off mid-Norway (the Møre, Vøring and Lofoten basins) correlate with the basins off NE Greenland (the Danmarkshavn and Thetis basins). However, the Norwegian basins are up to 15 km thick, while the maximum depth off Greenland seems to be only 8 km. The seismic refraction constraints within the basins off NE Greenland are restricted to the southern edge of the Danmarkshavn Basin (line AWI 20030300) and the eastern edge of the Thetis Basin (line AWI 20030200) (Fig. 3). In addition, the basement on these two lines is only poorly resolved by the seismic data and is therefore approximated by the 5.7 km s 21 velocity contour. Newly released seismic reflection data indicate a maximum sediment thickness of 18 km (Hopper et al., this volume, in prep), which is in better agreement with the mid-Norwegian basins and also with the SW Barents Sea.
While the gravity constraints used in the kriging procedure could outline the general structure of the basins not covered by seismic refraction data, this method seems to underestimate the basement depth, as seen in the two examples from Greenland. Initially, gravity constraints were also tested for the kriging of the Moho depth. In particular, filtered versions of the Bouguer gravity anomaly were used for this purpose. However, these resulted in some features, such as crustal roots, that were difficult to explain in some cases. This is why the final kriging of the Moho was performed without gravity constraints. Instead, regional and global datasets (Grad et al. 2009;Laske et al. 2013) were used to constrain the surrounding areas. Assuming isostasy, the basins on the continental shelves should be associated with a shallowing of the Moho. While this is the case for basins covered with seismic data (e.g. the Rockall Trough), the basins off Greenland (the Ammassalik, Danmarkshavn and Thetis basins) do not show a corresponding expression in the Moho depth. In these areas, the Moho depth obtained from seismically controlled gravity inversion (Haase et al., this volume, in review) provides greater structural detail.

Thickness of oceanic crust
The oceanic crust in the NE Atlantic Ocean displays variations in thickness ranging between 2 and 40 km (Fig. 9) compared to an average of 7 km for normal oceanic crust (White et al. 1992). Areas with thick oceanic crust can generally be related to an increased magmatism associated with the Iceland plume. The Greenland-Iceland -Faroe Ridge is characterized by a 28 -40 km-thick crust, which White (1997) interpreted as the interaction of a rising mantle plume with a spreading ridge. He suggested that the thickness variations were related to variations in the temperature of the mantle or mantle flow rates, or both.
South of Iceland, the initial oceanic crustal thickness at break-up is generally greater than 15 km (Fig. 9) and decreases with increasing distance from the plume (cf. Holbrook et al. 2001). Over time, the thickness of the oceanic crust decreased to values of around 8 km at a distance of 250 km from the continent-ocean boundary. Refraction seismic experiments on the Reykjanes Ridge away from Iceland indicate variations in the crustal thickness that ranged from 4 to 9 km (Bunch & Kennett 1980;Smallwood et al. 1995;Navin et al. 1998;Jacoby et al. 2007).
North of Iceland, the initial spreading after break-up was along the Aegir Ridge, which lasted until 30 Ma when spreading there became extinct ). At that time, the Kolbeinsey Ridge started to develop from the south, with final detachment of the Jan Mayen microcontinent from East Greenland occurring at 20 Ma (Chron C6b) Peron-Pinvidic et al. 2012). The thickness of the oceanic crust that was formed in the northern portions of the Aegir and Kolbeinsey ridges differs significantly (Fig. 9). Between the Jan Mayen microcontinent and mid-Norway, the initial crustal thickness at break-up was around 11 km (Breivik et al. 2006), which is less than that observed in SE Greenland at a similar distance from the Iceland plume (19 km: Hopper et al. 2003). Close to the extinct Aegir Ridge, the crustal thickness is as little as 4 km (Breivik et al. 2006). In contrast, the crust at the Kolbeinsey Ridge has a thickness of between 7 and 10 km (Kodaira et al. 1997). Breivik et al. (2006) speculated that the thin oceanic crust between the Jan Mayen microcontinent and mid-Norway is caused by interaction with the Iceland plume. They suggested that the construction of the magmatic Greenland -Iceland -Faroe Ridge to the south depleted the mantle. Asthenospheric flow transported this depleted mantle to the Aegir Ridge, giving a lower than normal magma productivity. Howell et al. (2014) employed three-dimensional (3D) numerical models that simulated a plume interacting with rifting continents and spreading ridges. Their results support a plume with a relatively low flux (95-128 m 3 s 21 ) to explain the restriction of the relatively thick crust in the southern part of the Aegir Ridge.
The Kolbeinsey Ridge terminates at the West Jan Mayen Fracture Zone. Northwards of the fracture zone, the crustal thickness decreases markedly (Fig. 9). The Mohns Ridge produced thicker than normal oceanic crust just after break-up, but most of the younger crust is only between 4 and 6 km thick. The main exception is the southernmost part of the Mohns Ridge, which is affected by the volcanism on the Jan Mayen islands. Rickers et al. (2013) could identify two distinct low-velocity zones in the mantle: one centred beneath Iceland and one beneath the northern Kolbeinsey Ridge. At depth, the latter anomaly covers the whole length of the Kolbeinsey Ridge and extends beyond the West Jan Mayen Fracture Zone to the southern Mohns Ridge. Just to the north of the West Jan Mayen Fracture Zone, close to Jan Mayen, the crustal thickness is 10 km (Kandilarov et al. 2012). Based on the distinct mantle velocity anomalies, Rickers et al. (2013) see two seperate hotspots beneath Iceland and Jan Mayen, which can be also supported by isotope studies (Schilling et al. 1999).
On a series of seismic refraction lines on the eastern flank of the Mohns Ridge, Klingelhöfer et al. (2000) modelled a mean crustal thickness of 4 km that was produced at a full spreading rate of 14 mm a 21 . For spreading rates greater than 20 mm a 21 , there is little variation in crustal thickness: however, the thickness drops rapidly for lower rates (Dick et al. 2003). Crust that was produced at the Knipovich Ridge (a full spreading rate of 14 mm a 21 : DeMets et al. 1990DeMets et al. , 1994 is often just 3 km thick (Hermann & Jokat 2013), in particular to the west of the present location of the ridge. To the east, the crust is generally slightly thicker. Even though there are some asymmetries in crustal accretion at the Knipovich Ridge ), a systematic difference in crustal thickness on either side of the ridge is difficult to explain. This is why it is worthwhile revisiting the datasets on which the crustal thickness compilation is based.
The main profile that is constraining the crustal thickness west of the Knipovich Ridge is line AWI 20090200 (labelled as number 14 in Fig. 2) (Hermann & Jokat 2013). This line displays a 2-3 km-thick oceanic crust (Fig. 10) that lacks a distinct oceanic layer 3 (Hermann & Jokat 2013). Some 30 km to the south, another profile extends from the Knipovich Ridge towards the NE (Barents 98 line 8, labelled as number 43 in Fig. 2). Close to the ridge, this line exhibits 6-7 km-thick crust ( Fig. 10) with distinct oceanic layers 2 and 3 (Ljones et al. 2004). A further profile in close proximity is Knipovich 02 line 1 (labelled as numbers 48 in Fig. 2) (Kandilarov et al. 2008). Here, the crust is 4 km thick adjacent to the ridge (Fig. 10), which is in better agreement with line AWI 20090200. However, Kandilarov et al. (2008) also interpreted the presence of an oceanic layer 3, in contrast to the model of line AWI 20090200 (Hermann & Jokat 2013), on the western side of the Knipovich Ridge. Although crust formed at ultraslow-spreading ridges can display large variations in thickness and velocity structure Minshull et al. 2006), some differences may also result from different approaches in the velocity modelling and the subsequent interpretation of the model. This is why the three lines will be briefly reviewed here.
The total thickness of oceanic layer 2 is similar on lines AWI 20090200 and Barents 98 line 1 (Fig. 10), even though the level of detail in the models is different. On the AWI line, Hermann & Jokat (2013) differentiated three sublayers (2A, 2B and 2C), while no such subdivision is seen on Barents 98 line 8 (Ljones et al. 2004). However, the velocity range observed within layer 2 is not too dissimilar. The main difference here is that the AWI line resolves velocities as low as 2.7 km s 21 at the top of the oceanic crust that are not resolved on Barents 98 line 8.
The potentially critical issue on Barents 98 line 8 is the interpreted oceanic layer 3 (Ljones et al. 2004) with velocities of 6.4-6.9 and 7.2 km s 21 in layers 3A and 3B, respectively (Fig. 10). Ljones et al. (2004) stated that there were a total of 107 travel time picks for reflections between layers 2 and 3A, while only 11 Moho reflections (PmP) were picked. With a shot spacing of 200 m, these 11 picks correspond to a 2 km-long segment along which the PmP is observed. With such a limited number of observations, the Moho appears not to have been mapped reliably by reflections. In contrast, the numerous observed reflections between layers 2 and 3 are also unusual, as this is commonly not a reflective boundary. Minshull et al. (2006) did not report a single such reflection in their dataset from the ultraslow-spreading SW Indian Ridge, and the same is true for the Mohns Ridge (Klingelhöfer et al. 2000). However, both of these other studies show frequent reflections from the Moho. This is why an alternate explanation for the interpreted layer 3 should be looked into. Instead of having a gabbroic composition, a partially serpentinized mantle rock may explain the seismic observations. The Moho interpretation on line AWI 20090200 is also unclear. Despite a modelled velocity contrast of .1.5 km s 21 (Fig. 10), not a single Moho reflection is observed in a 140 km-wide zone adjacent to the Knipovich Ridge (Hermann & Jokat 2013). Hence, there might not be a sharp Moho, as indicated in the model, but rather a gradual velocity transition from layer 2C into the mantle. In this case, the interpreted layer 2C or parts of it could, in fact, be a highly serpentinized mantle. Nevertheless, there are some large misfits between the observed and calculated travel-time curves along the AWI line. For example, ocean-bottom seismometer (OBS) 216 close to the Knipovich Ridge ( fig. 2c in Hermann & Jokat 2013) shows that the calculated travel times of the mantle refraction are up to 300 ms too fast, which could mean that the Moho locally could be up to 1.8 km deeper than shown in the velocity model.
When the AWI profile is compared with the Knipovich 02 line 1 (Fig. 10), there is a good match between the combined layers 2A and 2B on line AWI 20090200 (Hermann & Jokat 2013), and oceanic layer 2 on the Knipovich 02 line 1 (Kandilarov et al. 2008). This applies to both layer thickness and velocities. The underlying layer is interpreted as layer 2C on the AWI profile (4.8-6.1 km s 21 ) and as layer 3 on the Knipovich 02 line 1 (5.6-7.0 km s 21 ). Hence, based on the modelled velocities, the different interpretations seem to be justified. However, as discussed earlier, there is also the possibility that part of the interpreted layer 2C on the AWI line could be highly serpentinized mantle.
The definition of the Moho along the Knipovich 02 line 1 seems to be based mainly on mantle refractions (Pn). Kandilarov et al. (2008) showed only one station that recorded a Moho reflection (PmP) and here the misfit is up to 200 ms. Close to the Knipovich Ridge, the Moho is difficult to model with Pn phases alone, as the model indicates only a small velocity increase across the Moho. With that, there is no distinct change in the phase velocity of the first arrivals between the crustal and mantle arrivals that would allow for an unequivocal phase interpretation. From the previous discussion, it can be seen that the differences in the crustal structure of the lines at the Knipovich Ridge may, in fact, be less than that indicated by the velocity models. In order to decide whether the differences are real or not, the lines should be reanalysed together.

Conclusions
The maps of the basement and Moho depth are useful tools to use in the discussion of first-order crustal features in the NE Atlantic realm. Although the region is not evenly covered with seismic refraction data and receiver function analyses, there is an adequate coverage of the large-scale tectonic structures. Data used in this compilation were acquired over a timespan of more than 40 years, and are therefore very variable in the shot and receiver spacing, as well as in the modelling technique. In addition, our understanding of rifting processes at continental margins has considerably changed over this period, which, of course, is reflected in the conceptual models on which the interpretation of the velocity models are based. Nevertheless, the basement and Moho depth are generally fairly robust features of velocity models, with the exception of areas with thin crust. Here, the mantle rock can be partially serpentinized and display velocities as low as 4.8 km s 21 at a serpentinization rate of 100% (Christensen 2004), and, hence, may erroneously be interpreted as crustal rock. One important measure of quality control for the basement and Moho compilation was the fit at cross-points to check for internal consistency of the datasets. While this helped to eliminate some erroneous data from the compilation, the maps are only as good as the underlying velocity models.
Using seismic refraction data to map the basement depth has two advantages compared to seismic reflection data. One is that no time-to-depth conversion is necessary as the velocity models are developed in the depth domain. The other is that the basement is often better defined on seismic refraction data, in particular in areas with deep sedimentary basins or where basalts are interbedded with sedimentary rocks. Many seismic refraction lines benefit from coincident reflection data by incorporating detailed basement relief from there. The disadvantage is that the data coverage with seismic refraction lines is not nearly as good as that with reflection lines. However, using the kriging method with gravity as a constraint, a rather detailed basement map could be compiled from the seismic refraction data (Fig. 3). That way, even basement lows that are not covered by seismic data are imaged, such as the Ammassalik Basin off SE Greenland. The basin depths of these seismically unconstrained features may be erroneous, but at least allow a qualitative assessment and comparison with the conjugate margin to be made (Fig. 8).
Our results show that the continuous rift basins extending from the southern Rockall Trough to the Lofoten Basin carry on to the Danmarkshavn and Thetis basins of NE Greenland (Fig. 8). Similarly, the Hatton Basin off Ireland seems to continue along with the Ammassalik Basin in SE Greenland. These correlations are most prominent on the basement map (Figs 3 & 8), while details of the Moho geometry beneath the Greenlandic basins are not resolved. This is related to a lack of seismic refraction data in these regions. However, 3D gravity modelling based on the seismic constraints is able to show a continuity of these features on the crustal thickness and Moho maps (Haase et al., this volume, in review).
The crustal thickness that is compiled from the basement and Moho maps can be used to calculate crustal stretching factors and first-order crustal strength profiles, which has done by Kimbell et al. (this volume, in press) for the NW European margin. In addition, deformable plate reconstructions depend on good estimates of crustal thickness. Apparent inconsistencies in the crustal thickness map, such as the generally thinner oceanic crust west of the Knipovich Ridge compared to the east, can help to identify datasets that it might be worthwhile remodelling/reanalysing.