Transport of mafic magma through the crust and sedimentary basins: Jameson Land, East Greenland

Igneous sheet-complexes transport magma through the crust, but most studies have focused on single segments of the magma transport system or have low resolution. In the Jameson Land Basin in East Greenland, seismic reflection data and extensive outcrops give unparalleled constraints on mafic intrusions down to 15 km. This dataset shows how sill-complexes develop and how magma is transported from the mantle through sedimentary basins. The feeder zone of the sill-complex is a narrow zone below a basin, where a magmatic underplate body impinges on thinned crust. Magma is transported through the crystalline crust through dykes. Seismic data and published geochemistry indicate that magma is supplied from a magmatic underplate without perceptible storage in crustal magma chambers and crustal assimilation. As magma enters the sedimentary basin, it forms distributed, bowl-shaped sill-complexes throughout the basin. Large magma volumes in sills (4–20 times larger than the Skaergaard Intrusion) and the presence of few dykes highlight the importance of sills in crustal magma transport. On scales smaller than 0.2 km, host-rock lithology, and particularly mudstone tensile strength anisotropy, controls sill architecture in the upper 10 km of the basin, whereas sills are bowl-shaped below the brittle–ductile transition zone. On scales of kilometres and towards basin margins, tectonic stresses and lateral lithological changes dominate architecture of sills. Supplementary material: An uninterpreted and unwarped version of the seismic line in Fig. 4 (DR1), a spreadsheet showing thickness of sills and width of dykes in the study area (DR2), and assumptions and calculations of magma volume (DR3) are available at https://doi.org/10.6084/m9.figshare.c.5670470

Sill-complexes are interconnected magma conduits that are broadly parallel to stratigraphy, transgress upwards and transfer magma through the crust and in certain cases to eruption (e.g. Lorenz and Haneke 2004;Cartwright and Møller Hansen 2006;Jaxybulatov et al. 2014;Eide et al. 2017;Schofield et al. 2017). Interconnected sill-complexes appear to be the primary way in which mafic magma is transported through sedimentary basins, in contrast to more viscous silicic magmas, which are more likely to form plutons (Cartwright and Møller Hansen 2006;Muirhead et al. 2014;Schofield et al. 2015;Magee et al. 2016Magee et al. , 2019Hafeez et al. 2017;Svensen Henrik et al. 2018). Studies on sill-complexes are often focused on data types that highlight specific segments of the magma transport pathway (seismic reflection data, outcrops of sillcomplexes, active volcanoes, geochemistry) or have low resolution (earthquake data, seismic tomography, seismic refraction data). Insights into basic principles of crustal magma transport and intrusion geometries can be gained from outcrop examples (e.g. Hansen et al. 2011;Schofield et al. 2012;Eide et al. 2017), seismic data (e.g. Planke et al. 2015;Magee et al. 2016;Schofield et al. 2017), and through analogue and numerical modelling studies (e.g. Galland 2012;Kavanagh et al. 2017;Walker and Gill 2020), yet linking the complete plumbing architecture from deep in the crust to the surface is a challenge (Jerram and Bryan 2015). Not only is understanding the geometry of the magma transport system and physical controls on development of sill-complexes valuable to map out the route of magma within upper crustal and basin settings, but also understanding the propagation and architecture of sillcomplexes is important to interpret earthquake (e.g. Sigmundsson et al. 2010) and ground deformation data (e.g. Pedersen and Sigmundsson 2004;Galland 2012;Magee et al. 2017a, b) related to volcanic unrest. Another area where understanding intrusion networks is important is the utilization of geological resources in intruded basins, such as hydrocarbons (e.g. Senger et al. 2017), groundwater (e.g. MacDonald et al. 2001, mineral resources (e.g. Hayman et al. 2021) and geothermal heat (e.g. Elders et al. 2014).
The Jameson Land Basin in East Greenland ( Fig. 1) is an onshore basin that contains a large amount of mafic igneous intrusions. This basin offers a unique opportunity to understand magma transport through the crust and sedimentary basin on scales from centimetres to tens of kilometres, for the following reasons: (1) about 3 km of post-emplacement erosion (Mathiesen et al. 2000;Hansen et al. 2001) has dissected the mafic intrusions, leading to extensive unvegetated outcrops where dykes (subvertical igneous sheets) and sills (broadly layer-concordant igneous sheets) can be readily investigated in large tens of kilometres scale outcrops (Figs 2 and 3); (2) there is a well-constrained stratigraphy owing to widespread outcrops and decades of surface and subsurface geological investigations (e.g. Surlyk et al. 1973;Dam and Surlyk 1998;Ahokas et al. 2014;Eide et al. 2016Eide et al. , 2017Eide et al. , 2018aGuarnieri et al. 2017;Brethes et al. 2018); (3) extensive seismic refraction surveys have been performed directly offshore from the outcrops, leading to good control on the crustal structure and P-wave velocities in the basin (Fig. 1b;Weigel et al. 1995;Mandler and Jokat 1998); (4) an onshore 2D grid of seismic reflection data (Fig. 2a) makes it possible to investigate both structure of the basin and distribution of intrusions at depth.
Integration of available data from the Jameson Land Basin allows us to bridge the observational gap between field and seismic studies in three dimensions, investigate a mafic volcanic plumbing system in an area that is 60 × 120 km in extent and more than 15 km deep, and link those observations to deeper crustal structure from seismic refraction data extending down to c. 30 km. Because the thick cover of shallow-level intrusions and lava flow sequences, which reflect seismic energy and can commonly make sub-basalt imaging difficult (e.g. Ziolkowski et al. 2003;Jerram et al. 2010;Eide et al. 2018b), is eroded in the Jameson Land Basin, the study site offers a rare opportunity of a window to image deeper intrusions and link them both to occurrences of the same intrusions in outcrop and to the overall crustal structure. No studies have yet investigated all these different datasets from the Jameson Land Basin together, and little is therefore known about intrusion geometries throughout the basin, what controls these geometries, and what it means for our understanding of magma transport in sedimentary basins.
In this contribution, we show that two sill-complexes shaped as nested bowls occur in the Jameson Land Basin and that at least the most well-imaged of these are fed from a narrow zone along the axis of the basin. We investigate the basin-scale controls on the shape of these sill-complexes, and what these insights tell us about controls on magma transport and sill emplacement through sedimentary basins. In particular, this study offers insights into how magma behaves as it propagates from the crystalline crust into a sedimentary basin, and how igneous intrusions are controlled by lithology, far-field stresses and position within a sedimentary basin. The goals of this paper are to (1) present the structure of the Jameson Land Basin and the igneous intrusions within it, (2) investigate the controls on the architecture of  Weigel et al. 1995), and igneous sills mapped as part of this study. It should be noted that the sills occur above where a lower-crustal high-velocity body (magmatic underplate) impinges on the crust at the base of the deep Jameson Land Basin. (c) P-wave velocities measured from seismic refraction data in the Jameson Land Basin (Mandler and Jokat 1998), compared with seismic velocities in deep wells on the Norwegian Continental Shelf (see (a) for location).
the Jameson Land intrusive complex at shallow to deep crustal levels and (3) discuss how the observations from the Jameson Land Basin might be relevant to understanding mafic sill-complexes in general.

Geological history of the Jameson Land Basin
The Jameson Land Basin in East Greenland is a north-southtrending, c. 15 km deep (Figs 1a and 4; Mandler and Jokat 1998) basin that formed after collapse of the Caledonian orogeny. The stratigraphic succession in the Jameson Land Basin consists of three main packages (e.g. Guarnieri et al. 2017), which are discernible in the seismic data ( Fig. 4a): (1) an upper unit in the uppermost 1.5 s two-way travel time (twt) (c. 2 km) characterized by a gently concave-upward, unfaulted succession; (2) a middle unit with rotated fault blocks and associated growth wedges (0.7-1.5 km thick); (3) a poorly imaged lower part below the growth wedges  Black boxes show location of (d)-(h). Noteworthy features are the large proportion of sills compared with dykes (c. 99% sills by outcrop area), the overall upwards-towards-the-north stepping of sills (green dashed lines), obvious feeding relationships between dykes and sills, the tendency for sills to transgress stepwise upwards towards the north in the southern part, and to follow extensive mudstones in the northern part, and turn from sills to dykes without any changes in the host-rock ( pinchout of sedimentary bodies, faults). (b) Synthetic seismogram (from Eide et al. 2018a) showing what (a) would look like in seismic data at c. 6 km depth. It should be noted that dykes are not detected, that reflections from several sills interfere and appear as one, and that actually stepped sills appear oblique sheets. (c) Synthetic seismogram (from Eide et al. 2018b) showing what (a) would look like in seismic data at c. 3 km depth. It should be noted that dykes are in some cases possible to detect, and that geometries of sills c. 10 m thick are well constrained at these shallower depths. (d) Feeding relationships between dykes and sills, and sills stepping upwards-towards-the-north through the stratigraphy. (e) Abrupt transition from sill to dyke without any clear structural or lithological c. 10 km thick. Stratigraphic interpretation in the upper two units is straightforward because the different seismic units crop out and can be compared with units in the geological map ( Fig. 2; e.g. Bengaard and Henriksen 1982;Pedersen et al. 2013;Guarnieri et al. 2017;Brethes et al. 2018). The upper unit corresponds to the Late Triassic-Early Cretaceous post-rift succession, the middle unit corresponds to the synrift Triassic Pingo Dal Formation, and the lower unit consists of undifferentiated pre-rift deposits and basement. It is worth noting that what is termed 'pre-rift' here are deposits of several pre-Triassic rift-and post-rift events, but none of these can be differentiated in the present seismic data. It is seldom possible to pick confidently the boundary between the pre-rift sediments and basement, but the top of a basement block is imaged in the seismic data towards the eastern part of the basin (Fig. 5), and this interpretation is supported by extensive outcrop, aeromagnetic and electromagnetic data (Guarnieri et al. 2017). Post-Triassic rift episodes in East Greenland left few marks in the Jameson Land Basin, and deposits younger than the Cretaceous are eroded, as the basin has been subject to c. 3 km of erosion since the Eocene (Mathiesen et al. 2000;Hansen et al. 2001).
The opening of the North Atlantic gave rise to the North Atlantic Igneous Province and to emplacement of large amounts of igneous rocks along what are now the North Atlantic margins during the Paleocene and Eocene (e.g. Saunders et al. 1997;Hansen et al. 2009;Brooks 2011;Larsen et al. 2014;Horni et al. 2017). Opening of the North Atlantic and seafloor spreading on the AEgir spreading cause. (f ) Image showing one vertically stable sill emplaced in a regional mudstone bed, and one less stable sill emplaced in poorly cemented sand. (g) Thin sills that seek out very small mudstone drapes in crossbed foresets to intrude into, indicating very strong host-rock controls on sill emplacement. (h) A 3D composite image of virtual outcrop and Google Earth imagery (Image © 2021 Maxar Technologies) showing that dykes can be observed both in vertical outcrops and the nearly vegetation-free plateaux in Jameson Land, making them easy to map over large areas. Fig. 4. Large-scale 3D view at depth showing the geometry of the Jameson Land Basin, basin fill and igneous intrusions. It should be noted that the sills show nested bowl shapes at depth and propagate obliquely upwards, broadly concordant with the overall basin shape, and that the sills become flatter in the upper parts of the basin fill; also that the bowl-shaped sills appear to be fed from a magma entry zone along the centre of the basin, and that steep southern sills appear to be supplied from an out-of plane magma feeder zone. The magma entry zone broadly coincides with where the lower-crustal body impinges on the base of the Jameson Land Basin (Figs 1b and 2a). Inset labelled 'amplitude' shows colour scale for seismic data. (For location, see Fig. 2.) Uninterpreted and unwarped versions of the seismic data are given in Supplementary material DR1. ridge ( Fig. 1a) commenced at c. 55 Ma. Several kilometres of flood basalts were also emplaced in central East Greenland at the same time as onset of seafloor spreading, constrained to 55-53.5 Ma (e.g. Larsen et al. 2014). Importantly, around this time significant mafic underplating of the crust occurred along East Greenland. This magma was supplied from the nearby palaeo-Iceland plume, which fed the flood basalt succession (Hopper et al. 2003;Voss and Jokat 2007;Voss et al. 2009). This high-density underplated region and showing relationships between bedding architecture, lithology and igneous intrusions at depth. It should be noted that the angle of sills is relatively constant at depth, and that sills cross-cut basement (1) and bedding in pre-rift (2) and synrift (3). In some areas, sills follow the trends of nearby faults (4), probably indicating that some minor faults parallel to the major faults were exploited by sills for short distances. Sills in the post-rift and the synrift in the basin centre (5a, 5b) are generally concordant with layering, but transgress upwards towards the basin margins. (b) Uninterpreted and unwarped view of the seismic line on the south face of the illustration in (a). (c) The 1D seismic response from a tuned (thin) highimpedance reflector. Bar labelled 'amplitude' shows colour scale for seismic data. the mantle beneath it represent the original source regions for the melt generation.
The studied sills and dykes in outcrops in the Jameson Land Basin are interconnected, geochemically high-Ti tholeiitic basalts dated to 53 Ma using Ar-Ar and palaeomagnetism, and, based on the geochemistry, are interpreted to have been emplaced in a single intrusive episode (Hald and Tegner 2000). Thus, the studied intrusive rocks slightly postdate the main flood basalts in East Greenland, and they are not geochemically similar to any preserved flood basalts in the area (Hald and Tegner 2000). The outcropping intrusions were emplaced into Jurassic sandstones and mudstones at a depth of 3 km during maximum burial. This estimate is based on analysis of vitrinite reflectance data, apatite fission-track data and reconstructions of eroded stratigraphy (Mathiesen et al. 2000;Hansen et al. 2001). This implies that the intrusions in surface outcrops today were emplaced at c. 3 km depth. The intrusions at c. 15 km depth today, which are imaged in the seismic reflection data in Jameson Land, were emplaced at c. 18 km. These constraints are taken into account for our models for sill emplacement presented below. Any volcanoes, lava sequences and shallow intrusions (which may have intruded into uncemented, overlying host-rocks) have now been eroded from the study area.
A change in plate movement in the Oligocene led to northward propagation of the North Alantic, a ridge-jump from the AEgir to the Kolbeinsey Ridge and separation of the Jan Mayen Microcontinent from Liverpool Land during 40-25 Ma ( Fig. 1a; Talwani and Eldholm 1977;Mjelde et al. 2008;Gaina et al. 2009). This event is believed to have led to emplacement of a small number of narrow alkaline dykes that cross-cut the studied tholeiitic basalts, which have been reported in Jameson Land (Hald and Tegner 2000;Eide et al. 2017) and also in many places in central East Greenland (Brooks 2011;Larsen et al. 2014). It should be noted that older alkaline magmatic events do occur that are coincident in age with the North Atlantic Igneous Province sensu stricto, and seem to be spatially linked with the Jan Mayen fracture zone (see Hafeez et al. 2017), so without absolute dating the exact age of the alkali basalts rocks remains unresolved.

Dataset and methods
The study area comprises the areas of Jameson Land covered by 2D seismic data and a virtual outcrop dataset (Fig. 2a). Sedimentological logs were acquired in a field campaign in 2012 (Eide et al. 2016).
The virtual outcrop dataset ( Fig. 3; available for study at https:// v3geo.com/model/61) was acquired from the sea cliffs on the west side of Hurry Inlet using helicopter-mounted lidar scanning (Buckley et al. 2008). It has a resolution of c. 7 cm and is 32 km long and c. 400 m high. The data were visualized using LIME (Buckley et al. 2019), and 22 km of this dataset is shown in this paper (Fig. 3a) as the rest has only minor amounts of intrusions. Igneous intrusions in this outcrop are clearly visible as up to 17 m thick, most commonly 10 m thick, dark-coloured resistant rocks emplaced within the Early Jurassic sandstones and mudstones ( Fig. 3; Eide et al. 2016Eide et al. , 2017; for additional statistics on sill and dyke geometries see spreadsheet in Supplementary material DR2). These thickness measurements are created by measuring the vertical thickness of all outcropping sills every 1 km along the outcrop, and the width of every exposed dyke. The geometries of igneous intrusions in the virtual outcrop were used to generate synthetic seismograms that help to constrain interpretation of seismic data in the area ( Fig. 3b and c). These seismograms were presented in, and are reproduced from, Eide et al. (2018a).
The studied 2D seismic data were acquired by Atlantic Richfield Company (ARCO) from 1986 to 1989 in a hydrocarbon exploration campaign in onshore East Greenland (Larsen and Marcussen 1992), and are available through the Geological Survey of Denmark and Greenland (GEUS). Most of these seismic lines were acquired using a dynamite source, leading to deep penetration, high frequencies preserved to great depth (dominant frequencies from sill reflectors at 15-20 Hz at 5 s twt, c. 14 km depth), and in general good imaging of sills and locally also sedimentary geometries. The lines are spaced 5-10 km apart (Fig. 2a). P-wave velocities are taken from seismic refraction experiments published by Mandler and Jokat (1998; Fig. 1b and c). The studied host-rocks have been subject to c. 3 km of erosion since deposition, and therefore show a high P-wave velocities at the surface and a low velocity-depth gradient ( Fig. 1c; Mandler and Jokat 1998). The observed velocities are consistent with patterns observed in deep wells on the Norwegian Continental shelf when 3 km of erosion are accounted for (Fig. 1c). In seismic data (Fig. 4), sills appear as high-amplitude, hard-kick (increase in acoustic impedance) reflectors, which correspond to seismic responses where sills have a red-coloured upper reflection and a blue-coloured bottom reflection with the chosen colour scale (Fig. 5c). The reflections from interpreted sills are in the vast majority of cases tuned, which means they are so thin that they do not give separate reflections from top and base ( Fig. 5c) and their thickness at depth can therefore not be confidently estimated.
It must also be noted that it is not possible to seismically distinguish between the 'high-titanium basalts' (emplaced at c. 53 Ma), and the cross-cutting alkali basalts (assumed to be emplaced later at c. 30 Ma) (Hald and Tegner 2000). However, fewer than 10% of the sills investigated by Hald and Tegner (2000) consist of alkali basalts, indicating that these are rare compared with the main, high-titanium basalts.

Architecture of sills and dykes in outcrops
The igneous intrusions in the Jameson Land Basin are well exposed in the sea cliffs on the west side of Scoresby Sund (Fig. 2), where they can be observed as dark, transgressive intrusive rocks commonly 10 m thick, within the lighter Jurassic host rock (Fig. 3). Dykes and sills are mainly interconnected, and sills in many cases turn into dykes and vice versa (Fig. 3a,d and e). Sills are mainly emplaced within weak lithologies: on the large scale, sills are preferentially emplaced within the large regional mudstone beds, within more restricted mudstones overlying flooding surfaces and within poorly cemented sandstones ( Fig. 3a and f; Eide et al. 2016). This is true both on the scale of hundreds of metres to several kilometres ( Fig. 3a) and on the scale of centimetres to metres, as thin sill 'splays' exploit mudstone-draped cross-bed foresets in tidal dunes (Fig. 3g). Eighty-four per cent of the sills (by area of outcropping sills in Fig. 3a) occur in mudstone-dominated lithologies (Eide et al. 2017). This indicates that mudstones must have a fundamental control on the architecture of sill-complexes. This will be discussed below.
Host-rocks mainly consist of thinly bedded (centimetres to decimetres) sandstones and mudstones in the southern part of the outcrop, and thick (tens of metres) sandy packages interbedded with regional mudstones (1-20 m thick) in the northern part (Fig. 3a).
Most sills are stepped sills in the sense that they are horizontally segmented by broken bridges (sensu Schofield et al. 2012) and longer transgressive dykes, and the segments are generally vertically offset and show an upwards-to-the-north trend (green dashed lines in Fig. 3a and d). The length of the sill segments increases from c. 0.2 km in the southern part of the outcrop ( Fig. 3d), through c. 1 km in the middle part, to c. 6 km in the northern part (Fig. 3f ). This increase in segment spacing coincides with an increase in sandstone content towards the north in the outcrop (Fig. 3a). This indicates that the segment spacing and jog height are strongly controlled by host-rock properties, and this is discussed in detail below. In most of the outcrop, two sills occur within 50 m vertical distance from each other, and these occasionally cross-cut each other. Thus, the studied outcrop exposes part of a sill-complex.

Basin-scale architecture of sills
In the seismic data from the Jameson Land Basin, the intrusions occur as high-amplitude reflections that show hard-kicks (i.e. an increase in acoustic impedance), which can be seen to transgress stratigraphy and show layer-parallel, layer-parallel oblique and rarely saucer-shaped geometries in the uppermost parts (Fig. 4). These reflections are interpreted as the same mafic igneous intrusions that are seen in outcrop because of their similarity to igneous intrusions in seismic data elsewhere (e.g. Cartwright and Huuse 2005;Reynolds et al. 2017;Eide et al. 2018b), documented presence in the basin (Larsen and Marcussen 1992;Eide et al. 2017;Larsen 2018) and attitude towards outcrops where they are observed and sampled (Figs 3 and 4). Some of these reflections are clearly stepped and are thus interpreted as stepped sills, similar to the sills observed in outcrop (e.g. Fig. 3d). Other high-amplitude reflectors at greater depths are clearly oblique, and it is unclear whether these represent truly oblique igneous sheet intrusions or if they are highly transgressive stepped sills, as it is not possible to distinguish between these two geometries in lower quality seismic data because of limited lateral resolution (compare Fig. 3b and c). Based on the lack of truly oblique sheets in the outcrop (Fig. 3a), we speculate that these dipping seismic reflectors are mainly highly transgressive stepped sills. Because of this, and because of the low dip of oblique high-amplitude reflections (commonly 6°, maximum 15°), we refer to all broadly subhorizontal subsurface igneous sheets as sills.
Sills are present at all levels in the basin (i.e. from 0 to 15 km depth), and no particular depth intervals or stratigraphic intervals appear to have accumulated a greater proportion of sills than other intervals (Fig. 6). It should be pointed out that the seismic resolution decreases with depth, so it is possible that a constant amount of intrusions imaged in seismic data might indicate an increase in amount of intrusions with depth. Both in outcrop (Fig. 3) and in similar basins around the North Atlantic (e.g. Faroe-Shetland Basin; Mark et al. 2018), mafic sill intrusions preferentially exploit mudstone beds. This implies that mud-dominated lithologies occur at many levels throughout the basin fill.
Sills observed in the seismic data are generally flat-lying and concordant to stratigraphy in the upper 4 km of the basin, and in the central parts of the basin at great depth (4-6 s twt; c. 9-15 km). Towards the basin margins, sill reflectors are oblique with regard to stratigraphy, and transgress upwards away from two distinct zones at the base of the basin, one area in the north and one area in the south of the basin (Figs 4, 5a and 6). This indicates that there are two main sill-complexes in the Jameson Land Basin, and that these have two distinct feeder zones at the base of the basin (Fig. 7). We term these the Southern and Northern Sill-complexes.
For the Southern Sill-complex, the seismic data show that the sills occur as oblique sheets or stepped transgressive sills towards the basin margins. When these oblique reflectors are extrapolated towards their origin in the SW, they appear to originate from a linear feeder zone in the centre of the basin (Fig. 4). This area is where the underlying crust has been stretched the most, and also where a lower-crustal high-velocity body both terminates and impinges on this thinned crust (Fig. 1b). These critical observation lead us to interpret that the sills were fed from this lower-crustal high velocity body, and that the magma moved from the zone of melting in the mantle, through highly stretched crystalline crust, and entered the basin along a linear north-south-oriented feeder zone.
For the Northern Sill-complex, fewer seismic lines cross the origin of the sills, and imaging is poorer at depth compared with other parts of the basin. The orientation and shape of the northern magma feeder zone can therefore not be determined in the present dataset.
A 3D view of interpreted sills (Fig. 7) shows that the sills occur as two main shapes. The first type is sets of nested bowls sourced from either of the two distinct magma feeders in the northern and southern parts of the basin. These sills have flat centres and show gradual transgression upwards through the basin towards the basin margins. These nested bowls can be up to 15 km deep and have a radius of 60 km, making them among the largest sills described globally (see Wrona et al. 2019). The second type of sills are flatter, up to 50 km long, follow stratigraphy to a greater degree, but also often transgress upwards towards the closest basin margins, and commonly occur in the centre of the basin and in the shallower parts (upper 4 km) of the basin (Fig. 6). This indicates that the sills follow the same stratigraphic intervals for great distances in the centre of the basin and towards the upper few kilometres in the basin, and that the sill segments that are deep and/or occur towards the margins of the basin have a greater tendency to transgress upwards through stratigraphy. This agrees with the upwards-towards-the-north stepping observed in the outcrop (Fig. 3a). Towards the SE margin of the study area, sills show a tendency to maintain their oblique trajectory even when intruding through tilted post-rift sediments and even basement. This important observation will be discussed below.  Fig. 2). A small number of these dykes cross-cut the sills, and this population of cross-cutting dykes is clearly younger than the sills and is believed to have been emplaced during the Oligocene plate-tectonic reorganization in the North Atlantic (Hald and Tegner 2000). However, the vast majority of observed dykes are interconnected with, and geochemically similar to, the sills and are believed to be emplaced at the same time as the sills (Hald and Tegner 2000;Eide et al. 2017). The dykes are 1-10 m wide, and the median width is 2.3 m (see Supplementary material DR2 for more statistics). More dykes are observed in the virtual outcrop, where the exposure is excellent (Fig. 3a), than on the plateau (Fig. 2). Many of the larger dykes observed in the virtual outcrop can be followed onto the plateau for several kilometres (Fig. 3h), showing that the dykes observed in the outcrop correspond to dykes mapped in the terrain.
From analysis of the virtual outcrop data, about 10% of the outcrop consists of intrusive rocks. Of the intrusive material, 1% consists of dykes and 99% of sills (by area of outcrop in Fig. 3a; Eide et al. 2017; see also Supplementary material DR2). The dykes in the study area are consistently oriented east-west to ESE-WNW, and no observed dykes in the Jameson Land Basin are oriented in the north-south direction (Fig. 2b). This indicates that the magma pressure was great enough to form both sills and dykes at the same time, and that the east-west compressive stress was greater than the north-south compressive stress, which was then greater than the vertical compressive stress.
The dykes are most common in an area towards the south of the study area, and we term this the Southern Dyke Swarm (Fig. 2a). An area with elevated dyke density also occurs in the northern part of the basin, and we term this the Northern Dyke Swarm (Fig. 2a). The Southern Dyke Swarm occurs at the eastern edge of the lowercrustal high-velocity body (Figs 1b and 2), and appears to radiate away from the origin of the Southern Sill-complex in the basin (i.e. as the mapped dykes are extrapolated towards the west, they intersect the interpreted origin of the Southern Sill-complex). The dykes in the Northern Dyke Swarm have a similar east-west to ESE-WNW orientation to the Southern Dyke Swarm, and this dyke swarm occurs in the middle between the Southern and Northern Sill-complexes, without any clear relation to any mapped geological feature.

Relationship between sill-complexes and dyke swarms in the Jameson Land Basin
Vertical and steeply dipping features such as dykes are difficult to image in seismic reflection data (e.g. Lecomte et al. 2016), and their presence in the subsurface may be under-represented. However, dykes can in some instances be observed in seismic data nonetheless because (1) they may create discontinuities in otherwise tabular layers, (2) they may have irregularities and thin associated sill intrusions that can be imaged, or (3) they deform the host-rock in ways that can be imaged (e.g. Wall et al. 2010;Eide et al. 2018a;Minakov et al. 2018;Magee and Jackson 2020). Because dykes in the Jameson Land Basin have been well mapped throughout the available outcrop data (Figs 2 and 3h), it is possible to investigate their expression in seismic data by carefully studying the seismic data beneath and away from zones where dykes are mapped.
In seismic lines crossing the Southern Dyke Swarm, areas with several mapped dykes correspond to subvertical zones in the seismic data with a larger amount of noise and lower degree of reflector continuity (Fig. 8a). We interpret that the poor imaging in these zones is related to the presence of dykes in these locations, and can assume a degree of understanding of the dyke distribution and architecture using the surface examples as a guide. These subvertical disturbance zones appear to originate at the edge of dipping intrusive sheets (1 in Fig. 8a) and the edges of relatively flatlying sills (2 in Fig. 8a). Thus, it appears that the Southern Dyke Swarm is fed from transgressive segments from large sills at great depth (10-15 km) in the basin. This again also implies that sills have a very important role in magma transport in the crust (Cartwright and Møller Hansen 2006;Muirhead et al. 2014;Schofield et al. 2015).
In seismic data from the Northern Sill-complex (Fig. 8b), no clear subvertical zones of disturbance can be observed in the areas where dykes are mapped, and the imaging is also poorer at depth (>3 s twt) in general, which may preclude observations of dykes. No correspondence between dykes and sills can therefore be demonstrated in this part of the basin.

Vertical connectivity of sills
The sills observed in the outcrop along Scoresby Sund (Fig. 3a) are locally transgressive, and do not show cross-cutting relationships but are rather interconnected with each other (Eide et al. 2017). In the subsurface, sills are also transgressive, strongly interconnected and appear to originate mainly from other sills and from short transgressive segments (steep oblique sheets, dykes) from underlying sills (Fig. 9). This shows that significant magma transport through the crust can occur through interconnected sill networks rather than through dykes or dyke swarms alone. Although dykes may be underestimated in the subsurface seismic data, the low proportion of dykes compared with sills in the outcrop dataset offers an indication that dykes are in fact much less dominant in the basin than sills.

Thickness of sills
The sills in the studied outcrops along Hurry Inlet are 1-10 m thick ( Fig. 3a; Eide et al. 2017). Geochemically similar sills up to 50 m thick have been reported from Ørsted Dal in the northern part of the basin (Hald and Tegner 2000; Fig. 2). In the seismic data, the vast majority of interpreted sills are reflections that are 'tuned' (e.g. Widess 1973), meaning that they come from high-impedance bodies that are so thin that reflections from top and bottom interfere with each other and their thickness cannot be determined (e.g. Fig. 9c). The tuning thickness varies from c. 25 m just below the present-day surface to 80 m at 6 s twt (c. 15 km), using velocities from Mandler and Jokat (1998) and frequencies measured from the seismic reflection dataset. This means that most of the sills are thinner than 25 m in the upper part of the basin, and thinner than 80 m in the deepest part of the basin.
A small number of interpreted intrusions show a positive amplitude on top and a negative reflection at the base, and the thickest of these has an apparent thickness of 30 ms, which would correspond to a thickness of 100 m (Fig. 9a). However, based on the complex geometries of these reflections, it is possible that these reflections represent several closely spaced thinner intrusions rather than a single 100 m thick intrusion (see Eide et al. 2018b). Sills throughout the North Atlantic Igneous Province commonly range in thickness from 10 to c. 100 m thick (e.g. Hansen et al. 2011;Schofield et al. 2016;Fyfe et al. 2021). Known sill thickness within other large igneous provinces can be up to c. 300 m (e.g. Jerram et al. 2010;Svensen Henrik et al. 2018), and this was earlier believed to be the case also within the deeper parts of Jameson Land (Larsen and Marcussen 1992). However, within the current dataset we have not seen evidence of sills as thick as 300 m, as interpreted earlier by Larsen and Marcussen (1992). This highlights the importance of data quality in subsurface estimation of sill thicknesses (Planke et al. 2015).

Magma transport through the crust and sedimentary basins
Based on the observations made from the Jameson Land Basin in East Greenland, we present a model for magma transport from the mantle, through the thinned crystalline crust and through the Jameson Land Basin (Fig. 10). We suggest that this model is relevant for understanding transport of magma from the mantle to the surface also in other parts of the world where predominantly mafic magma is emplaced into thick sedimentary basins, and that transport of mafic magma through sedimentary basins is distinctively different from magmatism that transits through crystalline crust. This is mainly due to the low viscosity of basaltic magma, the weak and stratified nature of sedimentary host rocks (see the section  Figure 2. (a) In the Southern Dyke Swarm, areas with several dykes correspond to near-vertical zones of more noise on the seismic images. These zones appear to have origins at the tips of (1) high-and (2) low-angle sills at great depths in the basin, indicating that dykes are mainly sourced from transgressive edges of sills deep in the basin. (b) The seismic lines from the Northern Dyke Swarm show in general few vertical features that can be attributed to the dykes mapped in the surface, and no particular architecture of sills that can be related to the dykes. It is unclear whether this is real or if it is an effect of poorer data quality here. (c) Map showing location of seismic lines and dyke swarms. Legend is shown in Figure 2.
'Why do sills follow mudstones'), and the progressive changes in mechanical properties of sedimentary host-rock with depth.

Transport of magma into a sedimentary basin
Seismic refraction data from the Jameson Land have identified the presence of a lower-crustal high-velocity body at the base of the crust c. 20 km below the present-day surface of the Jameson Land Basin ( Fig. 1b; Weigel et al. 1995;Mandler and Jokat 1998). This body is inclined and its highest point lies directly beneath the base of the Jameson Land Basin, where the continental crust is highly thinned. The seismic data image sills down to the base of the sedimentary basin c. 15 km below the present-day surface, and the architecture of sills in the south of the basin shows that they are supplied from a north-south-trending zone in the centre of the basin (Figs 4 and 7). This strongly suggests that the mafic intrusive rocks in the basin were supplied from magma chambers below the crust represented by the lower-crustal high-velocity body (Fig. 10). This lower-crustal body and other examples of such high-velocity bodies along the East Greenland margin and on the conjugate mid-Norway margin are interpreted to result from voluminous magmatic underplating (e.g. Voss and Jokat 2007;Neumann et al. 2013). Such magmatic underplating is commonly thought to result from the addition of mafic magma to the lower crust and uppermost mantle around the Moho (see Thybo and Artemieva 2013), and is likely to include the development of extensive magma bodies, where  (1) the considerable vertical connectivity of sills, indicating that dykes are not necessary to transport large amounts of magma through the crust, and (2) the absence of thick or outsized sills, indicating that magma transport was through a complicated sill network, and that any crustal magma chambers were absent. (c) Explanation of the tuning effect. (See Fig. 2 for location.) fractionation processes can result in more evolved magmas than the primary high-temperature mantle melts. The evolved basaltic melts we find in the Jameson Land Basin (with 5-7 wt% MgO; Hald and Tegner 2000) are likely to be the result of such underplating processes (e.g. Cox 1993), with the resulting melts transiting the crust where the crystalline basement is thinnest and weakest. The architecture of intrusive rocks in the area where the sills transit through the crystalline crust is not constrained by seismic data, but the narrowness of the magmatic feeder zone and the lack of any perceptible offset between the feeder zone and the top of the magmatic underplate suggests that magma transport through the crystalline crust must have been through localized, near-vertical dykes along the basin axis (Fig. 10b). As magma transport through the sill-dominated intrusive network in the sedimentary basin is much more distributed than through this narrow zone in the crystalline crust, magma supply through this zone must have been very large per unit volume compared with that in the sedimentary basin.

Sills and volcanoes significantly offset from magma sources and main feeder zones
The deeper sills fed from the magma feeder zone and emplaced in the lower parts of the Jameson Land Basin are bowl-shaped and (b) Compilation of key observations of the relationship between magmatic plumbing networks, their host-rocks and regional tectonics made in the Jameson Land study area. 1, Accumulation of residual melt from localized mantle melting. 2, Transit of magma through narrow zone of highly stretched crystalline crust above where mantle melting impinges on crust. 3, Formation of deep-seated intrusions that are flatter in the basin centre and transgress obliquely towards basin margins. 4, Magma transport in basin mainly through interconnected sills. A small proportion of intrusive material occurs as dykes sourced from deep sills. 5 and 6, sills in upper 5 km of basin mainly strata-concordant (5), but highly transgressive towards basin margins (6). Sills occasionally exploit faults. 7, Formation of relatively isolated dyke swarms close to magma sources and in areas where sill-complexes interact. radiate outwards and upwards away from the feeder zone (Figs 4 and  7). Sills observed in outcrops today originated from magmas fed from at least 50 km away (Fig. 2), and this implies that the majority of sills and possibly volcanoes in similar settings may be offset by many tens of kilometres away from where the magma was generated in the upper mantle and injected into the basin (Fig. 10). This supports earlier work on lateral magma transport in sedimentary basins (e.g. Magee et al. 2016). The seismic data presented here clearly show two geographically distinct magma feeder zones for the Southern and Northern Sill-complexes (Fig. 7). Although the geochemistry of igneous intrusive rocks in the Jameson Land Basin has been studied (Hald and Tegner 2000), there is no evidence yet of distinct geochemical signatures of magma that correspond to these two feeder zones identified in the seismic data. Investigating geochemical and geochronological data from this area with the aim of characterizing these two complexes would be highly interesting, particularly to see if they come from a similar or different source region.

Lack of magma chambers and volcanic centres within the basin
Geophysical and geochemical data show that many large modern volcanoes are underlain by complex zones of low seismic velocities that are interpreted as magma chambers (e.g. Jaxybulatov et al. 2014;Huang et al. 2015). Such magma chambers could be analogous to plutons and laccoliths in ancient systems, but plutons and laccoliths appear to be more common in silicic and andesitic systems (Jerram and Bryan 2015). Although mafic laccoliths and other large-scale intrusive mafic bodies occur in sedimentary basins and upper crust (e.g. Jackson and Pollard 1988;Monreal et al. 2009;Holness et al. 2017;Walker et al. 2021), they are smaller in volume and extent and less common compared with sills in large sedimentary basins studied in large outcrops and extensive seismic datasets (e.g. Eide et al. 2016;Polteau et al. 2016;Coetzee and Kisters 2017;Reynolds et al. 2017;Schofield et al. 2017;Svensen Henrik et al. 2018;Gilmullina et al. 2021). For example, the volcanic centre of Rum in Scotland is c. 200 km 2 in extent, and its associated Little Minch Sill-complex covers an area 20 times larger (c. 4000 km 2 ; Fyfe et al. 2021).
Volcanic centres associated with large igneous provinces can be large, and are commonly characterized by multiple pulsed magmatic events and transitions from mafic to silicic compositions (Jerram and Bryan 2015). In some offshore sequences along volcanic rifted margins such complexes are recognized through complex seismic zones, often with significant gravity and magnetic anomalies, and highlight the location of candidate volcanic centres within the magmatic plumbing systems (e.g. Emeleus and Bell 2005;Kilhams et al. 2021;Walker et al. 2021). Many large isolated centres are known within the North Atlantic Igneous Province along the British Paleogene sections (e.g. Emeleus and Bell 2005;Schofield et al. 2017), and some centres have also been noted recently along the Norwegian Margin (e.g. Kilhams et al. 2021). Many of these volcanic centres are known to be associated with long-lived, pre-existing, crustal-scale faults and lineaments (e.g. Fyfe et al. 2021). In Central East Greenland, there are also several large igneous centres, such as the Werner Bjerge Complex at the NW margin of the Jameson Land Basin (Fig. 2a), but these are also associated with magma intrusion along basin margins and crustalscale faults, and these are associated with younger and more felsic magmas (Brooks 2011).
In contrast, the Jameson Land Basin itself is dominated by sills that are tens of kilometres long and wide, and ( probably much) less than 80 m thick. No candidates for large-scale magma chambers are observed within the Jameson Land Basin. Further along this same margin, the Danmarkshavn and Tethis basins within offshore East Greenland also show extensive sill-complexes with, as yet, no clear igneous centres identified (Reynolds et al. 2017). This, combined with observations from other sedimentary basins described above, suggests that such large-scale crustal mafic magma chambers are somewhat limited compared with sill-complexes within thick sedimentary sequences, and that there are large-scale geological controls on the development of sill-complexes compared with localized igneous centres: mafic magmas are likely to create distributed sill-complexes within sedimentary basins owing to the low viscosity of basaltic magma (e.g. Shaw et al. 1968;Wada 1994) and the weak and layered nature of host-rocks. Large-scale subsurface mafic magma chambers may be more likely to form when large amounts of magma are supplied along structural and possibly major lithological discontinuities. If such magma chambers are present at the bases of sedimentary basins near structural entry-points through the underlying crystalline crust (such as the feeder zone below the Jameson Land Basin; Fig. 10), it would indeed be difficult to detect them in seismic data as these would be located under thick sedimentary piles and close to basement rocks with little seismic impedance contrast.
Using simple assumptions for the shape of the Southern Sillcomplex and the individual sills within it (the sill-complex is symmetrical across the north-south axis, the sill-complex can be approximated as a cone, the proportion of sills at depth is comparable with what is observed in the outcrop dataset in Fig. 3a), we calculate the volume of magma in the Southern Sill-complex to be between 110 and 5600 km 3 (see Supplementary material DR3 for details). This large uncertainty is due to the uncertainty in estimating thickness of sills at depth. This calculated magma volume in the Southern Sillcomplex is 4-20 times the volume of the well-known Skaergaard Intrusion (Nielsen 2004). This shows that the volume of sillcomplexes can be significant and that in sedimentary basins, sills may make up much more volume than igneous centres.

Relation of studied igneous intrusions to regional tectonics
Paleogene dykes are widespread across most of the Jameson Land Basin (Fig. 2a), and occur as two separate dyke swarms, one in the north and one in south of the basin (Fig. 2a). The dykes are oriented east-west to ENE-WSW, and no dykes occur in the north-south direction (Fig. 2b). In outcrop data, 100 times more magma has been emplaced as sills compared with dykes (by area of outcrop), and the vertical extension by sills is 100 greater than the lateral extension by dykes (Fig. 3a). Based on the large amounts of sills imaged in seismic data at all basin depths, we assume that these relationships also hold throughout the basin. This implies that the stress-state at time of emplacement had the axis of least compressive stress (σ 3 ) oriented vertically (to allow for a large amount of sills to be emplaced), intermediate compressive stress (σ 2 ) oriented northsouth (to allow for a smaller amount of dykes to open in the east-west direction), maximum compressive stress (σ 1 ) oriented east-west (to prevent north-south dykes from opening), and magma pressure comparable with σ 2 (to account for the smaller amounts of intrusive rocks in east-west dykes).
It is likely that the interpreted east-west-oriented maximum compressive stress at time of emplacement was due to ridge-push from the AEgir spreading ridge in the Norwegian Sea, which started spreading at c. 55 Ma (Fig. 10a; e.g. Blischke et al. 2017), 2 myr before the age of emplacement for the Jameson Land intrusive rocks (Hald and Tegner 2000). The 53 Ma igneous complex has been interpreted by Hald and Tegner (2000) to represent rifting associated with a failed attempt at a westwards ridge jump, similar to the ridge jump that occurred during the late Eocene-Oligocene and rifted off the Jan Mayen microcontinent (Talwani and Eldholm 1977;Mjelde et al. 2008;Gaina et al. 2009). However, the dyke orientations in Jameson Land are not consistent with rifting in the east-west direction. The exact mechanism that led to the emplacement of the sill-complexes in Jameson Land is therefore unclear. Only one sill and one dyke have been dated hitherto using Ar-Ar and palaeomagnetism (Hald and Tegner 2000), and clearly a more robust and comprehensive age dataset (ideally with more precise U-Pb dates if possible) would be beneficial to improve the understanding of this intrusive complex and its relation to the regional tectonics.
The great distance (c. 150 km) between the magmatic underplate below Jameson Land and the continent-ocean transition at time of breakup is unusual (Voss and Jokat 2007;Voss et al. 2009). The crustal thinning below the Jameson Land Basin was mainly set up through post-Caledonian to Triassic extension (Fig. 10a), and this probably led to weakening of the crust that led to the capture of magma in the Jameson Land Basin. This is thus an example of deepseated basin structure influencing magma input a long time after the crustal structure was made.

Why do sills follow mudstones?
Sills observed in the outcrop dataset from Jameson Land are preferentially emplaced within, and propagated along, mudstones (Fig. 3). Similar trends are also seen in other areas (e.g. Parsons et al. 2017;Spacapan et al. 2017;Mark et al. 2018), and here we offer an explanation of why this happens.
Brittle host-rocks in contact with an inflating magma-filled conduit will rupture and an intrusion will propagate, when the following condition is satisfied (Gudmundsson 1990): where p l is the lithostatic pressure, p e is the magma overpressure with regard to lithostatic pressure at time of failure, σ 3 is the least compressive stress or maximum tensile stress, and T 0 is the in situ tensile strength at the failure location. The stress-state has often been discussed in studies of igneous intrusions (e.g. Barnett and Gudmundsson 2014), whereas the tensile strength of host-rock has commonly received much less attention. Tensile stress is often regarded as a property that varies relatively little (0.5-6 MPa), perhaps because these models were originally developed for basaltic lava piles (Gudmundsson 1990(Gudmundsson , 2006. In sedimentary basins, lithologies are variable and mudstones are common both as thick (metres to kilometres thick) layers and as millimetre-to centimetre-scale drapes (Fig. 3). Clay-rich mudstones, and in particular mudstones with a well-developed fissility ('shales'), split easily along bedding planes (Fig. 11). The tensile strength (in MPa) of mudstones has been measured to be c. 300 times lower parallel to the bedding plane compared with perpendicular to the bedding plane (Gao et al. 2015). Many mudstones thus have a strong tensile stress anisotropy oriented along the bedding planes. It follows from equation (1) that if tensile strength is small enough, fractures can develop in directions that are not parallel to σ 3 . A high tensile stress anisotropy can explain why mudstones are so easily intruded, and why nearly all dyke-to-silltransitions in the Hurry Inlet outcrop occur within mudstones ( Fig. 3a; Eide et al. 2017). It is also worth noting that σ 3 does not have to be determined by far-field tectonic stress alone, but can be local, as significant local stresses occur in and around propagating sill-complexes (e.g. Schofield et al. 2012;Poppe et al. 2020).

Description of overall sill architecture
On the scale of kilometres, sills mapped in the Hurry Inlet outcrop show a prominent upwards-towards-the-north stepping through the stratigraphy at an angle of c. 6°(green arrows in Fig. 3a). A similar upwards-towards-basin-margins stepping of sills is observed deep in the basin in seismic data (orange arrows in Figs 4, 5 and 7). In most of the basin, particular in the basin centre and in the upper post-rift deposits, the sills are strongly concordant to host-rock layering. In other places, particularly towards basin margins, oblique intrusive sheets cross-cut flat-lying bedding, tilted synrift bedding and even crystalline basement without changing their attitude (Fig. 5a).
Although sills often appear as oblique sheets in the seismic data, it is likely that many of the sills at depth are in fact stepped and segmented rather than truly oblique (compare Fig. 3b and c), as no oblique sheets are observed in the outcrops. The transgressive segments of the sills in the outcrop dataset (Fig. 3a) are not controlled by any apparent host-rock features such as faults, fractures or discontinuous sedimentary bodies. The host-rock in the outcrop is laterally homogeneous apart from a gradual increase in sandstone content towards the north, and it is generally unfaulted and has few fractures (Eide et al. 2016). We therefore believe that the organized nature of transgressive sill segments and upwards stepping of sills, from injection centres in the deep basin towards the surface at basin margins, is mainly controlled by stresses in the basin interacting with the host-rock and local stresses from the propagating sill intrusions.

Controls on metre-to kilometre-scale sill architecture
The lateral variation in lithology in the outcrop (Fig. 3a) allows us to investigate how changing host-rock properties influence architecture of sills. The proportion of sandstone decreases towards the south in the sand-rich stratigraphic units ( Fig. 3a; Eide et al. 2016), leading to a high-contrast succession in the north, with layers with strong difference in sandstone content, and a low-contrast succession in the southern part of the outcrop, where there is little difference in sand content in the stratigraphy (Fig. 12). In the southern parts of the outcrop, with mud-rich deposits throughout, sills consist of c. 200 m wide segments that gradually transgress upwards and are linked by vertical jogs 1-10 m high (Fig. 12). In the middle of the outcrop, with intermediate sandstone content, sills segments are c. 1 km long and separated by c. 25-35 m upward-tothe-north jogs. In the high-contrast part in the north, sills follow mudstone intervals for much longer distances, up to 6 km, and transgress out of the sand-rich intervals over short distances. Sill segments here are joined by a single, c. 150 m high jog.
The observations above have a set of important implications (Figs 12 and 13), as follows.
(1) On the outcrop scale (<200 m), the intrusions in Jameson Land predominantly follow bed boundaries and especially mudstones (Fig. 3). This implies strong host-rock control on sill architecture. Any sill transgressions observed in smaller outcrops (less than a few hundred metres) could easily be interpreted to be random, and large-scale stress-related controls on sill intrusions are likely to be underestimated from such smaller outcrops.
(2) On the large scale (>5 km), sills rather follow trajectories concordant with the overall structure of the sills in the basin, upwards towards the basin margins (Fig. 13). This implies that the regional stress field has a strong control on sill intrusions at larger scales (Stephens et al. 2017;Walker et al. 2017;Walker and Gill 2020), that the stress controls are largest towards the basin margins, and that stress controls diminish towards the basin centre. It is possible that the basin fill is more sandstone rich near the basin margins, because of proximity to the clastic sediment source (Fig. 13). This further promotes stress-related controls on intrusion architecture near basin margins, as weak mudstones are rarer close to basin margins owing to increased sand supply.
(3) On intermediate scales (0.5-5 km), the controls on sill architecture are more variable. In areas of low host-rock contrast, sills consist of c. 200 m long segments and show regular steps upwards through stratigraphy, leading to a stepped oblique sill that goes upwards at c. 6°. In areas of high host-rock contrast, sills occur in regional mudstones and transgress, but the segment lengths and step heights are much greater (6 km and 150 m respectively; Fig. 12).

Depth control on large-scale sill morphology
On even larger scales, sill geometries in the Jameson Land Basin change with depth. From the surface down to c. 10 km depth in the Jameson Land Basin, sills appear to be mainly layer-concordant or obliquely transgressive, and appear to have been emplaced and accommodated mainly by brittle processes. This present-day depth interval corresponds to depths of 3-13 km at time of emplacement. In brittle, cemented host rocks, the lithological controls on sill emplacement are strong, and sills generally follow weak layers in host-rock stratigraphy instead of developing saucer shapes (e.g. Schofield et al. 2012;Eide et al. 2017), which are common in uncemented, shallowly buried sediments (Fig. 13).
In the deep (>10 km) parts of the basin (corresponding to depths of more than 13 km at time of emplacement) the sills form large bowl-shaped geometries (Figs 6 and 7). These bowl shapes are similar to shallow saucers, but they have much larger diameters: a few tens of metres for shallow saucers (Polteau et al. 2008) compared with up to 50 km for the deep sills in the Jameson Land Basin. The brittle-ductile transition zone generally occurs around a temperature of about 300°C (e.g. Aharonov and Scholz 2019). Using a palaeo-geothermal gradient in the Jameson Land Basin of 30°C km −1 (Green and Japsen 2018), the brittle-ductile transition would have been at c. 10 km depth at time of emplacement of the Fig. 12. Conceptual diagram showing the interplay between effects of regional stress field and host-rock heterogeneity on sill-complex architectures in areas where sill emplacement is also controlled by stress. In the left part, vertical host-rock heterogeneity is low and stepwise sills transgressions occur at short (c. 200 m) intervals. In the right part, where vertical host-rock heterogeneity is high, sills make much greater (up to 3 km) deviations from the directions set up by the regional stress field. sills, which fits well with the observed gross changes in sill geometries with depth (Fig. 13). It is therefore likely that the large bowl-shaped geometries are related to a change in emplacement mechanism of magma within partly ductile rocks. Polteau et al. (2008) showed that the diameter of saucer-shaped sills increases with depth, something that fits well with the very large diameter of the deep-seated sills in Jameson Land. Thus, it seems that saucershaped sills are common in host-rocks that may behave in a nonbrittle manner (ductile rocks and uncemented sediments), but that development of saucer-shaped sills is inhibited in the depth interval where clastic sediments are commonly cemented and brittle from c. 2.5 to 10 km depth. Different emplacement mechanisms and architectures below the brittle-ductile transition zone have been reported before from igneous dykes (Kjøll et al. 2019).

Conclusions
The Jameson Land Basin is a unique place to study the evolution of mafic sill-complexes on the basin scale because of the combination of c. 3 km of erosion that exposes intrusive elements in excellent exposures, deep seismic reflection lines and abundant previous studies on stratigraphy, geochemistry and crustal structure. The following are our main findings from investigating these datasets together.
(1) At least one of the two sill-complexes in the Jameson Land Basin is fed from where a lower-crustal high-velocity body impinges on highly stretched crystalline crust below the basin.
(2) Sills propagate upwards and outwards into the basin, away from linear magma feeder zones at the base of the basin. Sill-complexes are shaped as nested bowls that have a maximum radius of c. 60 km and depth of 15 km. (3) On scales from 0.2 to 6 km, the architecture of sill-complexes is strongly controlled by host-rock lithology, and sills generally follow the sedimentary layering. On scales from centimetres to 0.2 km, sill architecture is mainly determined by weak mudstone beds, which the sills follow. The sills follow mudstones because of the strong tensile strength anisotropy of mudstone, which makes it possible to open fractures in mudstones regardless of the tectonic stress state. (4) Sills are generally layer-concordant in the basin centre and transgress obliquely towards basin margins. The main reason for this is probably the distribution of stresses in the basin and the higher sandstone content near the basin margin that yields fewer weak mudstone beds for possible intrusion. (5) Sills that are emplaced below the brittle-ductile transition zone show a change in geometry to larger bowl-shaped morphologies. (6) Any magma chambers in the form of laccoliths or stocks are not observed in the Jameson Land Basin. It appears likely that magma transport mainly through interconnected sills is common for mafic systems within sedimentary basins, whereas large intrusive magmatic bodies are associated with crustal-scale faults through crystalline rocks. (7) The WNW-ESE orientation of igneous dykes makes us doubt the existing geodynamic interpretations for the origin of the studied sills in Jameson Land, where they were emplaced because of east-west-directed rifting associated with a failed early attempt at a westwards ridge jump in the area.
In summary, these observations and ideas highlight that mafic igneous systems that intrude thick sedimentary basins are different from traditional models of magma transport through the crust. In particular, the Jameson Land Basin is characterized by a lack of crustal magma chambers, magma transport largely through interconnected sills, and strong host-rock and strong stress-related controls on sill-complex architecture. Furthermore, this study shows that there is great potential to generate further understanding of basin-scale magma transport through integrated, targeted geochemical and geochronological studies of this well-exposed sill-complex. C. Tegner for thorough reviews that significantly improved this paper. We acknowledge Geological Survey of Denmark and Greenland (GEUS) for access to, and permission to publish images of, onshore seismic data acquired by Atlantic Richfield Company (ARCO). Schlumberger is acknowledged for an academic licence for Petrel, which was used for seismic interpretation and visualization. G. Henstra, B. Nyberg and S. Buckley are thanked for assistance with fieldwork and data processing. The virtual outcrop was visualized and interpreted using LIME (http://virtualoutcrop.com/lime; Buckley et al. 2019).